Efficient Control of Potable Water Distribution

Transcription

1 Efficient Control of Potable Water Distribution Raed Al Nuaimi 1 *, Rajendra Kumar Ramakrishnan 1, and Mohammed Barghothi 1 1 AECOM Middle East Ltd., P.O. Box 1419, Al Ain, United Arab Emirates (*correspondence: raed.nuaimi@aecom.com) FORMAT: 6-12 page paper and 35 minute PowerPoint presentation KEYWORDS DCS, VFD, SCADA, District Metering Area, Flow Control ABSTRACT Service providers struggle to control potable water pump stations that are distributing to public consumer networks with varying demands while maintaining pumping efficiency, especially while the consumer points are physically located at different elevations. Many service providers have tried to introduce Variable speed drives to control the pump discharge and thereby trying to achieve energy efficient operations. A DCS pumping station control system was designed based on flow; it utilizes local control system at selected points of the network, i.e. DMA (District Metering Areas) where a flow control valve is installed to control flow, in addition to logging of other data like water quality parameters. The summation of the individual actual flows at each DMA is used to manipulate the speed and/or the number of running pumps, so that the pump station outlet pressure is always kept at the exact value required to meet the demand at that instant of time, regardless where the pumped water is consumed. The system eliminates the need for different pressure head discharge headers in areas where consumer points are at different elevations and provides a built-in leak detection capability. Model Driven control can be achieved utilizing logged data collected from the network. This efficient control system optimizes energy usage, limits water consumption and helps for faster detection and rectification of water leakage. Introduction Many conventional pumping stations use the pressure at the header or at pre-identified critical/control points along the water network as the main target control data input point, while others depend on the tanks levels. These systems will operate without real appreciation to the exact consumer needs during the day, mainly in terms of flow and pressure; compromising the system efficiency. In many of the cases it will be required to install a bypass from the pump back to the reservoir, where the pump target pressure at the set point results in more than demanded flow. Another physical complication in the operation of pumping stations is the control over the number of pumps running at a time, notwithstanding the decrease of pump efficiency, with the increase of number of pumps

2 Al Nuaimi et al 2 running. In a defined network two pumps might not give double the flow to the system even though they consume almost double the power. Similarly, a third pump will not triple the flow to the system while consuming not less than three times the power of a single pump (Ertin, N. Dean, L. Moore, & L. Priddy). This inefficiency can be, to a certain degree blamed for the soaring operation costs of water systems reaching up to 65% of the water services yearly operational cost (F. Boulos, Wu, & Hou Orr, 2001) There have been similar attempts in recent years to develop optimal control algorithms to assist in operation of complex water distribution systems. The various algorithms were oriented towards determining least-cost pump scheduling policies (proper on-off pump operation) using different programs (F. Boulos, Wu, & Hou Orr, 2001). Many of these studies remained as theories without actual application in the operation process due to complexity, which might require experts dedicated for operation, or due to dependency on the system of which the mathematical solutions were derived from. Another reason could be the inadaptability of the proposed schedulers to the dynamic nature of the water systems operation, mainly the water demand which shall directly affect the pressure. In many cases, the scheduler will be used by designers based on planning data/input, and may not be calibrated with actual demand and pressure data during commissioning or after some time in operation.. Even in cases where the scheduler is applied as a model and satisfactory results are obtained, an advanced control system/automation is still required to depict the model (theory) in actual operation. In this paper, a control system for a water distribution system utilizing variable speed pumps is considered. The system supplies multiple consumption points with different hydraulic characteristics (elevation, distance, flow, etc.). The control system is based on flow scheduler, utilizing the flow control, on-line telemetry, Instrumentation and Control apparatus aiming at reducing the operating costs through maintaining the pumping station operating in the best possible efficiency (minimum number of running pumps, efficiency of each pump, delivering the exact pressure required to deliver the actual demand). Study Drivers The figure below shows a pumping station serving several consumption points. A consumption point, is defined as any demand point equipped with local control facility. It can be a bulk consumer tank; a sub distribution zone where water is stored and pumped later or a District Metering Area. Consumption points are located at different distances and elevations. In case of a conventional pressure control system, the highest consumption point or critical point is considered as the main driver. The pumps should target a certain pressure required to deliver demand at hydraulically remote points. This pressure value needs to be estimated from hydraulic modeling and further verified on site. This scenario will work efficiently with the estimated number of pumps in case the system is operating exactly as depicted in the model, or during testing and commissioning. In real world, some tanks will be filled before others (for tank based consumer points), demand varies with time (though average is fixed), in this case, the target pressure will remain the same yet more than required, and therefore excess flow will be discharged from the pump and needs to be bypassed back to the tank. Another example for inefficiency is

3 Al Nuaimi et al 3 the operation of pumps to fill the different tanks without actual realization of whether this water will be consumed or stored. The only limitation for the flow supplied is the hydraulic capacity of the system. This situation might lead to the running of several pumps simultaneously, while only one pump may suffice for the actual demand. Figure-A System Design In order to have better water management capability with multiple consumer points and ensure that the pumps are at optimum power consumption, the flow data at the consumer can be utilized in the control system. The system is based on the flow scheduler model where the pumping capacity is shared between consumption points. Each consumption point is equipped with a local flow control system receiving a remote set point from pumping station scheduler. According to conservation of mass, all points shall receive their share of the total pumped flow if it is maintained at the summation of the instantaneous values of flow required at consumption points (assuming no leakage). The local control system clips the flow, at low hydraulic resistance points, restricting them from receiving more than their share allocated by the scheduler. This will cause the pressure to rise, thus allowing water to reach the high hydraulic resistance points.

4 Al Nuaimi et al 4 When the actual flow at a certain consumption point is less than the allocated share, the system evaluates the cause of the difference and accordingly updates the scheduler. The scheduler reacts to all signals from consumer points and adjusts the pumps speed accordingly. System Description Each consumption point is equipped with its own local control system as part of the distribution network DCS. It comprises of the following instrumentation as minimum: Flow Control Valve, Flow Meter, and Downstream Pressure Transmitter. The instrumentation is hooked up to a Remote Terminal Unit (RTU), the RTU performs all local control activities in addition to handling the communication with the pumping station central control system. According to the pumping scheme for a certain instance of time, the local control system receives the allocated share of pumping capacity available to the consumption point; this value is fed as a set point to the flow PID controller at the RTU. The PID control utilizes the flow meter signal as a feedback signal to the PID controller which in turn manipulates the flow at the set point. The nature of this flow control system is demand driven thus the PID controller works in clipping mode i.e. it manipulates the flow at/or below the given set point. If the flow is less than the set point (due to lower demand), the controller will drive the flow control device to fully opened position, but obviously can t force flow into the consumption area. The system communicates the following signals back to the central pumping station control system. The Actual Flow fed to the consumer (flow meter reading). Pressure downstream of the control valve. Required Flow Signal. The first two signals are self explanatory while the third will have different values. The summation of all Required Flow signal from consumer points is used as a set point to the pumping station flow control system. The Required Flow signal is interpreted by the control system as the value of flow the particular consumption point flow control valve is supposed to pass, while the Actual Flow is the actual value passing. If the summation of all required flow signals is equal to the pumping station discharge (i.e. the flow control system at the pumping station reached steady state), and one or more of the consumption points are reporting less Actual Flow than Required Flow supposed to pass, the control system reports the difference as leak, assuming all instruments are calibrated and speed-up the pumps to satisfy the starving consumption points, provided the total leak is less than a predefined limit. This is an add-on feature to the design. The Required signal is processed at the local control system prior to sending to the scheduler. The below table summarize how the signal is locally processed at the Local RTU. Table 1 Actual Flow with Respect to Allocated Flow Control Valve % Opened Pressure Low Returned signal Remarks/Action

5 Al Nuaimi et al 5 Less than Equal Greater than 100 Any 0 Value - x - - x - - Actual - - x - x - - Allocated Transient State otherwise malfunction x - Actual Malfunction x x - - Allocated Transient State otherwise malfunction x - - x - - x Allocated More pumping needed x - - x Actual Local consumer demand is covered- no more than the current actual flow can be consumed. System Optimization When the system is first commissioned, it follows the pumping scheme fed to the scheduler, based on a hydraulic model or old consumption data. In both cases it is far from being optimum, as hydraulic models need to make many assumptions and the actual friction losses may differ due to ageing or in case of any change in pipework from the designed route. The system shall be programmed to utilize the logged data to develop a model and predict the consumption scheme. The first step towards optimization is to use the average consumption as the demand for the scheduler. As data is accumulated and the consumption scheme approaches the real representation of the demand, the system manipulates pumping schedule towards optimum utilization of resources. Once the scheduler is optimized and the system is stable, the pumped quantity to a certain consumption point will be governed by the actual consumption, thus reducing the retention time in consumer tanks (if any), this will lead to better water quality. System Scalability The above description shows the minimum configuration required for the system to operate at reasonable efficiency. Depending on the implementation, the system can be programmed to utilize extra resource available to improve system performance. As an example, a system implemented in a transmission network, where all the water is supplied to storage tanks from which water will be distributed further. The signal from local level transmitter shall be utilized to determine the actual consumption scheme. Once this is determined the scheduler utilizes it to cover the actual demand and keeps the water level in the tank at a minimum level according to the demand. Keeping the minimum required level helps in two aspects. The first is, a better water quality resulting from minimum retention time within the tank. The second is, spare storage for use by the scheduler as a flow relief, when the system requires to run at certain speed to meet certain pressure, while the actual demand is less than the optimum flow for this pressure.

6 Al Nuaimi et al 6 Conclusions This new system design has high potential to be very effective and Green, especially when coupled with the new communication technologies available. It represents a new field for research and development especially on the model driven control and optimization algorithm. The system was adopted by a leading utility company in the gulf region. Design of an 8 MIGD potable water pumping station, utilizing this new approach to supply bulk quantity of potable water to 16 consumers with about 15 miles of pipe-work, is completed and is under tendering at the moment. Further research will be carried out when the system is constructed and put in operation. Bibliography C, B., M, M., M-V, L. L., CA, B., & CJ, B. (2003). Optimal Operation of Water Distribution Networks by Predictive Control Using MINLP. ISSN = Water SA Vol. 29, Ertin, E., N. Dean, A., L. Moore, M., & L. Priddy, K. (n.d.). Dynamic Optimization for Optimal Control of Water Distribution Systems. Batelle Memorial Institute. F. Boulos, P., Wu, Z., & Hou Orr, C. (2001). Optimal Pump Operation of Water Distribution Systems Using Genetic Algorithms. Pasanda, USA Camarillo, USA Kansas City, USA Markham, Canada. M. Waski, T., V. Chase, D., A. Savic, D., Grayman, W., Stephen, B., & Edmundo, K. (2009). Advanced Water Distribution Modeling and Management. Bentley Institute Press. Marino, M., & Afshar, M. (2005, October). A Convergent Genetic Algorithm for Pipe Network Optimization. Scientia Iranica, pp About the Authors: Raed Al Nuaimi has been in design, construction and commissioning of control systems serving refineries, power plants, water and waste water facilities for over 25 years. He is currently heading the Instrumentation and Control Department at AECOM Middle East Ltd. in United Arab Emirates. He is also the Honors & Awards chair of ISA UAE section. Rajendra Kumar Ramakrishnan has over 22 years of experience in project design, site execution including commissioning of water and waste water systems including treatment, transmission and distribution projects. Rajendra is currently heading the Mechanical Department of AECOM Middle East Ltd, Al Ain in the UAE. Mohhamed Barghothi has been in design of water and drainage systems for almost 8 years. He is currently working as a Senior Design Engineer AECOM Middle East in the United Arab Emirates for water supply systems at AECOM Middle East in the United Arab Emirates. Mainly responsible for hydraulic and surge studies for various potable water and reclaimed water pressure and gravity systems.