基于MSIBOA-BiGRU的海上风电机组功率预测研究OA

To improve the accuracy of power prediction for offshore wind turbine,a prediction model based on a multi-strategy improved butterfly optimization algorithm(MSIBOA)optimized bidirectional gated recurrent unit(BiGRU)was proposed.First,the structural mechanism of the BiGRU network and its advantages in time-series modeling were thoroughly analyzed.Meanwhile,several strategies,including good point set initialization,tent chaotic perturbation,and Gaussian perturbation were introduced to improve the traditional butterfly optimization algorithm(BOA),thereby enhancing its global search capability and convergence performance.Second,MSIBOA was employed to adaptively optimize key hyperparameters of the BiGRU model,including the number of hidden neurons,time step,regularization coefficient,and dropout rate,thus constructing the MSIBOA-BiGRU offshore wind power forecasting model.Finally,experiments were conducted based on measured data from a domestic offshore wind farm.The results demonstrate that compared with the BiGRU,PSO-BiGRU,and BOA-BiGRU models,the proposed model achieves significant improvements in both prediction accuracy and stability.Specifically,the MAE,MAPE,and RMSE are reduced by 51.6％,66.6％,and 61.0％,respectively,while the coefficient of determination(R2)is increased to 0.98.The proposed model can effectively capture the complex variation patterns of offshore wind power,improve accuracy and convergence performance of offshore wind power prediction,and exhibits strong potential for engineering applications.