The sectoral electricity load forecasting by using deep learning techniques

Forecasting of electricity load is a critical issue because of the increasing global demands for energy and the need for its efficient management in grids. This study aims to utilize Deep Learning (DL) models, including Long Short-Term Memory (LSTM), Convolutional Neural Network (CNN), and a hybrid CNN-LSTM model, for forecasting electricity load. Monthly data from 2016 to 2023 is used to predict the electricity load in the provinces of Adana, Mersin, and Antalya in different sectors, concerning lighting, residential, industry, agriculture, commercial, and total load. The Matlab/Simulink model has been used to observe the 24-hour electrical load profile. The main contribution of the thesis is to examine and propose a comparative analysis of DL approaches and their hybrid applications. Comparison results illustrated noteworthy information in terms of the effectiveness of DL models in capturing complex patterns, which focused on the importance of DL models in strategic energy planning. It was concluded that concerning Root Mean Square Error (RMSE), R2, Mean Absolute Percentage Error (MAPE), and Mean Absolute Error (MAE), the hybrid model provides superiority to the LSTM and CNN models. The results show that the capability of CNN-LSTM hybrid models in capturing the temporal dynamics of sectoral electricity loads is more effective than the LSTM and CNN as the case.