A multi-agents and occupancy based strategy for energy management and process control on the room-level Labeodan, T.M.

Transcription

1 A multi-agents and occupancy based strategy for energy management and process control on the room-level Labeodan, T.M. Published: 28/06/2017 Document Version Publisher s PDF, also known as Version of Record (includes final page, issue and volume numbers) Please check the document version of this publication: A submitted manuscript is the author's version of the article upon submission and before peer-review. There can be important differences between the submitted version and the official published version of record. People interested in the research are advised to contact the author for the final version of the publication, or visit the DOI to the publisher's website. The final author version and the galley proof are versions of the publication after peer review. The final published version features the final layout of the paper including the volume, issue and page numbers. Link to publication Citation for published version (APA): Labeodan, T. M. (2017). A multi-agents and occupancy based strategy for energy management and process control on the room-level Eindhoven: Technische Universiteit Eindhoven General rights Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights. Users may download and print one copy of any publication from the public portal for the purpose of private study or research. You may not further distribute the material or use it for any profit-making activity or commercial gain You may freely distribute the URL identifying the publication in the public portal? Take down policy If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim. Download date: 08. Oct. 2018

2 A Multi-agents and occupancy based strategy for energy management and process control on the room-level Timilehin Moses Labeodan / Department of the Built Environment bouwstenen 229 finalcover-mh3.indd :56:51

3 A multi-agents and occupancy based strategy for energy management and process control on the room-level PROEFSCHRIFT ter verkrijging van de graad van doctor aan de Technische Universiteit Eindhoven, op gezag van de rector magnificus prof.dr.ir. F.P.T. Baaijens, voor een commissie aangewezen door het College voor Promoties, in het openbaar te verdedigen op woensdag 28 juni 2017 om 16:00 uur door Timilehin Moses Labeodan geboren te Ilesha, Nigeria

4 Dit proefschrift is goedgekeurd door de promotoren en de samenstelling van de promotiecommissie is als volgt: voorzitter: Prof. dr. ir. B. de Vries 1 e promotor: Prof. ir. W. Zeiler 2 e promotor: Prof. dr. -Ing. A.L.P. Rosemann leden: Prof. dr. ir. R. Yao (University of Reading) Prof. dr. -Ing. D. Müller (RWTH AACHEN University) Prof. ir. P.G. Luscuere (Technische Universiteit Delft) Prof. dr. ir. G.P.J. Verbong Prof. dr. I.G. Kamphuis Het onderzoek of ontwerp dat in dit proefschrift wordt beschreven is uitgevoerd in overeenstemming met de TU/e Gedragscode Wetenschapsbeoefening.

5 A multi-agents and occupancy based strategy for energy management and process control on the room-level Timilehin Moses Labeodan

6 This research was funded by the Province of Noord Brabant, the Netherlands Cover photo - Bagiuiani Constantin Dreamstime.com A catalogue record is available from the Eindhoven University of Technology Library ISBN: NUR: 955 Published as issue 229 in the Bouwstenen series of the Department of the Built environment of the Eindhoven University of Technology. Copyright T Labeodan 2017 All rights reserved. No part of this publication may be reproduced or transmitted in any form or by any means, electronic, mechanical, including photocopy, recording, or any information storage and retrieval system, without prior written permission of the owner.

7 Acknowledgment To Chief Emmanuel, my lovely wife Wuraola, and my parents.

8 SUMMARY Energy consumption in buildings accounts for up to 40% of total energy consumption. In an effort to improve energy and environmental sustainability, buildings have been the target of a number of energy demand reduction strategies. Occupant behaviour in buildings is a key factor that significantly influences the energy consumption of buildings. Whilst researchers have focused considerable effort into understanding and developing strategies that address the impact of occupant behaviour on building energy consumption, the high-energy demand of buildings is considered a useful source of demand-side energy flexibility for the smart-grid. In addressing the influence of occupant behaviour in buildings and in harnessing the energy flexibility potentials of buildings, building energy management systems play a vital role. Traditional building energy management systems are however static and lack input of dynamic factors such as occupancy as well as occupant preference. In addition, traditional building energy management systems are limited in their ability to optimally coordinate the use of buildings energy flexibility without compromising occupant comfort on the room-level. This research thus aims to explore the use of increased sensor data and computational support for enhancement of energy management and energy flexibility in buildings. Specifically, the research focuses on the application of fine-grained information on building occupancy as well as multi-agent systems in enhancing the performance of traditional building energy management systems in office buildings. The research includes a thorough survey of existing systems in use in buildings for collection of occupancy information as well as a survey on multi-agent systems application in buildings for energy management and coordination of building s energy flexibility for grid support services. Field experiments were conducted in an office building to evaluate alternative methods for obtaining fine-grained occupancy information that enhance energy management. Furthermore, a multi-agent system coordination framework was developed and assessed in the test-bed office building for coordination of occupant behaviour on the room-level and the building s energy flexibility. The results from the research indicate that the availability and utilization of finegrained building occupancy information particularly on the building room-level can facilitate worthwhile energy savings. In addition, it also indicates that the use of multiagent systems in buildings can enhance coordination of occupant behaviour to achieve improved energy management and facilitation of buildings energy flexibility for grid support while maintaining comfort.

9 TABLE OF CONTENTS TABLE OF CONTENTS LIST OF FIGURES... I LIST OF TABLES... III LIST OF ABBREVIATIONS... IV BUILDINGS, THE SMART GRID AND, ENERGY AND ENVIRONMENTAL SUSTAINABILITY... 2 BUILDING ENERGY MANAGEMENT SYSTEM & MULTI-AGENT SYSTEM... 7 SCOPE, CONTEXT AND CONTRIBUTION OF THIS RESEARCH RESEARCH METHODOLOGY THESIS OUTLINE PUBLICATION OVERVIEW OCCUPANCY INFORMATION IN BUILDING OPERATION OCCUPANCY DETECTION SYSTEMS DISCUSSION MULTI-AGENT APPLICATION IN BUILDING MANAGEMENT DISCUSSION OCCUPANCY DETECTION SYSTEM DESIGN AND EVALUATION EXPERIMENT I- CHAIR DETECTION SYSTEM DESIGN AND EVALUATION RESULTS AND DISCUSSION EXPERIMENT SUMMARY EXPERIMENT II- EVALUATION OF DIFFERENT CHAIR SENSING TECHNOLOGIES RESULTS AND DISCUSSION EXPERIMENT SUMMARY EXPERIMENT III- SENSOR FUSION APPLICATION IN ENERGY MANAGEMENT RESULTS AND DISCUSSION EXPERIMENT SUMMARY MAS APPLICATION AND AGENT DESIGN FRAMEWORK EXPERIMENT I - MAS COORDINATED OCCUPANCY BASED LIGHTING CONTROL RESULTS AND DISCUSSION EXPERIMENT SUMMARY EXPERIMENT II: MOTION DETECTION SYSTEM TIME-DELAY REDUCTION RESULTS AND DISCUSSION EXPERIMENT SUMMARY... 81

10 MAS APPLICATION IN BUILDING FLEXIBILITY COORDINATION RESULTS AND DISCUSSION EXPERIMENT SUMMARY STUDY OUTCOMES STUDY LIMITATIONS CONCLUSION FUTURE WORK BIBLIOGRAPHY APPENDIX A: - TRIAS ENERGETICA APPENDIX B: - EVE- AGENTS APPENDIX C: - PV AND BATTERY ENERGY STORAGE SYSTEM INSTALLATION AT TEST-BED

11 LIST OF FIGURES LIST OF FIGURES FIGURE 1-1: THE TRIAS ENERGETICA METHOD AND THE FIVE STEP METHOD[38]... 4 FIGURE 1-2: ENERGY RELATED COMPONENTS OF OCCUPANT BEHAVIOUR IN BUILDINGS... 6 FIGURE 1-3: THE BEMS FUNCTIONS AS INDICATED BY EN FIGURE 1-4: INCREASING INPUT AND OPERATIONAL DOMAIN OF BEMS... 8 FIGURE 1-5: ENERGY USE OF COMMERCIAL BUILDINGS [EIA[153]] FIGURE 1-6: BUILDING FUNCTIONAL ABSTRACTION LEVELS FIGURE 1-7: MAJOR ELECTRICITY END USERS IN COMMERCIAL BUILDINGS[EIA[153]] FIGURE 1-8: THESIS OUTLINE FIGURE 2-1: COMPONENTS OF FINE-GRAINED OCCUPANCY INFORMATION FIGURE 3-1: EXPERIMENT OUTLINE FIGURE 3-2: CHAIR DESIGN COMPONENTS FIGURE 3-3: CO 2 SENSOR[282] FIGURE 3-4: TEST-BED CONFERENCE ROOM, CHAIRS AND WIRELESS DOOR CONTACT SENSOR FIGURE 3-5: SYSTEM SET-UP FIGURE 3-6: DIFFERENCE BETWEEN OUTDOOR AND INDOOR CO 2, CHAIR OCCUPANCY AND CO 2 COMPUTED OCCUPANCY (27 TH ) FIGURE 3-7: DIFFERENCE BETWEEN OUTDOOR AND INDOOR CO 2, CHAIR OCCUPANCY AND CO 2 COMPUTED OCCUPANCY (28 TH ) FIGURE 3-8: DIFFERENCE BETWEEN OUTDOOR AND INDOOR CO 2, CHAIR OCCUPANCY AND CO 2 COMPUTED OCCUPANCY (29 TH ) FIGURE 3-9: DIFFERENCE BETWEEN OUTDOOR AND INDOOR CO 2, CHAIR OCCUPANCY AND CO 2 COMPUTED OCCUPANCY (30 TH ) FIGURE 3-10: DIFFERENCE BETWEEN OUTDOOR AND INDOOR CO 2, CHAIR OCCUPANCY AND CO 2 COMPUTED OCCUPANCY (31 ST ) FIGURE 3-11: DOOR POSITON ON THE 28TH AND 30 TH FIGURE 3-12: STRAIN GAUGE OPERATION[286] FIGURE 3-13: CHAIR WITH MECHANICAL-SWITCH, VIBRATION AND STRAIN SENSORS FIGURE 3-14: STRAIN SENSORS CONNECTION FIGURE 3-15: SENSOR CONNECTION FIGURE 3-16: CHAIR SENSORS SET-UP FIGURE 3-17: VIBRATION, STRAIN AND CHAIR SENSORS (DAY 1) FIGURE 3-18: VIBRATION, STRAIN AND CHAIR SENSORS (DAY 2) FIGURE 3-19: VIBRATION, STRAIN AND CHAIR SENSORS (DAY 3) FIGURE 3-20: VIBRATION, STRAIN AND CHAIR SENSORS (DAY 4) FIGURE 3-21: Z-WAVE CONTROL NODE AND MOTION SENSOR[297] FIGURE 3-22: OCCUPANCY DETECTION SYSTEMS PLACEMENT FIGURE 3-23: SYSTEM CONNECTION FIGURE 3-24: TEST WITHOUT OCCUPANCY-BASED CONTROL FIGURE 3-25: CHAIR SENSOR CONTROLLED TEST FIGURE 3-26: MOTION SENSORS CONTROLLED TEST FIGURE 3-27: CONTROL SEQUENCE FIGURE 3-28: FUSION OF ALL THREE SENSORS FIGURE 3-29: PIR SENSOR FALSE ERRORS (17 TH ) FIGURE 3-30: PIR SENSORS FALSE ERRORS (18TH) i

12 LIST OF FIGURES FIGURE 3-31: LIGHTING ENERGY USE FOR EACH TEST DAY FIGURE 4-1: EXPERIMENT SCHEMATIC FIGURE 4-2: BUILDING ABSTRACTION LAYER AND AGENT REPRESENTATION FIGURE 4-3: TEST-BED OFFICE BUILDING FIGURE 4-4: THE EVE AGENT DESIGN PLATFORM FIGURE 4-5: TEST BED SPACE WITH ARMATURE PLACEMENT AND OCCUPANTS WORKPLACE FIGURE 4-6: CHAIR SENSORS, MOTION SENSORS AND WIRELESS CONTROL NODE FIGURE 4-7: SYSTEM CONNECTION AND MAS STRUCTURE FIGURE 4-8: USER AGENT CONTROL SEQUENCE FIGURE 4-9: SPACE OCCUPANCY DURING FIRST WEEK OF THE EXPERIMENT FIGURE 4-10: DAILY LIGHTING ENERGY USE IN TEST SPACE FOR THE THREE-WEEK PERIOD FIGURE 4-11: ROOM COUNT INFORMATION FROM ROOM AGENT (WEEK TWO) FIGURE 4-12: ROOM COUNT INFORMATION FROM ROOM AGENT (WEEK THREE) FIGURE 4-13: CHAIR SENSOR, OBSERVED AND MOTION SENSOR RECORDED TRANSITIONS (WEEK TWO) FIGURE 4-14: CHAIR SENSOR, OBSERVED AND MOTION SENSOR RECORDED TRANSITIONS (WEEK THREE) FIGURE 4-15: TEST-BED FLOOR FIGURE 4-16: SYSTEM CONNECTION FIGURE 4-17: A COMMUNICATION SEQUENCE FIGURE 4-18: PRINT ROOM OCCUPANCY AND LIGHTING ENERGY USE WITH 10 MINUTES TIME-DELAY FIGURE 4-19: PRINT ROOM OCCUPANCY AND LIGHTING ENERGY WITH TIME DELAY REDUCED FIGURE 5-1: MAS FRAMEWORK FIGURE 5-2: AGENT PLATFORM AND BEMS CONNECTION FIGURE 5-3: AGENT COMMUNICATION SEQUENCE TEST DAY FIGURE 5-4: PRINTER AND COFFEE MACHINE START UP CONSUMPTION FIGURE 5-5: AGENT COMMUNICATION SEQUENCE TEST DAY FIGURE 5-6: MAS COORDINATED VENTILATION AND APPLIANCE FANS DEMAND REDUCTION ON THE 24 TH FIGURE 5-7: MAS COORDINATED VENTILATION FANS, APPLIANCE DEMAND REDUCTION, AND PRINT ROOM OCCUPANCY 25 TH ii

13 LIST OF TABLES LIST OF TABLES TABLE 1-1: OVERVIEW OF PUBLICATIONS TABLE 2-1: PIR SENSORS TABLE 2-2: ULTRASONIC SENSORS TABLE 2-3: CO 2 SENSORS TABLE 2-4: IMAGE BASED DETECTION SYSTEMS TABLE 2-5: SOUND BASED DETECTION SYSTEMS TABLE 2-6: DEVICE BASED LOCALIZATION WIRELESS SYSTEMS TABLE 2-7: DEVICE FREE DETECTION TABLE 2-8: APPLIANCE LOAD TABLE 2-9: COMPUTER APPLICATIONS TABLE 2-10: SENSOR FUSION TABLE 2-11: MAS APPLICATION IN OFFICE BUILDINGS FOR ENHANCING ENERGY MANAGEMENT AND USER COMFORT.. 32 TABLE 2-12: FLEXIBILITY SOURCES AND STRATEGIES IN COMMERCIAL BUILDINGS TABLE 2-13: MAS APPLICATION IN OFFICE BUILDINGS FOR ENERGY FLEXIBILITY COORDINATION TABLE 3-1: OUTLOOK MEETING TABLE 3-2: CO 2, CHAIR AND GROUND TRUTH OBTAINED OCCUPANCY COUNT INFORMATION TABLE 3-3: SUMMARIZED LOGS TABLE 3-4: LOGGED AND SENSOR REGISTERED TRANSITIONS TABLE 3-5: TEST SCHEDULE FOR TEST DAYS TABLE 4-1: ARMATURE MAPPING TABLE 4-2: MEASURED TRAVEL TIME TO PRINT ROOM TABLE 5-1: FLEXIBILITY SOURCES AND STRATEGIES IN TEST-BED TABLE 5-2: TEST-BED AFT iii

14 LIST OF ABBREVIATIONS LIST OF ABBREVIATIONS BEMS DFL DSM EIA EISA GHG LED µchp EBC IEA Bu2Bu Bu2G HVAC HVAC&L MAS PIR PMV PROSA UWB USEF V2G Wi-Fi Building Energy Management Systems Device Free Localization Demand Side Management Energy Information Administration Energy Information and Security Act Greenhouse gases Light Emitting Diode Micro Combined Heat and Power Energy in Buildings and Communities International Energy Agency Building-to-Building Building-to-Grid Heating Ventilation and Air Conditioning Heating Ventilation Air-Conditioning and Lighting Multi-agent systems Passive infrared Predictive Mean Vote Product-Resource-Order-Staff Architecture UltraWideBand Universal Smart Energy Framework Vehicle-to-Grid Wireless Fidelity iv

15 INTRODUCTION The chapter provides the background and context of this research. It provides details of the scope, goals, objectives and outline of the thesis. The particular focus of this investigation is on the use of fine-grained occupancy information and multi-agents systems in enhancing energy management and utilization of buildings energy flexibility. 1

16 INTRODUCTION 1.1. Buildings, the Smart Grid and, energy and environmental sustainability Growing population and the corresponding increasing demand for electricity[1], climate change[2][3], rapid depletion of non-renewable energy sources and price volatility[4] are few amongst a number of factors that have fueled the drive for global improvement in energy and environmental sustainability within the built environment. These factors have in recent times compelled policy makers, engineers and researchers to rethink the structure, composition and future of the entire energy supply and demand infrastructure[5][6][7]. Hence in this direction, there has been growing focus on the development of strategies aimed at reducing emissions and improving energy efficiency in the built environment in the past few decades[8][9][10][11][12][13]. Given the energy-environment nexus and its impact on economic development, social development and environmental protection, various governments across the globe have also rolled out a number of initiatives and legislations to support the fight against environmental degradation and rapid depletion of non-renewable energy sources. In the United States for example, the Energy Information and Security Act (EISA) was enacted in The primary aim of EISA was to move toward greater energy independence and security, increased production of clean renewable fuels, increased efficiency of products, buildings, and vehicles, [14]. Similarly, the European Union in 2009 enacted a similar set of legislations contained in the 2020 climate and energy package that aims to achieve 20% cut in greenhouse gas emissions (from 1990 levels), 20% renewables and 20% improvement in energy efficiency by 2020[15]. More recently in April of 2016 in Paris[10], in a bid to combat climate and in support of environmental sustainability measures, 175 governments agreed to a long-term goal of keeping the increase in global average temperature to well below 2 C above pre-industrial levels. Though these legislations by itself do not improve energy and environmental sustainability, it has prompted actions that are contributing towards improved delivery of across board energy-efficiency measures. It has also contributed towards the proliferation of renewable and distributed energy resources within the existing energy infrastructure. 2

17 INTRODUCTION Within the built environment, buildings are responsible for a significant amount of energy consumption. As noted by the Energy Information Administration (EIA), 40% of total energy consumption in the United States in 2015 was in residential and commercial buildings, while in the European Union; about the same amount was also consumed in buildings[16][17][18][19]. Furthermore, with anticipated population and economic growth in developing and developed countries respectively, the energy demand of buildings is projected to increase by up to 50% if concrete steps are not taken to reduce energy demand and increase energy efficiency[20][21][22]. The building sector has thus being the target of a number of energy and environmental sustainability improvement strategies[23][24][25]. Toward efforts to improve energy performance and sustainability in the built environment, there has been growing emphasis on the application of decentralized renewable energy sources, storage sources as well as Demand Side Management (DSM) strategies in buildings. DSM as described by Palensky and Dietrich(2011)[26] and by the authors in [27] includes everything that is done on the demand side of an energy system to improve energy efficiency. From the perspective of buildings, these include measures such as: innovative building design[28], improved building insulation[29][30], use of energy efficient systems and strategies such as Light Emitting Diodes (LED), heat pumps, low temperature heating, high temperature cooling[31][32][33][34], as well as the use of sophisticated building control and management tools[35][36][37]. These measures also form the individual component part of the five step-method depicted in Figure 1-1[38]. The five-step method is being promoted as an improved building design methodology over the Trias Energetica[39][38]. The Trias Energetica is a model commonly used in the Netherlands as a guide when pursuing energy efficiency and sustainability goals in the built environment. As depicted in Figure 1-1, the Trias Energetica method is composed of three steps in which various initiatives connected to efficient and sustainable use of energy in buildings are organized in order of priority. The proposed five-step method introduces two additional steps namely[38][40]: 3

18 INTRODUCTION Figure 1-1: The Trias Energetica method and the Five step method[38] I. Implement energy exchange and storage The proliferation of distributed energy resources and storage systems on the demand side of the energy infrastructure and in buildings in particular, has given rise to the concept of the smart grid. Within the smart grid, prosumers (a consumer who also produces energy)[41] are expected to able to trade and exchange energy with the electrical grid. Though the concept of the smart-grid differs for different discipline, within the context of energy and environmental sustainability, it is considered the future electrical power infrastructure. It is also considered a solution to the current environmental as well as energy sustainability challenges[42][43]. The so-called smartgrid in this context would need to accommodate and support more renewable and cleaner generators with stochastic output, distributed energy resources, storage systems and new loads such as electric vehicles[43][44]. The smart-grid would also facilitate Building-to-Building (Bu2Bu), Building-to-Grid (Bu2G) and Vehicle-to-Grid (V2G) integration. The transition to the smart-grid is however not without its own challenges. In this new era, keeping the balance between demand and supply, as well as using transmission and distribution capacity as economically as possible without overloading and blackouts become major challenges[44][45]. Guaranteeing the reliability and stability of the grid in the face of these challenges thus demands additional flexibility in the power system, 4

19 INTRODUCTION both in terms of transmission and distribution capacities as well as flexible production[44][46][47]. Flexibility in the context of power systems engineering as described by Ulbig 2014[48] and Lund et al. 2015[49] is the ability to cost effectively balance electricity supply and demand continually while also maintaining an acceptable service quality to connected loads. In the traditional electrical grid, flexibility is mostly obtained from supply side sources and large industrial consumers, aggregators and electrical storage systems[49][50][51]. Nevertheless, with growing demand for power flexibility due to the increasing share of renewable energy sources in the new/future greener and cleaner energy infrastructure, buildings are being considered as viable sources of energy flexibility[52][53][54][55]. The interest in buildings is centred on its significant energy demand and its consideration as a cost-effective and/or environmentally beneficial means for sourcing the required additional energy flexibility for the smartgrid[56][57][58]. Buildings, both residential and commercial are thus increasingly being promoted as viable sources of useful and essential flexibility for support of the smart-grid through participation in demand response programs and other grid support programs[53][57]. Demand response is one of the mechanisms within demand side management[59]. As described by Siano 2014[43], it refers to changes in electricity use by customers from their normal consumption patterns in response to changes in the price of electricity over time, or to incentive payments designed to induce lower electricity use at times of high wholesale market prices or when system reliability is jeopardized. Buildings are considered a viable source of energy flexibility due to its huge energy demand and buildings are also increasingly becoming the target for demand response interventions[59][60][61][62]. As noted by the Universal Smart Energy Framework(USEF)[62], demand response is beneficial to energy providers as it: reduces investment in facilities specifically designed for peak-demand management, reduces grid congestion, avoids expensive grid upgrades, limits any penalties for failing to balance supply and demand. In addition, it is also beneficial to the consumer because it results in reduced electricity costs[63]. 5

20 INTRODUCTION II. adapt building energy demand to its users Building occupants and their behaviour in buildings is considered a key factor that significantly influences buildings energy demand[64][65]. Occupants presence, actions and inactions in buildings can lead to a reduction or increment of up to one-third in the design energy performance of buildings[66][67]. The actions of occupants in buildings is also considered as one of the most significant sources of uncertainty in the prediction of building energy use[68]. In recent years, strict legislations and advancements in building envelope design technology has resulted in improved energy efficiency of buildings in most advanced countries[23][69]. Occupancy Presence Location Activity Identity Count Track Use of building Controls Use of thermostats Use of lighting controls Use of operable windows Use of shadings Use of Appliances Use of equipment Plug loads Figure 1-2: Energy related components of occupant behaviour in buildings Occupants actions and behaviour in buildings nonetheless are considered the major cause of large discrepancies between predicted and actual building energy performances. This typically averages around 30% and reaches as high as 100% in some cases[70][71]. Although having no specific definition, occupant behaviour in buildings is often described in relation to activities that influence building energy consumption[72][73]. User presence, the use of building appliances and systems as well as the use of building controls as depicted in Figure 1-2 are typical activities used in describing occupant behaviour[64][74]. These actions directly and indirectly influence the energy use and energy performance of a building. 6

21 INTRODUCTION Occupant behaviour in buildings is in general a complex issue and has been the subject of a large body of academic research in the past few years[73][74][75][76]. The Energy in Buildings and Communities (EBC) annex 66 programme of the International Energy Agency (IEA)[77] was specifically established to investigate and provide better understanding of the influence of occupant behaviour in buildings. The Annex also aims to formulate solutions that address the negative impact of occupant behaviour on building energy performance. Even though a number of solutions such as provision of occupants with feedback[78][79], use of interventions and encouraging behavioural changes[80][81] have been proposed in literature, solutions that make buildings less sensitive to negative occupant behaviours have drawn more attention[38][82][83]. This is because unlike solutions that aim to change occupant behaviour in buildings[79][84][85], these solutions adapt building operations in such a manner that tend to diminish/counter the negative effect of occupant behaviour on energy consumption with little or no additional burden on occupants. This solutions also eliminate rebound effects which is common with solutions that tend to try to encourage behavioural changes in occupants. In this context, the role of the Building Energy Management System (BEMS) is quite significant as the general objective of the BEMS is to satisfy occupants requirements for comfort while reducing energy consumption during building operation[66][82]. Building Energy Management System & Multi-agent system Building Energy Management systems noted by the EN 15232[37] standard, can lead to increased operational and energy efficiencies in buildings. BEMS provides information for operation, maintenance and management of buildings especially for energy management. BEMS as depicted in Figure 1-3 enables complex and integrated energy saving functions and routines that can be configured based on the actual use of a building subject to real user needs. Hence, avoiding needless energy use and carbondioxide emissions[36][66][86]. Given the significance of occupant behaviour on building energy consumption, BEMS are in particular a useful building tool that facilitates adaptation of building energy demand to users as depicted in Figure

22 INTRODUCTION User User User BEMS User User Figure 1-3: The BEMS functions as indicated by EN Although initially limited in scope to only the core building domains, with advancement in building technology and adoption of smart grid technologies, today s BEMS provides coordination of other aspects such as: distributed energy resources, thermal and electricity storage systems[87][88](see Figure 1-4). These additional domains increase buildings value and energy flexibility potential. Traditional BEMS functions Lighting Airconditioning Heating-boiler Cooling/ /heat-pump refrigeration Weather Future BEMS functions Lighting Airconditioning /heat-pump Heating-boiler Cooling/ refrigeration Weather Electricity generation Electricity Storage Electric vehicles Thermal Smart meters Energy storage Plug Loads/ Appliances User input Figure 1-4: Increasing input and operational domain of BEMS Notwithstanding the introduction of these additional functions in buildings, the core functionalities of traditional BEMS has largely remained the same and undeveloped over the years[89][90][91][92]. While the communication and hardware technology of building controls has improved over the years, the control functions are still rudimentary, with very little use of supervisory control or embedded intelligence[25][86][90]. Traditional BEMS in most buildings are thus static and limited in their ability to process constantly changing and dynamic inputs such as occupant 8

23 INTRODUCTION behavior and user preferences. In addition, traditional BEMS are limited in their ability to effectively coordinate buildings interaction with the smart-grid [83][89][93][94]. A distinctive and common example is the use of fixed schedules and coarse-grained occupancy information in the operation of comfort systems in buildings[95][96]. Heating Ventilation Air-conditioning and Lighting (HVAC&L) systems account for more than 50% of the energy consumed in buildings[97][98]. Worthwhile energy savings can however be obtained from buildings when its operation is tailored to actual building occupancy information[99][100]. The utilization of fine-grained occupancy information for control of lighting systems in buildings for example, has been shown to have the potential to facilitate reduction in lighting energy use in by up to 50%[101][102]. Similarly, the application of fine-grained occupancy information for heating, ventilation and air conditioning systems has been demonstrated to have the potential to facilitate reduction in HVAC energy use by up to 40%[103][104]. Furthermore, in equipping buildings for their role within the smart-grid, BEMS is required to facilitate the integration and interoperation of the various information sources within and outside buildings that have to interoperate as one homogenous system. As noted by Becker et al. (2015)[105], the BEMS is a key tool that can enhance buildings participation in grid support activities such as demand response. The low level of sophistication of current BEMS does however limit the scope and level of participation in grid support and interaction with the smart-grid[106]. Hence, toward efforts to improve the operational scope and level of sophistication of traditional BEMS, there has been emphasis recently on the application of increased sensor-data and computational intelligence tools in building energy management[83][107][108]. Advancement in information and communication as well as low-cost sensor technologies such as wireless sensors[109][110][111] have accelerated interest in the application of increased sensor data and computational intelligence in building energy management systems[89][107][112][113][114]. The application of increased sensor-data on one hand facilitates better monitoring and management of buildings demand for energy based on inputs from actual building occupants and users[56][109][115]. In building operation, information on occupants presence is one of the key sensory data 9

24 INTRODUCTION that is essential for energy efficient operation of building systems[116]. Occupancy information is also one of the key components of occupant behavior that influences the energy performance of buildings[117][116]. Building occupancy varies throughout the day and it varies from building to building and organization to organization. In commercial office buildings in particular, occupants do not arrive and leave at the same time and occupants move through floors and spaces in a stochastic manner[116][118]. The use of increased sensor data and fine-grained building occupancy information can therefore lead to tailored delivery of comfort to occupants on the room level and enhancement of building energy management. The application of computational intelligence tools on the other hand facilitates near real-time intelligent reasoning, processing and application of the obtained information to achieve optimal building performance with limited operator input[25][119][120]. The application of computational intelligence tools in building operation can in general be grouped under three main categories[35][107][121]: Learning and prediction- These include tools such as expert systems[122], artificial neural networks[123][91][124], support vector machines[125], fuzzy tools[126] and Multi-agent based solutions[127] Process optimization- These include a number of linear and non- linear techniques[128][129][130][131] as well as Multi-agent based solutions[132] Energy management, control and diagnostics- This includes techniques such as fuzzy systems[133],model predictive systems[134][135], rule-based and supervisory control systems[136][137], as well as Multi-agent based solutions [138] Multi agent systems Multi-agent system (MAS) is a paradigm derived from the distributed artificial intelligence. MAS allows an alternative way to design distributed control systems based on autonomous and cooperative agents, exhibiting modularity, robustness, flexibility, adaptability and re-configurability[139]. Although there is no universally accepted definition of MAS, it is in general described as a loosely coupled network of software agents that interact to solve problems that are beyond the individual capacities or 10

25 INTRODUCTION knowledge of each problem solver[140]. Russel and Norvig[141] define an agent as an entity that can be viewed as perceiving its environment through sensors and acting upon its environment through effectors. Wooldridge and Jennings[142] state that an agent is a hardware and/or software-based computer system displaying the properties of autonomy, social adeptness, reactivity, and proactivity. Agents are highly distributed and have been suggested as a viable tool for solving complex control and coordination problems such as the coordination of the various elements of the smart-grid in a rational and natural way[143][144]. The concept of multi-agent systems is at an intriguing stage in its development as commercial strength intelligent agent applications have been developed in a number of sectors including buildings[108][140][145][146]. MAS has been proposed in buildings for enhancement of energy management and occupant comfort[89][147][78][148]. As noted by Huber et al 2015[149], the MAS approach guarantees an automated and self-learning commissioning process that could help improve buildings energy efficiency on a large scale. Furthermore, MAS is promoted as a viable tool for overall management of the smart-grid[143][150] and its use has been explored in buildings for coordination of buildings energy flexibility for grid support services[93][105][151][152]. Even though MAS is considered a viable computational intelligence tool for use in buildings for enhancing energy management and facilitating integration with the smart-grid, there are still very limited empirical studies that have demonstrated its added-value in buildings for enhancing energy management and coordination of building processes during participation in grid support activities. Scope, context and contribution of this research Within the commercial building sector, office buildings as depicted in Figure 1-5 have the highest relative floor space and energy use in most countries[24][66][93]. In addition, commercial buildings are also usually equipped with more actuation and control possibilities. Furthermore, commercial office buildings are increasingly becoming host to a large number of renewable, electrical as well as thermal storage 11

26 INTRODUCTION systems thereby further increasing their value and importance within the smart-grid for demand-side energy flexibility. Figure 1-5: Energy use of Commercial buildings [EIA[153]] Therefore, the scope of this study is limited to commercial office buildings, particularly the traditional BEMS in office buildings. As noted earlier, the performance of traditional BEMS is limited by a number of inadequacies such as[86][89][132]: Lacking input of dynamic information on occupant behaviour, articulate user preference and system behaviour - occupant behaviour, which includes building occupancy, occupants use of controls and appliances, are key parameters that influence the energy performance of buildings. Compared to the other attributes of occupant behaviour, building occupancy is a dominant factor in building energy evaluation because use of controls, lighting and plug loads, HVAC equipment utilization, fresh air requirements and internal heat gain greatly depend on the level of occupancy within a building[117]. The use of detailed fine-grained occupancy information in the energy management of buildings systems can thus yield energy savings of between 20-50%[154]. In addition, the availability of detailed information on building occupancy can also be exploited by the BEMS in the determination and utilization of a buildings useful available flexibility[155][156][157]. 12

27 INTRODUCTION Lacking level of sophistication required for coordination of building systems and occupants in relation to buildings interactions with the smart-grid- The use of buildings flexibility for grid support services such as demand response often leads to conflicting and contradictory requirements. This is because buildings demand for energy primarily serves the purpose of providing a comfortable, healthy and safe indoor environment. Grid support services such as demand response do however often require reduction in the energy demand of building comfort systems, which might degrade comfort[157] i.e., visual, thermal and indoor air quality. In commercial office buildings in particular(which is the focus of this research),where a strong link between user comfort and productivity has been established[158], participation in grid support might not be beneficial to the building when considered directly in terms of user comfort, cost and efficient operation of the building[50]. Research methodology Architectural systems and functional abstraction levels Built Environment Building Floor/Zone Room Workplace Users Figure 1-6: Building functional abstraction levels A typical office building equipped with a traditional BEMS was selected as test-bed for the study with the room-level as the core focus. This is because the room-level as depicted in Figure 1-6[159], is where occupants needs and behavior, system response and orientation effects give rise to contradictory requirements. Furthermore as noted by the 13

28 INTRODUCTION authors in [160], balance between occupant comfort and energy efficiency can as a first step be effectively achieved on the room-level. The hypothesis of this research is thus: a combination of fine-grained occupancy information and multi-agents coordination can enhance energy management and preservation of occupants comfort during buildings participation in demand response. Given the stated aim of this research, the methodology was divided into two parts: The first part was focused on occupancy detection techniques in building energy management, while the second part focused on the use of multi-agent coordination for enhancing energy management and buildings interaction with the grid. As depicted in Figure 1-7, lighting is the singular major electricity end user in commercial office buildings. Therefore, this study focuses on the use of fine-grained occupancy information for control of lighting to enhance energy management. Figure 1-7: Major electricity End users in commercial buildings[eia[153]] Both analytical and empirical methods were applied in testing the research questions posed in each part outlined below: 14

29 INTRODUCTION I. Occupancy detection systems application in enhancing building energy management What are the limitations of current existing and commonly used occupancy detection systems? - This question was addressed through a detailed and analytical review of literature; the strengths and limitations of commonly used occupancy detection systems in building operations were outlined and summarized. What alternative methods are available and applicable in buildings for obtaining fine-grained occupancy information? - This question was addressed using the outcome of the literature survey as well as through design and experimental evaluation of occupancy detection systems in an office building. Based on the literature review outcome, three experiments were defined to test specific methodologies to obtain fine-grained occupancy information. II. Multi-agent systems application in enhancing building energy management What are the benefits of multi-agent systems as a computational support tool in building operation and interaction with the smart-grid? This question was addressed through a detailed and analytical review of literature; benefits of multi-agent systems in enhancing energy management and utilization of buildings flexibility for demand response were outlined and summarized. Based on the outcome of the survey, three experiments were formulated to evaluate the advantages of MAS coordination for energy management and demand response in a test-bed office building. Thesis outline The thesis as shown in Figure 1-8 is comprised of seven chapters. This chapter has presented a general introduction, the scope, contribution as well as the aim and objectives of this study. Chapter two provides a detailed review on occupancy detection systems commonly used in commercial buildings for occupancy detection. Chapter two also provides a review on the application of multi-agent systems in buildings as a computational support tool. Emphasis was placed on the application of MAS in buildings for enhancing energy management and facilitating buildings integration with the smart-grid while maintaining occupant comfort. 15

30 INTRODUCTION Chapter One Chapter Two Introduction Scope, aim, hypothesis research questions and outline Literature review Occupancy detection systems Multi-agent systems application in building operation Chapter Three Methodology part I Occupancy detection system design and evaluation Chapter Four Methodology part II MAS design and evaluation in occupancy-based lighting Chapter Five Chapter Six Chapter Seven Methodology part III MAS design and application in building energy flexibility coordination Results and discussion Effectiveness of fine-grained occupancy information in building operation Benefits of MAS in building operation and for flexibility coordination Conclusion Answers to research questions Outcome of hypothesis Figure 1-8: Thesis Outline Chapters three, four and five are the methodology chapters. Chapter three provides details of the experiments on the design and evaluation of occupancy detection systems in the test-bed office building. In chapter four, details of the multi-agents frameworks, the performed experiments, the test-bed office building s BEMS and the added-value for occupancy-based lighting control is presented. In chapter five, MAS application in coordination of the building s energy flexibility is provided. Chapter six contains a discussion on the result from the experiments performed while the research outcome, conclusions and future research direction are provided in chapter seven. 16

31 INTRODUCTION 1.6 Publication overview A substantial part of the text and results in this thesis was extracted from already published academic journals. Table 1.1 provides an overview of the various published articles. The contributions are to be considered joint with the respective co-authors. Table 1-1: Overview of publications Chapter Journal/Citation 2 T. Labeodan, W. Zeiler, G. Boxem, and Y. Zhao, Occupancy measurement in commercial office buildings for demand-driven control applications - A survey and detection system evaluation, Energy Build., vol. 93, pp , Apr Y. Zhao, W. Zeiler, G. Boxem, and T. Labeodan, Virtual occupancy sensors for real-time occupancy information in buildings, Build. Environ., vol. 93, pp. 9 20, Nov T. Labeodan, K. Aduda, G. Boxem, and W. Zeiler, On the application of multi-agent systems in buildings for improved building operations, performance and smart grid interaction A survey, Renew. Sustain. Energy Rev., vol. 50, pp , T. Labeodan, K. Aduda, W. Zeiler, and F. Hoving, Experimental evaluation of the performance of chair sensors in an office space for occupancy detection and occupancydriven control, Energy Build., vol. 111, pp , Jan T. Labeodan, C. De Bakker, A. Rosemann, and W. Zeiler, On the application of wireless sensors and actuators network in existing buildings for occupancy detection and occupancydriven lighting control, Energy Build., vol. 127, pp , K. O. Aduda, T. Labeodan, W. Zeiler, G. Boxem, and Y. Zhao, Demand side flexibility: Potentials and building performance implications, Sustain. Cities Soc., vol. 22, pp , Apr K. O. Aduda, T. Labeodan, W. Zeiler, and G. Boxem, Demand side flexibility coordination in office buildings: A framework and case study application, Sustain. Cities Soc., vol. 29, pp ,

32 LITERATURE SURVEY This chapter provides a review of occupancy detection systems commonly used in buildings to enhance energy management as well as a review of multi-agents application in buildings for enhancing energy management. The chapter aims to provide answers to the first research question by identifying the drawbacks and limitations of commonly used occupancy detection systems. The chapter also discusses the benefits of MAS in building operation and interaction with the smartgrid. A significant part of the content of this chapter is published in the following articles: T. Labeodan, W. Zeiler, G. Boxem, and Y. Zhao, Occupancy measurement in commercial office buildings for demand-driven control applications - A survey and detection system evaluation, Energy Build., vol. 93, pp , Apr T. Labeodan, K. Aduda, G. Boxem, and W. Zeiler, On the application of multi-agent systems in buildings for improved building operations, performance and smart grid interaction A survey, Renew. Sustain. Energy Rev., vol. 50, pp ,

33 LITERATURE SURVEY 2.1 Occupancy information in building operation Traditional BEMS as noted in the preceding chapter lacks real time input of dynamic factors such as occupancy, occupant preference and information on use of appliances. In addition, if these information is available, traditional BEMS often lack the intelligent reasoning required for optimizing building processes based on these dynamic and distributed input[89][93]. Consequently, as noted in the preceding chapter, towards improving the level of sophistication of traditional BEMS, there has been focus on the use of increased sensor data providing particularly fine-grained occupancy information[21][101][161][162]. Moreover, the use of fine-grained occupancy information is not only limited to the management of building processes for improved building energy performance. Finegrained occupancy information is also essential for the determination and exploitation of buildings energy flexibility for use in grid support services such as demand response[155][156][157][163]. Because of the growing importance of building occupancy information on both energy efficiency and flexibility related processes in buildings, researchers have been actively exploring possibilities for obtaining fine-grained information on building occupancy[99][154][164][165]. Occupancy information can be obtained from various multiple sources in buildings and this includes explicit occupancy detection systems such as motion sensors and Carbon-dioxide(CO 2 ) sensors[166][167], as well as implicit systems such as digital computers, Wireless fidelity(wi-fi) access points as well as temperature measurement systems[100][164][168]. Explicit occupancy detection systems though described in detail subsequently, in this context refer to systems installed in buildings for the primary purpose of occupancy detection. Implicit occupancy detection systems on the other hand, refer to systems not installed primarily for occupancy detection, but from which information on building occupancy can be determined. Irrespective of the source, the occupancy information provided by most occupancy systems is composed of one or more of six properties depicted in Figure 2-1 namely[169][170][171]: presence, location, count, activity, identity and track. 19

34 LITERATURE SURVEY Figure 2-1: Components of fine-grained occupancy information Presence- this relates to information on when occupants are present or absent in a particular space or room in a building Location-refers to Where occupants are present in a building Count- refers to information on the number of occupants present in a particular space, room or location within the building at any particular time Activity- refers to information on the activity being carried out in a particular location, space or room at any particular time in a building Identity- refers to information on who is present at particular location or space at any particular point in time Track- refers to information on occupants' trajectory or movement history through locations within a building In addition to classification of occupancy detection systems based on the granularity of occupancy information provided, occupancy detection systems can also be classified using the following properties[96]: Method- This property refers to the detection technique utilized i.e., whether the detection system is device-free or not. Some occupancy detection systems such as those based on wireless signals including devices equipped with Wi-Fi, Bluetooth, Ultra- Wideband (UWB)[172][173][174] and others, usually require occupants to have a device on their person at all time[120][175]. On the other hand, detection systems such as Passive Infrared (PIR) and carbon dioxide systems do not require building occupants to be in possession of any device. The former systems are termed terminal based, while the latter are termed non-terminal or device free systems[176]. Function- This property relates to the level of information provided by the occupancy sensor. While some detection systems are able to provide information on the occupants' identity and exact coordinates within the building, other detection systems are only able 20

35 LITERATURE SURVEY to provide information on occupants presence and location. PIR sensors for example are only able to provide information on occupant s presence and fixed coordinates whereas Radio Frequency Identity (RFID) tags are able to provide information on occupants identity as well as occupants coordinates in a mapped building[96][170].the former systems in this case are termed individualized systems while the latter systems are termed non-individualized systems[96]. Infrastructure- This property relates to the primary function of the detection system. Some detection systems are installed for the sole purpose of occupancy detection in buildings and this includes PIR and ultrasonic sensors[101][166]. On the other hand, some systems such as smart-phones [100][177][178] provide the secondary function of occupancy detection. Systems with the primary function of occupancy detection are termed implicit systems[154]. In the following sub-sections, using the descriptors discussed above, a review of occupancy detection systems commonly used in buildings to enhance energy management is analyzed and discussed. 2.2 Occupancy detection systems I. Passive infrared sensors All objects with a temperature above absolute zero emit heat energy in the form of radiation. The emitted infrared radiation is invisible to the human eye but can be detected by electronic devices such as PIR sensors. PIR sensors detect thermal radiation from the human body in the wavelength range 8μm to approximately 14μm[179] when a human object arrives in its proximity. In principle, PIR sensors detect the energy given off by objects within its line of sight[180]. Table 2-1: PIR sensors PIR sensors Presence Location Count Activity Identity Track Method Function Infrastructure Non-terminal Terminal Nonindividualized Individualized Explicit Implicit 21

36 LITERATURE SURVEY PIR sensors, due to their ease of installation, low maintenance and power consumption, small form factor, as well as the unobtrusive and privacy-preserving interaction, are one of the commonly used occupancy detection system in buildings[166][181][182]. Despite their popularity, PIR sensors however have a few drawbacks that hamper their use in buildings for occupancy detection[183][184]. A common drawback of PIR sensors is False-off or False-negative errors[101][103]. These errors occur mostly when occupants are still for a long-time and do not make any movement. Furthermore, false motion alarms may be caused by moving masses of air, heat as well as trees blowing in the wind and reflection from cars passing. Secondly, PIR sensors are only capable of providing information on occupants presence and absence in a space[185] as heterogeneous detection systems. Although additional information such as track[181][166] and count[186] as outlined in Table 2-1 can be obtained from the sensors, it does however requires the use of multiple sensors and advanced computational support techniques. II. Ultrasonic and sonar based detection systems Ultrasonic detectors are another technology used in commercial buildings. Unlike their PIR counterparts, ultrasonic sensors are active devices: they emit ultrasonic sound waves, and receive the sound energy reflected back to the sensor from the environment[187][188][189]. The time of arrival, strength and angle of arrival of the reflections in the environment are some of the parameters used by the sensor to determine occupancy[190][191]. Even though these qualities offer improvements over PIR sensors, the sensors also have some drawbacks. Table 2-2: Ultrasonic sensors Ultrasonic sensors Presence Location Count Activity Identity Track Method Function Infrastructure Non-terminal Terminal Nonindividualized Individualized Explicit Implicit The emitted signals from the sensors are attenuated significantly by obstructions like walls and may create difficulties for estimating actual location of human object[180]. 22

37 LITERATURE SURVEY Ultrasonic detection systems are also susceptible to false positives errors triggered by vibrations such as the air turbulence from HVAC systems or printers[192]. In addition, as outlined in Table 2-2, the sensors are also only able to provide occupancy information on user presence and cannot provide information on track, count and identity. III. CO 2 based detection systems Exhalation of CO 2 is a natural phenomenon for humans. The concentration of CO 2 in a space can hence provide an indication of occupancy in a space provided humans are the only major source of CO 2 production[193][194]. CO 2 sensors are also used in buildings as an indicator for indoor environmental quality as it is one of the most commonly available gas in polluted air[98][195]. The measurement of the CO 2 concentration in a space as outlined in Table 2-3 can hence be used in determining occupancy information on presence[167][196], count[197] as well as user activity level[176][195]. CO 2 based occupancy detection systems do however have a few drawbacks as well. The main drawbacks of CO 2 sensors are its susceptibility to environmental factors and its slow reaction time[180]. The authors in [198] showed the reaction time of CO 2 sensors as typically between minutes. In addition, factors such as the space layout, ventilation rate, sensor location, wind, pressure and temperature difference can adversely affect the reaction time of the sensors[176][199]. Table 2-3: CO 2 sensors CO 2 sensors Presence Location Count Activity Identity Track Method Function Infrastructure Non-terminal Terminal Nonindividualized Individualized Explicit Implicit IV. Image detection systems With growing interest in the development of detection systems having the ability to provide fine-grained occupancy information in buildings, the use of imaging devices such as cameras have also been investigated by a number of researchers [200][201][202][203]. The authors in [200], developed an image-based method for unsupervised real-time room occupancy detection with room. The developed system 23

38 LITERATURE SURVEY was able to detect and track multiple persons of various heights moving in different directions within the camera s field of view. In another study, the authors in [202] developed an image-based occupancy detection system using a low-cost thermal infrared sensor co-located with a wide-angle camera. The developed system provided improvement to the tracking capability of image-based systems as it was concluded to be capable of robustly tracking people in indoor environments over long periods. Table 2-4: Image based detection systems Image sensors Presence Location Count Activity Identity Track Method Function Infrastructure Non-terminal Terminal Nonindividualized Individualized Explicit Implicit Though image based systems as outlined in Table 2-4 are able to provide occupancy information on presence, count, location, identity, activity and track, its application in buildings for process control is hampered due to issues bordering on ethics and user privacy[96]. In the indoor environment, some factors, such as serious occlusions and rapid change of occupant scales, can cause errors in the tracking. Furthermore, advanced computational support tools are required for obtaining occupancy information from large data output from these systems[168][184][204]. V. Sound detection systems The measurement of sound waves in a space is another phenomenon that is currently been studied for use in buildings for occupancy detection[205][165]. The authors in [205] developed a network embedded acoustic processing system for estimating the occupancy level in rooms and buildings. Making use of 8 microphones, an estimation algorithm was also developed for determining the number of people in a space. Table 2-5: Sound based detection systems Sound sensors Presence Location Count Activity Identity Track Method Function Infrastructure Non-terminal Terminal Nonindividualized Individualized Explicit Implicit 24

39 LITERATURE SURVEY Although impressive results were recorded, audio-based systems are also seldom used in buildings due to the need for expensive equipment and advanced signal processing techniques[206]. Sound based detection systems as outlined in Table 2-5 are able to provide occupancy information on user presence and location. VI. Device-based wireless detection systems Wireless enabled devices such as mobile phones, laptops, and smart watches have also been evaluated in buildings for occupancy detection[172][177][207]. Device based wireless detection systems are usually composed of a transmitting device (usually carried by the user) and a receiving node (usually static), which allows for either the measurement of the energy or timing of the response echo of a transmitted signal by the receiving node[174][208]. The transmitted signal may consist of short series of pulses or a modulated radio signal. Device based wireless detection systems can in addition to providing occupancy information on presence and locations provide information on user identity. Furthermore, as shown in Table 2-6, device based wireless detection systems can also be explicit and implicit detection systems. Table 2-6: Device based localization wireless systems Nondevice free wireless systems Presence Location Count Activity Identity Track Method Function Infrastructure Non-terminal Terminal Nonindividualized Individualized Explicit Implicit Device based wireless systems do however have a few drawbacks that limit their application in buildings. Wireless signals such as Wi-Fi for example, are able to penetrate through walls therefore making them prone to false reads[100][178]. Additionally, being device based, occupants are required to have on their person a device from which occupancy is determined. Occupants can however not be completely relied upon to constantly be in possession of such a device. VII. Device free wireless based systems Given some of the limitations of device based wireless detection, particularly the challenging task of getting occupants to constantly be in possession of a device in order 25

40 LITERATURE SURVEY to be detected, Device Free Localization(DFL) systems were introduced[184]. DFL exploits the fact that the radio signals propagate across different environments where the signals can be reflected, diffracted, scattered or absorbed by many objects in the propagation paths[180]. The Human body thus partially absorbs a radio signal, thus decreasing the received signal strength (RSS)[183][209].The authors in [210] evaluated the performance of DFL in a building and observed that DFL systems purely aimed at indoor applications could be prone to give feedback related to events that take place outside the desired area. In addition, the number of people present in a space is still a challenge for the DFL technology. Though DFL detection systems as outlined in Table 2-7 are able to provide occupancy information on user presence, location, track, activity and identity, more research, it is still a developing technology and seldom used in buildings for occupancy detection. Table 2-7: Device free detection Device free wireless detection VIII. Building appliances With increased awareness for energy conservation and advancement in information and communication technology, electrical power measurement sensors, which were hitherto costly devices, are now increasingly used in buildings[115][186][211]. Power meters are at times installed on building appliances such as printers, personal computers and display monitors. Presence Location Count Activity Identity Track Method Function Infrastructure Non-terminal Terminal Nonindividualized Individualized Explicit Implicit Table 2-8: Building appliances Presence Location Count Activity Identity Track Appliances Method Function Infrastructure Nonterminaindividualized Terminal Non- Individualized Explicit Implicit 26

41 LITERATURE SURVEY The analysis of the energy consumption profile of these meters as outlined in Table 2-8 can be used to obtain occupancy information on occupants' presence and location. The application of power measurement systems in demand driven control application is however still limited due to the coarse grained nature of the provided occupancy information. IX. Computer applications The use of computer and mobile phone applications for occupancy detection in buildings has also being explored by a number of researchers[97][212][213]. Table 2-9: Computer applications Computer applications Presence Location Count Activity Identity Track Method Function Infrastructure Nonterminaindividualized Terminal Non- Individualized Explicit Implicit Personal computers as outlined in Table 2-9 are implicit building appliances and as building occupants often spend significant amount of time on them it s a source from which occupancy information can be deduced[97]. By installing an application that can detect keyboard and mouse usage, the authors in [212] and [213] were able to obtain occupancy information on presence and location. Though a low-cost means of obtaining occupancy information in buildings, its use is still limited due to user privacy and computer data security, as well as false negative errors. X. Sensor fusion As noted from the preceding sections, most occupancy detection systems commonly used in buildings suffer from one drawback or the other that limits their use in buildings. Towards effort to address the limitations of individual heterogeneous detection systems, researchers have often explored the use of fusion of multiple sensors[99][109][120][168][214]. Composed of multiple different sensors or same sensors, the resulting detection system compensates for the drawback of each individual 27

42 LITERATURE SURVEY sensor that makes up the system therefore facilitating the availability of fine-grained occupancy information. Table 2-10: sensor fusion Sensor Fusion Presence Location Count Activity Identity Track Method Function Infrastructure Non-terminal Terminal Nonindividualized Individualized Explicit Implicit As a proof of concept, the author in [214] designed an occupancy detection sensor network in an open-plan space of an office building using a combination of CO 2, acoustic, and PIR sensors as well as a thermal imaging camera. The resulting detection system was shown to be able to provide occupancy information on presence, count, location, track and activity. Using three PIR-based modules, each consisting of two pairs of PIR sensors orthogonally aligned, the authors in [166] were able to obtain information on track and identity. Similarly, using a pressure, PIR as well as acoustic sensors, the authors in [212], were able to obtain fine-grained occupancy information on occupants' presence as well as user activity. Similarly using a combination of sound, motion sensors, carbon dioxide, and door state sensors, the authors in [147] were able to determine occupancy information on count and user presence. In addition, to facilitate use of the information provided by the multiple sensors in the system by the BEMS, computational intelligence based on the MAS paradigm was also applied for process control in real-time and for prediction. Sensor fusion as outlined in Table 2-10 can thus facilitate the availability of fine-grained occupancy information on user presence, location, count, activity, track, and identity for enhancing energy management in buildings Discussion Although fine-grained occupancy information plays a key role in enhancing the energy performance of buildings through tailored delivery of lighting, heating, ventilation and cooling to spaces based on actual building occupancy information, most common sensors used in buildings for occupancy detection have inherent drawbacks. These limitations as outlined in Table 2-1 to Table 2-10 in relation to the sensors ability 28

43 LITERATURE SURVEY to provide occupancy information on user presence, count, location, track, activity and identity significantly hamper their performance in buildings for energy management. PIR motion sensors for example are practically one of the most commonly used occupancy detection system in buildings for enhancement of energy management [166][206]. Due to their susceptibility to errors resulting from still occupants, timedelays which typically varies between 5 to 20 minutes are often utilized when applied in buildings for occupancy based control of lighting systems[101][192]. Higher energy savings can be realized with a low time delay, but a low time delay value could also lead to false triggers where the lights are switched off when spaces are still occupied. False triggers could lead to user dissatisfaction and discomfort[215]. Sensor fusion as can be deduced from the survey is a cost effective way of addressing the shortcomings of heterogeneous occupancy detection systems Moreover, with growing advancement in wireless sensor technology, multiple common occupancy detection systems can be easily fused to form a robust occupancy detection network[109][216]. Though wired solutions are preferred in majority of cases, wireless devices are becoming more prevalent due to improvement in communication link speed, security, and battery technology[217][218]. Wireless sensors can now boast of batteries having average life spans of up to six years. Wireless sensors as a result nowadays replace, or in some cases augment, traditional hard-wired solutions, therefore facilitating the realization of sensor fusion in buildings. Sensor fusion in addition to enabling the availability of detailed and fine-grained occupancy information also facilitates the application of computational support tools such as data mining[180], prediction[185], learning and intelligent building monitoring, control as well as diagnostics[147][117][162] Multi-agent application in building management Agents as noted earlier are sophisticated computer programs that act autonomously on behalf of their users, across open and distributed environments, to solve complex problems. A number of computational support techniques such as expert systems[122][219], genetic algorithms[130][220], artificial neural networks[123][132], 29

44 LITERATURE SURVEY fuzzy logic[83] and other offshoots of artificial intelligence[25][107] have been applied in buildings for data-mining, learning, prediction and control. Computational tools based on the MAS paradigm are nonetheless gaining widespread application based on its distributed, parallelism, autonomous and scalability capabilities[83][108]. Furthermore, MAS is considered advantageous because[108][140][221][222]: It distributes computational resources and capabilities across a network of interconnected agents. It allows interconnection and interoperation of multiple existing legacy systems It models problems in terms of autonomous interacting component-agents, which is proving to be a more natural way of representing generation, distribution, and supply It efficiently retrieves, filters, and globally coordinates information from sources that are spatially distributed and provides solutions in situations where expertise is spatially and temporally distributed It enhances overall system performance, specifically along the dimensions of computational efficiency, reliability, extensibility, robustness, maintainability, responsiveness, flexibility, and reuse. In building operation, the application of MAS has followed two main directions [108][223][224][225][226]: I. Enhancing energy management Although the concept of multi-agents has intuitive appeal for dealing with complex systems and has received considerable commercial and academic interests over an extended period[140], its first empirical application in buildings was demonstrated by Huberman & Clearwater[227] in 1995 at the Xerox-PARC laboratory. The designed MAS system was based on a market/auction based framework. As noted by the authors in [227], prices and auctions greatly facilitate resource management in human societies. Auctions are an effective mechanism for redistributing goods based on the ability of participating agents to pay for them. The same concept could however be considered 30

45 LITERATURE SURVEY applicable in managing resources for other applications such as building services systems as depicted in Figure 1-3. Within a market/auction framework, agents competitively negotiate and interact in a purely electronic market through auctions[228][58]. Agents submit competitive bids for a shared resource of commodity and an auctioneer computes a general equilibrium (i.e. a set of prices such that supply meets demand for each commodity). In a highly distributed environment, a number of auctioneers are used in a hierarchical fashion so as to reduce the computational burden[229]. Using this market concept[227], heat and cold were offered as the commodity in the test-bed office building. Individual rooms fitted with variable air volume boxes were represented by room agents who were able to make bids for lesser or more air depending on the set-point temperature in the rooms. The auctioneer agent was the system s umpire. It was tasked with collecting bids from all agents and determining the sell price. A few conclusions can be drawn from this study by Huberman & Clearwater[227]; (i) it points to the inadequacies of traditional BEMS in that the available actionable information from the rooms were not intelligently exploited (ii) it demonstrated the viability of MAS in addressing these limitations (iii) though efficient exchange of air was hindered by the location of the air ducts, the MAS coordinated system was shown to reduce deviations between the set point and room temperatures. However in a comparative study by the authors in [230], it was demonstrated that the primary driver of the performance gain was the ability of the MAS to access and harness both global and local information, which the conventional control system could not access. While not diminishing the significance of the pioneering work by Huberman and Clearwater, Ygge et al.(1999)[230] further demonstrated that the conventional control system could even outperform the auction-based MAS if it had access to the same global information. Notwithstanding, following the noteworthy results from the study by Huberman and Clearwater[227], a significant number of researchers have also exploited its use in buildings as outlined in Table 2-11[108][223]. 31

46 LITERATURE SURVEY Table 2-11: MAS application in office buildings for enhancing energy management and user comfort Study Coordination method Building type Implementation Year Huberman & Clearwater[227] Market based Commercial Field-test 1995 Davidsson and Boman[231] Non-market based Commercial Field-test 1999 Mo[232] Non-market based Commercial Conceptual 2002 Qiao et al.[233] Non-market based Commercial Conceptual 2006 Zeiler et.al. [234] Non-market based Commercial Field-test 2006 Wall et al[235][236] Non-market based Commercial Field-test 2007 Duan & Lin [237] Non-market based Commercial Simulation 2008 Abras[238] Non-market based Residential Simulation 2010 Bin et al[239] Non-market based Commercial Conceptual 2010 Spirleanu & Diaconescu[240] Non-market based Commercial Conceptual 2011 Fortino & Guerrieri[241] Non-market based Institutional Field-test 2011 Liu et al[242] Non-market based Institutional Simulation 2011 Mamdi et al[147] Non-market based Institutional Field-test 2011 Klein et al. [89][243] Non-market based Institutional Simulation 2012 Wang et al.[244][245] Non-market based Institutional Simulation 2012 Lacroix et al.[246] Non-market based Commercial Simulation 2012 Kelly and Bushby[247] Non-market based Commercial Simulation 2012 Mokhtar et al[132] Non-market based Institutional Simulation 2013 Yu et al[78] Non-market based Commercial Field-test 2013 Urieli & stone[127] Non-market based Commercial Simulation 2013 Lin et al.[248] Non-market based Commercial Conceptual 2015 Al-Daraiseh[222] Non-market based Institutional Field-test 2015 Huber et al Davidsson and Boman[231][249] for example, developed a MAS coordination system, which unlike the MAS framework developed by Huberman and Clearwater[227] was non-market based. The goal of the agents was to achieve energy saving while ensuring occupants satisfaction. In achieving this, the MAS relied on multiple information sources within the building including occupancy, user preference and appliances energy use. In addition, the developed MAS also incorporated a real-time decision module that 32

47 LITERATURE SURVEY provided advice in generic situations, and uses information provided by the agents themselves at query time, e.g., in the form of decision trees with probability and utility assessments. The MavHome[250] project is another MAS-oriented approach applied in buildings. The MAS design was composed of a set of agents able to communicate through a hierarchical interconnection schema for control and information flow. Zeiler et al. 2006[234] as part of the SMART (Smart Multi Agent internet Technology)/IIGO (Intelligent Internet mediated control in the built environment) developed an intelligent agent based system for optimization of user perceived comfort and energy in buildings. In this case, a utility function that weighed occupant comfort deviation with respect to the preferred comfort against cost and energy use was defined. Through field tests conducted in an office, the authors[234] showed that the agent based system yielded positive improvement in the energy efficiency of the test bed. Yang and Wang[245][244] also proposed a MAS based systems for improving the level of sophistication of BEMS. Composed of three types of agents namely: central, local and personal agents. Personal agent communicates with local agent to provide information on occupant s preferences and receive feed- back from local agent concerning environmental changes and the interactions between humans and their environment. Local agents are designed to control each subsystem while the central controller continuously interacts with multiple local agents for achieving the overall control goals in the ever-changing operating conditions. The agents were also able to learn using artificial neural networks. The authors in [242] also designed a similar MAS based system which incorporated an occupant agent with negotiation functionalities. The authors in [78] also developed a MAS based building process coordination system that utilizes user centred information such as user preferences and feedback in order to continuously reconfigure itself. The developed MAS was modelled according to the PROSA (Product-Resource-Order-Staff Architecture) architecture. PROSA models agents in a manner similar to a manufacturing environment where product, resource, order and staff agents all have to interact to ensure optimal service delivery. The authors[78] were able to demonstrate the feasibility of implementing a MAS based on 33

48 LITERATURE SURVEY actual user preferences and feedback for comfort and energy management in an actual office building. The authors were also able to identify through actual experimentation some implementation challenges, one of which was congestion in the communication channel due to many devices transmitting over a shared bus. The study also identified the impact occupant behaviour had on a buildings energy performance. The authors reported that users tend to maximize lighting power even when there was abundance of daylight illumination, hence prompting the authors to recommend a certain regression of the set-point brightness value for improved energy efficiency. With a large number of studies on multi-agents being simulations, this study does demonstrate the feasibility of implementing MAS that function with actual user preferences and feedback in an office building. Users in the environment as reported by the authors indicated a higher degree of satisfaction after the MAS deployment. The authors in [147] also designed a MAS coordination strategy that optimizes user comfort and building energy use. The core component of the system was the adaptive estimation and prediction agents that utilized data from multiple buildings sensors and learned patterns of occupant behaviour. By modelling occupancy patterns, the developed system could conserve energy by constraining heating and cooling policies to be active only during the hours when occupants were actually in specific thermal zones. In the same vein, the authors in [89]and [243] also developed a similar system using the MAS paradigm. The developed MAS system had the ability to negotiate with users to optimize the use of spaces based on occupancy density. Similarly, Huber et al.(2015)[149] developed an agent based control system for use in buildings for selection of the optimal heating system. Utilizing cost functions as a basis for the negotiation process, agents in the system was able to decide on the most efficient heating component at any time. II. Enhancing buildings integration with the smart grid The transition to the smart grid as noted in chapter one demands for additional electrical energy flexibility induced by the uncertainties from the growing numbers of distributed renewable energy sources that are being introduced in the electrical 34

49 LITERATURE SURVEY grid[46][251]. Although, the current practice in power systems is to provide flexibility with power plant capacity, this results in tying up expensive capital investment, operating the generators at potentially lower efficiency, and increasing wear and tear from continually adjusting their output in response to the immediate balancing needs of the grid[50][58]. Distributed demand response assets can however provide these equivalent services potentially at a lower cost by adjusting load rather than power plant output. Buildings are considered to be capable of providing faster response than power plants and can effectively reduce the need and lower the cost of regulation[53][58][252]. Due to the significant energy demand of buildings, there has been recent focus on the development of innovative coordination techniques to harness the energy flexibility of buildings[53][163][253][254][255]. Although the concept of harnessing the energy flexibility of buildings for is not entirely new[252][256][257], commercial buildings, both individually and collectively are nonetheless considered to hold the greatest potential for energy flexibility in the near term[53][54][163]. This is because in commercial buildings HVAC systems and other office equipment can be used to provide grid support services at different timescales. The supply air fan has fast dynamics, and is suitable for high frequency ancillary services, while heat pumps with variable speed drives are another potential resource. Chillers, even those without variable speed drives, can also be used to provide ancillary services by indirectly varying the load on them[50][52][54][163]. Moreover, most commercial buildings are equipped with BEMS making the task of implementing additional control algorithms easy and inexpensive. In commercial office buildings, the common sources of flexibility are highlighted in Table 2-12[258][255]. Table 2-12: Flexibility sources and strategies in commercial buildings System Strategy HVAC Precooling Global temperature set point adjustment Exhaust and inlet air fan speed reduction Chiller quantity reduction Light Turning off lighting in auxiliary spaces Turning off lighting in perimeter zones Dimming ballasts Appliances and plug-loads Turning off non-essential appliances 35

50 LITERATURE SURVEY As can be observed from Table 2-12, flexibility in office buildings is often sourced from systems tasked with providing comfort in building. This includes visual, thermal and indoor air quality. The quality of life in buildings (comfort conditions) is determined by these three parameters[83]. Thermal comfort is determined by the index Predictive Mean Vote (PMV)[259]. Visual comfort is determined by the illumination level (measured in lux)[260]. Indoor air quality can be indicated by the carbon dioxide (CO2) concentration in a building[98]. These systems are often operated based on given standards and recommendations[98][259][260] that ensure the indoor conditions are maintained at levels that guarantees a comfortable and healthy indoor environment for occupants. Hence, in harnessing the energy flexibility of these systems for grid support, it is essential that occupants demand for energy for comfort be satisfied while delivering a reliable resource to the power system when required[256][261]. This is particularly essential in commercial buildings where a strong link between occupants comfort and productivity has been demonstrated[158][262]. Moreover, compared to residential buildings, the complexity of commercial buildings makes participation costly and challenging. This is because in addition to requiring occupants to actively participate, the capacity for error in manually executing load reductions can lead to even greater costs, due to facility operation problems, unexpected losses in productivity and occupant comfort degradation[157][263][264][265]. Therefore, as described by the EBC annex 67[266], the utilization of the energy flexibility of buildings must be subject to local indoor climate conditions such as thermal, visual and indoor air quality requirements as well as user needs and preferences. The challenge is therefore achieving effective management and coordination between functions in conflict, i.e. to uphold an acceptable indoor comfort conditions while achieving the required system wide demand reduction for grid support[267]. Traditional BEMS as noted are limited in their ability to coordinate these interactions[105]. The added intelligence, flexibility, autonomous and communication capabilities of MAS does however make it a suitable tool for addressing these shortcomings[152][268][269]. 36

51 LITERATURE SURVEY The authors in [152] using a m 2 office building as test-bed, demonstrated the application of MAS in coordinating the flexibility of the office building for use in demand response. Using the energy flexibility of the ventilation, lighting and appliances achieved through reduction on in the electrical demand of these loads, a peak electricity reduction of between 14-24% was achieved without any noticeable deterioration in occupants comfort. Load reduction was mostly achieved through reduction in the ventilation air supply and zone temperatures, as well as automated dimming off lights and unused appliances. To facilitate the use of the building flexibility however, in addition to extensive use of information on occupancy and appliance use, functionalities that enabled detailed analysis and nimble access to data in real-time were also incorporated in the BEMS. Similarly, the authors in [270] developed a dual agent-based management system for the inter-operation of the smart grid and smart building. The abilities of the developed agent system were evaluated in a virtual multi-zone buildings connected to a LV distribution feeder. It was shown that with the use of basic operation rules, the proposed system could effectively improve the voltage profile of the feeder, while ensuring acceptable comfort levels. The authors in [271] also proposed a multi-agent control approach and demonstrated its effectiveness in solving the demand response problem in a multi-zone office building. A key element in the MAS framework as noted by the authors[271] was the integration of a near-optimal heuristic rule within the multi-agent control formulation that aided convergence. A number of studies have also applied MAS in residential buildings for building flexibility coordination[272][273][274][275][276]. These MAS frameworks can nonetheless be applied in office buildings as well. Notable amongst these applications is the concept of the PowerMatcher[273]. The PowerMatcher operates as a competitive MAS based coordination system. It implements supply and demand matching using a multi-agent system and a market-based control approach. The PowerMatcher is in particular able to manage conflicts between interacting entities within the smart grid. The authors in [267] also developed a MAS coordinated system consisting distributed units with communication ability that can measure sensor data and control the heating 37

52 LITERATURE SURVEY system in individual buildings. Realizing that occupants participation in demand side flexibility diminishes over time, participation was achieved via a MAS system installed in each individual home to control the heating system. This enabled the system to coordinate the heating behaviour based on indoor climate limits on a system wide scale and energy. Utilizing the thermal mass of the building for energy flexibility, besides reducing peak load, the system also showed clear ability to lower the energy usage in the participating buildings, without any noticeable difference in the perceived indoor climate[267] Discussion Multi-agent systems as can be deduced from Table 2-11 and Table 2-13 have received considerable attention in building energy management applications in recent years. As noted from the reviewed articles, the application of MAS in buildings was shown to have the potential to facilitate improvement of the energy management of buildings through the utilization and processing of actionable building processes information in near realtime. Table 2-13: MAS application in office buildings for energy flexibility coordination Study Coordination method Building type Implementation Year US DOE [152] Non-market Commercial Field -test 2013 Hurtado et al[270] Non-market Commercial Simulation 2015 Cai et al [271] Non-market Commercial Simulation 2016 Although most of the studies on the application of MAS were conceptual and theoretical, a few empirical studies demonstrated its unique distributed decisionmaking, autonomous and cooperative capabilities. These capabilities of MAS were exploited in solving the complex interaction and sometimes-conflicting interest between buildings demand for energy for comfort and utilization in grid support. 38

53 OCCUPANCY DETECTION SYSTEMS EVALUATION This chapter provides details of three experiments conducted on occupancy detection systems for answering the second research question. The chapter aims to evaluate alternative occupancy detection systems that can be applied in buildings to obtain finegrained occupancy information for enhanced energy management. The results contained in this chapter are published in the following articles: T. Labeodan, W. Zeiler, G. Boxem, and Y. Zhao, Occupancy measurement in commercial office buildings for demand-driven control applications - A survey and detection system evaluation, Energy Build., vol. 93, pp , Apr T. Labeodan, K. Aduda, W. Zeiler, and F. Hoving, Experimental evaluation of the performance of chair sensors in an office space for occupancy detection and occupancy-driven control, Energy Build., vol. 111, pp , Jan

54 OCCUPANCY DETECTION SYSTEMS EVALUATION 3.1. Occupancy detection system design and evaluation In the preceding chapter, the key strengths and drawbacks of occupancy detection systems commonly used in building operations were discussed and outlined. A key point that can be deduced from the survey is that most systems commonly used in buildings for occupancy detection have inherent limitations that affect their performance when used in buildings for energy management. The use of a fusion of multiple sensors was however identified as a viable means of obtaining fine-grained occupancy information for use in buildings to enhance energy demand management. Therefore, in testing the hypothesis and sub questions outlined on the use of increased sensor data on building occupancy information, three experiments where conducted and are outlined in Figure 3-1. The first experiment aims to evaluate the use of chair sensors as heterogeneous occupancy detection systems for use in building operation. Specifically, it seeks to evaluate the detection systems ability to provide information on the components of fine-grained occupancy information as depicted in Figure 2-1. The choice of chair sensors was based on two main reasons: Users Chair detection system design and test Chair based occupancy detection PIR motion detection Users Chair detection systems evaluation Occupancy based lighting control Figure 3-1: Experiment outline 1. office workers depending on task and function typically spend between 65% to 75% of their working hours seated[277]. The determination of occupant s presence at the workspace via sensors embedded in the chairs can thus provide useful information on occupants presence and location[212][278]. 40

55 OCCUPANCY DETECTION SYSTEMS EVALUATION 2. unlike other sensors such as PIR sensors, chair sensors are not susceptible to false a negative error, which typically occurs when occupants are still for an extended period of time. In addition, chair sensors have been shown by a few authors[212][278] as easy to install and low maintenance occupancy detection technique. For the second experiment, the performance of three different chair sensors using a mechanical contact switch, strain and vibration sensors were evaluated in an office space. Lastly, for the third experiment, due to observed limitations of the chair sensors, a fusion of the chair sensors and motion sensors developed and evaluated in an office space for occupancy-based lighting control Experiment I- Chair detection system design and evaluation The goal of this experiment was to evaluate the performance of alternative heterogeneous occupancy detection systems, in this case chair sensors in building operation. Two different occupancy detection sensors were utilized in this experiment; chair and carbon dioxide sensors. Carbon dioxide sensors were chosen over PIR motion sensors because of its ability to provide occupancy information on count as noted from the preceding chapter and Table 2-3. I. Chair-based occupancy detection system Seat cushion Z-wave wireless transmitter switch chair Figure 3-2: Chair design components For this experiment, the chair sensor was developed from low cost devices and components. This included as depicted in Figure 3-2 a conference room chair, a cushion with dimensions 42cm x 4cm x 5cm, 8 switches and a wireless analogydigital transmitter box using the z-wave[279]. z-wave is a growing wireless communication protocol that is commonly used for control, monitoring and status 41

56 OCCUPANCY DETECTION SYSTEMS EVALUATION reading applications in residential and light/small commercial environments[280][281]. Eight switches connected in parallel were used in the design to ensure presence detection even when users are improperly seated. II. Carbon dioxide sensor The utilized CO 2 sensor is depicted in Figure 3-3. The sensors have a measurement range of ppm and accuracy of greater than 1% of the span. The CO 2 sensors were placed at the inlet and outlet air ducts of the test space. The sensors were also recalibrated with air containing 1000 ppm of CO 2 before they were installed. Figure 3-3: CO 2 sensor[282] III. The test space Figure 3-4: Test-Bed conference room, chairs and wireless door contact sensor The test was conducted in the conference room of a university building depicted in Figure 3-4. The test space has a seating capacity of 25. The detection systems were also connected as depicted in Figure 3-5. The chair sensors were connected wirelessly, while the CO 2 sensors were connected physically to a data-logger 42

57 OCCUPANCY DETECTION SYSTEMS EVALUATION situated in the test space. The use of the room for meetings is coordinated via a Microsoft outlook meeting planner that is accessible by students and staffs. In addition to the occupancy sensors described above, a wireless door contact sensor was also fitted to the test room entrance as depicted in Figure 3-4. The designed sensors were placed on all the chairs in the conference room as depicted in Figure 3-4. Chair Detection Gateway Users CO2 Detection Data Logger Computer Figure 3-5: System set-up The data from each of the chair sensors was sent to the z-wave gateway and then transferred to an online database. On the other hand, data from the CO 2 sensors was collected locally via a data logger. The data from the CO 2 sensors were logged in steps of 1 minute. The test was conducted for a week between the 27 th and 31 st of October Ground truth data was recorded at intervals of 15 minutes during periods the room was indicated reserved on Microsoft outlook meeting planner Results and discussion Table 3-1 provides details of the test space usage schedule for 5 days between the 27 th October and 31 st of November 2014 as provided by Microsoft outlook meeting planner. Table 3-1: outlook meeting Time (HH:MM) Time(HH:MM) Time(HH:MM) Time(HH:MM) Monday 27 th October 9:30-10:00 11:00-12:00 No. of indicated attendees 4 12 Tuesday 28 th October 9:00-12:00 13:00-14:00 14:00-15:00 15:00-17:30 No. of indicated attendees Wednesday 29 th October 10:00-13:00 16:00-17:00 No. of indicated attendees 2 1 Thursday 30 th October 10:00-11:00 12:45-13:15 14:00-16:30 No. of indicated attendees Friday 31 st October 14:00-15:00 No. of indicated attendees 2 43

58 OCCUPANCY DETECTION SYSTEMS EVALUATION Figure 3-6: Difference between Outdoor and Indoor CO 2, chair occupancy and CO 2 computed occupancy (27 th ) Figure 3-7: Difference between Outdoor and Indoor CO 2, chair occupancy and CO 2 computed occupancy (28 th ) 44

59 OCCUPANCY DETECTION SYSTEMS EVALUATION Figure 3-8: Difference between Outdoor and Indoor CO 2, chair occupancy and CO 2 computed occupancy (29 th ) Figure 3-9: Difference between Outdoor and Indoor CO 2, chair occupancy and CO 2 computed occupancy (30 th ) 45

60 OCCUPANCY DETECTION SYSTEMS EVALUATION Figure 3-10: Difference between Outdoor and Indoor CO 2, chair occupancy and CO 2 computed occupancy (31 st ) Results from the CO 2 and the chair sensors for the five test days are depicted in Figure 3-6 through Figure The difference between the indoor and outdoor CO 2, the chair detection recorded occupancy and the CO 2 computed count was presented. Count information was obtained from the CO 2 detection system using the steady state relation provided in[283][284]: Where; NNNN. OOOOOOOOOOOOOOOOOO = SSSS (NN CC ii) GG 10 6 (1) N = the space CO 2 concentration at the present time step in ppm SA = the supply air flow rate in m 3 /s C i =the CO 2 concentration in the supply air in ppm G = the CO 2 generation rate per person in m 3 /s The value of the supply airflow rate SA, was determined using an air-flow-matching measurement device (Acin FlowFinder). This was determined to be approximately 0.11 m 3 /s The estimated value of the amount of CO 2 generated per person G was assumed as 5.3 x 10-6 m 3 /s[283][284]. 46

61 OCCUPANCY DETECTION SYSTEMS EVALUATION Table 3-2: CO 2, chair and ground truth obtained occupancy count information Test on the 28 th Test on the 30 th Time (hh:mm) Co 2 estimated Chair recorded Groundtruth Time (hh:mm) CO 2 estimated Chair recorded Groundtruth 9: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : Table 3-2 presents the data from two test days (28 th and 30 th ) with the highest space occupancy. As can be observed from Figure 3-7 (the 28 th ), 4 occupants were present in the space at 9:00 am as indicated by the chair detection system, but the ground-truth data indicated on 5 people were present at the time. The 5 th occupant at the time the ground truth data was recorded was present in the room, but not seated. For the CO 2 detection system, presence was indicated approximately 25 minutes after the first 47

62 OCCUPANCY DETECTION SYSTEMS EVALUATION occupant was seated. The same observation was made on the 30 th when presence was observed from the CO 2 detection system after approximately 60 minutes. The observable elongated response time of the CO 2 detection system in this particular case can be linked to open doors as depicted in Figure 3-11 and the low number of occupants. Figure 3-11: Door positon on the 28th and 30 th As can be noted from Table 3-2, with the exception of periods when occupants were standing in the space, the chair detection system was able to provide the required occupancy information on user presence. Occupancy information on count was thereby determined by a summation of the values obtained from each individual sensor. The chair detection system s measurement error in relation to occupancy information on count for both test days was determined using the values of the ground truth data outlined in Table 3-2. The ground-truth recorded values between 9:00 and 16:30 hr on the 28 th and the ground-truth values recorded between 10:00 and 17:00 hr on the 30 th were used in computing the detection system s error. For the 28 th, a comparison of the recorded ground-truth occupancy information on count and the chair detection system recorded occupancy as indicated in Table 3-2 gives an error measurement of approximately (17/31 x 100) = 55%. While on the 30 th the computed error was 48

63 OCCUPANCY DETECTION SYSTEMS EVALUATION approximately (8/29 x 100) = 28%. For the CO 2 detection system, the error with respect to occupancy information on count was 100% for both days Experiment summary In the preceding, commonly used heterogeneous occupancy-detection systems used in buildings were noted as having drawbacks that hamper their ability to enhance building energy management. In this experiment, the application of chair based occupancy detection was evaluated in a conference room. In addition, CO 2 sensors were also installed in the test-bed for occupancy detection as well. The drawback of the CO 2 occupancy detection system verified through this experiment. The sensors had a slower response and factors such as opening of doors was also noted as having significant influence on the performance of CO 2. The chair detection system on the other hand performed better but with errors of only up to 55%. The error was largely due to the presence of standing occupants, which visibly the sensors are unable to detect. Confining the use of the system to work spaces or in combination with other sensors can however improve the systems performance. In addition to the sensors inability to detect standing occupants, another drawback of the system was its sensitivity to changing occupant-seating position. Occupants constantly change their seating position, which caused the switches to momentarily flip between the ON and OFF positons. These movements by occupants could trigger a change in state of connected building systems. As the change in state was observed to be in the magnitude of seconds, the system was configured to return the average value from 6 measurement samples every minute. This ensured such changes in occupants seating position did not negatively influence the system s performance Experiment II- Evaluation of different chair sensing technologies In the preceding experiment, a chair detection system based on the use of a contact switch and a wireless transmitter was utilized in the conference room of a university building. As noted from the experiment, the chair detection system exhibited measurement errors of up to 55% due to its in ability to detect standing occupants. As mentioned earlier, office workers depending on task and function typically spend 49

64 OCCUPANCY DETECTION SYSTEMS EVALUATION between 65% to 75% of their working hours seated[277]. Therefore, in this experiment, the chair detection system was modified for use in the workplace and its performance evaluated alongside two other chair detection system using strain gauge and vibration sensors. I. Strain sensor A resistance strain gauge consists of a resistive element whose physical dimensions changes on application of force. As depicted in Figure 3-12, the application of force to the object on which the strain gauge is mounted, causes tension or compression leading to a change in length and consequently a change in the resistance of the element[285]. Figure 3-12: Strain Gauge Operation[286] Strain gauge Mechanical switch Sensor Vibration sensor Figure 3-13: Chair with mechanical-switch, vibration and strain sensors The strain gauge and its circuitry were embedded in a chair as shown in Figure Occupants presence or absence on the chair causes a change in the resistance of the 50

65 OCCUPANCY DETECTION SYSTEMS EVALUATION strain gauge; the resistance value was converted to a voltage, which was then transmitted via a wireless low power Bluetooth[287] transmitter to the gateway. The data was subsequently transmitted via a local network connection as shown in Figure 3-14 to a local database. Four strain sensors connected in parallel were used for the setup. Strain sensor Strain sensor Strain sensor Users Strain sensor A/D converter Wireless transmitter Gateway Computer Figure 3-14: Strain sensors connection II. Vibration Sensor The vibration sensor as depicted in Figure 3-13 was composed of a microcontroller-based tri-axial accelerometer with the ability to measure acceleration in three directions. Triaxial accelerometers are often used in applications involving monitoring and assessing human movements and physical activity[288][289]. Although the sensors measure acceleration, other derivative properties such as vibration, shock, and tilt can be determined from the measured signal[289][290]. By measuring the vibrations in the chairs, the sensors are able to determine occupants presence or absence. The vibration sensor s output was also connected via a low-power Bluetooth wireless transmitter from which the collected data was sent to a local database. III. Mechanical-switch sensor The mechanical-switch sensor also depicted in Figure 3-13 is similar in operation to that described earlier. The major difference in this experimental set-up was the modification to the sensor s construction and placement. Unlike in the previous design where a 51

66 OCCUPANCY DETECTION SYSTEMS EVALUATION cushion was used, the switch was in this case embedded in the chair s seating pan. In addition, only one contact switch was used as against 8 used in the previous experiment. Strain Sensor Vibration Sensor Gateway Users Computer Contact Sensor Gateway Figure 3-15: Sensor Connection Figure 3-16: Chair sensors set-up All three sensors were connected on a typical office chair with rollers as depicted in Figure The test was conducted in an office space depicted in Figure 3-16 for a period of four days between the 7 th and 10 th of September Occupants mainly carried out desk related tasks such as typing, reading and writing. Participants were provided with a log book on which entries were made on their first arrival time, last departure time as well as transition time. 52

67 OCCUPANCY DETECTION SYSTEMS EVALUATION Results and discussion Figure 3-17: Vibration, strain and chair sensors (day 1) Figure 3-18: Vibration, strain and chair sensors (day 2) Figure 3-17 to Figure 3-20 depicts the result from the evaluation for the four test days. Table 3-3 also provides details of occupant s first arrival and last departure time as well 53

68 OCCUPANCY DETECTION SYSTEMS EVALUATION as the logged transitions. Table 3-4 provides a comparison of the sensors registered transitions and occupants logged transitions. Figure 3-19: Vibration, strain and chair sensors (day 3) The sensors errors were computed using these values. The vibration sensor on the first test day had errors of (((7-5)/5) x 100) = 40%, while the strain and contact sensors had errors of (((6-5)/5) x 100) = 20% and 0% respectively. On the third test day, the strain and contact sensors as indicated in Table 3-4 both recorded errors of 0%, while the vibration sensor had errors of 20%. On the third test day, the test subject failed to register one transition state. This transition was however captured by all three sensors. Table 3-3: Summarized logs Test day/events First arrival time Number of transitions Last departure time Test day 1 08: :30 Test day 2 08: :45 Test day 3 08: :50 Test day 4 08: :00 54

69 OCCUPANCY DETECTION SYSTEMS EVALUATION Table 3-4: Logged and sensor registered transitions Test Day Logged transitions Vibration sensor Strain sensor Contact sensor Test day Error 40% 20% 0% Test day Error 20% 20% 0% Test day 3 4(5)* Error 40% 0 0 Test day Error 20% 20% 0 * Actual number of transitions was 5 as occupant missed one record which was captured by the sensors Figure 3-20: Vibration, strain and chair sensors (day 4) Experiment summary In the first experiment with the chair detection system, errors of up to 55% were observed when the sensors were applied in a conference room. The recorded errors were mainly as a result of standing occupants in the test-space. In this second experiment, in addition to the contact switch chair detection system, two additional chair detection systems using strain and vibration sensors were evaluated in an office set-up. As observed from the experiment results, errors of up to 40% and 20% were noticeable from the vibration and strain chair sensors respectively. The contact chair sensors however had 100% accuracy when compared with the transitions logged by occupants. 55

70 OCCUPANCY DETECTION SYSTEMS EVALUATION The vibration sensors were observed to be highly susceptible to movement, which resulted in higher errors. Though susceptibility to movement resulted in false positive errors, i.e. the detection system indicating presence when there was none, this was also advantageous. The vibration sensors could as a result detect occupancy when occupants make movements around the chairs. This enabled presence to be determined before occupants were actually seated Experiment III- Sensor fusion application in energy management The availability, access and use of improved occupancy information in building operation as noted in the preceding chapters can enhance energy management in buildings[291]. Lighting energy use in buildings for example accounts for up to 40% in building energy consumption[19][102]. The use of fine-grained occupancy information has however shown to have the potential to reduce this demand significantly[101][292]. Sensor fusion as noted from the preceding section is one of the cost effective means of obtaining fine-grained occupancy information for enhancing energy management in building operation. In the preceding experiment, chair detection system was introduced and its performance evaluated for use in buildings. Though the sensors have been shown to be capable of providing detailed information on occupancy, it still has a few drawbacks in particular when occupants are standing at their workspace. Occupants often stand at their workplace, while settling in at the beginning of the day, clearing out at the end of the day and while fetching items from their workplace. The use of chair detection would however lead to false negatives during these periods. Therefore, to address the drawback a fusion of chair sensors and PIR sensors was developed and evaluated in the test-space depicted in Figure 3-16 for occupancy-based lighting control. PIR motion detection systems as noted in chapter two are one of the commonly used occupancy detection systems in buildings for energy management[293][294]. The detection system like others outlined in chapter 2, have a few drawbacks one of which is susceptibility to false-negatives[103]. Given that the chair detection systems on the other hand are only susceptible to false-positives, the combination of both detection systems forms a robust occupancy system for energy management in buildings. 56

71 OCCUPANCY DETECTION SYSTEMS EVALUATION The experiment was conducted in the workspace depicted in Figure 3-16 for two days between the 15 th and 18 th of September For controllability, wireless control nodes using the z-wave communication protocol were utilized. The armatures were connected to the mains supply via the z-wave nodes depicted in Figure The lighting armatures consisted of 2 x 45 watt TL florescent lamps. The wireless node in addition to enabling controllability of the armatures also able measures the energy consumption of the connected load. The node has a measurement range of W, a connectivity range of 100 meters and up to 300 meters when connected in a mesh[295][296]. The motion sensor also depicted in Figure 3-21 has a detection angle of 60 degrees and range of 7 meters. Figure 3-21: z-wave control node and motion sensor[297] Chair Detection Figure 3-22: occupancy detection systems placement Two motion detection systems and the switch sensor based chair detection system as depicted in Figure 3-22 were installed at the test space. One motion sensor was focused 57

72 OCCUPANCY DETECTION SYSTEMS EVALUATION directly on the participant s upper torso area and the other was focused on occupant s hands. Wall switch Figure 3-23: System Connection The placement of the sensors at very close proximity to participants was to ensure participants were constantly in the direct line of sight of the sensor to reduce false negative errors. All components of the test were connected as depicted in Figure 3-23 and control applications designed in a Java programming suite. The test as cited earlier was conducted over a period of four days. On the first test day, the lighting system was operated by the occupant using the wall switch and without occupancy-based control. On the second test day, the chair detection system alone was used for control of the lighting system. On the third test day, only the motion sensors were used. In order to prevent the system from turning off the lighting in the space during periods of inactivity, a time delay setting of 5 minutes was incorporated in the control suite. On the fourth test day, a fusion of all three sensors (two motion sensors and one chair sensor) was used for control of the lighting system. To enable comparison of the test results, testsubjects followed the same schedule as indicated in Table 3-5. Table 3-5: Test schedule for test days Event Arrival Transition 1 Transition 2 Transition 3 Departure Time 08:30 10:00-10:30 12:00-13:00 14:30-15:00 16:30 58

73 OCCUPANCY DETECTION SYSTEMS EVALUATION Results and discussion Figure 3-24 depicts the result from the first test day without the use of any occupancybased control. The measured electrical energy consumption of the lighting system on this test day was approximately kwh. Figure 3-24: Test Without occupancy-based control Figure 3-25 and Figure 3-26 depicts the result from the test on the second and third test days with the chair and motion sensors respectively. The total lighting energy use as recorded by the wireless power nodes on both days was approximately kwh and kwh respectively. From the first test with the chair-controlled lighting, the test subject expressed dissatisfaction with the system due absence of light on arrival and the rapid switching of the light on departure. These shortcomings were however addressed on the third test day with the use of both motion sensors. The use of both motion sensors and the time delay for the third test also reduced the occurrence of false negative triggered errors. 59

74 OCCUPANCY DETECTION SYSTEMS EVALUATION Figure 3-25: Chair sensor controlled test Figure 3-26: Motion sensors controlled test 60

75 OCCUPANCY DETECTION SYSTEMS EVALUATION Start Turn Off Get sensor data Database Is armature ON? NO Is motion or chair true? YES NO Turn ON light Do nothing NO Is motion true & chair false YES Turn off light after 2 mins End Figure 3-27: Control sequence For the fourth test with all three sensors, the control algorithm depicted in Figure 3-27 was utilized. The control sequence was designed to ensure the lighting system remained on while occupants were seated and when occupants were standing. The control sequence was also designed to turn off the lighting system presence was not detected by the chair sensors after two minutes of presence being detected by either of the two motion sensors. This ensures occupants are provided with sufficient desk lighting while settling in at the beginning of the day, clearing out at the end of the day or while fetching items from their workplace while standing. Figure 3-28 depicts the test result from the use of all three sensors. The measured energy consumption of the lighting system on this test day was approximately kwh. As can be observed from Figure 3-25 and Figure 3-28, the chair detection system had errors of 0%. This was because the 61

76 OCCUPANCY DETECTION SYSTEMS EVALUATION number of observed transitions by the test-subject (3) was equal to the number of transitions recorded by the chair sensors. Figure 3-28: fusion of all three sensors Figure 3-29: PIR sensor false errors (17 th ) 62

77 OCCUPANCY DETECTION SYSTEMS EVALUATION Figure 3-30: PIR sensors false errors (18th) Figure 3-31: Lighting energy use for each test day On the contrary, as can also be observed from Figure 3-26 and Figure 3-28, the motion detection systems individually recorded substantial false negative errors. The false negative errors recorded by each individual motion sensor on the third and fourth test days are depicted in Figure 3-29 and Figure Due to the positioning of the sensors at different locations, the use of a 5 minutes time delay during the third experiment was observed to be sufficient though periods of non-activity as high as 12 minutes was 63

78 OCCUPANCY DETECTION SYSTEMS EVALUATION observed from the individual PIR motion sensors. For the fourth experiment, the use of the combination of all three sensors eliminated the need for the use of the time delay Experiment summary In this experiment, a fusion of two motion sensors, a chair sensor and a fusion of the chair and PIR motion sensors were separately applied in a test-bed office space for occupancy based lighting control. On the first test day, which was the control experiment, the total lighting electrical energy use was measured as approximately kwh, and on the second, third and fourth test days the electrical energy consumption of the lighting system as depicted in Figure 3-31 was kwh, kwh and kwh respectively. Compared to the first/control test, the total reduction for the second through the fourth test was approximately 28%, 23% and 27% respectively. The test on the third day contributed the least savings due to the use of a 5 minutes time delay. On the other hand, the chair detection system contributed the highest savings. The high reduction of the chair sensors did however result in abrupt switching of the light while occupants were in the process of exiting their workplace as well as non-availability of light when within close proximity to the space. The fusion of both sensors, i.e. the motion and chair sensors did however address this drawback. Furthermore, each individual motion sensor was observed to have still exhibited significant false negative errors during periods the test subjects were seated still. The chair sensors on the other hand did not produce such errors. Although the use of two PIR motion sensors focussed and positioned differently reduced the errors, a time delay was still required. Furthermore, it was also observed that the use of a fusion of one PIR sensor and the chair sensor would provide the same as the use of a fusion of two PIR motion sensors and the chair sensors. This is because the chair sensors as noted from the experiment produced errors of 0% but only exhibited errors during occupants arrival and departure from the space. The use of one PIR motion sensor can however be utilized to obtain this information. 64

79 MULTI-AGENT SYSTEMS APPLICATION IN ENERGY MANAGEMENT The chapter aims to evaluate the added value of multiagent systems as a computational support tool for use in buildings to enhance building energy management. This chapter aims to provide answers to the third research question. Some of the results in this chapter are published in the article listed below: T. Labeodan, C. De Bakker, A. Rosemann, and W. Zeiler, On the application of wireless sensors and actuators network in existing buildings for occupancy detection and occupancy-driven lighting control, Energy Build., vol. 127, pp ,

80 MULTI-AGENT SYSTEMS APPLICATION IN ENERGY MANAGEMENT 4.1. MAS application and agent design framework As noted in chapter two, there is growing interest in the application of computational intelligence in buildings for enhancing energy management as well as buildings interaction with the smart-grid. Multi-agent systems due to its flexibility, autonomous and communication properties, is considered a suitable computational intelligence tool that can be applied in buildings to enhance energy management as well as buildings interaction with the smart-grid[108][298]. In this chapter, two experiments as depicted in Figure 4-1, in which the application of MAS in a test-bed office building for coordination of occupancy-based lighting control for enhanced energy management was evaluated is presented. Experiment One Users Chair based occupancy detection PIR motion detection... Occupancy based lighting control... Experiment Two... Agents Users PIR motion detection Time delay setting reduction Figure 4-1: Experiment Schematic Architectural systems and functional abstraction levels MAS representation of each Level Built Environment... Operational Scope of typical traditional BEMS Building Floor/Zone Room Workplace Users & Devices Figure 4-2: Building abstraction layer and agent representation The utilized MAS framework as depicted in Figure 4-2 is representative of the functional abstract and architectural systems levels. 66

81 MULTI-AGENT SYSTEMS APPLICATION IN ENERGY MANAGEMENT As noted by the authors in [159], in Building Evaluation Models, office buildings are divided into six levels of scale: (1) work station/user, (2) workplace, (3) Room (4) floor/zone, (5) building, and (6) built-environment. These architectural system functional and abstraction levels represent the level of architectural dimension or detail in which decision-making takes place. These levels of abstraction are also reflected in the design of building systems particularly HVAC & L systems as well as building energy management systems. As noted from Figure 4-2, the agents represent all levels within the building including users and devices. This enables input from users as well as all appliances within the building, which as depicted in Figure 4-2 is seldom utilized in traditional BEMS. A hierarchical organization was used in the design of the MAS to aid management and coordination. Based on the hierarchical agent structure, more agents are concentrated at the lower levels comprising mainly the workplace and user levels. The agents in the system decreases up the hierarchy as illustrated in Figure 4-2. As opposed to a market based coordination mechanism mentioned in chapter two, the developed agent system was rule-based. In the first experiment, a fusion of the chair and motion sensors utilized in the preceding chapter (section 3.4) was installed in an open plan office space of the test bed office building depicted in Figure 4-3. Figure 4-3: Test-bed office building The test-bed building is an office building located in the city of Breda in the Netherlands. The building as depicted in Figure 4-3 is a 2-storey building with a floor area measuring approximately 1500 m 2. For possible interaction with the smart-grid, photovoltaic and electrical storage systems were also installed in the building (Appendix 67

82 MULTI-AGENT SYSTEMS APPLICATION IN ENERGY MANAGEMENT C). The building is composed of 4 large open-plan spaces, a conference room, a lunch room and limited numbers of singular and double cell office spaces. The building is equipped with web based BEMS tasked with control and management of the building comfort systems. The building has a peak occupancy density of approximately 60, but daily average occupancy ranges between 25 and 30. As with a significant number of office buildings as noted by the authors in [89], the test-bed operates based on assumed occupancy profiles and schedules. The HVAC and lighting systems are operated continually during the building operational hours, which is typically between 7 am and 6 pm. Occupancy sensors are only present in the rest and print rooms. The occupancy sensors are PIR motion sensors and are integrated with the lighting systems in the rooms. Figure 4-4: The EVE agent design platform The detection systems in these rooms were configured with a time delay of 10 minutes. The 10 minutes time delay setting was introduced by the building managers to reduce unnecessary switching of the lights when occupants were present in these rooms but seating or standing still. Time delays are commonly used with PIR controlled lighting systems in office buildings. This reduces the risk of the system becoming a nuisance to 68

83 MULTI-AGENT SYSTEMS APPLICATION IN ENERGY MANAGEMENT occupants. Time delay settings of minutes are typical and often results in unnecessary energy being utilized for lighting[192][299]. For control of lighting systems with the fused sensors positioned in the test openplan space, a MAS framework was developed in the EVE agent design platform. The EVE agent design platform as depicted in Figure 4-4 is a multipurpose, web-based agent platform developed in the Netherlands and based on the JAVA programming language[300]. Though there are other comparable agent development platforms based on JAVA such as JADE [144][301], the choice of the EVE agent platform was based on it scalability and massive parallelization In addition, compared to other agent development platforms, the eve-agent platform is locally supported, stable, easy to use and web-based. For the second experiment, the designed MAS framework was applied in the test-bed office building to enhance the operation of the installed motion sensors. Specifically, MAS coordination and information exchange between agents was utilized in effecting significant reduction in the configured time delay setting of the PIR sensor in the print room Experiment I - MAS coordinated occupancy based lighting control This experiment was conducted in the open-plan workspace with the highest occupancy density in the test bed office building. The open plan space as depicted in Figure 4-5 has 12 workspaces. The open plan workspace is illuminated with 31 x (2 x 36W) TL luminaries that are centrally controlled. To achieve controllability of the lighting armatures, wireless sensors nodes with switching and measurement functions described in section 3.4 and depicted in Figure 4-6 were utilized. For occupancy detection, a fusion of the chair sensors and wireless PIR motion sensor also described in section 3.4 and depicted in Figure 4-6 were utilized. From the results from section 3.4, the fusion of the chair sensor and the motion sensors was observed to have produced errors of 0%. Therefore, each of the 12 workspaces in the test-bed open-plan workspace was fitted with chair detection and a motion detection system. As against the use of two PIR motion sensors, only one PIR motion sensor was utilized. The use of 69

84 MULTI-AGENT SYSTEMS APPLICATION IN ENERGY MANAGEMENT one PIR motion sensor as noted in section 3.4.2, would also provide the same result since the chair detection system were observed to have produced 0% false negative and positive errors. For this experiment, the motion sensor was positioned directly above occupants workplace. Data was collected for a period of three weeks between the 22 February and 11 March 2016 in the test-bed. At the end of the three-week period, occupants were interviewed to obtain their feedback wp11 wp wp10 wp wp2 wp3 wp6 wp wp1 wp4 wp5 wp Figure 4-5: Test bed space with armature placement and occupants workplace The chair sensors, motion sensors and control nodes all communicated via the z-wave protocol and were connected as depicted in Figure 4-7. Figure 4-6: Chair sensors, motion sensors and wireless control node 70

85 MULTI-AGENT SYSTEMS APPLICATION IN ENERGY MANAGEMENT Figure 4-7 also depicts the structure of the designed multi-agent system. Although the MAS has a hierarchical structure, the agents are loosely coupled and the position of agents does not limit the agents ability to communicate with agents positioned layers above or beneath. Switch/Meter Gateway Database Computer Motion sensor Users Chair sensor Room agents Detection system Admin agent... User agents... Device agents Figure 4-7: System connection and MAS structure The designed MAS as depicted in Figure 4-7 is composed of four agent types described below: Device agents- Within the MAS framework, devices such as motion sensors, the chair sensors and the controllable armatures are represented by device agents. The main function of the device agents is to provide sensors within the framework state and sensory information. These include information on occupancy, the state of each represented armature as well as the energy consumption of each represented armature. User agents- Every occupant in the test bed was represented with a user agent within the MAS framework. Using occupancy and state information provided by the devices agent, the user agent controlled the armatures at the represented occupants workplace. Room agent- The test room was also represented by a room agent in the MAS framework. The task of the room agent was to determine the total number of occupants in the room at every point in time using information provided by users and device agents. 71

86 MULTI-AGENT SYSTEMS APPLICATION IN ENERGY MANAGEMENT Admin agent- The admin agent as the name implies performs is responsible for management of all agents within the MAS framework. The admin agent contains the location/address and function of each agent. Table 4-1: armature mapping Workplace Primary Armatures Aux. Armatures WP1 1, 2 3 WP2 9, 20 WP3 10,19 WP4 2,3 WP5 4,5 WP6 11,12,17 WP7 13,14,16 WP8 6 5 WP WP10 25,26 WP WP , 27 Start Turn Primary & Aux.armature OFF YES Turn Primary armature OFF NO Get sensor data (device agent) Aux. unoccupied & ON? YES Aux. armature present? YES Is primary Armature ON? NO Is motion or chair true? NO YES NO Do nothing Turn ON primary & Aux. armatures NO Is motion true & chair false Do nothing YES NO Turn OFF Primary after 2 mins NO Is Aux present? YES Aux. unoccupied & ON? YES Turn AUX & Primary OFF after 2 mins Figure 4-8: User agent control sequence The EN lighting standard[260], recommends a minimum of 500 lux for desk related activities such as writing, typing and reading. In the test-bed open-plan office space, the required minimum lighting intensity as indicted by Table 4-1 is provided with 72

87 MULTI-AGENT SYSTEMS APPLICATION IN ENERGY MANAGEMENT a combination of armatures. The primary armatures are those, which provide lighting to only one user while the auxiliary armatures provide lighting to multiple users. The user agent utilized the mapping outlined in Table 4-1 and the control sequence depicted in Figure 4-8 for control of the armatures Results and discussion Figure 4-9: Space Occupancy during first week of the experiment Figure 4-9 depicts the space occupancy during the first week of the experiment and Figure 4-10 depicts the lighting energy consumption of the test-bed lighting system through the duration of the experiment. During the first week of the experiment, the measured lighting electricity use was measured to be approximately 113 kwh. During the second and the third week of the experiment, when the MAS controlled system was applied in the test-bed the measured lighting energy use for the test-space was approximately 81 kwh and 90 kwh respectively. This represents a reduction of approximately 28% and 20% reduction in the space lighting energy used when compared with the total lighting energy use during the week preceding the use of the occupancy-based control measures. The yearly reduction based on the use of this system in the test-bed office building was also estimated using the average savings from both 73

88 MULTI-AGENT SYSTEMS APPLICATION IN ENERGY MANAGEMENT weeks, i.e kwh. Assuming the annual yearly consumption for the test bed office space is 1356 kwh (113 x 12), the estimated annual saving would be ( )/1356= 24%. Figure 4-10: Daily lighting energy use in test space for the three-week period Figure 4-11: Room count information from room agent (week two) 74

89 MULTI-AGENT SYSTEMS APPLICATION IN ENERGY MANAGEMENT Figure 4-12: Room count information from Room agent (week three) Figure 4-13: Chair sensor, observed and motion sensor recorded transitions (week two) Count information as noted in chapter two is also an essential occupancy information that can enhance the application of demand-controlled ventilation in buildings. The 75

90 MULTI-AGENT SYSTEMS APPLICATION IN ENERGY MANAGEMENT room agents collected occupancy count information from the individual user agents in the open-plan space. The test-space count occupancy information for the two weeks duration of the experiment as provided by the room agent is depicted in Figure 4-11 and Figure During the test, the count information was also observed to be at times understated when occupants were at locations in the test-bed where occupancy sensors were not located. The accuracy of the individual occupancy sensors was also observed during the test as depicted in Figure 4-13 and Figure Figure 4-13 depicts the chair sensor, motion sensor and observed occupancy of one of the test participants on one test day during week two of the experiment. As can be observed from Figure 4-13, the test participant had 7 transition states in between arrival and departure times. Figure 4-14: Chair sensor, observed and motion sensor recorded transitions (week three) The chair detection system can be observed from Figure 4-13 to have captured these transitions, i.e. 0% positive and negative error, whereas the motion sensors displayed substantial false negative errors and transitions. Similarly, an observation of measurement result for the same test subject on a test day during the third week of the test also produced the same results. The test subject as can be observed from the measurement result in Figure 4-14 made 8 transitions between arrival and departure 76

91 MULTI-AGENT SYSTEMS APPLICATION IN ENERGY MANAGEMENT times. These transitions were recorded by the chair detection system. The motion sensors on the other hand recorded far more transitions. The fusion of both sensors however resulted in improved performance, as the chair sensors were able to counter the negative drawbacks of the PIR motion sensors. At the end of the test, all 12 participants expressed satisfaction with the occupancy detection systems and the occupancy based lighting control system in the test-bed. Participants particularly expressed satisfaction with the energy savings obtained from the use of the system but cited some dissatisfaction with the non-uniformity of lighting in the space Experiment summary In this experiment, the application of multi-agents was evaluated in an open-plan place for coordination and control of the lighting system in the open plan space of an office building using occupancy information from a fusion of chair and motion sensors. The application of wireless sensors nodes and sensor fusion facilitated the availability of fine-grained occupancy information thereby facilitating worthwhile energy reduction of up to 28% in lighting energy use during one-week application and an estimated annual energy savings of 24%. The application of the MAS framework also facilitated the availability occupancy information on count in the test bed Experiment II: Motion detection system time-delay reduction Single user office Rest-Room Open-plan Space Single user office Test-bed Open-plan workspace Open-plan workspace Print Room Stairs Rest-Room Single user office Single user office Single user office Figure 4-15: Test-bed floor. 77

92 MULTI-AGENT SYSTEMS APPLICATION IN ENERGY MANAGEMENT In the experiment described earlier, the use of a fusion of the chair and motion sensors was utilized in combination with a MAS framework to enhance energy management in the test bed open-plan workspace. In transitional spaces, such as print and rest rooms, the application of a fusion of both sensors may however not be feasible. In the next experiment, the functionalities of the designed MAS framework in section 4.2 was further extended and used in reducing the lighting energy demand of the print room based on fine-grained occupancy information. The print room highlighted in Figure 4-15 as mentioned uses occupancy based control for control of the lighting system. The room is fitted with 2 x (2 x 36W) TL luminaries and they are turned off after 10 minutes of inactivity. This experiment therefore aims to make use of the occupancy information available in the open-plan test space also highlighted in Figure 4-15 and the occupancy information from the print room to achieve worthwhile reduction in the time-delay setting of the detection system. Installed PIR System Switch/Meter Gateway Database Computer Users Motion sensor Z-wave devices Room agents Admin agent... User agents... Device agents Figure 4-16: System connection For this experiment, in addition to the occupancy sensors already installed at the open-open work space described in section 4.2, additional wireless PIR motion sensors also utilizing the z-wave wireless communication protocol were installed in the print room. Wireless control nodes having measurement capabilities used in section 4.2 were also installed in the print room. The wireless motion sensor and control nodes were 78

93 MULTI-AGENT SYSTEMS APPLICATION IN ENERGY MANAGEMENT connected in series with the already installed PIR motion sensor as depicted in Figure The MAS framework as also depicted in Figure 4-16 was utilized for control of system. Each user in the open-plan space was represented as a user agent and devices such as sensors and armatures represented as device agents. Both rooms, i.e. the open-plan space and print room were represented as room agents. The communication between the agents in the system is depicted in Figure The user agent as depicted communicates users presence or absence to the room agent. When absence is established, the user agent request presence information from other room agents in the framework. In this case, the print room agent returned user presence information. User Agent Work Room: Room Agent Print Room: Device agent Print Room: Room Agent user absent user present user present user present user Absent Figure 4-17: A communication sequence Table 4-2: Measured travel time to print room User Estimated travel time to print room (sec) WP 1 21 WP 5 14 WP 6 12 WP 7 14 WP 8 20 WP 9 18 WP For this experiment, the user agent utilized the average travel time in seconds between occupant s workplace and the print room outlined in Table 4-2 for determination of occupant s presence at the print room. Since information on user-identity was not collected, to identify particular occupants the user agent used the time stamp of 79

94 MULTI-AGENT SYSTEMS APPLICATION IN ENERGY MANAGEMENT occupants arrival time to establish user identity. For this experiment, the average travel time of occupants with fixed workstations in the open-plan workspace was determined by observing the test subjects over a period of one week. The result of the observation is summarized in Table 4-2. The user agent used this estimated travel time to establish occupants presence in the print room. Due to the availability of occupancy sensors only in the largest open-plan workspace, only a selected occupant in the open-plan workspace participated in the test. In normal operating mode, the lighting system in the print room is triggered on when an occupant enters the room and is triggered off after approximately 10 minutes. Using the MAS configuration, when the occupant returns to his workplace, the occupant s location is updated and made available to the print room agent and the lighting system is immediately turned off if no occupant is present in the room. If the test subject does not return to his workspace, the default configuration of 10 minutes delay holds. This experiment was conducted over a period of two days between the 14 th and 15 th of July Results and discussion Figure 4-18: Print room occupancy and lighting energy use with 10 minutes time-delay 80

95 MULTI-AGENT SYSTEMS APPLICATION IN ENERGY MANAGEMENT Occupancy and the lighting energy use in the print room on the 14 th are depicted in Figure The total electricity energy consumption of the lighting system was measured as approximately 0.24 kwh. On the second test day, when the MAS system was activated, the measured lighting energy use in the print room was measured as 0.12 kwh. As can be observed from Figure 4-19, the time delay of the motion sensor was reduced by approximately 9 minutes from 10 minutes when the user agent was able to establish occupants presence in the print room. This resulted in a reduction of approximately 90% in the time delay-setting configuration of the installed PIR motion sensor. Figure 4-19: Print room occupancy and lighting energy with time delay reduced As can be also noted from Figure 4-19, the original time delay was still operational during periods the user agent could not establish occupants presence at the print room Experiment summary In this experiment, the MAS framework was applied in the test-bed to improve the performance of the existing occupancy based lighting control through reduction in the time delay setting. Time delays between occupants departure from a space and the 81

96 MULTI-AGENT SYSTEMS APPLICATION IN ENERGY MANAGEMENT instant the light is turned off leads to unwanted use of energy particularly infrequently occupied spaces. In a large number of cases, the use of a longer time delay in buildings is informed by the need to ensure the occupancy detection system does not become a nuisance to users. This as noted in the experiment results in unnecessary use of energy in such spaces and by allowing exchange of information by existing sensors in the building as demonstrated by the experiment worthwhile performance improvement can be achieved. Effective use of these system does however demands additional occupancy information such as track and user identity which due to ethical and privacy concerns are still not commonly used in buildings for occupancy detection. 82

97 MULTI-AGENT SYSTEMS APPLICATION IN FLEXIBILITY COORDINATION The chapter evaluates the benefit of multi-agent systems as a computational support tool for use in buildings coordination of flexibility and demand response. This chapter also aims to provide answers to the third research question posed in this thesis. Some of the results presented in this chapter are presented in the articles listed below: K. O. Aduda, T. Labeodan, W. Zeiler, G. Boxem, and Y. Zhao, Demand side flexibility:potentials and building performance implications, Sustain. Cities Soc., vol. 22, pp , Apr K. O. Aduda, T. Labeodan, W. Zeiler, and G. Boxem, Demand side flexibility coordination in office buildings: A framework and case study application, Sustain. Cities Soc., vol. 29, pp ,

98 MULTI-AGENT SYSTEMS APPLICATION IN FLEXIBILITY COORDINATION 5.1. MAS application in building flexibility coordination Buildings as noted in chapters one and two are being considered as viable sources of useful demand side energy flexibility to address the energy flexibility deficit introduced by increased integration of renewable and distributed energy resources within the smart grid. Buildings are in particular now considered to hold the potential for grid support services such as demand response in the near term[53][61][302]. As noted in chapters one and two, in commercial office buildings in particular, the utilization of buildings energy flexibility for grid support service, could lead to disruptions that impair optimal building performance and user comfort[157]. Therefore, in this chapter the use of MAS is explored in the test bed office building described in section 4.1 for coordination of the buildings energy flexibility. Particular focus is placed on the use of MAS to minimize degradation in occupants comfort. This is because the utilization of a building s energy flexibility for grid support such as demand response often demands partial or complete reduction in the energy demand of certain building systems and appliances. Such actions could result in degradation of occupants indoor environmental comfort conditions. Degradation of occupants comfort can reduce productivity or prompt occupants to take actions to restore their comfort[158][215][303]. It is therefore imperative that in the utilization of buildings energy flexibility for grid support such as demand response, occupants comfort and actions during participation should be prioritized. Table 5-1: flexibility sources and strategies in test-bed System Strategy Response (min) Flexibility (kw) Condition Availability HVAC Exhaust and inlet air fan speed reduction <1 2 kw - 4 kw PPM < MIN Chiller quantity reduction & pre-cooling >1 7 kw T O C < MIN Light Turning off lighting in auxiliary spaces <1 0.4 kw User needs MIN plug-loads Turning off non-essential appliances <1 1.3 kw User needs MIN The energy flexibility of the test bed office building described in section 4.1 was quantified in an experiment detailed in [50][302][304][305] and the results summarised in Table 5-1. Although additional sources of energy flexibility in the form of photovoltaic and battery storage systems were installed in the test-bed office building during the 84

99 MULTI-AGENT SYSTEMS APPLICATION IN FLEXIBILITY COORDINATION course of the study, the scope of this experiment was limited to only the energy flexibility achievable from the building comfort systems on the room-level. Figure 5-1: MAS Framework The main sources of flexibility in the test-bed as outlined in Table 5-1 are the ventilation fans, the chiller, auxiliary lighting and appliances. The speed of the inlet and outlet ventilation fans was reduced from a nominal setting of 100% to 70% and 60% to achieve reduction of 2 kw and 4 kw respectively while ensuring the indoor CO 2 concentration remained within the acceptable band (< 1000 ppm). For chiller, the building was precooled few hours before it was completely turned off to achieve a reduction of 7 kw. The auxiliary lighting in combination with the building plug appliances such as printers, coffee machines and refrigerators also provided energy flexibility of approximately 1.3 kw. While the operation of the ventilation and cooling systems is subject to indoor 85

100 MULTI-AGENT SYSTEMS APPLICATION IN FLEXIBILITY COORDINATION thermal and indoor air quality conditions, the operation of the auxiliary lighting and appliances is subject to user needs. In office buildings, appliances such as printers and coffee machines complement building comfort systems in the provision of a conducive and productive environment for users. For this experiment, a MAS framework depicted in Figure 5-1 and composed of device, user, room, zone, building, grid, interface, service and admin agents was designed. The function of each of the agent is outlined below: Device Agent- The function of the device agent is similar to that described in section 4.4. Device agents represent all sensors, appliances and actuators in the test-bed office building. The main function of the device agents is to provide sensors within the framework state and sensory information. These include information on occupancy, temperature, carbon dioxide and the state of each represented actuator. User Agent- The user agent represents each building occupant. It communicates with the relevant devices agent to obtain information about its operating environment. The user agent as described in section 4.2 and 4.3 exchanges occupancy information with the room agents. Room Agent- The room agent in addition to the functions described in section 4.2 and 4.3 also determines the quantity of available energy flexibility present in each room based on the information provided by the devices agents in the room. Appliances and HVAC&L systems in buildings as outlined in Table 2-12 are the major sources of energy flexibility in buildings. As noted by the authors in [160], striking of a balance between energy efficiency and occupant comfort can only be achieved at the room level where occupant needs and behaviour, system response and orientation effects give rise to contradictory requirements. The room level is therefore the ideal starting point for sourcing the available energy flexibility of buildings. Zone-agent- The zone agent in this MAS framework performs the function of an aggregator. As an aggregator, the zone agent computes the sum of useful energy flexibility available from each room agent in that building zone. In the design of building services systems, building spaces with similar heating or cooling loads or orientations exposed to the same amount of sunlight or daylight are usually designated a thermal 86

101 MULTI-AGENT SYSTEMS APPLICATION IN FLEXIBILITY COORDINATION zone. Thermal zones in buildings often have to share the use of certain building systems such as ventilation, cooling, heating and solar shading. Building agent- The building agent obtains the available building energy flexibility as provided by the rooms, zones and devices agents. The building agent utilizes the available building energy flexibility for negotiating participation in any grid support services. The building agent also maintains the buildings available service table outlined in Table 5-2. The table contains information on the response time, duration, quantity, hourly frequency, location as well as recovery time of the building s available flexibility from the building equipment. The device, room and zone agents constantly update the table. Table 5-2: test-bed AFT Agent ID Quantity (kw) Duration (min) Response (min) Hourly frequency Recovery (min) Device ID Lunch-Room agent <1 2 <5 CoffeeM Fridge1 Print -Room agent <1 2 <5 P1,P2,P3 AL Ventilation 4 15 <1 2 <5 VentFan fan agent Cooling agent 7 15 >1 2 >5 Chiller *Al- auxiliary lighting, P1- printer one *Print room agent is composed of all printers and the auxiliary lighting In addition, given that certain building services systems such as the HVAC are centralized, the building agent within this framework determines the availability and participation of the centralized systems based on information provided by the device agents. The grid side agent- The grid agent in this framework represents the interface between buildings and the utility. It negotiates with the building agent participation in grid support based on the grid requirement. Interface agent- The interface agent functions as a translator between the MAS design platform and the BEMS. Given that all the actuators in the building are connected via the BEMS, the agents are only able to send and receive control signals via the BEMS to the actuators. For this experiment, the agent platform was connected with the installed BEMS in the test bed office building via a web-server as depicted in Figure

102 MULTI-AGENT SYSTEMS APPLICATION IN FLEXIBILITY COORDINATION InsiteView BEMS Web Clients Priva Benext z- wave Philips Hue LON InsiteServer InsiteViewNext NodeJs Client InsiteViewNext Server InsiteViewEve NodeJs Proxy EVE Agents (Java) Client Figure 5-2: Agent platform and BEMS connection Services agent- The services agents introduces more task distribution in the agent structure. As it is with daily human interaction where specialized tasks are often assigned to specialist, the services agent provides dedicated specialized services to the agents within the system. Within this framework, the services agent can be used to perform tasks such as data mining and decision support functions. Figure 5-3: Agent communication sequence Test day 1 The designed MAS framework was evaluated in the test bed office building on the 24th and 25 th of august For this experiment, the office appliances utilized were the coffee machine and the printers. Wireless control nodes using the z-wave communication protocol and employed in previous experiments were also used. PIR motion sensors using the z-wave communication protocol and described in section 3.4 were installed in the print and coffee machine rooms for occupancy detection. The 88

103 MULTI-AGENT SYSTEMS APPLICATION IN FLEXIBILITY COORDINATION open-plan workspace, print room depicted in Figure 4-15 located on the first floor and the coffee room located on the ground floor of the test-bed office building were represented as room agents. Devices and users in the rooms were also represented as devices and user agents. For this set-up only the zones and services agent were inactive. The zone agent was inactive due to the centralized nature of the building HVAC system and no specialized function was required from the services agent. On the first test day, a request for demand reduction by the grid-side agent for a period of 15 minutes was initiated and the performance of the proposed agent framework evaluated. The agent sequence diagram for this test is illustrated in Figure 5-3. For this test the flexibility of the ventilation fans, auxiliary lighting and plug loads were completely utilized. Figure 5-4: Printer and coffee machine start up consumption On the second day, the same test was repeated using half the capacity of the ventilation fan i.e. 2 kw and all other appliance loads completely turned off. As occupants could require the use of these appliances during their use for flexibility the remaining available flexibility of the ventilation fans was used to compensate for occupants use of these appliances. As noted earlier, building occupants often take actions to restore their 89

104 MULTI-AGENT SYSTEMS APPLICATION IN FLEXIBILITY COORDINATION comfort when it is compromised. This means occupants might turn on appliances already turned off. The unused ventilation fan flexibility was hence used to compensate for these actions of occupants. This way, the building can provide the required grid support service without any inconvenience to building users though at the cost of reduced energy flexibility. The printer and coffee machines as depicted in Figure 5-4 both draw approximately 2.1 kw of electrical power when powered on, which is approximately equal to the unused capacity of the ventilation fans. The agent sequence diagram for this test is depicted in Figure 5-5. Figure 5-5: Agent communication sequence Test day Results and discussion The results from both tests are depicted in Figure 5-6 and Figure 5-7. As depicted in Figure 5-6, the agents were able to coordinate the operation of the ventilation and appliances to achieve required level of demand reduction. The measured energy consumption of the ventilation fans and the appliances on the first test day was approximately 1.87 kwh 15 minutes (9:45-9:59) before the initiation of the demand reduction. 90

105 MULTI-AGENT SYSTEMS APPLICATION IN FLEXIBILITY COORDINATION Pre-participation Post-participation Participation Figure 5-6: MAS coordinated ventilation and appliance fans demand reduction on the 24 th Pre-participation Participation Post-participation Figure 5-7: MAS coordinated ventilation fans, appliance demand reduction, and print room occupancy 25 th During the 15 minutes interval the reduction lasted (10:00-10:15), the measured energy consumption was approximately 0.51 kwh and 1.95 kwh for the following 15 minutes (10:15-10:30) after normal operation was restored. This resulted in load reduction of approximately 1.3 kwh during participation. On the second test day as depicted in Figure 5-7, the pre-reduction (9:45-9:59) consumption was approximately 1.89 kwh, which was reduced to approximately 0.86 kwh (10:00-10:15) and increased to approximately 1.94 kwh afterwards. As noted from Figure 5-7, occupancy was registered in the print room during participation in the DR event. As a result of occupants need for the printer, the printer was turned-on while the ventilation fan speed was further reduced to maintain the agreed reduction of 1 kwh. 91

106 MULTI-AGENT SYSTEMS APPLICATION IN FLEXIBILITY COORDINATION Experiment summary Commercial office buildings are now considered a potential source of demand-side energy flexibility. In this experiment, MAS was applied in the test-bed office building for coordination of the buildings useful flexibility for grid support services. In the first test, the building agent maximized the available useful building energy flexibility for grid support delivering approximately 1.3kWh reduction over a 15 minutes interval. Although the indoor CO 2 concentration during the interval was maintained within acceptable levels (<1000 ppm), appliances such as printers and the coffee machine were completely unavailable for use by occupants. These appliances are however also crucial for creating a productive and comfortable environment for occupants. Therefore, in the second test, one of the printers was made available for occupants use during participation in grid support and a portion of the buildings available useful energy flexibility was created to compensate for occupants use of the printer. Although the use of the printer resulted in reduced flexibility as only approximately 1kWh reduction was achieved within the 15 minutes interval, it however ensured occupants did not have to tolerate any level of discomfort while the building serves as a source of energy flexibility. In both experiments, the developed MAS framework was demonstrated to have the potential to effectively complement the BEMS in ensuring effective coordination of building systems and processes for participation in grid support within the smart grid. The use of the MAS system was also shown to have the potential to reduce the burden of participation on building occupants in that occupants were not required to manually turn off appliances. In addition, when the use of building appliances is required, the developed MAS framework was shown to have the potential to effectively coordinate its use without defaulting. 92

107 RESULTS AND DISCUSSION This chapter provides a discussion of the results obtained in the preceding chapters of this thesis. Correlations with other similar studies as well as limitations of this research are discussed in this chapter. 93

108 RESULTS AND DISCUSSION 6.1. Study Outcomes To enhance energy management in building operation and to facilitate buildings interaction with the smart-grid, the use of increased sensor data and computational intelligence has been advocated for use in BEMS. Therefore, in this study, the use of increased sensor data and fine-grained information on building occupancy as well as the use of multi-agent systems was investigated. Three sub-research questions were formulated and are discussed below. I. What are the limitations of current existing and commonly used occupancy detection systems? This question was addressed through a thorough review of existing literature in the second chapter of this thesis. Commonly used occupancy detection systems in office buildings were discussed and the limitations that affect these systems ability to provide fine-grained occupancy information outlined. From the literature survey, PIR and CO 2 occupancy detection systems were identified as commonly used occupancy detection systems in buildings for demand-driven control of building services systems. It was also noted that these systems provide coarse-grained occupancy information, which limits their energy saving potentials when applied in buildings for demand driven control applications. PIR motion sensors, which are commonly used in buildings for occupancy based lighting control, were noted to be susceptible to false-negatives a phenomenon that occurs when occupants are present but reported not to be present by the sensors. This limitation of PIR detection systems limits their application in buildings to only lighting based control and is seldom used for control of ventilation systems. CO 2 sensors on the other hand are commonly used for demand controlled ventilation applications but were noted to have a slow response time. Although modern systems such as RFIDs and vision-based detection systems have been proposed for use in buildings for demand driven control applications, their use is still limited due to privacy, ethical, maintenance and cost related issues. As a result, there has been focus on the use of a fusion of multiple low cost sensors in buildings for occupancy-based control of building systems particularly lighting and ventilation systems. Having noted that commonly used occupancy detection systems provide 94

109 RESULTS AND DISCUSSION coarse-grained occupancy information that limits the energy saving potentials of existing building energy management systems, the second research question sought to explore alternative detection systems for use in buildings for occupancy detection to enhance energy management. II. What alternative methods are available and applicable in buildings for obtaining fine-grained occupancy information? Office workers typically spend between 65% and 75% of their working hours seated at their workspace. The use of occupancy information from chairs permanently positioned at occupants workplace can thus provide fine-grained occupancy information for occupancy-based control of building services systems. The concept of using chair sensors for occupancy-based control of building systems was also demonstrated by the authors in [212] and [278]. In [278], the authors developed a chair occupancy detection system using pressure sensors and a wireless transmitter. The chair detection system was applied in an office space for occupancy based lighting control and was shown to have facilitated savings of up to 30% in lighting energy use. The authors in [212] on the other hand designed a fusion of multiple sensors including, chair detection, acoustic, motion sensors as well as computer applications for occupancy detection. The authors reported false- negatives errors (which may lead to discomfort) of 0%. In both experiments by the authors in [212] and [278], only information on user presence and location were reported by the authors. Having established that chair sensors are a viable means of obtaining fine-grained occupancy information in buildings for occupancybased control applications, the first experiment in this thesis explored its use for obtaining occupancy information on count. For the first experiment, a chair based occupancy detection system was developed from off-the-shelf components and its performance evaluated in the conference room of a university building. Particular focus was paid to the ability of the sensors to provide occupancy information on user presence and count. In addition to the chair occupancy detection system, CO 2 detection systems were also installed in the test-space. The results from the experiment were found to have substantiated the results obtained by the authors in [212] and [278]. The results obtained from the CO 2 detection systems also 95

110 RESULTS AND DISCUSSION collaborated findings from other similar studies cited by the authors in [96][167][197]. The CO 2 detection system was observed to have demonstrated a relatively slow response. The slow response time of the CO 2 detection system was in part attributed to constant opening/closing of the test-room door, varying CO 2 production rate of occupants and the location of the sensors. These factors also affected the performance of the CO 2 sensors when its output was used to estimate the number of people in the room, i.e. occupancy information on count. The CO 2 occupancy detection system was shown to have exhibited 100% error when used to estimate the number of people in the room. The chair detection system on the other exhibited improved performance when compared with the CO 2 detection system. The chair sensors were observed to have had false negative errors of up to 55% in relation to occupancy information on count. The clearly noticeable error from the chair detection system was observed to have resulted from a fundamental drawback of the chair sensors; its inability to detect standing occupants. The observed error was also attributed to the application of the chair detection system in a conference room where some occupants were observed to spend considerable time standing. Therefore, in the second follow up experiment, the chair occupancy detection system was evaluated in an office space. Three different chair detection systems based on strain, vibration and a contact switch were evaluated during this experiment. The observed false negative errors for the vibration, strain and contact chair detection systems were 40%, 20% and 0% respectively. The noticeable reduced errors observed from the contact chair sensors in this experiment can be attributed to its application in an office space where occupants mainly carried out functions while seating. In the third experiment, the designed contact based chair occupancy detection system was applied in a cellular office space for occupancy-based control of the lighting system. For the third experiment, three tests were carried out. The first test involved the use of the chair sensors alone for control of the lighting system. For this test, the chair occupancy detection system facilitated the achievement of lighting energy savings of approximately 28%. Two PIR motion sensors with a time-delay of 5 minutes were utilized for the second test. For this test, the occupancy detection system yielded 96

111 RESULTS AND DISCUSSION lighting energy savings of 23%. For the third test, a fusion of the PIR and chair sensors was utilized for occupancy based lighting control. This facilitated the achievement of 27% reduction in the lighting energy use of the test-bed office space. Though the chair detection system yielded a reduction of 28%, which is also relatively close to the reduction obtained by the authors in [278], it also introduced some discomfort for occupants. This was due to its fundamental limitation, which comes into play when occupants arrive at their workplace or in the process of departing. The authors in [278] applied a time delay setting of 5 minutes to address this shortcoming. The time delay did not however take into consideration occupants arrival time at the workplace. As occupants often require some time to settle in on arrival at the workplace, the use of the chair sensors alone for control of the lighting system could result in visual discomfort during this period. This shortcoming was addressed with the use of the fusion of the PIR motion and the chair sensors. This enabled the availability of fine-grained occupancy information and facilitated energy reduction of 27% without false negative and positives for both standing and seating occupants. The result obtained from this experiment was also in line with the results obtained by the authors in [212], who also obtained false- negatives (which may lead to discomfort) of 0% when a fusion of the chair sensors and other commonly used occupancy detection system was evaluated. Through these experiments, the viability of chair occupancy detection systems for occupancy-based control has been demonstrated. In addition, the application of the fusion of chair sensors and PIR motion sensors was demonstrated as an alternative occupancy detection system that can be applied in buildings to facilitate the availability of fine-grained occupancy information for enhanced energy savings. III. What are the benefits of multi-agent systems as a computational support tool in building operation? The use of computational intelligence tools in building operation as noted in chapters one and two can in combination with increased sensor data also enhance energy management in building operation. In addition, the use of increased computational intelligence in building operation can also enhance buildings interaction and integration with the smart grid. Multi-agent system as cited in chapter one and two is a 97

112 RESULTS AND DISCUSSION growing computational support tool that has in recent years been the focus of academic research. It has been applied in buildings to enhance energy management and to facilitate buildings integration with the smart-grid[108]. Though a number of studies as noted in chapter two have demonstrated the added value of multi-agents in buildings for enhancing energy management and for integrating buildings with the smart grid[108][223], very few studies have been conducted to assess its added value in existing BEMS. The third research question of this thesis therefore sought to evaluate the added value of multi-agent application in building operation for enhancing energy management as well as integration with the smart-grid. For the first experiment with multi-agents, a MAS framework was designed and applied in an open-plan office space for coordination of the occupancy-based lighting control system. Compared to singular spaces, the application of occupancy-based lighting control strategies is more complex in existing office buildings. This is because occupants are not usually serviced by one lighting armature, but often have to share with other occupants in the room. This presents a resource sharing/allocation problem similar to that resolved with MAS by Huberman and Clearwater[227]. In this experiment, agents represented each room as well as occupants in the test-bed room. To enhance energy saving without compromising user comfort, user agents had to communicate with each other to establish user presence. A similar approach was applied by the authors in [89]. The authors[89] leveraged on the communication capabilities of MAS to improve energy savings in an office building as well as occupant comfort. Similarly, the authors in [78] also leveraged on the communication and autonomous properties of MAS to enhance energy management and user comfort in an office building. As with the MAS developed by the authors in [89], the developed MAS framework[78] was able to obtain fine-grained occupancy information from the occupancy sensors in the room and subsequently implemented policies focused on satisfying only users who were currently occupying respective rooms. For the first experiment, using the developed MAS framework, occupancy-based lighting control was easily implemented in the open-plan workspace of an existing office building at no cost to occupants comfort. Lighting energy savings of approximately 28% was also obtained from the test-bed office space. 98

113 RESULTS AND DISCUSSION Given that a key advantage of MAS as noted in the chapters one and two is its easy reconfigurable structure, the functions of the designed MAS framework was extended. The functionalities of the same MAS framework utilized in the first occupancy-based lighting control experiment was extended to other rooms in the test-bed office building and utilized in improving the energy reduction capabilities of the installed occupancy based lighting control system in the print room. Time delays as noted in the preceding chapters of this thesis are commonly used with PIR controlled lighting systems in office buildings. This reduces the risk of the system becoming a nuisance to occupants. Time delay settings of minutes are typical and as a result more energy is wasted than if occupancy were more accurately determined and energy use correspond to actual occupancy[192][299]. In [299], the authors proposed the use of subjective measurements to determine occupants preferred time-delay settings. The authors reported savings of 37.9% and 73.2% compared to a standard setting control baseline and a worst-case scenario. In [89], the authors developed a MAS framework in which agents exchanged messages to determine suitable rooms for meetings based on the number of participants. A similar concept was applied in the second MAS experiment and used to reduce the time delay of the occupancy detection system in the print room of the test-bed office building. The exchange of occupancy information between the agents in different rooms in the test-bed facilitated the reduction of the time-delay by approximately 90%. Though with impressive results, the achieved savings was only limited to a few occupants. This was because occupancy sensors were not installed in all workspaces in the test-bed office building. This limited the sensors ability to identify and localize occupants present at all workspaces within the test-bed office building. Notwithstanding, the application of the MAS provided improved energy savings and occupant tracking possibilities without the privacy and ethical concerns. The limited level of sophistication of traditional BEMS was also highlighted as a drawback hampering buildings interaction with the smart-grid. Buildings as noted by the authors in [40][60][306] play a key role within the smart grid and are a viable source of energy flexibility. Therefore, in the last experiment, the functionalities of the designed MAS were extended to include coordination and management of the 99

114 RESULTS AND DISCUSSION building s useful energy flexibility. Within the designed MAS framework, agents within the building constantly updated the available useful flexibility of the test-bed office building in a table. With the available information, the building agent was then able to negotiate participation in grid support services such as demand response. A similar concept was demonstrated by the authors in [152] in which the energy flexibility of an office building was utilized for demand response. In contrast to the studies by the authors in [152] and a number of similar studies on building flexibility[157][155], the developed MAS framework was designed to prioritize occupant comfort. This is because the main function of buildings is to make use of energy for achieving its comfort needs. This function of buildings does however, conflict with the goal of some grid support services such as demand response. In the MAS coordinated demand response experiment described in section 5.1, although the first approach resulted in a higher energy flexibility for demand response participation, the second approach ensured occupants were availed the use of essential building appliances during participation. As demonstrated by the authors in [89][147][78] and [152] and in the experiments in this study, the MAS paradigm was demonstrated as an easy to (re)configure computational tool that can facilitate enhanced energy performance and buildings interaction with the grid without compromising occupant comfort. However, the application of MAS in building operation is still a developing field as noted by a number of authors[108][140][307]. One of the key challenges facing its application in building operation is the fact that existing software technology is still considered adequate. Though current BEMS lacks input of factors such as occupancy and occupant preferences, current software technology used in traditional BEMS are also able to exploit these information for energy management when available. In the test bed office building for example, additional sensory and control points had to be installed to obtain information on building occupancy and appliances. This information can also be fed to the installed BEMS and optimal occupancy based control strategies implemented directly from the BEMS. 100

115 RESULTS AND DISCUSSION 6.2. Study Limitations Although the limitations of each of the experiments were discussed in the preceding chapters, in this section the limitations in relation to the outcomes and the study hypothesis are discussed. The hypothesis as provided in the introduction seeks to evaluate the use of fine-grained occupancy information and multi-agents for enhancing energy management and coordination of buildings energy flexibility. Through the experiments it was shown that the use of fine-grained occupancy information can yield worthwhile energy reduction in lighting energy use. While the same can be stated for both heating and ventilation systems as demonstrated by the authors in [21][104][161], this study was however limited in scope to just the lighting system. For the experiments described in sections 4.1 and 4.2 in chapter 4 as well as section 5.1 in chapter 5, only a few rooms in the test bed office building comprising majorly the open-plan workspace with the highest occupancy density and the print room were utilized. This was in part due to the absence of occupancy sensors in other rooms in the test-bed office building. This as noted in section limited the scope of the study to a few test subjects. Furthermore, whilst a number of studies[101][102][293] have also reached similar conclusions on the added-value of fine-grained occupancy information for control of building systems, its use in practice is still rather very limited particularly in open-plan workspaces. A contributory reason for the rather slow uptake and use of fine-grained occupancy information in buildings can in part be attributed to the installation cost and the return on such investment. As noted in the preceding chapters, PIR motion sensors are the de facto occupancy sensor commonly used in buildings for lighting control. As also noted, due to the fact that these sensors are highly susceptible to false negatives, they are often configured with time delays or alternatively used in a combination of two sensors as demonstrated in this study. The use of multiple sensors does however introduces additionally capital or renovation costs for building owners which can be recouped through reduced electricity cost. Using the experiment result from section 4.2 as an example, the yearly savings on lighting in terms of cost is estimated at approximately (0.084 x 1430) = 120 Euros. This estimate is based on an estimated 52 weeks in a year, savings of 27.5 kwh per week and euros per kwh[308]. Given the 101

116 RESULTS AND DISCUSSION cost of each wireless sensor and cost of installation, the return on investment (ROI) on the use of 12 wireless occupancy sensors used in this study would therefore be in approximately 7 years. Although wired sensors could provide shorter ROI, it also eliminates the increased placement flexibility introduced by wireless sensors. The high ROI on the wireless sensors can thus be considered a contributory factor limiting the use of fine-grained occupancy information in buildings. While the concept of utilizing buildings as viable source of energy flexibility for the grid is still an ongoing debate, the cost implication of such interventions to the building is also still a key discussion point. The use of buildings particularly commercial office buildings as energy flexibility sources as noted in the preceding chapters can lead to reduced worker productivity and reduction in the useful life of building appliances. A number of studies as documented by Kershaw & Lash(2013)[309] and Mofidi & Akbari(2016)[158] have demonstrated strong links between user productivity and indoor environmental quality conditions. A simple practical approach to consider productivity as provided by the authors in [309] is that if a person is distracted, then productivity will fall proportionally to the time off task. As this is quantifiable, a worker with billable hours of 8 hours and charge of 100 Euros per hour would result in a loss of 50 Euros if 30 minutes is spent being distracted or attending to distractions from the indoor environment. However as noted by the authors in [310], layout of individual workspaces, workplace colour schemes, interior plants, dust levels and biological contaminants, and many other factors also contribute towards occupant(s) productivity in buildings. In terms of useful life of building equipment, forced duty cycling of building systems often results in deviations from the design operation of these systems thereby shortening there useful life and resulting in increased maintenance and operational cost for building operators. Therefore in this study, in order to avoid distractions to building occupants which could lead to reduction in productivity, changes to the building systems for flexibility was constrained within the boundaries of established indoor environmental condition as outlined in table 5-1. Utilizing rules specific to the test-bed, agents were able to determine in real-time the available useful energy flexibility that 102

117 RESULTS AND DISCUSSION can be used for grid support without occupants having to tolerate any level of discomfort. Finally, aside from the cost benefits relating to reduced energy use during use of a buildings energy flexibility for demand response, other cost benefits such as incentives from the utilities for participation were not covered in this study. In addition, although photovoltaic and battery storage system as described in appendix C were installed and included in the agent framework, the contribution of the renewable energy system was not considered in this study. 103

118 CONCLUSION AND FUTURE WORK This chapter provides the conclusion of this research as well as direction for future research. 104

119 CONCLUSION AND FUTURE WORK 7.1. Conclusion Buildings have in recent times being the target of a number of energy conservation strategies due to its significant energy consumption. While considerable effort is being put into the development of strategies to improve the energy performance of buildings, the energy demand of buildings is considered a source of energy flexibility for the smartgrid. In this study, based on the hypothesis that the use of increased sensor data and computational support can enhance energy management and facilitate use of building energy flexibility without impairing performance, three research questions were defined. The first research question sought to identify the limitations of current approaches for occupancy detection in buildings. Emphasis was placed on occupancy detection tools because occupancy is one of the main elements of occupant behaviour in buildings. Furthermore, occupant behaviour is considered a major source of uncertainty that significantly influences the energy performance of buildings. This first research question was addressed through a thorough review of literature. From the survey, it was shown that commonly used occupancy detection systems provide coarse-grained occupancy information, which subsequently influences the magnitude of the obtainable savings when applied in demand-driven control applications. The second research question focussed on the evaluation of alternatives tools for obtaining fine-grained occupancy to enhance energy management in buildings. The concept of chair based occupancy detection was introduced and evaluated through a number of experiments. The result from the experiments showed chair-based occupancy detection to be a viable tool for obtaining occupancy information on user presence, count and location in buildings. The result from the experiments also showed that chairbased occupancy detection systems could be combined with existing occupancy detection systems to obtain improved occupancy detection and energy savings from building services systems. The third research question was aimed at identifying the benefits of MAS in buildings as a computational tool to facilitate improved energy management and interaction with the smart-grid. Through three experiments, MAS was demonstrated as an easily (re)configurable computational support tool that can facilitate enhanced energy 105

120 CONCLUSION AND FUTURE WORK management as well as coordination of building processes for optimal interaction with the smart-grid. Through the experiments conducted in this study as well as similar studies cited in this study, MAS has been demonstrated as a viable computational support tool that can enhance building operation and management. The uptake of MAS in building operation and management is however still is still rather very slow. This as noted is due to the adequacy of existing building energy management design software and lack of a clear business case for a transition to a more intelligent MAS platform. It is therefore imperative to explore its application in practical areas in buildings where it s potential strengths; flexibility, autonomy, and communication provide clear additional benefits. That being said, with the transition to the smart-grid and the increasing number of systems within the building optimal coordination of building processes without compromising occupant comfort with limited input from building operators would be overwhelming for existing BEMS. Moreover, in utilizing the energy flexibility of buildings for grid support services, such as demand response, energy flexibility in the magnitude of megawatts is often required, whereas buildings are only able to provide energy flexibility in the magnitude of kilowatts. Achieving the megawatt energy flexibility threshold thus demands organized and coordinated participation of a federation or aggregation of buildings. In this capacity, the MAS paradigm has been shown to be a viable and easy approach for managing and coordinating the interactions of several different participating buildings Future Work In the course of this study, a number of new questions were posed which should be further investigated. Use of existing building infrastructure for occupancy detection Most commercial office buildings today incorporate WI-FI access points distributed through various spaces in the building. In addition, most building occupants often have portable mobile devices on their person. The fusion of the 106

121 CONCLUSION AND FUTURE WORK information from WI-FI access points and mobile devices can in combination with other low-cost sensors such as PIR and chair sensors provide fine-grained occupancy information on presence, count, location, and track without invading occupants privacy. The fusion of the information from these multiple sources combined with computational support tools such as agent systems can provide fine-grained occupancy information that can yield worthwhile improvement in building energy consumption. Applying MAS for coordination of flexibility of groups of buildings For participation in grid support activities, the grid often requires flexibility in the magnitude of megawatts. Buildings such as the test bed office building are however able to provide flexibility in the magnitude of kilowatts. A federation or aggregation of similar buildings would therefore be required to provide the energy flexibility needed by the smart grid. The application of MAS in this direction for coordination of the flexibility from multiple buildings is therefore a research direction to explore. Exploring alternative solutions for building flexibility utilization without building comfort and productivity degradation Though buildings are a viable source of demand-side energy flexibility, quite often to maximize buildings useful energy flexibility, occupants often have to tolerate some degree of discomfort. Given that the primary function of buildings and building services is the provision of a comfortable and productive indoor environment for occupants, alternative and effective ways of utilizing the useful energy flexibility of buildings without compromising occupants comfort and productivity needs to be explored. Another alternative solution as noted in this thesis is the use of the growing number of distributed energy resources combined with energy storage systems being introduced in buildings. This possibility is currently being explored in the test bed office building with the installation of PV-generation as well as battery storage systems (appendix c). 107

122 BIBLIOGRAPHY BIBLIOGRAPHY [1] N. Kannan and D. Vakeesan, Solar energy for future world: - A review, Renew. Sustain. Energy Rev., vol. 62, pp , [2] R. A. Rosen and E. Guenther, The energy policy relevance of the 2014 IPCC Working Group III report on the macro-economics of mitigating climate change, Energy Policy, vol. 93, pp , [3] IPCC, Climate Change 2014 Synthesis Report, [4] T. Kousksou, A. Allouhi, M. Belattar, A. Jamil, T. El Rhafiki, A. Arid, and Y. Zeraouli, Renewable energy potential and national policy directions for sustainable development in Morocco, Renew. Sustain. Energy Rev., vol. 47, pp , [5] S. S. Reka and V. Ramesh, A smart survey on demand response potential in global energy market, Indian J. Sci. Technol., vol. 8, pp , [6] S. Barsali, M. Ceraolo, R. Giglioli, and D. Poli, Storage applications for Smartgrids, Electr. Power Syst. Res., vol. 120, pp , [7] G. P. J. Verbong, S. Beemsterboer, and F. Sengers, Smart grids or smart users? Involving users in developing a low carbon electricity economy, Energy Policy, vol. 52, pp , Jan [8] A. Trianni, E. Cagno, and A. De Donatis, A framework to characterize energy efficiency measures, Appl. Energy, vol. 118, pp , [9] A. Trianni, E. Cagno, and S. Farné, Barriers, drivers and decision-making process for industrial energy efficiency: A broad study among manufacturing small and mediumsized enterprises, Appl. Energy, vol. 162, pp , [10] P. Christoff, The promissory note: COP 21 and the Paris Climate Agreement, Env. Polit., vol. 25, no. 5, pp , Sep [11] K. V. Wong, Planning and Engineering Strategies to Mitigate Effects of Climate Change, J. Energy Resour. Technol., vol. 138, no. 1, pp , Sep [12] P. Villoria-Sáez, V. W. Y. Tam, M. del Río Merino, C. Viñas Arrebola, and X. Wang, Effectiveness of greenhouse-gas Emission Trading Schemes implementation: a review on legislations, J. Clean. Prod., vol. 127, pp , Jul [13] I. Shishlov, R. Morel, and V. Bellassen, Compliance of the Parties to the Kyoto Protocol in the first commitment period, Clim. Policy, vol. 16, no. 6, pp , Aug [14] U.S. Congress, Energy Independence and Security Act of 2007, U.S. Government Printing Office., [15] EU, Directive 2009/28/EC, Official Journal of the European Union, [16] U.S Energy Information Administration, International Energy Outlook 2016, [17] R. Becqué, E. Mackres, J. Layke, N. Aden, S. Liu, K. Managan, C. Nesler, S. Mazurstommen, K. Petrichenko, and P. Graham, Accelerating Building Efficiency: Eight Actions for Urban Leaders, [18] N. Aste, P. Caputo, M. Buzzetti, and M. Fattore, Energy efficiency in buildings: What drives the investments? the case of Lombardy Region, Sustain. Cities Soc., vol. 20, pp , [19] M. A. Riffat SB, Building Energy Consumption and Carbon dioxide Emissions: Threat to Climate Change, J. Earth Sci. Clim. Change, vol. s3, pp. 1 3, [20] International Energy Agency, Transition to Sustainable Buildings. IEA, [21] K. Ahmed, J. Kurnitski, and P. Sormunen, Demand controlled ventilation indoor climate and energy performance in a high performance building with air flow rate controlled chilled beams, Energy Build., vol. 109, pp , Dec [22] K. J. Chua, S. K. Chou, W. M. Yang, and J. Yan, Achieving better energy-efficient air 108

123 BIBLIOGRAPHY conditioning - A review of technologies and strategies, Appl. Energy, vol. 104, pp , [23] J. Lee, J. Kim, D. Song, J. Kim, and C. Jang, Impact of external insulation and internal thermal density upon energy consumption of buildings in a temperate climate with four distinct seasons, Renew. Sustain. Energy Rev., no. xxxx, pp. 1 8, Nov [24] A. Allouhi, Y. El Fouih, T. Kousksou, A. Jamil, Y. Zeraouli, and Y. Mourad, Energy consumption and efficiency in buildings: current status and future trends, J. Clean. Prod., vol. 109, pp , Dec [25] M. Manic, D. Wijayasekara, K. Amarasinghe, and J. J. Rodriguez-Andina, Building Energy Management Systems: The Age of Intelligent and Adaptive Buildings, IEEE Ind. Electron. Mag., vol. 10, no. 1, pp , [26] P. Palensky and D. Dietrich, Demand side management: Demand response, intelligent energy systems, and smart loads, IEEE Trans. Ind. Informatics, vol. 7, no. 3, pp , [27] M. Brown, Innovative energy-efficiency policies: An international review, Wiley Interdiscip. Rev. Energy Environ., vol. 4, no. 1, pp. 1 25, [28] R. C. G. M. Loonen, M. Trčka, D. Cóstola, and J. L. M. Hensen, Climate adaptive building shells: State-of-the-art and future challenges, Renew. Sustain. Energy Rev., vol. 25, pp , Sep [29] K. Wei, C. Lv, M. Chen, X. Zhou, Z. Dai, and D. Shen, Development and performance evaluation of a new thermal insulation material from rice straw using high frequency hotpressing, Energy Build., vol. 87, pp , [30] F. Asdrubali, F. D Alessandro, and S. Schiavoni, A review of unconventional sustainable building insulation materials, Sustain. Mater. Technol., vol. 4, pp. 1 17, [31] A. Sheikhi, M. Rayati, and A. M. Ranjbar, Demand side management for a residential customer in multi-energy systems, Sustain. Cities Soc., vol. 22, pp , Apr [32] P. Principi and R. Fioretti, A comparative life cycle assessment of luminaires for general lighting for the office - Compact fluorescent (CFL) vs Light Emitting Diode (LED) - A case study, J. Clean. Prod., vol. 83, pp , [33] Z. Zhou, Z. Zhang, G. Chen, J. Zuo, P. Xu, C. Meng, and Z. Yu, Feasibility of ground coupled heat pumps in office buildings: A China study, Appl. Energy, vol. 162, pp , [34] P. Ovchinnikov, A. Borodiņecs, and K. Strelets, Utilization potential of low temperature hydronic space heating systems: A comparative review, Build. Environ., vol. 112, pp , Feb [35] P. H. Shaikh, N. B. M. Nor, P. Nallagownden, I. Elamvazuthi, and T. Ibrahim, A review on optimized control systems for building energy and comfort management of smart sustainable buildings, Renew. Sustain. Energy Rev., vol. 34, pp , [36] N. Mohamed, S. Lazarova-Molnar, and J. Al-Jaroodi, CE-BEMS: A cloud-enabled building energy management system, in 3rd MEC International Conference on Big Data and Smart City (ICBDSC), 2016, pp [37] C. RHE/16, BS EN 15232:2012: Energy performance of buildings. Impact of Building Automation, Controls and Building Management, BSI, [38] K. Gvozdenovic, W. Maassen, W. Zeiler, and H. Besselink, Roadmap to nearly Zero Energy Buildings: Towards nzebs in 2020 in the Netherlands, REHVA J., vol. 2, no. March, [39] Eurima, Trias Energetica.. [40] D. Kolokotsa, The role of smart grids in the building sector, Energy Build., vol. 116, pp , [41] K. Kok, B. Roossien, P. MacDougall, O. van Pruissen, G. Venekamp, R. Kamphuis, J. 109

124 BIBLIOGRAPHY Laarakkers, and C. Warmer, Dynamic pricing by scalable energy management systems- Field experiences and simulation results using PowerMatcher, in 2012 IEEE Power and Energy Society General Meeting, 2012, no. July, pp [42] M. L. Tuballa and M. L. Abundo, A review of the development of Smart Grid technologies, Renew. Sustain. Energy Rev., vol. 59, pp , [43] P. Siano, Demand response and smart grids - A survey, Renew. Sustain. Energy Rev., vol. 30, pp , Feb [44] CENELEC, SG-CG / M490 / L Flexibility Management Overview of the main concepts of flexibility management, [45] H. Farhangi, The path of the smart grid, IEEE Power Energy Mag., vol. 8, no. 1, pp , Jan [46] L. Saarinen, N. Dahlb??ck, and U. Lundin, Power system flexibility need induced by wind and solar power intermittency on time scales of 1-14 days, Renew. Energy, vol. 83, pp , [47] S. O. Ottesen and A. Tomasgard, A stochastic model for scheduling energy flexibility in buildings, Energy, vol. 88, pp , [48] A. Ulbig, Operational Flexibility in Electric Power Systems, arxiv Prepr., vol. 72, p. 230, [49] P. D. Lund, J. Lindgren, J. Mikkola, and J. Salpakari, Review of energy system flexibility measures to enable high levels of variable renewable electricity, Renew. Sustain. Energy Rev., vol. 45, pp , May [50] K. O. Aduda, T. Labeodan, W. Zeiler, G. Boxem, and Y. Zhao, Demand side flexibility: Potentials and building performance implications, Sustain. Cities Soc., vol. 22, pp , [51] L. M. Pierpoint, Harnessing electricity storage for systems with intermittent sources of power: Policy and R&D needs, Energy Policy, pp. 1 7, [52] X. Xue, S. Wang, C. Yan, and B. Cui, A fast chiller power demand response control strategy for buildings connected to smart grid, Appl. Energy, vol. 137, pp , Jan [53] T. Samad, E. Koch, and P. Stluka, Automated Demand Response for Smart Buildings and Microgrids: The State of the Practice and Research Challenges, Proc. IEEE, vol. 104, no. 4, pp , Apr [54] Y. Lin, P. Barooah, and J. L. Mathieu, Ancillary services to the grid from commercial buildings through demand scheduling and control, in 2015 American Control Conference (ACC), 2015, pp [55] I. Beil, I. A. Hiskens, and S. Backhaus, Round-trip efficiency of fast demand response in a large commercial air conditioner, Energy Build., vol. 97, pp , [56] M. W. Ahmad, M. Mourshed, D. Mundow, M. Sisinni, and Y. Rezgui, Building energy metering and environmental monitoring - A state-of-the-art review and directions for future research, Energy Build., vol. 120, pp , [57] J. L. Mathieu, M. Dyson, and D. S. Callaway, Using Residential Electric Loads for Fast Demand Response : The Potential Resource and Revenues, the Costs, and Policy Recommendations, in Proceedings of the ACEEE Summer Study on Buildings, 2012, pp [58] K. Subbarao, J. Fuller, K. Kalsi, R. Pratt, S. Widergren, and D. Chassin, Transactive Control and Coordination of Distributed Assets for Ancillary Services, [59] F. Shariatzadeh, P. Mandal, and A. K. Srivastava, Demand response for sustainable energy systems: A review, application and implementation strategy, Renew. Sustain. Energy Rev., vol. 45, pp , May [60] N. Oconnell, P. Pinson, H. Madsen, and M. Omalley, Benefits and challenges of electrical demand response: A critical review, Renew. Sustain. Energy Rev., vol. 39, pp , 110

125 BIBLIOGRAPHY [61] S. Wang, X. Xue, and C. Yan, Building power demand response methods toward smart grid, HVAC&R Res., vol. 20, no. 6, pp , Aug [62] USEF Foundation, USEF: The Framework Explained, [63] P. Bradley, M. Leach, and J. Torriti, A review of the costs and benefits of demand response for electricity in the UK, Energy Policy, vol. 52, pp , [64] T. Hong, S. D Oca, S. C. Taylor-Lange, W. J. N. Turner, Y. Chen, and S. P. Corgnati, An ontology to represent energy-related occupant behavior in buildings. Part II: Implementation of the DNAS framework using an XML schema, Build. Environ., vol. 94, no. P1, pp , Dec [65] K. Schakib-Ekbatan, F. Z. Cakici, M. Schweiker, and A. Wagner, Does the occupant behavior match the energy concept of the building? - Analysis of a German naturally ventilated office building, Build. Environ., vol. 84, pp , [66] T. A. Nguyen and M. Aiello, Energy intelligent buildings based on user activity: A survey, Energy Build., vol. 56, pp , Jan [67] C. Lee and H. Yang, A context-awareness system that uses a thermographic camera to monitor energy waste in buildings, Energy Build., vol. 135, pp , Jan [68] T. Hong, S. C. Taylor-Lange, S. D Oca, D. Yan, and S. P. Corgnati, Advances in research and applications of energy-related occupant behavior in buildings, Energy Build., vol. 116, pp , Mar [69] W. Qingqin and Z. Miao, Introduction of the standard for energy efficient building evaluation, Sustain. Cities Soc., vol. 14, no. 1, pp. 1 4, Feb [70] M. S. Gul and S. Patidar, Understanding the energy consumption and occupancy of a multi-purpose academic building, Energy Build., vol. 87, pp , [71] E. Azar and C. C. Menassa, A comprehensive analysis of the impact of occupancy parameters in energy simulation of office buildings, Energy Build., vol. 55, pp , [72] D. Yan, W. O Brien, T. Hong, X. Feng, H. Burak Gunay, F. Tahmasebi, and A. Mahdavi, Occupant behavior modeling for building performance simulation: Current state and future challenges, Energy Build., vol. 107, pp , [73] T. Buso, V. Fabi, R. K. Andersen, and S. P. Corgnati, Occupant behaviour and robustness of building design, Build. Environ., vol. 94, pp , [74] A. Roetzel, Occupant behaviour simulation for cellular offices in early design stages - Architectural and modelling considerations, Build. Simul., vol. 8, no. 2, pp , Apr [75] T. Zaraket, B. Yannou, Y. Leroy, S. Minel, and E. Chapotot, An Occupant-Based Energy Consumption Model for User-Focused Design of Residential Buildings, J. Mech. Des., vol. 137, no. 7, p , [76] O. T. Masoso and L. J. Grobler, The dark side of occupants behaviour on building energy use, Energy Build., vol. 42, no. 2, pp , [77] D. Yan and T. Hong, EBC Annex 66 Proposal Definition and Simulation of Occupant Behavior in Buildings, [78] D.-Y. Yu, E. Ferranti, and H. Hadeli, An intelligent building that listens to your needs, in Proceedings of the 28th Annual ACM Symposium on Applied Computing - SAC 13, 2013, p. 58. [79] B. Orland, N. Ram, D. Lang, K. Houser, N. Kling, and M. Coccia, Saving energy in an office environment: A serious game intervention, Energy Build., vol. 74, pp , [80] K. Ackerly and G. S. Brager, Window signalling systems: control strategies and occupant behaviour, Build. Res. Inf., vol. 41, no. 3, pp , Jun [81] S. Wei, T. M. Hassan, S. K. Firth, and F. Fouchal, Impact of occupant behaviour on the 111

126 BIBLIOGRAPHY energy-saving potential of retrofit measures for a public building in the UK, Intell. Build. Int., vol. 8975, pp. 1 10, Feb [82] T. R. Whiffen, S. Naylor, J. Hill, L. Smith, P. A. Callan, M. Gillott, C. J. Wood, and S. B. Riffat, A concept review of power line communication in building energy management systems for the small to medium sized non-domestic built environment, Renew. Sustain. Energy Rev., vol. 64, pp , [83] A. I. Dounis and C. Caraiscos, Advanced control systems engineering for energy and comfort management in a building environment-a review, Renew. Sustain. Energy Rev., vol. 13, no. 6 7, pp , Aug [84] S. Karjalainen, Should we design buildings that are less sensitive to occupant behaviour? A simulation study of effects of behaviour and design on office energy consumption, Energy Effic., vol. 9, no. 6, pp , Dec [85] I. Gaetani, P. J. Hoes, and J. L. M. Hensen, Occupant behavior in building energy simulation: Towards a fit-for-purpose modeling strategy, Energy Build., vol. 121, pp , [86] P. Rocha, A. Siddiqui, and M. Stadler, Improving energy efficiency via smart building energy management systems: A comparison with policy measures, Energy Build., vol. 88, pp , [87] R. Missaoui, H. Joumaa, S. Ploix, and S. Bacha, Managing energy Smart Homes according to energy prices: Analysis of a Building Energy Management System, Energy Build., vol. 71, pp , [88] E. Tsampasis, L. Sarakis, H. Leligou, T. Zahariadis, and J. Garofalakis, Novel Simulation Approaches for Smart Grids, J. Sens. Actuator Networks, vol. 5, no. 3, p. 11, Jun [89] L. Klein, J. Y. Kwak, G. Kavulya, F. Jazizadeh, B. Becerik-Gerber, P. Varakantham, and M. Tambe, Coordinating occupant behavior for building energy and comfort management using multi-agent systems, Autom. Constr., vol. 22, pp , Mar [90] E. Arens, C. C. Federspiel, D. Wang, and C. Huizenga, How ambient intelligence will improve habitability and energy efficiency in buildings, in Ambient Intelligence, Berlin/Heidelberg: Springer-Verlag, 2005, pp [91] S. Ikeda and R. Ooka, A new optimization strategy for the operating schedule of energy systems under uncertainty of renewable energy sources and demand changes, Energy Build., vol. 125, pp , [92] I. Georgievski, V. Degeler, G. A. Pagani, T. A. Nguyen, A. Lazovik, and M. Aiello, Optimizing Energy Costs for Offices Connected to the Smart Grid, IEEE Trans. Smart Grid, vol. 3, no. 4, pp , Dec [93] W. Zeiler, K. Aduda, and T. Thomassen, Integral BEMS controlled LVPP in the smart grid: An approach to a complex distributed system of the built environment for sustainability, Int. J. Des. Sci. Technol., vol. 22, no. 2, pp , [94] B. Onose, Control optimization for increasing energy performance of existing buildings, in 2016 Eleventh International Conference on Ecological Vehicles and Renewable Energies (EVER), 2016, pp [95] J. Kuutti, K. H. Blomqvist, and R. E. Sepponen, Evaluation of visitor counting technologies and their energy saving potential through demand-controlled ventilation, Energies, vol. 7, no. 3, pp , Mar [96] N. Li, G. Calis, and B. Becerik-Gerber, Measuring and monitoring occupancy with an RFID based system for demand-driven HVAC operations, Autom. Constr., vol. 24, pp , Jul [97] L. V. Thanayankizil, S. K. Ghai, D. Chakraborty, and D. P. Seetharam, Softgreen: Towards energy management of green office buildings with soft sensors, in th International Conference on Communication Systems and Networks, COMSNETS 2012, 2012, pp

127 BIBLIOGRAPHY [98] B. Chenari, J. Dias Carrilho, and M. Gameiro Da Silva, Towards sustainable, energyefficient and healthy ventilation strategies in buildings: A review, Renew. Sustain. Energy Rev., vol. 59, pp , [99] N. Nasir, K. Palani, A. Chugh, V. C. Prakash, U. Arote, A. P. Krishnan, and K. Ramamritham, Fusing Sensors for Occupancy Sensing in Smart Buildings, in 11th International Conference on Distributed Computing and Internet Technology, vol. 8956, Bhubaneswar, India, 2015, pp [100] K. Christensen, R. Melfi, B. Nordman, B. Rosenblum, R. Viera, K. Christensen, and R. Melfi, Using existing network infrastructure to estimate building occupancy and control plugged-in devices in user workspaces, Int. J. Commun. Networks Distrib. Syst. J. Commun. Networks Distrib. Syst., vol. 12, no. 1, pp. 4 29, [101] M. A. U. Haq, M. Y. Hassan, H. Abdullah, H. A. Rahman, M. P. Abdullah, F. Hussin, and D. M. Said, A review on lighting control technologies in commercial buildings, their performance and affecting factors, Renew. Sustain. Energy Rev., vol. 33, pp , May [102] C. de Bakker, M. Aries, H. Kort, and A. Rosemann, Occupancy-based lighting control in open-plan office spaces: A state-of-the-art review, Build. Environ., vol. 112, pp , Feb [103] J. Brooks, S. Goyal, R. Subramany, Y. Lin, C. Liao, T. Middelkoop, H. Ingley, L. Arpan, and P. Barooah, Experimental evaluation of occupancy-based energy-efficient climate control of VAV terminal units, Sci. Technol. Built Environ., vol. 21, no. 4, pp , [104] Y. Fan, K. Kameishi, S. Onishi, and K. Ito, Field-based study on the energy-saving effects of CO2 demand controlled ventilation in an office with application of energy recovery ventilators, Energy Build., vol. 68, no. PARTA, pp , Jan [105] B. Becker, I. Mauser, H. Schmeck, S. Hubschneider, and T. Leibfried, Smart grid services provided by building energy management systems, in IEEE PES Innovative Smart Grid Technologies Latin America (ISGT LATAM), 2015, pp [106] F. Ygge, H. Akkermans, A. Andersson, and E. Boertjes, The H OME B OTS System and Field Test : A Multi-Commodity Market for Predictive Power Load Management Agent- Based Energy Management, in 4th Intl. Conf. on the Practical Application of Intelligent Agents and Multi-Agent Technology, [107] M. W. Ahmad, M. Mourshed, B. Yuce, and Y. Rezgui, Computational intelligence techniques for HVAC systems: A review, Build. Simul., pp. 1 40, [108] A. Windham and S. Treado, A review of multi-agent systems concepts and research related to building HVAC control, Sci. Technol. Built Environ., vol. 22, no. 1, pp , Jan [109] A. H. Kazmi, M. J. O grady, D. T. Delaney, A. G. Ruzzelli, and G. M. P. O hare, A Review of Wireless-Sensor-Network-Enabled Building Energy Management Systems, ACM Trans. Sens. Networks, vol. 10, no. 4, pp. 1 43, [110] A. Anand, R. Vasudevan, S. Bhattacharya, R. Arun, and A. Sivanantham, Retrofit control solutions for old buildings using WSN, in I4CT nd International Conference on Computer, Communications, and Control Technology, Art Proceeding, 2015, no. I4ct, pp [111] L. Lan and Y. K. Tan, Advanced building energy monitoring using wireless sensor integrated EnergyPlus platform for personal climate control, in 2015 IEEE 11th International Conference on Power Electronics and Drive Systems, 2015, pp [112] R. Yao and J. Zheng, A model of intelligent building energy management for the indoor environment, Intell. Build. Int., vol. 2, no. 1, pp , Jan [113] M. A. Mahmood, W. K. G. Seah, and I. Welch, Reliability in wireless sensor networks: A 113

128 BIBLIOGRAPHY survey and challenges ahead, Comput. Networks, vol. 79, pp , [114] F. Robol, F. Viani, E. Giarola, and A. Massa, Wireless sensors for distributed monitoring of energy-efficient smart buildings, Mediterr. Microw. Symp., vol Janua, pp. 1 4, [115] B. Lasternas, J. Zhao, R. Yun, C. Zhang, H. Wang, and A. Aziz, Behavior Oriented Metrics for Plug Load Energy Savings in Office Environment, in Proceedings of 2014 American council for an Energy-Efficient Economy (ACEEE) Summer Study on Energy Efficiency in Buildings, 2014, pp [116] A. Mahdavi and F. Tahmasebi, The deployment-dependence of occupancy-related models in building performance simulation, Energy Build., vol. 117, pp , [117] S. D Oca and T. Hong, Occupancy schedules learning process through a data mining framework, Energy Build., vol. 88, pp , Feb [118] C. Duarte, K. Van Den Wymelenberg, and C. Rieger, Revealing occupancy patterns in an office building through the use of occupancy sensor data, Energy Build., vol. 67, pp , Dec [119] F. Xiao and C. Fan, Data mining in building automation system for improving building operational performance, Energy Build., vol. 75, pp , [120] K. Akkaya, I. Guvenc, R. Aygun, N. Pala, and A. Kadri, IoT-based occupancy monitoring techniques for energy-efficient smart buildings, in 2015 IEEE Wireless Communications and Networking Conference Workshops (WCNCW), 2015, pp [121] D. Kolokotsa, C. Diakaki, E. Grigoroudis, G. Stavrakakis, and K. Kalaitzakis, Decision support methodologies on the energy efficiency and energy management in buildings, Adv. Build. Energy Res., vol. 3, no. May 2015, pp , Jan [122] S. Y. Yang, Developing an energy-saving and case-based reasoning information agent with Web service and ontology techniques, Expert Syst. Appl., vol. 40, no. 9, pp , [123] S. S. K. Kwok, R. K. K. Yuen, and E. W. M. Lee, An intelligent approach to assessing the effect of building occupancy on building cooling load prediction, Build. Environ., vol. 46, no. 8, pp , Aug [124] S. S. K. Kwok, A study of the importance of occupancy to building cooling load in prediction by intelligent approach, Energy Convers. Manag., vol. 52, no. 7, pp , Jul [125] B. Dong, B. Andrews, K. P. Lam, M. Hoynck, R. Zhang, Y. S. Chiou, and D. Benitez, An information technology enabled sustainability test-bed (ITEST) for occupancy detection through an environmental sensing network, Energy Build., vol. 42, no. 7, pp , Jul [126] U. Rutishauser, J. Joller, and R. Douglas, Control and learning of ambience by an intelligent building, IEEE Trans. Syst. Man, Cybern. Part ASystems Humans., vol. 35, no. 1, pp , [127] D. Urieli and P. Stone, A Learning Agent for Heat-Pump Thermostat Control, in Proceedings of the 12th International Conference on Autonomous Agents and Multiagent Systems (AAMAS 2013), 2013, no. May, pp [128] S. Wang and Z. Ma, Supervisory and Optimal Control of Building HVAC Systems: A Review, HVAC&R Res., vol. 14, no. 1, pp. 3 32, Jan [129] R. Yang and L. Wang, Multi-zone building energy management using intelligent control and optimization, Sustain. Cities Soc., vol. 6, no. 1, pp , Feb [130] R. Li and R. Ooka, Multi-variable optimization of HVAC system using a genetic algorithm, J. Energy Power Eng., vol. 8, pp , [131] J. Counsell, O. Zaher, J. Brindley, and G. Murphy, Robust nonlinear HVAC systems control with evolutionary optimisation, Eng. Comput., vol. 30, no. 8, pp ,

129 BIBLIOGRAPHY [132] M. Mokhtar, M. Stables, X. Liu, and J. Howe, Intelligent multi-agent system for building heat distribution control with combined gas boilers and ground source heat pump, Energy Build., vol. 62, pp , Jul [133] Z. Wang and L. Wang, Intelligent Control of Ventilation System for Energy-Efficient Buildings With Predictive Model, IEEE Trans. Smart Grid, vol. 4, no. 2, pp , [134] F. Oldewurtel, A. Parisio, C. N. Jones, D. Gyalistras, M. Gwerder, V. Stauch, B. Lehmann, and M. Morari, Use of model predictive control and weather forecasts for energy efficient building climate control, Energy Build., vol. 45, pp , Feb [135] F. Oldewurtel, D. Sturzenegger, and M. Morari, Importance of occupancy information for building climate control, Appl. Energy, vol. 101, pp , Jan [136] H. Doukas, K. D. Patlitzianas, K. Iatropoulos, and J. Psarras, Intelligent building energy management system using rule sets, Build. Environ., vol. 42, no. 10, pp , Oct [137] H. Abbas, S. Shaheen, and M. Amin, Simple, Flexible, and Interoperable SCADA System Based on Agent Technology, no. August, pp , [138] S. D. Ramchurn, P. Vytelingum, A. Rogers, and N. R. N. N. R. Jennings, Agent-Based Homeostatic Control for Green Energy in the Smart Grid, ACM Trans. Intell. Syst. Technol., vol. 2, no. 4, p. 35, [139] P. Leitao, Multi-agent Systems in Industry: Current Trends& Future Challenges, in Beyond Artificial Intelligence-Topics in Intelligent Engineering and Informatics, vol. 4, J. Kelemen, J. Romportl, and E. Zackova, Eds. Berlin, Heidelberg: Springer Berlin Heidelberg, 2013, pp [140] D. Jarvis, J. Jarvis, R. Rönnquist, and L. Jain, Multi-Agent Systems, vol Cham: Springer International Publishing, [141] S. Russell and N. Peter, Intelligent Agents, in Artificial Intelligence: A Modern Approach, vol. 27, no. 2, Prentice-Hall, Inc., 2006, pp [142] M. Wooldridge and N. R. Jennings, Intelligent agents : theory and practice, Knowl. Eng. Rev., vol. 10, pp , [143] S. Al-Agtash, Electricity agents in smart grid markets, Comput. Ind., vol. 64, no. 3, pp , Apr [144] A. Kantamneni, L. E. Brown, G. Parker, and W. W. Weaver, Survey of multi-agent systems for microgrid control, Engineering Applications of Artificial Intelligence, vol. 45. Elsevier, pp , [145] E. Shakshuki and M. Reid, Multi-Agent System Applications in Healthcare: Current Technology and Future Roadmap, Procedia Comput. Sci., vol. 52, no. Ant, pp , [146] D. B. Upadhyay, A review paper on Multi agent base intelligent manufacturing system, Int. J. Adv. Eng. Res. Sci., vol. 1, no. 3, [147] S. Mamidi, Y. Chang, R. Maheswaran, and M. Rey, Adaptive Learning Agents for Sustainable Building Energy Management, in AAAI 2012 Spring Symposium on Game Theory for Security, Sustainability and Health, 2012, pp [148] R. Yang and L. Wang, Multi-agent based energy and comfort management in a building environment considering behaviors of occupants, IEEE Power Energy Soc. Gen. Meet., pp. 1 7, Jul [149] M. Huber, S. Brust, T. Schütz, A. Constantin, R. Streblow, and D. Müller, Purely Agent Based Control of Building Energy Supply Systems, in PROCEEDINGS OF THE 28TH INTERNATIONAL CONFERENCE ON EFFICIENCY, COST, OPTIMIZATION, SIMULATION AND ENVIRONMENTAL IMPACT OF ENERGY SYSTEMS, 2015, no. June, pp [150] J. Hu, A. Saleem, S. You, L. Nordström, M. Lind, and J. Østergaard, A multi-agent system 115

130 BIBLIOGRAPHY for distribution grid congestion management with electric vehicles, Eng. Appl. Artif. Intell., vol. 38, pp , Feb [151] L. Hernandez, C. Baladron, J. Aguiar, B. Carro, A. Sanchez-Esguevillas, J. Lloret, D. Chinarro, J. Gomez-Sanz, and D. Cook, A multi-agent system architecture for smart grid management and forecasting of energy demand in virtual power plants, IEEE Commun. Mag., vol. 51, no. 1, pp , [152] D. Auslander, D. Culler, P. K. Wright, Y. Lu, and M. A. Piette, A Distributed Automated Demand Response Building Management System, [153] EIA, Energy Use in Commercial Buildings, [Online]. Available: mercial. [Accessed: 17-Mar-2017]. [154] W. Shen and G. Newsham, Implicit occupancy detection for energy conservation in commercial buildings: A review, in IEEE 20th International Conference on Computer Supported Cooperative Work in Design (CSCWD), 2016, pp [155] C. D. Korkas, S. Baldi, I. Michailidis, and E. B. Kosmatopoulos, Occupancy-based demand response and thermal comfort optimization in microgrids with renewable energy sources and energy storage, Appl. Energy, vol. 163, pp , [156] V. L. Erickson and A. E. Cerpa, Occupancy based demand response HVAC control strategy, in Proceedings of the 2nd ACM Workshop on Embedded Sensing Systems for Energy-Efficiency in Building - BuildSys 10, 2010, pp [157] F. Sehar, M. Pipattanasomporn, and S. Rahman, Integrated automation for optimal demand management in commercial buildings considering occupant comfort, Sustain. Cities Soc., vol. 28, pp , [158] F. Mofidi and H. Akbari, Integrated optimization of energy costs and occupants??? productivity in commercial buildings, Energy Build., vol. 129, pp , [159] P. G. S. Rutten, Strategisch bouwen. Eindhoven: Technische Universiteit Eindhoven., [160] J.. Clarke, M. Janak, and P. Ruyssevelt, Assessing the overall performance of advanced glazing systems, Sol. Energy, vol. 63, no. 4, pp , Oct [161] X. Lin and J. Lau, Demand controlled ventilation for multiple zone HVAC systems: CO 2 -based dynamic reset (RP 1547), HVAC&R Res., vol. 20, no. 8, pp , Nov [162] P. Zhou, G. Huang, and Z. Li, Demand-based temperature control of large-scale rooms aided by wireless sensor network: Energy saving potential analysis, Energy Build., vol. 68, no. PARTA, pp , Jan [163] I. Lymperopoulos, F. A. Qureshi, T. Nghiem, A. A. Khatir, and C. N. Jones, Providing ancillary service with commercial buildings: The Swiss perspective, IFAC Proc. Vol., vol. 48, no. 8, pp. 6 13, [164] L. M. Candanedo and V. Feldheim, Accurate occupancy detection of an office room from light, temperature, humidity and CO2 measurements using statistical learning models, Energy Build., vol. 112, pp , [165] M. A. A. H. Khan, H. M. S. Hossain, and N. Roy, SensePresence: Infrastructure-Less Occupancy Detection for Opportunistic Sensing Applications, in th IEEE International Conference on Mobile Data Management, 2015, vol. 2, pp [166] J. Yun and S.-S. Lee, Human movement detection and identification using pyroelectric infrared sensors., Sensors (Basel)., vol. 14, no. 5, pp , [167] D. Calì, P. Matthes, K. Huchtemann, R. Streblow, and D. Müller, CO2 based occupancy detection algorithm: Experimental analysis and validation for office and residential buildings, Build. Environ., vol. 86, pp , [168] M. Amayri, A. Arora, S. Ploix, S. Bandhyopadyay, Q.-D. Ngo, and V. R. Badarla, Estimating occupancy in heterogeneous sensor environment, Energy Build., vol. 129, pp. 116

131 BIBLIOGRAPHY 46 58, Oct [169] T. Teixeira, G. Dublon, and A. Savvides, A Survey of Human-Sensing: Methods for Detecting Presence, Count, Location, Track, and Identity, ACM Comput. Surv., vol. 5, pp. 1 35, [170] F. Manzoor, D. Linton, and M. Loughlin, Occupancy Monitoring Using Passive RFID Technology for Efficient Building Lighting Control, in 2012 Fourth International EURASIP Workshop on RFID Technology, 2012, pp [171] T. Labeodan, W. Zeiler, G. Boxem, and Y. Zhao, Occupancy measurement in commercial office buildings for demand-driven control applications - A survey and detection system evaluation, Energy Build., vol. 93, [172] M. Choi, W.-K. Park, and I. Lee, Smart office energy management system using bluetooth low energy based beacons and a mobile app, 2015 IEEE Int. Conf. Consum. Electron., vol. 8, no. 4, pp , [173] H. M. Khoury and V. R. Kamat, Evaluation of position tracking technologies for user localization in indoor construction environments, Autom. Constr., vol. 18, no. 4, pp , Jul [174] S. Depatla, A. Muralidharan, and Y. Mostofi, Occupancy Estimation Using Only WiFi Power Measurements, IEEE J. Sel. Areas Commun., vol. 33, no. 7, pp , [175] H. M. A. Fahmy, WSNs Applications, in Wireless Sensor Networks, Signals and Communication, vol. 67, Hoboken, NJ, USA: John Wiley & Sons, Inc., 2016, pp [176] S. Lee, K. N. Ha, and K. C. Lee, A pyroelectric infrared sensor-based indoor locationaware system for the smart home, IEEE Trans. Consum. Electron., vol. 52, no. 4, pp , Nov [177] W. Shen and G. Newsham, Smart phone based occupancy detection in office buildings, in 2016 IEEE 20th International Conference on Computer Supported Cooperative Work in Design (CSCWD), 2016, pp [178] B. Balaji, J. Xu, A. Nwokafor, R. Gupta, and Y. Agarwal, Sentinel: Occupancy Based HVAC Actuation using Existing WiFi Infrastructure within Commercial Buildings, in Proceedings of the 11th ACM Conference on Embedded Networked Sensor Systems - SenSys 13, 2013, pp [179] G. Liu, a R. Dasu, and J. Zhang, Review of Literature on Terminal Box Control, Occupancy Sensing Technology and Multi-zone Demand Control Ventilation (DCV), PNNL Rep. PNNL , no. March, [180] R. Priyadarshini and R. M. Mehra, Quantitative Review of Occupancy Detection Technologies, Int. J. radio Freq. Des., vol. 1, no. 1, pp. 1 19, [181] Y. Li, D. Li, Y. Cheng, G. Liu, J. Niu, and L. Su, Demonstration Abstract: A Novel Human Tracking and Localization System Based on Pyroelectric Infrared Sensors, in th ACM/IEEE International Conference on Information Processing in Sensor Networks (IPSN), 2016, vol. 1, pp [182] L. D. Tran, A. Stojcevski, T. C. Pham, T. de Souza-Daw, N. T. Nguyen, V. Q. Nguyen, and C. M. Nguyen, A smart meeting room scheduling and management system with utilization control and ad-hoc support based on real-time occupancy detection, in 2016 IEEE Sixth International Conference on Communications and Electronics (ICCE), 2016, pp [183] F. Erden, A. Z. Alkar, and A. E. Cetin, A robust system for counting people using an infrared sensor and a camera, Infrared Phys. Technol., vol. 72, pp , Sep [184] C. Spataru and S. Gauthier, How to monitor people smartly to help reducing energy consumption in buildings?, Archit. Eng. Des. Manag., vol. 10, no. 1 2, pp , Apr [185] A. A. Adamopoulou, A. M. Tryferidis, and D. K. Tzovaras, A context-aware method for 117

132 BIBLIOGRAPHY building occupancy prediction, Energy Build., vol. 110, pp , [186] F. Wahl, M. Milenkovic, and O. Amft, A Distributed PIR-based Approach for Estimating People Count in Office Environments, in 2012 IEEE 15th International Conference on Computational Science and Engineering, 2012, pp [187] P. Liu, S. K. Nguang, and A. Partridge, Occupancy Inference Using Pyroelectric Infrared Sensors Through Hidden Markov Models, IEEE Sens. J., vol. 16, no. 4, pp , [188] A. Pellegrino and L. Blaso, Lighting Control Strategies and Energy Efficiency Benefit, in Sustainable Indoor Lighting, P. Sansoni, L. Mercatelli, and A. Farini, Eds. London: Springer London, 2015, pp [189] D. Caicedo and A. Pandharipande, Distributed ultrasonic zoned presence sensing system, IEEE Sens. J., vol. 14, no. 1, pp , [190] J. Hightower and G. Borriello, Location systems for ubiquitous computing, Computer (Long. Beach. Calif)., vol. 34, no. 8, pp , [191] D. Caicedo and A. Pandharipande, Ultrasonic arrays for localized presence sensing, IEEE Sens. J., vol. 12, no. 5, pp , [192] X. Guo, D. Tiller, G. Henze, and C. Waters, The performance of occupancy-based lighting control systems: A review, Light. Res. Technol., vol. 42, no. 4, pp , Dec [193] D. Müller, A. Monti, S. Stinner, T. Schlösser, T. Schütz, P. Matthes, H. Wolisz, C. Molitor, H. Harb, and R. Streblow, Demand side management for city districts, Build. Environ., vol. 91, pp , [194] J. Lau, CO2-Based Demand Controlled Ventilation for Variable Air Volume Systems Serving Multiple Zones, Indoor Built Environ., vol. 22, no. 5, pp , [195] T. Kalema and M. Viot, Methods to reduce the CO2 concentration of educational buildings utilizing internal ventilation by transferred air, Indoor Air, vol. 24, no. 1, pp , Mar [196] N. Nassif, A robust CO 2-based demand-controlled ventilation control strategy for multizone HVAC systems, Energy Build., vol. 45, pp , Feb [197] C. Jiang, M. K. Masood, Y. C. Soh, and H. Li, Indoor occupancy estimation from carbon dioxide concentration, Energy Build., vol. 131, pp , Nov [198] S. Meyn, A. Surana, Y. Lin, S. M. Oggianu, S. Narayanan, and T. A. Frewen, A sensorutility-network method for estimation of occupancy in buildings, in Proceedings of the 48h IEEE Conference on Decision and Control (CDC) held jointly with th Chinese Control Conference, 2009, pp [199] R. S. Risuleo, M. Molinari, G. Bottegal, H. Hjalmarsson, and K. H. Johansson, A benchmark for data-based office modeling: challenges related to CO2 dynamics, IFAC- PapersOnLine, vol. 48, no. 28, pp , [200] S. Petersen, T. H. Pedersen, K. U. Nielsen, and M. D. Knudsen, Establishing an imagebased ground truth for validation of sensor data-based room occupancy detection, Energy Build., vol. 130, pp , [201] V. L. Erickson, S. Achleitner, and A. E. Cerpa, POEM: Power-efficient Occupancy-based Energy Management System, in Proceedings of the 12th international conference on Information processing in sensor networks - IPSN 13, 2013, pp [202] S. Kumar, T. K. Marks, and M. Jones, Improving Person Tracking Using an Inexpensive Thermal Infrared Sensor, in 2014 IEEE Conference on Computer Vision and Pattern Recognition Workshops, 2014, pp [203] M. Berger and A. Armitage, Room occupancy measurement using low-resolution infrared cameras, in IET Irish Signals and Systems Conference (ISSC 2010), 2010, pp [204] D. Liu, X. Guan, Y. Du, and Q. Zhao, Measuring indoor occupancy in intelligent buildings using the fusion of vision sensors, Meas. Sci. Technol., vol. 24, no. 7, pp. 1 13, Jul

133 BIBLIOGRAPHY [205] B. Kelly, D. Hollosi, P. Cousin, S. Leal, B. Iglár, and A. Cavallaro, Application of acoustic sensing technology for improving building energy efficiency, Procedia Comput. Sci., vol. 32, pp , [206] M. Milenkovic and O. Amft, Recognizing Energy-related Activities Using Sensors Commonly Installed in Office Buildings, Procedia Comput. Sci., vol. 19, no. Seit, pp , Jan [207] E. Vattapparamban, B. S. Ciftler, I. Guvenc, K. Akkaya, and A. Kadri, Indoor occupancy tracking in smart buildings using passive sniffing of probe requests, in 2016 IEEE International Conference on Communications Workshops (ICC), 2016, pp [208] G. Morral and N. A. Dieng, Cooperative RSSI-Based Indoor Localization: B-MLE and Distributed Stochastic Approximation, in 2014 IEEE 80th Vehicular Technology Conference (VTC2014-Fall), 2014, pp [209] P. Rashmi, D. A, A. Kumar, R. S, and M. R, Smart building automation using device free localization technology Smart Building Automation Using Device Free, Int. J. Appl. Eng. Res., vol. 10, no. 10, pp , [210] E. Naghiyev, M. Gillott, and R. Wilson, Three unobtrusive domestic occupancy measurement technologies under qualitative review, Energy Build., vol. 69, pp , Feb [211] D. Brunelli, I. Minakov, R. Passerone, and M. Rossi, Smart monitoring for sustainable and energy-efficient buildings: A case study, 2015 IEEE Work. Environ. Energy, Struct. Monit. Syst. EESMS Proc., pp , Jul [212] T. A. Nguyen and M. Aiello, BEYOND INDOOR PRESENCE MONITORING WITH SIMPLE SENSORS, in Proceedings of the 2nd International Conference on Pervasive Embedded Computing and Communication Systems, 2012, pp [213] Y. Zhao, W. Zeiler, G. Boxem, and T. Labeodan, Virtual occupancy sensors for real-time occupancy information in buildings, Build. Environ., vol. 93, no. P2, pp. 9 20, [214] T. Ekwevugbe, N. Brown, V. Pakka, and D. Fan, Advanced occupancy sensing for energy efficiency in office buildings, Proc. Inst. Mech. Eng. Part I J. Syst. Control Eng., vol. 230, no. 5, pp , [215] W. O Brien and H. B. Gunay, The contextual factors contributing to occupants adaptive comfort behaviors in offices A review and proposed modeling framework, Build. Environ., vol. 77, pp , Jul [216] A. Peruffo, A. Pandharipande, D. Caicedo, and L. Schenato, Lighting control with distributed wireless sensing and actuation for daylight and occupancy adaptation, Energy Build., vol. 97, pp , Jun [217] T. Grimard and L. Kieran, Wireless Options Becoming More Prevalent with BAS, Building Automation. [218] C. de Farias, H. Soares, L. Pirmez, F. Delicato, I. Santos, L. F. Carmo, J. de Souza, A. Zomaya, and M. Dohler, A control and decision system for smart buildings using wireless sensor and actuator networks, Trans. Emerg. Telecommun. Technol., vol. 25, no. 1, pp , Jan [219] A. I. Dounis, Artificial intelligence for energy conservation in buildings, Adv. Build. Energy Res., vol. 4, no. 1, pp , [220] P. H. Shaikh, N. M. Nor, P. Nallagownden, and I. Elamvazuthi, Intelligent Optimized Control System for Energy and Comfort Management in Efficient and Sustainable Buildings, Procedia Technol., vol. 11, no. Iceei, pp , [221] R. Roche, B. Blunier, A. Miraoui, V. Hilaire, and A. Koukam, Multi-agent systems for grid energy management: A short review, in IECON Proceedings (Industrial Electronics Conference), 2010, pp [222] A. Al-Daraiseh, E. El-Qawasmeh, and N. Shah, Multi-agent system for energy 119

134 BIBLIOGRAPHY consumption optimisation in higher education institutions, J. Comput. Syst. Sci., vol. 81, no. 6, pp , [223] T. Labeodan, K. Aduda, G. Boxem, and W. Zeiler, On the application of multi-agent systems in buildings for improved building operations, performance and smart grid interaction A survey, Renew. Sustain. Energy Rev., vol. 50, pp , [224] D. Kolokotsa, D. Rovas, E. Kosmatopoulos, and K. Kalaitzakis, A roadmap towards intelligent net zero- and positive-energy buildings, Sol. Energy, vol. 85, no. 12, pp , Dec [225] M. Agrawal and A. Mittal, Overview of Multi-Agent Approach for Micro-Grid Energy Management, IOSR J. Electr. Electron. Eng., vol. 9, no. 2, pp , [226] P. Booij and V. Kamphuis, Multi-agent control for integrated heat and electricity management in residential districts, in Technologies for Energy, [227] B. Huberman and S. H. Clearwater, A Multi-Agent System for Controlling and Building Environments, in Proceedings of the First International Conference on Multiagent Systems, 1995, pp [228] C. J. Warmer, M. P. F. Hommelberg, B. Roossien, J. K. Kok, and J. W. Turkstra, A field test using agents for coordination of residential micro-chp, 2007 Int. Conf. Intell. Syst. Appl. to Power Syst. ISAP, pp. 1 4, Nov [229] F. Ygge, H. Akkermans, A. Andersson, and E. Boertjes, The HOMEBOTS System and Field Test: A Multi-Commodity Market for Predictive Power Load Management, in 4th Intl. Conf. on the Practical Application of Intelligent Agents and Multi-Agent Technology, 1999, pp [230] F. Ygge and H. Akkermans, Decentralized Markets versus Central Control: A Comparative Study, J. Artif. Intell. Res., vol. 11, pp , [231] P. Davidsson and M. Boman, Distributed monitoring and control of office buildings by embedded agents, Inf. Sci. (Ny)., vol. 171, no. 4, pp , May [232] ZhengChun Mo, Intelligent buildings and intelligent agents - a human-centered framework for building controls, in Proceedings of the 41st SICE Annual Conference. SICE 2002., 2002, vol. 5, pp [233] B. Qiao, K. Liu, and C. Guy, A multi-agent system for building control, Proc IEEE/WIC/ACM Int. Conf. Intell. Agent Technol. (IAT 2006 Main Conf. Proceedings), IAT 06, pp , [234] W. Zeiler, R. Van Houten, R. Kamphuis, and M. Hommelberg, Agent Technology to Improve Building Energy Efficiency and Occupant Comfort, in Proceedings of the Sixth International Conference for Enhanced Building Operations, Shenzhen, China, [235] J. Wall, G. Platt, G. James, and P. Valencia, Wireless Sensor Networks as Agents for Intelligent Control of Distributed Energy Resources, in nd International Symposium on Wireless Pervasive Computing, 2007, pp [236] G. Platt, J. Wall, P. Valencia, and J. K. Ward, The tiny agent- wireless sensor networks controlling energy resources, J. Networks, vol. 3, no. 4, pp , [237] J. D. J. Duan and F. L. F. Lin, Research of intelligent building control using an agentbased approach, th IEEE Int. Conf. Ind. Informatics, pp , Jul [238] S. Abras, S. Pesty, S. Ploix, and M. Jacomino, Advantages of MAS for the Resolution of a Power Management Problem in Smart Homes, in 8th International Conference on Practical Applications of Agents and Multi-Agent Systems, 2010, pp [239] Shen Bin, Z. Guiqing, Zhang Lin, and Wang Ming, Multi-agent system design for room energy saving, in th IEEE Conference on Industrial Electronics and Applications, 2010, pp [240] C. Spirleanu and E. Diaconescu, Distributed control system for robotic cells and intelligent building automation using agents, 15th WSEAS Int. Conf. Syst., pp , 120

135 BIBLIOGRAPHY [241] G. Fortino and A. Guerrieri, Monitoring building indoors through clustered embedded agents, 2011 Fed. Conf. Comput. Sci. Inf. Syst., pp , [242] K. Liu, J. Duangsuwan, and Z. Huang, Intelligent agents enabling negotiated control of pervasive environments, Chiang Mai J. Sci., vol. 38, no. 2 SPEC. ISSUE, pp , [243] J. Kwak, P. Varakantham, R. Maheswaran, Y.-H. Chang, M. Tambe, B. Becerik-Gerber, and W. Wood, TESLA: an extended study of an energy-saving agent that leverages schedule flexibility, Auton. Agent. Multi. Agent. Syst., vol. 28, no. 4, pp , Jul [244] L. Wang, Z. Wang, and R. Yang, Intelligent multiagent control system for energy and comfort management in smart and sustainable buildings, IEEE Trans. Smart Grid, vol. 3, no. 2, pp , Jun [245] R. Yang and L. Wang, Development of multi-agent system for building energy and comfort management based on occupant behaviors, Energy Build., vol. 56, pp. 1 7, Jan [246] B. Lacroix, L. Ines, and D. Mercier, Multi-Agent Control of Thermal Systems in Buildings, in Agent Technology for Energy Systems workshop, [247] G. E. Kelly and S. T. Bushby, Are intelligent agents the key to optimizing building HVAC system performance?, HVAC&R Res., vol. 18, no. 4, pp , [248] L. Zhang, T. Chang, H. Cui, X. Tian, and Z. Wang, Multi-agent System Design for Energy Saving in Intelligent Building, th Int. Symp. Comput. Intell. Des., pp , [249] P. Davidsson and M. Boman, A multi-agent system for controlling intelligent buildings, Proc. Fourth Int. Conf. MultiAgent Syst., pp. 9 10, [250] D. J. Cook, M. Youngblood, and S. K. Das, A Multi-agent Approach to Controlling a Smart Environment, in Designing Smart Homes, 2006, pp [251] A. Ulbig and G. Andersson, Analyzing operational flexibility of electric power systems, Int. J. Electr. Power Energy Syst., vol. 72, pp , [252] H. T. Haider, O. H. See, and W. Elmenreich, A review of residential demand response of smart grid, Renew. Sustain. Energy Rev., vol. 59, pp , [253] J. Shen, C. Jiang, and B. Li, Controllable load management approaches in smart grids, Energies, vol. 8, no. 10, pp , [254] E.-K. Lee, Experiencing Commercialized Automated Demand Response Services with a Small Building Customer in Energy Market, Int. J. Distrib. Sens. Networks, vol. 11, no. 11, p , Nov [255] Joyce Jihyun Kim, R. Yin, and S. Kiliccote, Automated Price and Demand Response Demonstration for Large Customers in New York City using OpenADR, [256] S. Nolan and M. O Malley, Challenges and barriers to demand response deployment and evaluation, Appl. Energy, vol. 152, pp. 1 10, Aug [257] B. P. Esther and K. S. Kumar, A survey on residential Demand Side Management architecture, approaches, optimization models and methods, Renew. Sustain. Energy Rev., vol. 59, pp , [258] S. Kiliccote and M. A. Piette, Advanced Control Technologies And Strategies Linking Demand Response And Energy Efficiency, in Procs of Int l Conf. for Evaluation of Building Operations (ICEBO 05), 2005, pp [259] R. J. De Dear, T. Akimoto, E. A. Arens, G. Brager, C. Candido, K. W. D. Cheong, B. Li, N. Nishihara, S. C. Sekhar, S. Tanabe, J. Toftum, H. Zhang, and Y. Zhu, Progress in thermal comfort research over the last twenty years, Indoor Air, vol. 23, no. 6, pp , [260] BSI, Light and lighting Lighting of work places (BS EN :2011). BSI, [261] D. S. Callaway and I. A. Hiskens, Achieving Controllability of Electric Loads, Proceedings of the IEEE, vol. 99, no. 1, [262] A. Feige, H. Wallbaum, M. Janser, and L. Windlinger, Impact of sustainable office 121

136 BIBLIOGRAPHY buildings on occupant s comfort and productivity, J. Corp. Real Estate, vol. 15, no. 1, pp. 7 34, [263] R. Badawy, A. Yassine, A. Hebler, B. Hirsch, and S. Albayrak, A novel multi-agent system utilizing quantum-inspired evolution for demand side management in the future smart grid, Integr. Comput. Aided. Eng., vol. 20, no. 2, pp , [264] K. Smith, Scaling Demand Response through Interoperability in Commercial Buildings, in Grid-Interop Forum, [265] P. Carreira, R. Nunes, and V. Amaral, SmartLink: A hierarchical approach for connecting smart buildings to smart grids, in Proceeding of the International Conference on Electrical Power Quality and Utilisation, EPQU, 2011, pp [266] IEA EBC Annex 67, Energy Flexible Buildings, [267] C. Johansson and P. Davidsson, Deployment of Agent Based Load Control in District Heating Systems, in 9th International Conference on Autonomous Agents and Multiagent Systems, 2010, pp [268] Z. Peng, S. Suryanarayanan, and M. G. Simoes, An Energy Management System for Building Structures Using a Multi-Agent Decision-Making Control Methodology, Ind. Appl. IEEE Trans., vol. 49, no. 1, pp , [269] H. A. Li and N. C. Nair, Multi-agent systems and demand response: A systematic review, in 2015 Australasian Universities Power Engineering Conference (AUPEC), 2015, pp [270] L. A. Hurtado, P. H. Nguyen, and W. L. Kling, Smart grid and smart building interoperation using agent-based particle swarm optimization, Sustain. Energy, Grids Networks, vol. 2, pp , Jun [271] J. Cai, J. E. Braun, D. Kim, and J. Hu, A multi-agent control based demand response strategy for multi-zone buildings, in 2016 American Control Conference (ACC), 2016, no. 1, pp [272] A. Cohen, P. Forghani, and W. Xing, Networked Distributed Optimization for Demand Response in Residential Load Control, Int. J. Innov. Technol. Explor. Eng., vol. 3, no. 4, pp , [273] J. K. Kok, C. J. Warmer, and I. G. Kamphuis, PowerMatcher: Multiagent Control in the Electricity Infrastructure, in Proceedings of the fourth international joint conference on Autonomous agents and multiagent systems - AAMAS 05, 2005, p. 75. [274] E. Karfopoulos, L. Tena, A. Torres, P. Salas, J. G. Jorda, A. Dimeas, and N. Hatziargyriou, A multi-agent system providing demand response services from residential consumers, Electr. Power Syst. Res., vol. 120, pp , [275] Z. Wang and R. Paranjape, A multi-agent system of evaluating residential demand response, Proc. Work. Commun. Comput. Control Resilient Smart Energy Syst. - RSES 16, pp. 1 6, [276] I. Dusparic, C. Harris, A. Marinescu, V. Cahill, and S. Clarke, Multi-agent residential demand response based on load forecasting, Technol. Sustain. (SusTech), st IEEE Conf., pp , [277] J. P. Buckley, A. Hedge, T. Yates, R. J. Copeland, M. Loosemore, M. Hamer, G. Bradley, and D. W. Dunstan, The sedentary office: an expert statement on the growing case for change towards better health and productivity, Br. J. Sports Med., vol. 49, no. 21, pp , Nov [278] M. Miki, H. Ikegami, R. Okunishi, K. Matsutani, and S. Fujimoto, An Intelligent Lighting System with a Seat Management Mechanism, in 2013 IEEE International Conference on Systems, Man, and Cybernetics, 2013, pp [279] A. Galajdová and D. Šimšík, Experimental Testing of Wireless Sensors Network Functionality, J. Autom. Control, vol. 3, no. 3, pp , [280] A. J. D. Rathnayaka, V. M. Potdar, and S. J. Kuruppu, Evaluation of wireless home 122

137 BIBLIOGRAPHY automation technologies, in IEEE International Conference on Digital Ecosystems and Technologies, 2011, vol. 5, no. June, pp [281] Zwave.. [282] CO2 Analyzer.. [283] B. D. Warden, Supply Air CO2 Control, ASHRAE J., no. October, pp , [284] S. a. Mumma, Transient Occupancy Ventilation By Monitoring CO2, IAQ Appl., vol. 5, no. 1, pp , [285] S. Ghosh, A. Mukherjee, K. Sahoo, S. K. Sen, and A. Sarkar, A novel sensitivity enhancement technique employing wheatstone s bridge for strain and temperature measurement, in Proceedings of the 2015 Third International Conference on Computer, Communication, Control and Information Technology (C3IT), 2015, pp [286] Strain Gauge Operation.. [287] M. Kuzlu, M. Pipattanasomporn, and S. Rahman, Review of communication technologies for smart homes/building applications, in 2015 IEEE Innovative Smart Grid Technologies - Asia (ISGT ASIA), 2015, pp [288] H. Vaha-Ypya, T. Vasankari, P. Husu, J. Suni, and H. Sievanen, A universal, accurate intensity-based classification of different physical activities using raw data of accelerometer, Clin. Physiol. Funct. Imaging, vol. 35, no. 1, pp , Jan [289] I. P. Machado, A. Luisa Gomes, H. Gamboa, V. Paixao, and R. M. Costa, Human activity data discovery from triaxial accelerometer sensor: Non-supervised learning sensitivity to feature extraction parametrization, Inf. Process. Manag., vol. 51, no. 2, pp , [290] R. Spinney, L. Smith, M. Ucci, A. Fisher, M. Konstantatou, A. Sawyer, J. Wardle, and A. Marmot, Indoor Tracking to Understand Physical Activity and Sedentary Behaviour: Exploratory Study in UK Office Buildings, PLoS One, vol. 10, no. 5, pp. 1 19, May [291] S. Goyal, P. Barooah, and T. Middelkoop, Experimental study of occupancy-based control of HVAC zones, Appl. Energy, vol. 140, pp , [292] L. L. Fernandes, E. S. Lee, D. L. Dibartolomeo, and A. McNeil, Monitored lighting energy savings from dimmable lighting controls in the New York Times Headquarters Building, Energy Build., vol. 68, no. PARTA, pp , Jan [293] A. Williams, A. Barbara, K. Garbesi, and P. Erik, Lighting Controls in Commercial Buildings, LEUKOS J. Illum. Eng. Soc. North Am., vol. 8, no. 3, pp , [294] N. Gentile, T. Laike, and M. C. Dubois, Lighting control systems in individual offices rooms at high latitude: Measurements of electricity savings and occupants satisfaction, Sol. Energy, vol. 127, pp , [295] S. Rohini and K. Venkatasubramanian, Z-wave based zoning sensor for smart thermostats, Indian J. Sci. Technol., vol. 8, no. 20, pp. 1 6, Aug [296] J. P. Carles Gomez, Wireless Home Automation Networks: A Survey of Architectures and Technologies, no. June, pp , [297] Benext Products.. [298] D. Vallejo, J. Albusac, C. Glez-Morcillo, J. Castro-Schez, and L. Jiménez, A multi-agent approach to intelligent monitoring in smart grids, Int. J. Syst. Sci., vol. 45, no. 4, pp , Apr [299] Z. Nagy, F. Y. Yong, M. Frei, and A. Schlueter, Occupant centered lighting control for comfort and energy efficient building operation, Energy Build., vol. 94, pp , May [300] J. De Jong, L. Stellingwerff, and G. E. Pazienza, Eve: A novel open-source web-based agent platform, Proc IEEE Int. Conf. Syst. Man, Cybern. SMC 2013, pp , [301] R. Leszczyna, Evaluation of Agent Platforms, [302] K. O. Aduda, T. Labeodan, W. Zeiler, and G. Boxem, Demand side flexibility coordination in office buildings: A framework and case study application, Sustain. Cities Soc., vol. 29, 123

138 BIBLIOGRAPHY pp , [303] L. Wang and S. Greenberg, Window operation and impacts on building energy consumption, Energy Build., vol. 92, pp , [304] K. O. Aduda, W. Zeiler, G. Boxem, and K. De Bont, On the use of electrical humidifiers in office buildings as a demand side resource, in Procedia Computer Science, 2014, vol. 32, pp [305] E. Mocanu, K. O. Aduda, P. H. Nguyen, G. Boxem, W. Zeiler, M. Gibescu, and W. L. Kling, Optimizing the energy exchange between the Smart Grid and Building Systems, in Proceedings of the Universities Power Engineering Conference, 2014, pp [306] Z. Wang and L. Wang, Adaptive Negotiation Agent for Facilitating Bi-Directional Energy Trading Between Smart Building and Utility Grid, IEEE Trans. Smart Grid, vol. 4, no. 2, pp , Jun [307] A. Rogers, S. Ramchurn, and N. R. Jennings, Delivering the smart grid: Challenges for autonomous agents and multi-agent systems research, Proc. 26th AAAI Conf. Artif. Intell., pp , [308] Eurostat, Electricity and gas prices, second half of year, (EUR per kwh). [Online]. Available: 15_(EUR_per_kWh)_YB16.png. [Accessed: 17-Mar-2017]. [309] T. Kershaw and D. Lash, Investigating the productivity of office workers to quantify the effectiveness of climate change adaptation measures, Build. Environ., vol. 69, pp , Nov [310] N. Kamarulzaman, A. A. Saleh, S. Z. Hashim, H. Hashim, and A. A. Abdul-Ghani, An overview of the influence of physical office environments towards employees, Procedia Eng., vol. 20, pp ,

139 APPENDIX APPENDIX A: - Trias Energetica The Trias Energetica is a very useful method in connection with energy optimization of a specific design after clarifying the client s wishes for the program and completing the Integrated Energy Design process. The Trias Energetica makes clear that energy savings have to come first on the path to environmental protection. Only when a building has been designed to minimise the energy loss, should the focus shift to renewable energy solutions, such as solar panels or heat exchange and recovery systems. The method was developed by Kees Duijvestein from the TU Delft in 1979 and internationally introduced by Erik Lysen of Nederlandse Onderneming voor Energie en Milieu (Novem) a preprocessor of Rijksdienst voor Ondernemend Nederland (RVO), Netherlands Enterprise Agency, the National funding organization which encourages entrepreneurs in sustainable, agrarian, innovative and international business. It helps with grants, finding business partners, know-how and compliance with laws and regulations. The Trias Energetica principle can be used as a superior principle under each step of the Integrated Energy Design process. The Trias Energetica method prioritizes various initiatives in connection with the use of energy. The method consists of 3 steps: 1. Reduce the demand for energy 2. Use sustainable sources of energy 3. Use fossil fuels as efficiently as possible 125

140 APPENDIX The starting point in a specific design process will always be user comfort, especially in commercial buildings. It is important to have a differentiated view on each comfort situation, and to consider simultaneity, dress-code, level of activity etc. The first step in the design process is to focus on minimising the energy need. This can be done by: Analyses of the site (solar and wind conditions, orientation etc.) Optimization of the building geometry Using good windows Optimize daylight use Sufficient insulation Minimising thermal bridges Tightening the building Using any free heat supplements Avoiding overheating (which produces cooling needs) Using energy efficient artificial lighting Avoiding standby losses Etc. The second step is about covering the building s energy need with as much sustainable energy as possible. In terms of design, it means focusing on e.g.: Using free solar heat supplements for heating Prioritizing low temperature heating like floor heating, which can be combined with \ solar heat Examining the possibilities for using biomass boilers Designing for the use of active solar heat and solar cell systems Etc Third step is about efficient use of fossil fuels to cover the remaining energy demands in the building. This is done in the design process by ensuring: An efficient energy supply Adjusting for the actual operating profile of the building Dividing into zones of similar functions and needs, which can be supplied from the same unit Etc. 126

141 APPENDIX APPENDIX B: - EVE- Agents Eve is a multipurpose, web-based agent platform. Eve envisions to be an open and dynamic environment where agents can live and act anywhere: in the cloud, on smartphones, on desktops, in browsers, robots, home automation devices, and others. The agents communicate with each other using simple, existing protocols (JSON-RPC) over existing transport layers (HTTP, XMPP), offering a language and platform agnostic solution. For a good understanding of Eve, it is important to look at its concept Agent. The basic definition of agent is: A software entity that represents existing, external, sometimes abstract, entities. Examples of such entities are: human beings, physical objects, abstract goals, etc. To be able to function, the agent needs to be able to run independently of the entity it represents. This autonomous behaviour requires a basic set of capabilities, which Eve provides for its agents. These features are: time independence, scheduling, independent of the represented entity. memory, the possibility to keep a model of the state of the world communication, a common language to communicate between agents The main reason for providing a separate memory capability to the agents is that, in most implementations, Eve agents have a request based life cycle. The agent is only instantiated to handle a single incoming request and is destroyed again as soon as the request has been handled. Only the externally stored state is kept between requests. The agent identity is formed by the address of its state, not by an in-memory, running instance of some class. The scheduling capability provides requests at scheduled times, thus instantiating the agent at that time. 127

142 APPENDIX This model mimics the way modern webservers handle servlets, allowing any servlet to become an Eve agent. Out of the box, a single agent (=single state) can therefore execute requests in parallel, multi-threaded, and (depending on the state-implementation) distributed among multiple servers. This model also allows for easier handling of asynchronous agent designs. The agent model is based on request driven instantiation. The requests, that trigger this instantiation, are JSON-RPC encoded. In the Protocol section, this protocol is further defined and described. Using JSON as its base, Eve is highly programming language independent; for most mainstream languages, there are existing JSON handling libraries. Any implementation that adheres to the described protocol and basic agent model, is therefore considered an Eve implementation. This high-level definition allowed effective reuse of existing tools and protocols in the implementation of Eve. Eve is defined by its agent model and its RPC protocol. But the JSON-RPC messages still need to be transported from one agent to another agent, possibly crossing to a completely different platform. The two implementations (on the different platforms) need to share a common transport service for this. Currently the Eve implementations support HTTP and XMPP as transport protocols, but it's relatively straight-forward to e.g. implement a ZeroMQ transport. If two implementations share a common transport service, its agents can work flawlessly together. For example, this allows a Java cloud agent to communicate with an in-browser Javascript agent, through JSON-RPC over an XMPP transport. An important part of any agent platform is its addressing and directory service. In line with maximum reuse of existing protocols and runtime environments, Eve also reuses existing addressing frameworks for its agents: The HTTP transport services assigns a normal public URL for each agent, allowing using normal DNS servers as directory services. Similarly, the XMPP transport uses the JIDs as agent addresses, allowing the server roster to be used as directory service. In the current implementations Eve is working on servlet based servers (e.g. Google App Engine, Amazon Web services, etc.), as standalone applications, on mobile devices (e.g. Android) and in-browser, through its JavaScript implementation. Eve defines an interaction protocol between agents. This protocol consists of a common RPC language and some expected behavioural patterns. This interaction protocol is described: Communication Protocol describes the RPC language for the agents. Required methods describes the standard methods of an agent. 128

143 APPENDIX Management methods describe optional methods that an implementation should make available to administrators and developers. Interaction patterns describe common behaviour that implementations should provide to agents, with associated methods. Documentation links to resources related to the JSON-RPC protocol. Eve agents communicate with each other using the JSON-RPC protocol. This is a simple and readable protocol, using JSON to format requests and responses. JSON (JavaScript Object Notation) is a lightweight, flexible data-interchange format. It is easy for humans to read and write, and easy for machines to parse and generate. JSON-RPC version 2.0 is the minimally required version, because Eve uses named parameters. A request from Agent X to agent Y can look as follows: Agent X addresses method add from agent Y, and provides two values as parameters. Agent Y executes the method with the provided parameters, and returns the result. Building EVE-Agent code package kropmantest; import java.io.file; import java.io.fileinputstream; import java.io.ioexception; import java.net.uri; import java.util.logging.level; import java.util.logging.logger; import org.joda.time.datetime; import com.almende.eve.agent.agent; import com.almende.eve.config.config; import com.almende.eve.config.yamlreader; import com.almende.eve.deploy.boot; import com.almende.eve.protocol.jsonrpc.annotation.access; import com.almende.eve.protocol.jsonrpc.annotation.accesstype; import com.almende.eve.protocol.jsonrpc.annotation.name; import com.almende.eve.protocol.jsonrpc.formats.params; import com.almende.util.callback.asynccallback; import com.fasterxml.jackson.databind.node.objectnode; 129

144 public class BuildingAgent extends Agent { //Start users agents public void startusers() throws IOException, InterruptedException{ call(uri.create(" "StartControl", null); call(uri.create(" "StartControl", null); call(uri.create(" "StartControl", null); } //Start Room agents public void startrooma() throws IOException, InterruptedException{ call(uri.create(" "StartControl", null); call(uri.create(" "StartControl", null); } //Start Gateway and databases public void startgateway() throws IOException{ call(uri.create(" "startgateway", null); } //Stop Gateway and databases public void exitgateway() throws IOException{ call(uri.create(" "exitprogram", null); } public class DRevent extends Agent { Kropmandecoder test = new Kropmandecoder(); String data; private static final Logger LOG= Logger.getLogger(DRAgent.class.getName()); 130

145 APPENDIX public void RequestData() { final Params params = new Params(); final Params subparams = new Params(); params.put("libraryname", "Datapoints"); //subparams.put("state", state); //params.set("parameters", subparams); try {call(uri.create("ws://tcc-kropman.dyndns.org:3000/agents/remoteguy"), "ExecLibrary", params, new public void onsuccess(objectnode result) LOG.info("Received result:" + (data= result.tostring()) ); } public void onfailure(exception exception) { LOG.log(Level.SEVERE, "Failed to call remoteguy:", exception); } catch (IOException e) { LOG.log(Level.SEVERE, "Failed to call remoteguy:", e); public void schedulespecific(){ schedule("schedulespecific",null,datetime.now().plus( DateTime.now().millisOfSecond().get())); RequestData(); test.databreaker(data); public void switchlamp(@name("state") int state){ final Params params = new Params(); final Params subparams = new Params(); params.put("libraryname", "Lib3056"); } }} }); } 131

146 APPENDIX subparams.put("state", state); params.set("parameters", subparams); try { "ExecLibrary", params, new AsyncCallback<ObjectNode>() { public void onsuccess(objectnode result) { LOG.info("Received result:" + (data= result.tostring()) public void onfailure(exception exception) { LOG.log(Level.SEVERE, "Failed to call remoteguy:", exception); } catch (IOException e) { LOG.log(Level.SEVERE, "Failed to call remoteguy:", e); }}} public static void main(string[] args) throws IOException, InterruptedException { }} } }}); final Config configfile = YamlReader.load( Boot.boot(configfile); new FileInputStream(new File(args[0]))); 132

147 APPENDIX APPENDIX C: - PV and battery energy storage system installation at test-bed Given that Photovoltaic and battery storage systems can serve as an additional source of useful energy flexibility in office buildings, Photovoltaic and battery storage systems were installed as depicted in the figures below in the test bed office building. Appendix 1: PV installation on test-bed building 133

148 APPENDIX Electrically conductive contact modules Controlled stack pressure Isolation plate Integrated temperature control Intelligent end pieces with terminals Appendix 2: Battery Appendix 3: battery Storage The design aimed to enhance self-consumption whilst also ensuring reliable delivery of power flexibility from the building. In particularly, the following were taken into account: 1. Self-consumption of generated energy: The battery was sized to ensure that the energy produced by the roof top installed PV generator in building was significantly self-consumed; the storage catered for excess generations whenever building is closed. 2. Demand side flexibility: This entailed improvement of building capacity to be used reliably as a demand side flexibility resource. Demand side flexibility is in this context defined as intelligently use of building based storage and other connected loads to enable occasional supply side power balancing without compromising occupants comfort, safety and productivity. Performance characteristics considered include length of cycles, cycle efficiency and number of cycle s energy capacity, power availability, power system response time requirements. The system was designed to ensure that at critical moments, participation of the building could be supported for short-term power flexibility services of up to one hour. Specification of the system chosen was as illustrated in Figures 4 to

149 APPENDIX Appendix 4: Line diagram of the EES with specification details Appendix 5: System connection with BEMS 135

150 The research presented in this dissertation bascially makes the argument for the use of fine-grained occupancy information and multi-agent systems in building operation for improvement of energy management. The conclusion from this research indicates that the availability and utilization of fine-grained building occupancy information particularly on the building room-level can facilitate worthwhile energy savings. In addition, it also indicates that the combination of both fine-grained occupancy information and multi-agent systems can enhance energy management and facilitation of buildings energy flexibility for grid support while maintaining comfort. / Department of the Built Environment finalcover-mh3.indd :54:19