Project 3.5 Aggregated Load Shedding Task 3.5.2 Detailed Measurement and Experimental Plans for LA County Test Buildings Deliverable 3.5.2(a) Report on Measurement and Experimental Plans Submitted to Architectural Energy Corporation Under the California Energy Commission’s Public Interest Energy Research (PIER) Program Energy-Efficient and Affordable Small Commercial and Residential Buildings California Energy Commission Contract 400-99-011 P. R. Armstrong and L.K. Norford Massachusetts Institute of Technology July 12, 2001 Report on Measurement and Experimental Plans for Load Shedding in LA County Buildings THIS REPORT WAS PREPARED AS A RESULT OF WORK SPONSORED BY THE CALIFORNIA ENERGY COMMISSION (COMMISSION). IT DOES NOT NECESSARILY REPRESENT THE VIEWS OF THE COMMISSION, ITS EMPLOYEES, OR THE STATE OF CALIFORNIA. THE COMMISSION, THE STATE OF CALIFORNIA, ITS EMPLOYEES, CONTRACTORS, AND SUBCONTRACTORS MAKE NO WARRANTY, EXPRESS OR IMPLIED, AND ASSUME NO LEGAL LIABILITY FOR THE INFORMATION IN THIS REPORT; NOR DOES ANY PARTY REPRESENT THAT THE USE OF THIS INFORMATION WILL NOT INFRINGE UPON PRIVATELY OWNED RIGHTS. THIS REPORT HAS NOT BEEN APPROVED OR DISAPPROVED BY THE COMMISSION NOR HAS THE COMMISSION PASSED UPON THE ACCURACY OR ADEQUACY OF THE INFORMATION IN THIS REPORT. This report documents building characteristics, NILM installation, and load shedding test plan and subtasks for CEC/AEC Task 3.5.2. The body of the report focuses on the load curtailment strategies, test plan and subtasks. Appendix A and B document the building characteristics obtained to date. Appendix C describes the HVAC motor start-up signatures obtained during field testing that will be used for NILM training. Test Locations and Objectives The objectives of the load control task are threefold: 1) characterize transient building cooling loads analytically and empirically, 2) identify load shedding potential and promising control strategies, 3) demonstrate and verify the resulting load shedding controls. Three LA County buildings will be used in the testing: The Internal Services Department is administered from one main three-story office building called the ISD Building. Floor areas are 45,646 sf office floor area, 8,511 sf mechanical rooms, 14,489 sf circulation areas, 10,388 sf storage, and 80,309 sf gross floor area. The ISD building is cooled by two 4-stage reciprocating chillers with constant-volume distribution. A built up fan system comprising a 60 hp supply fan and a 20 hp return fan serves the entire building. The Edmund Edelman Children’s Court (ECC) is a182,000 (net), 275,530 (gross) square foot, six-story building erected in 1992. The ISD and ECC are about one half mile apart, situated at opposite ends of the site. The ECC is cooled by two 500-ton centrifugal chillers with VAV distribution. A built up fan system comprising two 150 hp supply fans and two 50 hp return fans serves the entire building. The chillers, fans, cooling towers and boilers are all located at roof or penthouse level. The communications building (CM) is a one-story east-west oriented structure situated at a higher elevation than the ISD and ECC and about midway between them. HVAC plant and electrical load descriptions are not yet available. 1 Electrical load data and other details are tabulated in Appendix A for ISD and Appendix B for ECC. Because they are located on the same campus, coordination of load shedding among the three buildings should be possible. Coordination is important for achieving the best outcome with respect to electric supply reliability at a regional level. The County is currently installing a Cutler-Hammer monitoring system that will allow operators to track all significant loads, in all of the large buildings on the campus, by end-use. This information will be used by the operators to better coordinate their response to a utility curtailment and may also be used in our evaluations of the effectiveness of various electric load curtailment strategies. Load Shedding Control Strategies A building’s peak electrical demand may be reduced by 1) reducing non-HVAC loads (lighting and plug loads), and/or 2) reducing HVAC cooling capacity (fans, chillers, pumps and cooling tower). Reduction of base load is the first measure that should be considered because energy savings accrue for all hours of operation, are not subject to occupant, operator, or controller behaviors, and generally result in additional cooling plant and distribution system load reductions. Lighting is a significant base-load component in all three buildings. Office equipment is a significant base load in the ISD and ECC and other types of equipment may be significant in CM. Improvements in chiller plant and air distribution efficiency are also classified as baseload reduction measures. The ISD has definite potential for plant efficiency improvements. Short-term reductions in plug and lighting loads are generally the second most desirable curtailment measure because they produce additional cooling capacity reductions without occupant discomfort. The bonus, of course, is that curtailment of lighting and plug loads effects an immediate reduction in cooling load which the control system will sense and automatically respond to by effecting exactly the right capacity reduction needed to maintain room conditions at setpoint. Dispatchable loads may include such equipment as copiers and printers, two-level lighting, general lighting (if task lighting is available) or task lighting (if sufficient daylighting is available), and may even be extended to workstations of users who are able to indulge in non-computer-related work without undue disruption of their productivity. Most of these additional reductions would be controlled manually. Occupants could be notified by email, phone broadcast, or over a public address system. Some advance training and periodic "fire drills" will be necessary in most situations to implement this sort of curtailment strategy effectively. The ability to measure the response during such occupant-implemented load curtailment is probably crucial to its effectiveness. A third curtailment measure that can be implemented involves further reduction in cooling capacity, either by direct control or by raising room temperature (and humidity, if implemented) set points. The setpoint changes can be abrupt or gradual, depending on the curtailment response desired. The occupants will experience loss of comfort in either 2 case, and the amount of additional capacity reduction, and its duration, will be limited by occupants’ tolerances for elevated temperature and humidity levels. The fourth, and most difficult, curtailment measure requires control functions that anticipate curtailment by anywhere from one to sixteen hours, and increase cooling capacity modestly during this pre-curtailment period so that zone conditions are as cold and dry as can be tolerated immediately before the curtailment period and the thermal masses of building structure and contents are at or even below this minimum tolerable temperature. With such a favorable initial state, the reduction in cooling capacity and its duration can be made significantly larger—i.e., the amount of cooling capacity that can be shed during the curtailment event is significantly increased. There are at least three distinct schemes for implementing the fourth (precooling) measure as a retrofit: 1) Night cooling (outside air with or without chiller operation) 2) Extended period of reduced setpoint (starting the day with a lower setpoint or ramping it down gradually during the pre-curtailment period) 3) Short period of reduced setpoint (e.g. extra cooling at lunch hour when the cold will affect only the (presumably) small fraction of occupants who don’t go out to lunch1) Monitoring Requirements Baseload reductions have already been planned by ISD staff and the first increment is under way. The impact of ISD lighting retrofits will (if the timing makes it feasible) be assessed by the monitoring equipment to be installed as part of the load control study. Note that it is critical to have some way to observe temperature and humidity within the occupied zones during a curtailment event so that load shedding can be accomplished with a clear understanding of and regard for the comfort impacts. It is also critical to have continuous, local weather (temperature, humidity, and solar radiation) observations so that the impact on zone temperatures can be forecast minute-by-minute during the load shedding period for each day when load curtailment is required. However, most important is to have continuous observations of the HVAC and non-HVAC electrical loads. Without at least this level of end-use load tracking it will be impossible for an operator to implement any but the crudest of load curtailment strategies. Subtask 1: NILM Training Classification of loads is accomplished by matching load step changes and start transients. The step change and start transient associated with each load must be provided in a data base. Raw data are obtained by letting the NILM record while turning individual loads on and off in a documented sequence. This process and subsequent 1 For the lunch-hour scheme to be most effective, occupants should be trained to turn off lights and other unnecessary equipment before going out—at least on days when curtailment can be anticipated. If baseload is not so reduced a new building peak load may result. Moreover, if many buildings were to adopt lunch-hour precooling, the utility peak might actually be shifted from mid-afternoon to the lunch hour. 3 analysis of the record and entry of the resulting load characteristics into a database are known as training. The training data that were collected for the ISD building on 30 March are described in Appendix C The analysis of this data must be completed and the results entered into a classification database. The programs used to produce start transient templates (“v-sections”) are currently being revised to establish confidence intervals as well as mean trajectories for start transients observed in training. Better automation of the analysis process will make it practical to for the analyst to process larger training samples that will ultimately reduce classification errors. The current program upgrades are also a step along the path to more fully automated training. Subtask 2: Instrumentation Two NILM installations will be made in each building, one to track HVAC loads (chillers, fans, pumps) and the other to track whole building loads (lighting, plugs, other non-HVAC). Each NILM will be shadowed by a traditional end-use metering logger (C160E) with one-minute integration intervals. The C160E loggers serve two main purposes: 1) Verification of NILM end-use disaggregation. One of the overall project objectives is to demonstrate the ability of the NILM to detect and identify the status of individual loads and by analyzing changes in panel-level or building- level aggregate load. The most convenient and credible way to verify this NILM function is to install a traditional end-use metering logger to monitor as many individual loads as possible and operate it at its maximum time resolution. The C160E can monitor 16 end-use circuits at 1-minute time resolution. 2) Monitor non-power variables needed to properly control load shedding activity and analyze the effectiveness and comfort impacts of such actions2. The non- power variables monitored by the C160E loggers will include coil loads, return air conditions, and weather conditions, as discussed below. Return air temperature and humidity will be measured by a C160E in each building to estimate average zone temperature and humidity. The return temperature is typically a biased estimate of average zone temperature when zone air returns through ceiling plenums because most of the heat from lights is added to the air as or after it enters the ceiling plenum. This bias will be characterized by deploying room temperature microloggers in selected zones for one-week periods. The humidity of the return air (expressed as a mass ratio), on the other hand, usually is a good estimator of zone average humidity. Room temperature in selected zones will be monitored by unobtrusive battery-powered microloggers. The loggers sample temperature at 2 Hz and the sample averages are 2 The variables needed for control, including the end-use loads that will be output by future commercial versions of the NILM, would normally be monitored by the control system. Prototype NILMs may be interfaced a control systems in one or more of the project sites after the effectiveness of load shedding has been demonstrated. However, for the initial tests control and monitoring of load shedding actions and responses will be implemented independently. 4 recorded at selected intervals. For the load control tests, microloggers will be configured to record every minute. Data will be retrieved every week and the loggers will be rotated weekly so that, over the 5-week testing period, the relation between return air temperature and true zone-average temperature may be established with reasonable accuracy. We expect that eight to twelve micro-loggers will be deployed in each building with a total of about thirty loggers operating at all times. Coil delta-T and condensate flow rate may be measured in one or more of the three buildings. Air-side temperature difference across the coil will be used to determine sensible cooling capacity and will be measured by a thermopile with multiple upstream and downstream junctions distributed over the projected coil face area. Condensate flow rate will be used to determine latent cooling capacity and will be measured by tipping bucket rain gages plumbed into the condensate drain pipes. Both signals will be monitored by a C160 logger. A typical rain gage produces about 40 pulses per liter and the number of pulses sensed by a digital input channel will be accumulated and recorded every minute. The thermopile outputs a very low level signal (about 50 uV per Kelvin per pair of junctions) that will be amplified by an auto-zeroing op-amp configured for the appropriate gain (200 to 2000 depending on the number of junction pairs in the thermopile and the maximum temperature difference expected). The op-amp output is connected to a C160 analog input channel. (C160 input range is fixed at 0 to 5V). Weather will be monitored at the CM or ISD building. Outdoor temperature will be measured by an RTD mounted in a radiation shield and solar radiation will be measured by a LiCor pyranometer mounted on top of the radiation shield. Wind will not be measured since building thermal loads are not significantly affected by wind during occupied hours and other times when the fans are operating. Subtask 3: Characterization of Envelope Thermal Response Supply and return fan operation will be programmed so that transient thermal response of the occupied space can be characterized. Fans will be started about an hour before any heating or cooling (economizer or chiller) commences or allowed to run for an extra hour at day’s end with zero H/C capacity. Starting the fans early each day will ensure that the building is in thermal equilibrium when the chiller starts. The thermal response to a step change in cooling capacity will thus be obtained each time the chiller starts. To obtain a strong response, it may be necessary to delay the start until zone temperatures approach the comfort threshold and, at that point, start the chiller in stage two instead of stage one (or start both chillers simultaneously in stage one) to get twice the normal start up capacity. At the end of the day, chillers are normally shut off completely at some predetermined time. This results in a step change (albeit negative instead of positive) in cooling capacity. To obtain a strong signal in this mode it may be necessary to shut the chiller off early, while it is still operating at stage-two or stage-three capacity. It will then be possible to monitor the step response for at least one hour—possibly two—in order to observe the tail of the response while there is still a significant cooling load. Delay or early shut-down of chillers may not be permissible on days with very large cooling loads when the comfort impact is too great. Weekend step response tests may be 5 conducted instead. Nevertheless, the step-response method of characterizing load response is undesirable from the operator perspective because it entails planning, occupant notification, and manual intervention3. It is desirable to have a method of characterizing thermal response that does not rely on step-response tests. We will run the minimum number of tests needed to demonstrate repeatability of the result. The system will then be allowed to operate under its normal daily start and stop schedule and chiller staging algorithm while the data collection system measures thermal response to the time-variations in cooling load and chiller capacity changes that occur as the normal compressor staging sequences are played out. Model identification analysis methods will be applied to this data to obtain an alternate characterization of the building’s thermal response. The responses to typical load and chiller capacity changes can then be simulated using the two models and the adequacy of the model obtained without intrusive step-change tests will be assessed. Subtask 4: Assess Load Shedding Potential Load shedding potential is a function of many variables. However these may be reduced to three primary building characteristics and one control parameter: the aggregate magnitude of operating loads (primarily lighting and plug loads) that can be shut off at the time that load curtailment is called for; the cooling plant efficiency (that is, what is the reduction in HVAC plant power that can be realized per kW reduction in lighting and plug loads); and the potential for thermal storage within the conditioned space (determines what additional reduction in cooling capacity can be achieved). In the foregoing list, it is the third building characteristic that determines the relation between cooling capacity reduction, duration of curtailment, and comfort impact. The maximum sacrifice in comfort that can be tolerated will be decided in advance by managers and facility operators at the site. Thus the one control parameter that must be considered for each curtailment event is the duration of curtailment. The impacts of all gains will have been established in the Envelope Thermal Response task. The analysis will be carried further after it is determined which internal gains can be controlled. The thermal responses to the shedding of controllable loads will be estimated by DOE-2 simulation as well as by the models identified in Subtask 3. 3 The responses of return-air temperature and humidity to a step change in cooling capacity can be used to identify only one kind of thermal response transfer function. With sufficient variation in measured internal gains and weather variables the transfer functions governing the building’s thermal response to these other disturbances can be characterized only with the help of special identification methods because the necessary step change in weather variables are beyond our control. the system must behave as a linear time invariant (LTI) system during the observation period regardless of the nature of the excitations. The system is reasonably close to LTI only when the fans are operating continuously and after the initial response to fan startup has died away. 6 DOE-2 will be used to assess aggregate impacts on total load at the service entrance and the relation between comfort sacrificed and level of load reduction achieved. The sensitivity of load shedding potential to weather and occupancy will be established by modeling building thermal response with various levels of cooling capacity reduction. The relation between capacity reduction and duration will be established by tracking the simulated zone conditions over time until the comfort threshold is violated. The comfort and load reduction impacts will be determined under a range of typical weather conditions and occupancies. Subtask 5: Observe Operator Initiated Load Shedding Strategies The load shedding methods typically used by operators and their impacts on demand peaks need to be determined. Information about methods used in the LA County buildings will be obtained during the remaining weeks of the 2001 cooling season. The characterization will be based on operator interviews as well as continuous thermal and power monitoring. Through operator interviews MIT will learn how curtailment signals are received from the utility, how much time (advance notice) is given, typical curtailment durations. The amount of load reduction targeted for each building and the means used to shed lighting, plug and HVAC loads will be established. We will then ask the operators to keep a log of load shedding activities than can be correlated with NILM measurements to quantitatively assess the impacts in terms of both load reduction and loss of comfort for occupants. The NILMs will monitor electric load changes effected by occupants and plant operators during curtailment events and the C180s and microloggers will monitor weather and zone comfort conditions as well as coil loads. Analysis of the data will determine the following: Reductions, from baseload, achieved by curtailment of lighting and plug loads; Reduction, from model predicted HVAC load, achieved by the combination of Lighting and plug load curtailment, and Further reduction in cooling capacity; Disaggregation of HVAC load reduction by the two above-mentioned actions; Comfort impacts. Subtask 6: Test and Demonstrate Model-based Load Shedding Strategies Reductions in peak HVAC loads, as well as aggregate impacts on total load at the service entrance, will be measured under prevailing weather conditions through the Summer of 2002. We expect to experience at least two similar hot, humid days so that the peak loads with and without load shedding can be compared. In addition, we expect to be able to “pair up” days that have similar conditions. The objective will be to test capacity reduction on one of each of the paired days and not on the other. 7 Because weather cannot be accurately predicted even a day in advance, the sequence of daily tests will necessarily be iterative and a decision to reduce capacity will be made daily, based on what actions have been taken on previous days with similar conditions The operators will be asked to monitor weather conditions and be prepared to reduce cooling capacity every day. A prediction of peak cooling load, expressed as a percent of design load, will be made at about noon each day. Test Plan. Five bins will be established to represent 10% increments of design load starting at 50%. These are the 6th through 10th cooling load deciles4. The impact of capacity reduction will be tested on the first day where predicted load is more than 50% of design load. If the predicted load on the next day falls into the same bin as the previous predicted peak load, the cooling plant capacity will not be reduced. The data for these two same-bin days will thus let us compare peak electrical demand and thermal response of the building with and without reduction in cooling capacity. If the predicted load on the second day does not fall in the same bin as the previous predicted peak load, but is still more than 50% of design load, the impact of capacity reduction will be tested. On subsequent days, when the predicted peak cooling load is more than 50% of design load, the impact of capacity reduction will be tested if the predicted peak load falls in an empty bin and will not be tested if the bin is already populated. Continuing for several days in this manner we hope to obtain data for each 10% bin of predicted peak load under two conditions: with and without cooling capacity reduction. Once the two actions (rather action and inaction) have been tested for a given bin, the operator can alternate between reducing and not reducing capacity by keeping a record for each bin of the numbers of days with and without capacity reduction. Schedule: Subtask 1: NILM training: July 7 – 13 Subtask 2: Instrumentation: July 27 – August 17 Subtask 3: Model Thermal Response: July 30 – August 10 Subtask 4: Assess Potential and Strategies: August 10 – September 28 Subtask 5: Observe Operator Load Shedding Strategies: July 29 – September 7 First-year Report: September 30, 2001 Subtask 6: Test Model-based Load Shedding Strategies: June 1 – August 31, 2002 Second-year Report: September 30, 2002 4 i.e., 50%, 60%, 70%, 80%, and 90% cooling load bins where each 10% bin is denoted by it lower percentile limit based on design load being 100%. 8 Appendix A. ISD Building Description, Los Angeles County. Internal Services Department (ISD) Building BIS#7022; LACO#5863 1100 N Eastern Avenue Los Angeles, CA 90063 Function: Offices Completion date: 1 May 1973 Lighting retrofit: 29 November 1996 Floor area net/gross (sf): 58,826/45,646 Operating hours: 11 (M-Thursday) Heat: two 1.1 Mbtuh (input) gas-fired hot water boilers Cooling: two 2-stage reciprocating chillers Distribution: Dual-duct constant volume system; built-up AHU Electric Utility: SCE, I-6 (interruptible) rate Contact: Ron Mohr Table A-1. ISD Building HVAC Electrical Loads. Circuit & Motor Ratings Name Function Ckt Amps HP rpm P1 Chilled water pump 1 30 7.5 P2 Chilled water pump 2 30 7.5 P3 Condenser water pump 1 30 10 P4 Condenser water pump 2 30 10 P5 Hot water pump 1 15 1.5 P6 Hot water pump 2 15 1.5 CP1 Circulation pump 1 0.5 SF Supply fan 200 60 RF Return fan 60 20 C1.1 Chiller stage 1 compressors C1.2 Chiller stage 2 compressors C1.3 Chiller stage 3 compressors C1.4 Chiller stage 4 compressors 300 TF1A Toilet fan 1A TF1B Toilet fan 1B TF2 Toilet fan 2 EF1 Exhaust fan 1 0.5 EF2 Exhaust fan 2 0.5 EF3 Exhaust fan 3 EF4 Exhaust fan 4 0.33 EF5 Exhaust fan 5 0.25 EF6 Exhaust fan 6 0.75 CAC1 Control air compressor 1 1 CAC2 Control air compressor 2 1 Strip heater, 13kW 9