Residential Load Forecasting by Recurrent Neural Network on LSTM Model

Electric load forecasting for individual customers is a challenging task and highly uncertain due to high volatility. In this paper, a long short-term memory (LSTM) recurrent neural network (RNN) based framework is proposed and tested to predict the expected future load. The root mean square error (RMSE) is determined using the normalized data. The proposed model performed well with RMSE less than 1.