COVID-19 has upended assumptions about how buildings use energy.
Shutdowns, work-from-home mandates, and uneven economic recovery have made forecasting gas and electric consumption and traditional measurement and verification a relative crapshoot. Simultaneously, moving toward a clean energy future will increasingly require market participation from behind-the-meter resources.
For these markets to work, utilities need settlement grade measurement to separate the impacts of procured interventions from the broader outside influences. Exogenous effects such as the COVID stay at home orders, the ongoing changes in energy consumption due to electrification of buildings and transportation, and growing deployment of behind-the-meter storage and renewable generation, require new data-driven tools to enable reliable telemetry for planning, procurement, and settlement.
Traditional deemed rebate programs, based on studies from years before COVID, are simply no longer valid.
Energy models, which were always costly and hard to verify, now must rely on many layers of assumptions and non-routine adjustments with no source of truth.
Measuring only at the meter using past consumption as a predictor of future usage (IPMVP Option C) has broken down during this period, where the past is no longer a straightforward predictor of the future.
As demonstrated by the table to the right, different sectors have changed in starkly different ways, with significant variance and different recovery rates from one building to the next.
Many are asking if it is even possible to measure efficiency and demand flexibility in this new era.
We have good news. While historical site based engineering approaches are not up to the task, it is clear that with new methods and the power of big data, M&V is possible during COVID and into the future.
Recurve is pleased to announce the launch of the GRIDmeter.
Developed in a public process, the GRIDmeter has proven to accurately normalize for the impacts of COVID and other exogenous events to enable revenue-grade settlement quality measurement and verification to support the ever-increasing need for behind the meter demand flexibility.
Recurve worked with MCE, the Department of Energy, and dozens of colleagues from across the industry in a public process to research and identify the factors most critical for identifying and removing the impacts of COVID from the measurement of demand-side programs. The team developed and tested a set of novel advanced stratified comparison group methods with open-source code designed to get programs and markets back up and running with confidence with this information. These methods allow utilities to adjust for the dramatic changes in consumption wrought by COVID and provide the tools needed to account for longer-term exogenous effects on consumption.
The GRIDmeter methods leverage the CalTRACK 2.0 Methods and the OpenEEmeter to generate an hourly baseline for all treated and untreated customers in the population. Next, a novel application of stratified sampling selects population comparison groups for specific treated customer segments that provide a sound representation of those customers. A difference of difference calculation then normalizes for exogenous effects.
Recurve has released a full report, appendices of research results, and an open-source library to enable others to apply these learnings and contribute to the project’s ongoing improvement.
Recurve has also applied innovative differential privacy methods, developed with DOE-funding, to privacy protect population data. These methods are designed to ensure that comparison groups can be formulated while providing full protection to each customer’s privacy by applying differential privacy methods to all outputs.
Recurve is already working with utilities across the country to deploy the GRIDmeter to support pay-for-performance and savings claims with utilities across the country, even in these uncertain times.
Results from MCE Data: Understanding COVID Impacts
Here’s an example of how Recurve used GRIDMeter methods to give MCE insights into the energy use impacts of COVID on its customers.
The figures below show the results from a test run of the advanced stratified sampling methods. These figures show seasonal weekly load shapes for a comparison pool (red) and a “treatment” group (blue), which consists of a sample of customers with expensive load profiles. The orange curve shows the load profiles for the comparison group selected by the GRIDmeter. In the bottom panel, it is clear that the match between “treatment” and comparison groups continues into the COVID period, an essential attribute for stable measurements.
Recurve has shown that sampling based on residential customer usage patterns and geographic locations can substantially improve annual savings measurements’ accuracy. These strategies are critical for hourly measurements. For commercial customers, comparison group selection on business type should be the priority.
Figure 1 shows the observed and counterfactual daily load shape for an average meter in the COVID period. As expected, after stay-at-home orders went into effect in MCE territory, residential use increased. We measured a total increase in consumption of 7.9 percent due to COVID. The majority of this increase occurs in the mid-day hours.
But while the average customer experienced an increase in usage, we can also see wide distribution in COVID impacts at an individual customer level. Figure 2 shows the distribution of COVID impacts across the residential sector.
Figure 3 shows the average observed and counterfactual daily load shapes for non-solar Commercial customers in MCE territory. As would be expected, commercial use declined (overall by about 15 percent) as businesses were shut down due to COVID.
As in the Residential sector, the most considerable difference is seen in the middle of the day where most businesses have typical operating hours. As in the Residential sector, we saw a wide distribution of COVID impacts at an individual customer level among different businesses. However, unlike Residential, we observed that other Commercial sectors’ different segments exhibit drastically different responses to COVID.
As an example, Figure 4 shows distributions of COVID impacts for Grocery and Convenience stores (left) and Hotels and Lodging facilities (right).
While the Grocery/Convenience stores have seen an 8 percent decrease in consumption, Hotels’ business operations appear to have experienced far more significant impacts, with a 24 percent drop in electricity usage. While these are just two examples, across the commercial sector’s distinct economic segments, we observe a wide range of COVID impacts.
Understanding the differences in how residential and commercial use changed and how different customer segments were impacted allows a utility to target interventions to the customers who would both need and benefit the most--and those would help meet grid needs.
Recurve would like to thank our fabulous partners of DOE and MCE, who facilitated this timely research, along with everyone who joined and who provided valuable insights in our working group. We look forward to continuing to collaborate with you on this critical project.
Are you interested in learning more about how GRIDmeter comparison groups can make M&V viable again? Join our upcoming webinar to find out more.