Electrical load and solar power forecasting using machine learning techniques

Abstract

Abstract In contemporary power networks, short-term load forecasting (STLF) is essential for efficiently managing reserve requirements. During the power-balancing operation, it then helps the grid operator make wise and cost-effective decisions. This paper thoroughly examines STLF techniques including particle swarm optimization (PSO), enhanced particle swarm optimization (EPSO), and artificial neural network (ANN) methods. The benefits and drawbacks of these approaches are shown through in-depth mathematical and graphical analysis as well as a comparative assessment. In order to increase the model's effectiveness for temporal sequence prediction, a hybrid ANN-solar power model is proposed and is evaluated using extensive data from the Xingtai Power Plant in China. The investigation shows improved accuracy and performance in short-term load prediction in terms of root mean square error (RMSE), mean absolute error (MAE), standard deviation (σ), and mean absolute percentage error (MAPE) in 24-h forecasting for the Xingtai Power Plant. Furthermore, the model demonstrates superiority in improving reserve management and balancing supply and demand in a contemporary electrical networks by outperforming earlier models using the Xingtai Power Plant dataset.

Keywords

• Renewable energy
• Artificial neural network
• Particle swarm optimization
• Short-term load forecasting
• Enhanced particle swarm optimization
• Machine learning techniques (MLT)