基于MLP与iTransformer融合的中长期光伏发电功率预测OA

In order to improve the prediction accuracy of PV power in the medium-to-long-term,an MLPformer model based on the fusion of Multilayer Perceptron(MLP)and iTransformer is proposed,which can accurately and effectively predict PV power.MLPformer innovatively proposes to perform variable correlation,feature extraction,and temporal feature extraction for the time series separately.This approach solves the problem of insufficient feature extraction of time series by existing single models.The Multilayer Perceptron Feedforward(MLPforward)unit is designed to improve the ability of the iTransformer model to extract variable correlation features.Multilayer Perceptron Blcok(MLPBlock)is designed to extract time-dependent features and capture nonlinear relationships.It is experimentally verified that the model possesses higher accuracy in medium-to-long-term PV power prediction,and the model shows an average decrease of 4.49％in mean square error and 4.79％in mean absolute error compared with the current state-of-the-art iTransformer model on two publicly available PV power datasets.