Forecasting Electrical Demand for the Residential Sector at the National Level Using Deep Learning

Abstract

A fundamental element of power-system planning is estimating electricity demand at the national level. However, given the residential sector's trend of rapidly fluctuating energy consumption, it’s challenging to achieve these targets in the residential sector, which is the main source of demand. While deep learning methods have lately demonstrated success in a variety of time series studies, its relevance to forecasting monthly household energy demand has yet to be thoroughly investigated. The forecasting model for this paper used is long short-term memory (LSTM); it has proven itself to be successful in deep learning-based time series forecasting problems. A compilation of data on social and weather variables spanning 42 years in the United States of America was used to validate the proposed model. In addition, the performance of this model was compared to the performance of three benchmark models. According to all of the metrics used, the proposed model performed exceptionally well. This model will make power-system planning effective and improve grid efficiency by properly anticipating the future energy demands.