Evaluating forecasting methods in the context of local energy communities

Abstract

To push the boundaries of self-sufficiency, local energy communities may rely on load demand forecasts to schedule energy usage ahead of time. In the perspective of selecting a forecasting method, this work explores goodness of a forecast from two sides: its quality, and its value. Traditionally, forecasting methods are ranked based on quality metrics such as the Mean Absolute Percentage Error (MAPE). This work additionally considers the value of a forecast, quantifying practical outcomes for local energy communities such as self-sufficiency, cost of electricity, and fairness. Our contribution is twofold: first in creating a broader framework to evaluate forecasting performance with regards to energy communities, and second, in highlighting the relationship between quality and value metrics for energy communities ranging from 2 to 95 participants. When selecting a forecasting method, our results show that simply relying on quality metrics is misleading. This paper illustrates with a study case, the clear difference in considering value metrics rather than quality metrics, and the potential impact on the design of a coordination platform for local energy communities.