Machine Learning based Load Forecasting Tool for Smart Grid Enabling Technology

Abstract

With increase in population and an overall expansion of the energy infrastructure, electricity demand is increasing rapidly. Smart grids are used to manage this increasing demand effectively. Load forecasting is the basis of demand side management which is the cardinal feature of smart grids and allows concerned personnel operating the smart grid to make efficient and effective decisions. Extensive previous research using artificial neural networks (ANN), support vector machines (SVM), regression tree models etc exists. This study focuses on using long short term memory (LSTM) algorithms to improve the accuracy of short term load forecasting (STLF). In this report, several LSTM models will be introduced for the purpose of generating hourly forecasts of electricity demand based on data made publically available by Réseau de transport d'électricité (RTE) France from the year 2013 to 2017. These models were also tested on an open source data set provided by the Electrical Reliability Council of Texas (ERCOT) to compare the results of LSTMs with previous studies that performed electrical load forecasting using autoregressive Moving Average (ARMA) and Seasonal Autoregressive Integrated Moving Average (SARIMA). Previous studies conducted to analyze ARIMA and SARIMA provided a mean absolute percentage error (MAPE) of 9.13% and 4.36% respectively. The LSTM models introduced in this study bring this MAPE down to 1.975%.