LSTM-Based Electrical Load Forecasting for Chattogram City of Bangladesh

Load forecasting is one of the necessary tools for the modern energy management system. In the smart grid, electric load forecasting offers a useful duty in decision making for power system companies to ease the decision-making action of electricity production and consumption. Short-term load forecasting is the most practicable kinds of load forecasting method as it can match an unsteady growth of power demand with the increasing population and reliance on electric power. Due to the chaotic nature of the electric power demand, an artificial neural network (ANN) is the favored method for load prediction purposes. In this paper, an electrical load forecasting technique based on recurrent neural network (RNN-based model): Long Short-Term Memory Network (LSTM) is proposed for a planned smart grid system in Chattogram city, the commercial capital of Bangladesh. The suggested method is tested on the freely accessible dataset for load forecasting purposes in the Chattogram city. In addition, the method is compared with other traditional load forecasting techniques, such as support vector machine (SVM). The proposed LSTM method outperforms the SVM method with the least RMSE and MAE error rate of 127.682 and 92.448 respectively, which results in 52.24% and 21.67% error reduction in both MAE and RMSE respectively.