Forecasting the Electricity Hourly Consumption of Residential Consumers with Smart Meters using Machine Learning Algorithms

The progressive deployment of smart meters in Spain since 2015 has changed the retail electricity sector imposing a strong operational impact for agents participating in the wholesale market. Each residential customer is characterized based on its real hourly consumption instead of the monthly aggregated consumption. New models are needed to forecast the hourly consumption of residential customers to buy the energy in the markets. This paper presents a robust and scalable methodology to predict the household's hourly energy consumption based on smart meters' data using machine learning algorithms such as Neural Gas, Classification Trees, Multilayer Perceptron Networks and XGBoost. First, a novel clustering methodology to aggregate consumers is presented. Secondly, a model to forecast the hourly consumption in the day-ahead is proposed for every cluster. Finally, a case example is used to illustrate the results, accuracy and robustness of the methodology.