Short-Term Load Forecasting With Long Short-Term Memory: A Case Study of Java-Bali System

Abstract

Short-term Load Forecasting (STLF) plays an important role in power system operation. It will be used for manage power balance between the dynamic power demand and power supply. This research presents a Long Short-term Memory (LSTM) and Recurrent Neural Network (RNN) for short-term load forecasting Java-Bali power system. We compare the performance of these network architecture models using Mean Average Percentage Error (MAPE) and Root Mean Squared Error (RMSE) to choose the best models for development Java-Bali power system operationalization in the future. The result show LSTM can forecast better than RNN due to vanishing and exploding gradient condition. The best LSTM model has MAPE 5,67% and RMSE 1683,09MW.