Neural networks application in short-term load forecasting

Short-term load forecasting (STLF) is a fundamental procedure in power systems operation that underlies the most important decision-making processes, such as economic dispatch or equipment maintenance planning. Due to the high degree of uncertainties in demand variations, advanced techniques based on artificial intelligence are needed in order to obtain an accurate electrical load forecasting. In this paper, multiple forecasting methods based on neural networks, including the multilayer perceptron (MLP), convolutional neural networks (CNN), long short-term memory (LSTM) and gated recurrent unit (GRU), are applied to solve the STLF problem, using a real dataset provided by the Romanian TSO. In this regard, the Mean Squared Error (MSE), the Root Mean Squared Error (RMSE), the Mean Absolute Error (MAE) and the Mean Absolute Percentage Error (MAPE) are used as evaluation metrics for the day-ahead load forecasting results. © 2021, Politechnica University of Bucharest. All rights reserved.