Projects Funded for Kevin Novan
The Potential for Forecasting to Improve Energy Efficiency Policies in California
Aaron Smith and Kevin Novan
Specific Objectives of the Project
- Create an econometric forecasting model that uses spatially and temporally disaggregated electricity consumption data along with weather and economic variables to predict a baseline level of regional electricity consumption in California.
- Determine the feasibility of using the counterfactual baseline consumption as a tool for measuring the performance of energy efficiency programs.
Summary of Results
We have obtained detailed electricity consumption data from the Sacramento Municipal Utility District (SMUD) and merged these data with hourly weather data and block-level census data. These data include the following:
- Hourly consumption by almost all households in the SMUD region (approximately 500,000 premises) for all of 2012 and 2013.
- Monthly electricity bills for almost all households in the SMUD region (approximately 500,000 premises) for 2005-2013.
- County assessor information on house cha racteristics.
- Energy efficiency program participation (89,000 records)
After several delays, we obtained these data in March 2014. To preserve individual privacy, we do not know the address of any premises in our sample.
The strongest predictor of household electricity use is temperature, which follows from the fact that air conditioning is a major source of household electricity demand. However, there is a lot of variation in the response of electricity consumption to temperature depending on the day of the week, the size and age of the home, and other characteristics. We have formulated an econometric approach to incorporate this heterogeneity and are in the early stages of testing our model.
So far, in testing our model, we have focused on the e nergy efficiency programs related to air conditioning (AC). AC programs are the largest operated by SMUD in terms of number of participants, and they potentially have the largest effects on electricity use. AC programs constitute just over a quarter of our energy efficiency program observations (about 27,200 records).
In preliminary analysis, we observe large reductions in peak electricity use for homes that replace an old air conditioning unit with a new one under an energy efficiency program. We obtain these results by conditioning on temperature, e.g., we estimate how much electricity is used when the temperature is 100° F before and after the energy efficiency intervention. This finding suggests that a forecasting model that conditions on temperature ha s the potential to provide an effective estimate of counterfactual baseline consumption.
There is significant variation across homes in the response to temperature, the response to the energy efficiency program, and propensity to participate in the program. Our next steps are to model this heterogeneity so as to obtain precise quantitative estimates of the baseline, and to expand our analysis to all of SMUD's energy efficiency programs.