Multi-variate Time-series Load Forecasting using Deep Learning

Abstract

To ensure the stable and reliable operation of a power system, load forecasting is required. Accurate forecasting leads to efficient dispatch, unit commitment, and energy security. Smart power management in the generating, transmission, and distribution network, as well as the accompanying energy demand, can be realized with accurate forecasting approaches. This paper analyses the short-term load forecasting of the Bangladesh power system. Various deep neural network models- XGBoost, LSTM, Stacked LSTM, CNN, CNN-LSTM, Time Distributed MLP, and Encoder-Decoder are used to forecast the load. The load is predicted based on previous load data and various features like temperature, Weekdays, Weekends, and Peak Business Hours are taken to ensure the accuracy of the results. This study reports the advantages and disadvantages of each model.