Modelling of Energy Consumptions at IST TagusPark Building

Transcription

1 Modelling of Energy Consumptions at IST TagusPark Building Carlos Varela Raposo June 2015 Abstract Technical Institute of Lisbon, through Sustainable Campus project, has been developing models of dynamic thermal simulation of its campus, to optimize energy consumptions. The purpose of this work was the construction, calibration and validation of a model from the TagusPark building, with the characterization of its internal areas, so that the experiments were reliable. It was intended to simulate the annual consumption of the building and testing the application of consumption reduction methods in order to realize its efficiency. The modeling of the building and its geometric, environmental and technological characteristics were made from DesignBuilder, an EnergyPlus graphical software interface, used to model gains and losses of energy in buildings, lighting and ventilation, thermal comfort simulation, photovoltaic systems, air circulation and sub-hourly evaluations. The building model was calibrated with values measured by sensors placed in the building and information from an inquiry. The model was validated through statistical analysis, and the results showed a 5% error, staying within the range forecasted 2%-16%. After validation of the model is simulated, the annual consumption of the building, which were of 1,555 MWh, and tested two consumption reduction measures: the lighting power reduction, replacing 36W T8 bulbs with 21W LED, checking a reduction of 5,3kWh / week; and decreasing the set point temperature of 24ºC to 23ºC, checking 48 kwh / week It was concluded that the implemented model was able to reproduce the actual consumption of the building, and that the proposed energy efficiency measures were effective. Keywords: Building modelling; DesignBuilder; EnergyPlus software; Energy consumption analysis. 1. Introduction Since the end of the twentieth century the world population has experienced a large increase in numbers [1]. Consequently, there is an increase in primary energy consumption as well as a larger production of CO2 and other greenhouse effect gases, urging the reduction of consumptions. The building construction sector has been at the forefront of efforts to reduce consumptions. In the European Union, global energy consumption estimates show that 40% of final energy consumptions come from buildings [2], [3], whereas in Portugal is approximately 30% [4]. The use of building modeling software and energy consumption analysis software has a way to identify potential areas in need of intervention and to simulate the results of applying energy consumption reduction measures and has been a used method of research [5]. Modeling software and energy consumption analysis software allow for quantitative and qualitative analysis of specific areas in buildings. When well calibrated, they present values very close to measured energy consumption values. The purpose of this work is to create a model for IST TagusPark service building using EnergyPlus software. The model will be calibrated with data collected from several sensors that exist in the building. Afterwards, a validation process will be made, were data collected from a simulation run on the model will be compared to real data collected from the building, to assess the validity of the model. Two experiments will be made with the validated model. The first experiment will be a long-term simulation of energy costs of the building, to identify the biggest energy expenditures. The second experiment will be a simulation of the application of energy consumption reduction measures, to see if they can successfully reduce energy consumption. 2. Bibliographic Review Zhang and de Dear [5], from Sidney University, Austrália, studied the reduction in electric energy consumption of an university campus, given the elevated consumption and fiscal penalties imposed by local authorities. The reduction strategy used Direct Load Control in air conditionings, so that usage and power spikes could be monitored and minimized. DesignBuilder 1

2 software was used to simulate the building s consumptions, using a previously validated model, with an acceptable lack of accuracy of around 7% lack of precision when compared to measured consumptions. As a follow-up to a previous research [6], Goia and Cascone [7] studied the advantages that a variable window-to-wall ratio building (dynamic system) would have versus the fixed window-to-wall ratio system of the actual building. EnergyPlus software was used to simulate both conditions in models of a service building in Oslo, Norway. Monthly, weekly, daily and hourly variations were simulated to measure the difference between systems. The monthly scheduling was concluded to be the most efficient energy reduction strategy in the dynamic system, reducing primary energy consumption in 6% when compared to the static system. The use of plants in facades and rooftops aids in building sustentability. The improvements obtained by such systems were reported by Feng and Hewage [8]. Advantages of these systems include reductions of energy consumption related to cooling and heating, noise pollution and air quality improvements. good simulation results are not guarented. The calibration process relies heavily on user experience and knowledge, on numerous engineering aproximations and many trial-and-error attempts. There is no consensus in the scientific community on how to calibrate a building s model. 3. IST-Taguspark Campus Model Construction IST-Taguspark campus is situated in Oeiras, a municipality in the western part of Lisbon Metropolitan Area. It is a three storey building, mostly above ground, and a fourth storey underground, that is mostly used as a parking garage. Given the land s slope, the ground storey has the southeast facade above ground and the northen facade of the ground floor completely underground. The central area of the building is open on all three storeys, creating a common loby and it is covered by a skylight roof that allows sunlight exposition. It is worth mentioning the building s angled geometry (as seen in Figure 1), which makes the western facade bigger that the eastern facade. According to Zhao et al. [9], another important factor in energy consumption, particulary in service buildings, is its occupants behavior. The way energy is used is influenciated by people s professional activites, as much as by their bad habbits. Mustafaraj et al. [10] published a methodology for calibrating simulated models from measurements of the actual building. Two calibration methods were used, one developed by Raftery et al. [11], and the other by Bertagnolio [12]. Both calibration methods rely on manual calibration performed in iterations. Coakley, Raftery and Keane [13] revised methods for comparison of simulated building s energy values and actual building s measured values. The authors note the difficulty to reach accurate results from simulations is due to the amount of different parameters that need to be modeled to properly characterize a building. Even with good calibration of the model, Figure 1 - Building top view (Google Maps image) 3.1. Geometric information To model TagusPark campus, geometric parameters were gathered from blueprints, 2

3 such as surrounding facades measures, building geometry, exterior wall s thickness as well as the building s blocks and partitions. Volume (VAV), which can be seen in Figure Glazing After gathering geometric information, glazing (ceiling and windows) were added to the model. Particular attention was paid to modeling the library, that has an interior glazing facade that spans two storeys, the roof which is a skylight, and nucleous 14 th (on the 2 nd floor), since only the library and nucleous 14 th had a network of temperature sensors installed and all three elements cause the most relevant impact of building temperature. Using the building s blueprints, the external windows on the facades were characterized with double glazing and aluminium frames, covering 30% of the exterior walls of each termal area. Shading is made with Venetian blinds inside and three metallic blinds per window outside. The skylight roof was modeled identical, except it has no blinds and occupies a larger area Building materials The informations regarding walls, pavement and building ceiling were gathered also from the building plans. Despite the high number of available materials in DesignBuilder database, it was not always possible to have a match between the materials described in the plans and those chosen to the model. In those cases were selected the material with characteristics closer to the plan. Table 1 - Characteristics values of internal resistence (R) and heat transfer coefficient by convection (U) of the building components R (m 2. K W) U (W m 2. K) Width (m) Outside Walls 2,956 0,338 0,285 Inside Walls 1,175 0,851 0,15 Floor 0,617 1,620 0,490 Ceiling 4,039 0,248 0, HVAC The HVAC system chosen from DesignBuilder database was a Variable Air Figure 2 - HVAC VAV system working diagram The heating process is obtained through combustion of natural gas in boilers with atmospheric combustion chambers. These boilers were designed to have efficiency equal or higher to 90%. This value was obtained in average from the combustion efficiency measuring process. It was then assumed a global efficiency value of 85% and 80 o C to the circulating water. A chiller powered with electricity is used to provide cooling. Energy Efficenfy Ratio, or EER, determination used technical data from the Maintenance Area. Since this chiller s termal power is 635 kw and maximum electrical power is 302 kw, in a worst case scenario, its efficiency would be 2,1. This value was chosen as EER, since chillers gradually decrease efficacy with use over time. From November to March, these equipments function from 8h am to 7h pm. From April to October/November, chillers have two work modes: from 12h am to 8h am, were ice banks are charging, and from 8h am to 7h pm, in production. Given these work modes, the following schedule was elaborated: April, May, October and November (coolest months) from 12h am till 8h am June, July and September (warmest months) from 12h am till 7h pm In August, the facilities of TagusPark are closed Lighting Lighting parameters characterization is made in two separate tabs in DesignBuilder. In Activity tab, given the activity of the area being modeled, it is possible to define the level of target luminance of that area. Default values for each type of area exist in the 3

4 program. Values assigned to each area during modeling were estimated according to [14] or default values were used. In Table 2 the values used for each type of space are presented. Table 2 - Illuminance by partition type Type of partition Illuminance (lux) Cafeteria 150 Amphitheaters 300 Storage 150 Locker room 100 Library 500 WC 100 Circulation 100 Kitchen 500 Offices (2 nd floor) 500 Laboratories 800 Classrooms/ Meeting room 300 With luminace values modeled, it is possible to specify the type of luminaire, the associated power (W/m lux) and the fractions of radiant heat and visible heat, as well as their schedule. Table 3 - Visible and radiant fraction values by luminaire type Visible Radiant Luminaire type Fraction Fraction Suspended 0,42 0,18 Recessed 0,37 In the offices of Nucleus 14 th, the number of luminaires changes from two (for exterior offices) to three (for interior offices), having two 36W lightbulbs in each one, with electronic ballast. Some interior offices only use two of the three available luminaires. As such, illumination energy estimations for all offices were assumed equal, a set at 2W/m lux. The associated gain from the lighting is solely dependent on the level of luminance defined for each space Occupancy/Activity The information in these models cover areas such as occupation, power consumption in electrical equipment (such as computers, printers, etc.), setpoint temperature, illuminance levels and per person ventilation rates. Taking advantage of the existence of these models were chosen some to characterize most areas of the modeled building. The general characteristics of the amphitheater 4 are in Table 4. The estimated maximum occupancy was 130 people, that number was used to create a daily profile of occupancy for each day of the week depending on the classes taking. Table 4 - Activity data of amphitheater A4 A 4 Area (m 2 ) 110,9 8 Occup (pers./m 2 ) Equip (W/m 2 ) Setpoint Temp. ( o C) Heat Cool.. 1, The procedure was then the characterization of Nucleus 14 th. In order to collect the necessary data it was used an online questionnaire to assess what kind of equipment were working and how many people were present for office, in a particular week (16-22 March), and the usual time of occupation of their office. In addition to the data collected through the questionnaire was carried out a survey of several data for these offices (current temperature, setpoint temperature, on / off status of lighting and HVAC, global power consumption and fan speed). This survey was achieved due to the existence of sensors installed in the cabinets and appliances, which allow performing the reading by connecting to the server. The values were recorded every 15 minutes during the week mentioned above, through a routine developed in Matlab and then combined with the data obtained by the inquiry. In Table 5 are the final data, in Nucleus 14 th used for simulation. Table 5 - Activity data of N14 offices N14 room Occup. (pers./m 2 ) Equip. (W/m 2 ) Setpoint Temperature ( o C) Heat. Cool. N14.2 0, ,5 24 N14.4 0, N14.6 (meeting 0, room) N14.8 0, ,5 24 N , N , N , N , N , N ,

5 N , N , N , The characterization of the parameters described in this chapter will influence the results obtained in the simulations. Therefore, it is important to collect this data in some detail so that the obtained model is favored in its precision and reliability. Through the results that will be discussed in the next chapter, it is found that the parameters related to the activity carried out in the area to characterize have greater impact on the variation of the results obtained compared with the set of construction related parameters, including the exact materials used. As if we are to study the variation of the light level, either by replacing existing luminaires or adding / changing shading systems (such as blinds or flaps), the areas relating to lighting and openings become more important. 4. IST-Taguspark Campus Model Validation The different simulations taken in the terminal phase of model calibration were carried out for the same week of March it was made readings from sensors installed in the Nucleus 14 th, in the library and in amphitheater 4. The type of parameter to be measured varied with the partition under analysis, since the monitoring is not equal between them. In these locations it was possible to perform consumed power readings (lighting, ventilation and electrical equipment); however, only in Nucleus 14 th was possible to register the evolution of temperatures within the various offices. Thus, the analysis is necessarily different, as well as the calibration of each area was, motivating some disparities between the simulation results and measured by monitoring. N14 Offices The Nucleus 14 th is a set of offices located on east side, of block E, on the second floor of the building. It consists of 14 rooms, of which 12 are offices, individual or shared by two or three people, a meeting room and a storage room. Figure 3 - N14 offices location Following the simulation it was made the comparison between the obtained temperature values read by the program and the existing monitoring values for all rooms. Overall the results were very satisfactory since on average for the 14 rooms, the difference in the readings have a relative error around 5%. However, it is noted that the night period, where there is no occupancy, provided the greatest differences between the values of the monitoring and of the simulation. This event can be explained by the possible working of some equipment at this time, since the power consumption values obtained from monitoring are not zero, as will be shown below. Another reason may be the location of temperature sensors, as these are in the inner walls of the offices, not identifying the actual inside temperature decrease. The displays the temperature profile of one of the offices, the N14.16 office, during the week under study, March 16 (Monday), starting from 1:00 pm, until 22 March (Sunday) at 11:00 pm. The presented curves refer to the temperatures recorded by the meteorological station of Oeiras and inserted in the weather program file (Outdoor Temperature - yellow), the temperatures obtained by the software after simulation (Model - Profile in blue) and the actual temperature read by the system monitoring offices (Real - Profile in gray). Despite the gap recorded between 1:00 pm and 7:00 pm from Friday due to server problems, it can be seen the general trend of temperature change inside the office following the temperatures outside the building, with its peak in the early afternoon. It also can be seen the closer relationship between the gray line (actual values) and the blue line (model values) during the day, coinciding with occupation of the area, and the largest distance between them, with a 5

6 weekly maximum recorded 2,3 C at 7:00 am Sunday. 3, ,5 20 POWER (KW) 2 1,5 TEMPERATURES (ºC) ,5 0 13:00 1:00 13:00 1:00 13:00 1:00 13:00 1:00 13:00 5 Model Real 0 Model Real Outside Graphic 2 - Model and real consumption profile of N14 offices Graphic 1 - Outside and measured temperatures at N14.16 office After comparing the temperature lines was carried out identically the comparison of consumption behavior. In this case the analysis comprised the total of the offices, since the consumed power control is not individualized. Thus, after the simulation has been gathered power consumption of each case and subsequently added in order to obtain the intake time profile in simulated testing week (weekdays only). The actual consumption profile was obtained by monitoring, and compared to the profile of the simulated model. This comparison can be seen in Graphic 2. In blue is the power profile of the simulated model and in orange is the actual profile. Is possible to see immediately the difference between the minimum requirements set daily. While the estimated consumption simulated model is virtually nil, with regard to the monitoring profile the reality is different, there is a minimum consumption associated with this set of offices of approximately 500 W. This minimum consumption can be attributed to the control and monitoring system. With regard to registered maximum values, there is an approximation in two days this week, wednesday and thursday, and an average difference of about 750 W in the remaining days. This difference comes from the estimated consumption obtained during the investigation and is not an accurate process at achieving results. However, the behavior of both curves is identical estimating, on average, a difference of about 5.2% between the actual values and the simulation Amphitheater A4 The A4 amphitheater is located on the west side of the block B of the building. Has an entrance on level 0 extending to the floor -1 due to the slope of the ground. In this area of the floor of the building partitions are below ground. This implies that there are no windows to the outside of the building. Figure 4 - Amphitheater location As in the Nucleus 14 th, the power consumption of the amphitheater were recorded throughout the week. Since there aren't in place temperature sensors, the power consumed values are the only parameters used in the comparison between the model and the actual data. Given that the number of unknown variables, in this case, is higher than the previous case, 6

7 it was estimated the frequency of use of the equipment and lighting depending on class schedules and occupation of spaces in order to ascertain the effects on consumption profiles. POWER (KW) POWER (KW) 1,4 1,2 1 0,8 0,6 0,4 0,2 0 13:00 1:00 13:00 1:00 13:00 1:00 13:00 1:00 13:00 Model Graphic 3 - Model and real consumption profile of amphitheater A4 Of the chart analysis it is possible to check the existence of night consumption, not accounted in the model, equivalent to about 150W. These inputs are probably associated with the control and monitoring. The existence of such consumption translates into the difference between the minimum values observed in both profiles. It is possible to see generally peaks associated with periods of no occupation between classes, of their days. However, contrary to what happens to Tuesday where there is a break around 1:00 pm, this is estimated to happen due to lack of class, the same does not occur on the next day a reduction in comsumption would be predicted between am and 2:30 pm by the absence of class shown in the amphitheater schedule. This decrease, shown in the profile in blue, is not seen in the real profile and it is estimated the lighting remained on during this period. Generally the simulation consumption profile, except for the night period already mentioned, shows a more realistic behavior profile and closer to the orange one. This approach is proven statistically checking for the daytime where there is not zero consumption in the simulation; the relative error is about 5.4%. Real 4.2. Library The library is divided into two floors (0 and 1), being located on the west side of the block C of the building. On the outer side (west) there are windows allowing solar radiation input. The opposite side, on the inside, is connected between the two floors and a glazing facade to separate the interior space of the library to the common lobby of the building. This facade allows a lot of natural light. The library location in the building can be found in Figure 4. As studied in the amphitheater, monitoring at library provides only the power consumption, these being related to the study room of the ground floor. Thus, comparative data will be presented only in this area, which is also the greatest in this area. Figure 5 - Library location From the data collected from the monitoring and parameters entered into the software, also making use of the characteristics of this in relation to this type of areas, were traced the power consumption curves for the 5 days of the week indicated. The consumer profiles are illustrated in Graphic :00 1:00 13:00 1:00 13:00 1:00 13:00 1:00 13:00 Model Graphic 4 - Model and real consumption profile of library From the previous chart can be checked the consumption peaks recorded during the period between 8:00 am and 8:00 pm, interval where there are consumption of about 8kW. Contributing to these Real 7

8 consumptions are identified lighting, ventilation and electrical equipment plugged into outlets, including computers. It is observed mainly the consumptions profile obtained by the program accompanying the actual profile identically obtained from the sensor readings. The moments of rising and falling consumption present themselves well characterized, and the values obtained during the day fit with the real profile. Statistically the findings support the above analysis verifying that on average the relative error is less than 6%. 5. Model Validation In the preceding chapter was explored the characterization of some building areas, which had monitoring of several parameters allowing a more careful modeling and, in most situations, satisfactory results and more consistent with the reality. Among the four areas the Nucleus 14th disposed more attention, not only due to the higher number of data associated with the highest number of divisions, but also for being the only one with monitoring internal temperatures. Such monitoring, existing in each office, was based on analysis of the modeled results, which are complemented with the power values consumption for the entire nucleus. In the remaining areas only the power consumed was the basis for analysis of the values obtained in the simulation. In Table 6 are the summary statistics for each area modeled in greater detail. The absolute error and relative error values are represented on average, presenting the standard deviation associated with the absolute error of the different measurements. Table 6 - Statistics data of the studied areas Mean Absolute Error Percentage Error Standard Deviation N14 1,07 ( o C) 4,9% 0,53 ( o C) Offices 0,06 (kw) 5,2% 0,33 (kw) Amphith. A4* 0,04 (kw) 5,4% 0,12 (kw) Library 0,37 (kw) 5,8% 0,67 (kw) *Null values during nighttime not accountable The results show an error of around 5%. Of the various items found and described in the second chapter, it was found that most did not have a validation of the models used or presented a purely graphical analysis when comparing the simulated results and expected, as was the case of Feng and Hewage article [8], thereby hindering the comparative analysis between the statistical results of this study and the existing work in the simulation area in buildings. However, some of the work carried out, including Mustafaraj et al. [10] and Picco et al. [15], presented statistical results for validation of the models, verifying that the error range on the results varied between 2% and 16%. Given that the results of this work, it appears that these lie within the range of values usually observed in modeling studies. 6. Conclusion The elaboration of this work aimed at building a computer model simulation of energy consumption associated with the building IST - Taguspark in Oeiras. Characterized the model was intended to evaluate the reliability of that in order to be used as a way of predicting future implementation of the measures to improve energy efficiency in the building. In conclusion, in the first instance, that development of energy analysis of buildings models is likely to be implemented. The model developed focused some consumption areas identified by monitoring with sensor system, allowing the performance of tests whose results were compared with measurements made. The characterization of areas must be made with the greatest possible approximation to reality. To this end, data collection must be gathered as much as possible, in order to approximate the results obtained at the end of simulation with values obtained from the measurements. The greater the detail in the characterization, the greater the probability that the relative error between the simulation results and measurements be minimized. In the tests carried out, it is concluded that the choice of energy consumption calculation software was adequate; verifying the relative error associated with the model characterized zones was approximately 5%. 8

9 Not only the simulation program, EnergyPlus, lived up to the recommendations of other users, referred to in the literature review, as well as the chosen graphical user interface for model creation, DesignBuilder, proved a successful bet, even though it presents some limitations, particularly in construction of building facades. In conclusion although modeling tests entail precision errors and overcoming challenges that can derail its implementation. However, the difficulties documented in this and other documents should not deter potential users of its use. Not only it is concluded that the use of these tools has the potential to allow the characterization of areas, expressing the results an approximation to reality at low cost. References [1] P. R. Bureau, 2014 World Population Data Sheet, [2] European Parliament, , Off. J. Eur. Union, no. DIRECTIVE 2010/31/EU OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL of 19 May 2010 on the energy performance of buildings, pp , [3] Agence de l Environnement et de la Maîtrise de l'energie (ADEME), Energy Efficiency Trends in Buildings in the EU. 2012, pp [4] DGEG, Eficiência Energética - Edifícios. [Online]. Available: [Accessed: 14-Jan- 2015]. Office Buildings, Energy Procedia, vol. 58, no. 1876, pp , [8] H. Feng and K. Hewage, Energy saving performance of green vegetation on LEED certified buildings, Energy Build., vol. 75, pp , Jun [9] J. Zhao, B. Lasternas, K. P. Lam, R. Yun, and V. Loftness, Occupant behavior and schedule modeling for building energy simulation through office appliance power consumption data mining, Energy Build., vol. 82, pp , Oct [10] G. Mustafaraj, D. Marini, A. Costa, and M. Keane, Model calibration for building energy efficiency simulation, Appl. Energy, vol. 130, pp , Oct [11] P. Raftery, M. Keane, and A. Costa, Calibrating whole building energy models: Detailed case study using hourly measured data, Energy Build., vol. 43, no. 12, pp , [12] S. Bertagnolio, Evidence-based model calibration for efficient building energy services, [13] D. Coakley, P. Raftery, and M. Keane, A review of methods to match building energy simulation models to measured data, Renew. Sustain. Energy Rev., vol. 37, pp , Sep [14] D. DiLaura, K. Houser, R. Mistrick, and G. Steffy, The Lighting Handbook, 4th ed. Dornbirn, [15] M. Picco, R. Lollini, and M. Marengo, Towards energy performance evaluation in early stage building design: A simplification methodology for commercial building models, Energy Build., vol. 76, pp , Jun [5] F. Zhang and R. de Dear, Thermal environments and thermal comfort impacts of Direct Load Control airconditioning strategies in university lecture theatres, Energy Build., vol. 86, pp , Jan [6] H. Shen and A. Tzempelikos, Sensitivity analysis on daylighting and energy performance of perimeter of fi ces with automated shading, Build. Environ., vol. 59, pp , [7] F. Goia and Y. Cascone, The Impact of an Ideal Dynamic Building Envelope on the Energy Performance of Low Energy 9