Short-Term Electric Load Forecasting Using ESN Neural Networks

Electric load forecasting is pivotal for strategic planning in the electricity industry, with repercussions for resource allocation and network responsiveness. The study’s primary goals are identifying the variables affecting electrical energy usage and applying artificial neural networks to make forecasts come true. This study focuses on modeling and forecasting electric load, employing Echo State Network (ESN) neural networks known for their efficiency. Data from Tehran Province’s electricity distribution company and meteorological data are leveraged for precise predictions. Using MATLAB, the model’s accuracy is validated against real-world data. The study explores atmospheric conditions, historical consumption records, and calendar factors influencing electricity consumption dynamics. The research uses artificial neural networks to construct prediction models for daily solar usage in 2022. Using neural networks, four models were created using recorded load data, weather data, and calendar parameters covering 2018 through 2021. A prediction for the daily consumption during 2022 was obtained using these models and modeling atmospheric behaviors based on comparable conditions from 2018 to 2021 combined with the calendar characteristics unique to 2022. The main accomplishments of the research center on a comprehensive methodology for predicting power usage, emphasizing the amalgamation of many data sources and advanced modeling approaches.