Consumption estimation analysis in electricity distribution networks with deep learning

This thesis examines the use of deep learning models aimed at improving the accuracy and efficiency of energy consumption forecasts. It investigates how these techniques can be beneficial in areas such as the effective management of electricity distribution networks, integration of sustainable energy sources, and demand management. By enhancing the accuracy of electricity consumption predictions, the study offers suggestions for optimizing energy costs, increasing the use of renewable energy, and reducing energy losses. Moreover, it highlights the potential of deep learning and artificial intelligence technologies to make energy systems more efficient and sustainable. The study analyzes four popular deep learning models used in electricity consumption forecasting: Long Short-Term Memory (LSTM), Multilayer Perceptrons (MLP), Recurrent Neural Networks (RNN), Gated Recurrent Units (GRU), and the Transformer Deep Learning Model. Based on the results obtained, it was determined that the RNN model exhibited higher prediction accuracy compared to the others. Keywords: Multilayer Perceptrons, Deep Learning, Electricity consumption, Weather, Gated Recurrent Units, Recurrent Neural Networks, Long Short-Term Memory