In order to enhance the prediction accuracy,the robust sparsity broad learning system(RSBLS)prediction method is proposed.The model training is transformed into a difference-of-convex functions optimization problem based on the regularization method.The robustness of the estimation is improved by using the L1 norm as the data fidelity term.The performance of the model is enhanced by integrating the L1-2 norm into the objective function to ensure the sparsity of the output weights.The numerical method that integrates the advantage of the half-quadratic splitting algorithm is proposed to solve the training model efficiently.Experimental analysis based on the criteria of absolute error,relative error,mean absolute error,root mean square error,mean absolute percentage error,algorithm stability,percentage improvement of the performance,gray correlation analysis and DM test shows that the prediction quality of the new algorithm is better than popular prediction algorithms and has better robustness,which provides a new method for wind power prediction.
wind power predictionbroad learning systemregularizationsparsity optimization
NETL
NSTL
万方数据
