Enhanced Normalized Metered Energy Consumption Analysis with Rapid Interventions

This study evaluated two methods for estimating energy savings from heat pump water heaters, using data from 25 homes in two multifamily buildings located in San Jose, California, and Woodland, California. The first method, traditional normalized metered energy consumption analysis, compares long pre- and post-intervention periods using utility billing data. The second method, enhanced normalized metered energy consumption with rapid interventions, applies randomized on/off control intervals to estimate savings more quickly. The intervention system used forecasts of energy cost, comfort, and greenhouse gas emissions to decide when to slightly preheat the tank. For example, if it anticipated high electricity prices soon and low emissions around midday, it raised the temperature to avoid running the heat pump during expensive or high-emission times while keeping comfort as the top priority. Our objective was to determine whether enhanced normalized metered energy consumption with rapid interventions can produce savings estimates comparable to those from traditional normalized metered energy consumption. Many programs cannot support 12 months of pre- and post-intervention data collection due to schedule or budget constraints. Enhanced normalized metered energy consumption with rapid interventions addresses this by mixing baseline and intervention data throughout the study period, allowing models to adapt in near-real time to changes in behavior or external conditions. Because both traditional normalized metered energy consumption and enhanced normalized metered energy consumption with rapid interventions are estimates of an unknown quantity, device-level data analysis is compared to the actual heat pump hot water heater usage.The project team recommends using enhanced normalized metered energy consumption with rapid interventions for evaluations when rapid control switching is technically possible and when faster results are desired. The project team will continue to validate the method across larger samples and varying control technologies to confirm its accuracy and applicability in broader contexts.