Research on the relationship between the digital transformation of new energy enterprises in the context of electricity reform

Abstract

Accurate wind power prediction is essential for enhancing the dispatchability of wind power and the stability of power grid operation. However, traditional single-group intelligent algorithms for optimizing BP neural network modeling often suffer from large result fluctuations and poor stability. To address this, this paper proposes a wind power prediction model based on BP neural network optimized by a particle swarm–neighborhood gravitation–cuckoo collaborative optimization algorithm (PSGC). PSGC integrates the fast global convergence of particle swarm optimization (PSO), the Lévy flight characteristic of cuckoo search (CS), and the neighborhood gravitation mechanism of gravitational search algorithm (GSA) to maximize the advantages of each algorithm and overcome the limitations of single algorithms. This integration is applied to the initial optimization of the BP neural network’s weights and thresholds, followed by secondary optimization through the BP network algorithm, ultimately for wind power prediction. Results show that the PSGC-BP model performs well in prediction accuracy and stability. In the testing phase, the model’s root mean square error (RMSE) is 22.0127, mean absolute error (MAE) is 17.1045, and correlation coefficient (R) is 0.7903; in the prediction phase, RMSE is 45.2569, MAE is 27.9380, and R is 0.8408, with the smallest operating fluctuation, highest stability, and best comprehensive performance. This model provides a feasible method for wind power prediction, contributing to improved dispatchability of wind power and grid stability.