Xavier Farrell earns NAPS 2024 presentation award for work to predict household savings on electricity bills

Xavier Farrell, PhD student in Electrical and Computer Engineering (ECE), presented a talk on his paper, Evaluating Probabilistic Load Forecasting in Predicting Long-Term Household Savings from Electricity Rate Plan Recommendations, at the 2024 North American Power Symposium (NAPS) in El Paso, TX. The paper, coauthored with ECE Prof. Johanna Mathieu, Farrell’s PhD supervisor, won a third place presentation award.

In the presented work, Farrell and Mathieu employ load forecasting methods to predict the probability of long-term financial savings through a rate plan switch and show the importance of high accuracy at each forecasting stage.  Future iterations of their work could help residents, especially in lower-income households, save money, avoid financial risk and improve their overall financial planning.

In Michigan and many other states with deregulated or partially deregulated energy markets, electricity providers offer multiple rate plans for consumers to choose from. For example, consumers may choose from plans that encourage power use during off-peak hours or guarantee a percentage of electricity comes from renewable sources. Tools and algorithms  called “electricity rate plan recommender systems” have been developed to help consumers select a plan that will help them save money on their electricity bill; however, when these systems do not account for future uncertainties, they may expose consumers that adopt their recommendations to unexpected financial risk. Farrell’s work aims to alleviate this risk.

“This work was born out of a need to provide low-to-moderate income residents with a sense of confidence and transparency that these recommendations will help them save on their electricity bills,” Farrell said.

“Xavier’s research is inspired by our interdisciplinary study leveraging an energy case management approach to reducing energy burden (i.e., percent of income spent on energy),” added Mathieu. “Electricity rate plan recommendations could be one way to reduce energy burden, but the quality of recommendations will be important for energy burden outcomes and people’s trust in the method.”

Using data from 865 households in Detroit over 24 months, the presented research showed that inaccuracies in household consumption forecasts can bias the short-term savings models. These biases can compound over longer time periods, resulting in less accurate long-term savings forecasts for consumers to rely on. To avoid these types of issues, researchers should carefully consider fluctuations in electricity load.

With further improvements and adaptations to their existing methodology, Farrell envisions this work supporting a web application that helps households make more informed decisions to reduce their energy bills. His continued work will expand the current algorithm to accommodate as many electricity rate plans as possible, such that it can be employed to help mitigate the financial strain on low-to-moderate income households across the country.

“Being called to receive a presentation award was shocking, and I felt honored to be selected, given the high caliber of presentations that I saw at the conference,” said Farrell. “It was very encouraging to see that the broader research community found value in my work—and I am inspired to keep forging ahead with this research.”