Development of a neural network model for forecasting energy consumption in energy management systems

This paper explores the use of neural networks to predict energy consumption based on data on the production and use of various fuels, including oil, gas and coal. The aim of the study is to develop a model that can take into account the complex relationships between key parameters such as production volumes, production costs, CO₂ emissions, energy efficiency, inventory levels and transportation costs. The relevance of the study is due to the need to improve the accuracy of forecasting for effective planning and management of resources in the energy industry, especially in conditions of unstable market and environmental factors. In the course of the work, a multi-layered fully connected neural network model was developed, which was trained and tested on the prepared data. The evaluation of the model using the coefficient of determination and the average absolute error demonstrated its high accuracy and suitability for practical application. Additionally, an analysis of the importance of the signs was carried out, which showed that the most significant factors are the import/export of energy and the share of alternative sources. Visualization of the results confirmed that the model is able to make accurate predictions, minimizing errors. The results obtained confirm that the use of neural network models to predict energy consumption is a promising area that can significantly improve decision-making processes in the energy sector and increase resource efficiency.