Presentation on theme: "Consortium for Electric Reliability Technology Solutions CERTS-CAISO-CEC Demand Response Demonstration 2 nd Technical Advisory Committee."— Presentation transcript:
Consortium for Electric Reliability Technology Solutions http://certs.lbl.gov CERTS-CAISO-CEC Demand Response Demonstration 2 nd Technical Advisory Committee Meeting April 4, 2005
CERTS-CAISO-CEC Demand Response Demonstration Project – Meeting Objectives Update TAC on status of CERTS Demand Response Demonstration Project Present and discuss draft Test Plan, including preparatory statistical analysis Present and discuss proposed monitoring screens for real- time observation of the tests Receive guidance from CAISO and TAC on technical issues for completing the test plan, including monitoring/verification procedures
CERTS-CAISO-CEC Demand Response Demonstration Project - Agenda 9:00Welcome/Introductions – Dave Hawkins 9:05Purpose/Organization of this Meeting – Joe Eto 9:10Presentation of Test Plan – John Kueck 9:30Statistical Analysis – Roger Wright 10:00SCE Load Management Technologies – Mark Martinez 10:20Data Collection, Communication, and Presentation – Arup Barat 10:40Review of Technical Feedback and Direction from CAISO and TAC – Joe Eto 11:00Adjourn
CERTS-CAISO-CEC Demand Response Demonstration Project - Organization Project Sponsor:CEC Public Interest Energy Research Energy Systems Integration Ron Hofmann(510) 547-0375 Project Coordinators:Joe Eto, LBNL/CERTS(510) 486-7284 Dave Hawkins, CAISO(916) 351-4465 Technical Leads:John Kueck, ORNL( 865) 574-5178 Brendan Kirby, ORNL( 865) 576-1768 Bob Yinger, SCE(626) 302-8952 Mark Martinez, SCE(626) 302-8643 Carlos Torres, SCE(626) 302-8364 Roger Wright, RLW Analytics ( 707) 939-8823 Thomas Yeh/Arup Barat, Connected Energy (585) 697-3800 Dave Watson, LBNL(510) 486-5562
CERTS Demand Response Test Plan John Kueck and Brendan Kirby Oak Ridge National Laboratory email@example.com@ornl.gov 865-576-1768 firstname.lastname@example.org@ornl.gov
Objectives oDemonstrate that demand response can provide the ancillary service of spinning reserve for system contingencies in a manner that will be adopted by system operators: o Build operator confidence regarding the value of demand response as an alternative to traditional approaches for providing spinning reserve. o Set the technical basis for modifying reliability rules to allow utilization of demand response for spinning reserve. oDemonstrate and benchmark statistically the reliability of large numbers of small responsive loads; compare this to the current responsiveness of generation.
Objectives, Contd. oDemonstrate ability to target demand response to geographic sub-regions. oDemonstrate that demand response can provide spinning reserve economically and reliably through modest enhancements to existing centralized systems. oDemonstrate to IOU distribution system planners the ability of demand response to influence the timing of distribution system upgrades.
A Test of Demand Response oCurtailment of air conditioning loads on a SCE Distribution circuit. oThis is a “typical” circuit with residential and commercial loads. oThe test is to be a benchmark in establishing the reliability of demand response. oWe need your input in the test planning.
This test has two objectives: oDemonstrate that when load is curtailed by a dispatch signal, the available MW demand response of a specific circuit can be precisely predicted with a 90% statistical confidence level using three variables: time of day, day of week, and temperature. oDemonstrate that the load can be curtailed reliably and quickly on the issuance of a dispatch signal. The load shed is expected to start within 10 seconds of the signal and be fully implemented within two minutes.
Test Overview oThe distribution circuit power level will be sampled via SCADA every 16 seconds. oA group of local data loggers will also be installed during the test to increase the statistical precision of the data analysis. oThe rigorous, statistical approach taken for the design of this test will be discussed by Dr. Roger Wright. oThe Zin circuit has roughly 2350 residential accounts and 169 commercial accounts. oBased on seasonal load profiles, there is approximately 850 kW of air conditioning load on this circuit.
Residential and Commercial Customers oSCE plans to implement a special contract for the test with 400 to 500 residential customers and 50 to 100 commercial customers. oResidential customers will be curtailed with a local switch. oCommercial customers will be curtailed with a thermostat that will be programmed to actually curtail, not to just set back the temperature.
Curtailment Durations oThe curtailment durations will be set to twenty minutes. oTwenty minutes is ample time to verify the exact amount of load that has been shed. oInstalled metering will also show how quickly it was shed. oThere will be 60 tests over July, August and September at random times between the hours of 2 pm and 6 pm. 60 tests are needed for statistical rigor.
Curtailment Duration oIn the actual provision of reliability services, the curtailment durations would be longer than 20 minutes. oA longer curtailment test duration would make it difficult to contract for this large number of tests because customers would be concerned about multiple air conditioning outages. oLonger curtailment intervals could be provided by simply dispatching a longer curtailment since the load management methods being tested provide direct control of the load. oIn this test, we are not addressing the issue of appropriate payment to enlist participation in longer duration outages, but only the issues of predicting response accurately, and delivering it reliably once the dispatch controls have been installed.
Test Plan Steps oContract with residential and commercial customers and modify SCE Load Management and Demand Response systems to control air conditioning loads within the circuit. oInstall sufficient numbers of two types of load control devices on this feeder to ensure interruptions can be observed statistically at the feeder. At time of device installation, record the size of the air conditioning compressor. oInstall spot meters on a designated sample of the controlled units. oAssemble the data acquisition and communications system. oConduct a series of 20 minute duration demand response tests during July, August and September, 60 tests total.
Test Plan Steps oRetrieve the spot meters and their data. oCorroborate load changes observed at the feeder with spot metering placed on a sample of controlled loads within the feeder; express findings statistically. oMeasure load drop and response times, including latency in observability; express findings statistically. oCharacterize load drops as function of relevant influences (temperature, time of day, day of week) with an eye toward extrapolation of findings to other feeders that having differing saturations of controllable end-use loads; express findings statistically to determine response confidence level. oPresent and discuss results with stakeholders (IOU distribution system planners, IOU dispatchers, CAISO operators, relevant WECC/NERC committees) to confirm adequacy of present analysis and priorities for future research in this area.
Statistical Plan Roger Wright RLW Analytics email@example.com firstname.lastname@example.org 707/939-8823
Questions oHow many accounts with substantial AC load are served by the circuit? oWhat does the circuit’s load look like? oDoes the circuit have substantial AC? oHow accurately can we predict the circuit load in the absence of curtailment? oHow many units do we need to curtail? oHow many units do we need to monitor?
Accounts with AC Load For example, the circuit serves 1,958 residential accounts. 857 residential accounts have at least 250 kWh of estimated AC use per month.
Load on the Circuit The load varies from a low of 2 MW to a high of 9 MW. There are a few gaps and bad measurements but the data are remarkably clean.
Load during the Peak Day These are 2-minute spot measurements. The random variation is about 1% of load.
Smoothing the Load The smoothed value is the numerical average of the five original values, i.e., the centered, 10-minute moving average.
Original vs. Smoothed Load The graph shows one hour, centered at the 3:58 PM peak. There is still substantial variation in the smoothed load.
Energy Print of the Load This graph shows the load in color. The x-axis is the day and the y-axis is the hour of the day.
Load versus Temperature The upper graph shows the outside temperature. The bottom graph shows the load. Clearly the load is driven largely by temperature.
Peak Temperatures vs. Loads There is a high correlation between the peak temperature and peak load
Statistical Plan This is an example of a Smart Thermostat curtailment. The blue curve is the baseline load in the absence of curtailment. The red curve is the actual load of the units.
Assumptions o The average nonresidential account has 2 ACs o Curtailments are for 20 minutes between 2 pm and 6 pm on weekdays o Each AC unit is completely curtailed o A typical residential unit is 3 tons with 3 kW operating load and a duty cycle of 50%. o A typical commercial unit is 4 tons with 4 kW operating load and a duty cycle of 70%.
Predicting the Baseline Load oPredictors included: o Average circuit load on prior weekday o Actual load just prior to the curtailment o Current and prior temperature and enthalpy o Load on two adjoining feeders oAll feeder loads were smoothed using a 10-minute moving average oWe were generally able to predict the load with a standard error of 1% ten minutes into the curtailment.
Expected Statistical Precision This plan assumes 80% commercial participation and 50% residential. N is the potential number of units. n is the number of controlled units. The expected impact is 906 kW with an error bound of 159 kW, for 18% relative precision at 90% confidence.
Expected Statistical Precision This plan assumes 50% commercial participation and 25% residential. The expected impact is 487 kW with an error bound of 154 kW, for 32% relative precision at 90% confidence – not good enough. Clearly we need to very aggressive marketing.
End Use Monitoring o If each unit is 100% curtailed, the potential controllable load is equal to the load of the controlled units. o The controllable load will vary by temperature, time of day and possibly day of week. o We need to supplement the information from the curtailments (e.g., 60 calls) with added data. s Extended End Use Monitoring –5-minute battery-powered loggers installed on a sample of the controlled units can measure the actual load throughout the summer. –How many do we need? s Real Time End Use Monitoring –Time synchronized ‘Smart spot meters’ gathering real time data from a sample of controlled units can characterize the time based behavior and responsiveness of the controlled units –10 meters on residential units and 10 on commercial units
Expected Statistical Precision from End Use Monitoring A sample of 50 units gives 18% relative precision at 90% confidence. 75 gives 15%. 100 gives 13%. 150 gives 10%. Added precision will come from statistically modeling the repeated measurements throughout the summer.
Conclusions oWe need to control about 525 units to see the impact reliably in the SCADA data. oWe need virtually 100% signal reception and no take back from other units serving the site. oWe should make a 20-minute call each weekday. oThe calls should be pre-scheduled but staggered to avoid contaminating the baseline. oWe need to monitor 50 to 100 units on a 5-minute basis to help model the size of the curtailable load. oWe need to monitor 10 residential units and 10 commercial units in real time to understand the time responsiveness of the curtailment devices. oSome variation in the magnitude of the impact is unavoidable – but we should be able to demonstrate that the aggregate resource is responsive and reliable. oIn larger-scale, the impact should be increasingly predictable.
SCE Load Management Mark Martinez Southern California Edison (626) 302-8643
Summer Discount Plan oAlso known as AC cycling program o154MHz private network that continuously communicates one way to SCE devices oDevice is installed as part of D-APS tariff that allows SCE to curtail air conditioner via thermostat wiring for billing credit oVarious program participation levels from 50% to 100%, unlimited events oTriggered on Stage 2 or local emergency
Current SDP Resource oResidential o Base = 163.8 MW86,042 SA o Enhanced = 74.3MW40,847 SA oCommercial o Base = 35.8 MW2,000 SA o Enhanced = 6.7MW452 SA oTotal load = 280.6MW
Dispatch Procedures (ACCP) oInterrupt notice is sent from the Dispatcher to the Broadcast Master Controller (BMC) oDevices to be interrupted are sent via Daily and Weekly files, which are fed through the Gateway and saved in the ELMA database oBMC builds the appropriate device download commands and queues this to be sent to the Port Expander oPort Expander sends the commands to the modem on the Remote Site Controller, which then encodes, transmits, and verfies the appropriate message has been broadcasted (A/C or API switch)
Smart Thermostat Program (E$T) oSmall Commercial pilot program to test demand responsiveness with t-stat oTwo-way communicating thermostat on public paging network oThermostat is given to customer and can be controlled by SCE oCustomer receives cash bonus for participation, deductions for override oCurrently 8,250 thermostats installed
Dispatch Procedures (E$T) oDevice IDs are stored in Itron application REM and tied to group and account info oCurtailment command is initiated by logging into web-based application and scheduling event for initiation oPage is sent to devices (900MHz public network) and event is set up in receiver oAt scheduled time of event, device activates curtailment strategy
Data Monitoring Strategies 1.Real Time Feeder Monitoring Used for load profile projection and monitoring the result of a dispatch command at the aggregated circuit load level. Data collected by SCADA system at 4-8 sec intervals and aggregated at Connected Energy’s systems. Will be displayed in real time. 2.End Use Monitoring (on a representative sample) 1. Extended monitoring: Measure the actual end use load throughout the summer in order to analyze the load drop during curtailment periods and model the size of the load. Needs 5-minute single channel load data using data loggers (50-100). 2. Real Time monitoring. To characterize the time based behavior and responsiveness of the load shedding devices during a curtailment period. Smart spot meters and real time data (10 residential, 10 commercial)
Common data collection and correction strategies oTime Synchronization o All times clocks on disparate systems will be synchronized to the US Naval Atomic clock server o All real time data will be time stamped accordingly oData correction for systematic errors o Systematic errors in the extended end use monitoring data will be corrected at the end of the test period at the aggregated system. oBaseline Prediction model o Calculated baseline prediction model data will be continuously updated as new real time data is gathered. This will be available for real time display
Data Presentation On a Connected Energy hosted, dedicated website throughout the duration of the test oReal Time data o Aggregated Views – circuit and end use sample and calculated data o Live trends – real time updates. oSummarized data o Predefined Reports o Exported Data – available in multiple formats for further analysis
Reports Summarized reports o Dispatch event detailed report s After the completion of each dispatch event to summarize the load behavior. o Aggregated dispatch events report s At the end of test period to summarize behavior and statistical findings o Aggregated end use load responsiveness report s Aggregated time response behavior of AC units