Lawrence Berkeley National Laboratory

This study describes the results of a low-cost approach used to measure reported load reductions from a residential electric water heater (EWH) load control program operated as part of PJM Interconnection's Demand Response small customer pilot program. Lawrence Berkeley National Laboratory (LBNL) conducted this independent review of the engineering estimates for EWH load control reported by a Curtailment Service Provider (CSP) at PJM's request. LBNL employed low-cost measurement and verification (M&V) approaches that utilized existing interval metering equipment to monitor results for a series of load control tests. The CSP collected hourly load data for two substations and several hundred households over a six-week period in October and November 2003. During this time period, the CSP operated its electric water heater load control program during pre-specified test periods in the morning, afternoon and early evening. LBNL then analyzed substation and premise-level data from these tests in order to verify the diversified demand reductions claimed by the CSP for customers participating in the EWH load control program. We found that the observed load reductions for the premise-level data aggregated over all households in the two participating electric cooperatives were, respectively, 40 percent-60 percent less and 3 percent less-10 percent higher than the estimated diversified demand reduction values assumed by the CSP, depending on whether observed or normalized results are considered. We also analyzed sub-station level data and found that the observed load reductions during the test periods were significantly lower than expected, although confounding influences and operational problems signifiogram during pre-specified test periods in the morning, afternoon and early evening. LBNL then analyzed substation and premise-level data from these tests in order to verify the diversified demand reductions claimed by the CSP for customers participating in the EWH load control program. We found that the observed load reductions for the premise-level data aggregated over all households in the two participating electric cooperatives were, respectively, 40 percent-60 percent less and 3 percent less-10 percent higher than the estimated diversified demand reduction values assumed by the CSP, depending on whether observed or normalized results are considered. We also analyzed sub-station level data and found that the observed load reductions during the test periods were significantly lower than expected, although confounding influences and operational problems significantly limit our ability to differentiate between control-related and non-control related differences in substation-level load shape data. The usefulness and accuracy of the results were hampered by operational problems encountered during the measurement period as well as in sufficient number of load research grade interval meters at one cooperative. Given the larger sample size at one electric cooperative and more statistically-robust results, there is some basis to suggest that the Adjusted Diversified Demand Factor (ADDF) values used by the CSP somewhat over-state the actual load reductions. Given the results and limitations of the M&V approach as implemented, we suggest several options for PJM to consider: (1) require load aggregators participating in ISODR programs to utilize formal PURPA-compliant load research samples in their M&V plans, and (2) continue developing lower cost M&V approaches for mass market load control programs that incorporate suggested improvements described in this study.