Intelligent Forecasting of Electric Energy Demand with Artificial Neural Networks

Abstract

Energy demand forecasting is critically important for the effective planning and management of energy production and distribution. Accurate demand forecasts in the energy sector can help reduce costs and enhance the reliability of energy supply. In this study, data-driven methods are employed to predict future energy demand. Multidimensional datasets, including historical consumption data, weather conditions, economic indicators, and demographic information are utilized in the forecasting process. To select the most appropriate model and improve prediction accuracy, various time series modeling techniques and artificial neural network algorithms are tested. The results demonstrate that the RNN-based deep learning model outperforms other methods, such as LSTM and CNN, in terms of forecasting accuracy. Particularly during periods of high variability, such as seasonal transitions, RNN models provide predictions that are more reliable by reducing the Mean Absolute Percentage Error (MAPE) to 9%. This study contributes to the literature by offering a comparative analysis of different forecasting approaches using real-world data. Furthermore, it presents a repeatable and adaptable forecasting framework for energy suppliers and decision-makers, delivering tangible benefits in resource planning and mitigating operational risks