Short-term Wind Power Forecasting Model Based on BP Neural Network Algorithm Optimized by Particle Swarm Optimization

Abstract

Under the impetus of the "double - carbon" goal, wind power, as one of the main forms of new - energy power generation, has been growing in significance. However, the unpredictability of wind power output has presented difficulties for the secure and stable operation of the power system as well as real - time scheduling plans. Regarding the issues that the prediction of wind power output based on the traditional BP neural network has a slow convergence rate and is prone to getting trapped in local optima, this paper puts forward a hybrid wind - power prediction model (PSO - BP), where the BP neural network is enhanced by the particle - swarm optimization (PSO) algorithm. This approach optimizes the initial weights and thresholds of the BP neural network via the global search of the PSO algorithm. As a result, it enhances the model's convergence capacity and, to some degree, circumvents the problem of local optimality.Public wind - power datasets are utilized in the experiments. The PSO - BP and traditional BP models are trained and tested multiple times under the same circumstances. The outcomes indicate that, in comparison with the traditional BP model, the PSO - BP model, while ensuring the convergence speed, mitigates the local optimization issue of the neural network. It significantly improves the reliability and precision of the model's prediction results. Moreover, it offers robust technical backing for the short - term prediction of wind power, which is conducive to improving the power - system consumption plan and ensuring its safe operation.