Load Forecasting of Food Retail Buildings with Deep Learning

Abstract

Electrical grids are traditionally operated as multi-entity systems with each entity managing a geographical region. The current movement toward energy democratization and decarbonization is resulting in higher penetration of distributed energy resources (DERs) and intermittent, renewable generation. This process in turn is increasing the number of grid entities (agents). The paradigm shift is also fueled by increased adoption of intelligent sensors collecting data and actuators for advanced processing and computing. Predicting the future load of different consumers has become increasingly important for grids as they must balance intermittent generation to meet instantaneous demand. The main challenges in demand forecasting stem from the heterogeneity of loads and their data. Deep learning provides tools to utilize the collected data for predicting future load profiles and anticipating high-demand scenarios. This article presents a deep learning approach for load forecasting of commercial buildings with multiple refrigeration units. It then presents a case study demonstrating the efficacy of this approach for predicting refrigeration and freezer load in food retail stores.