To solve the problems of strong uncertainty and multiple influencing factors in photovoltaic power prediction,this paper proposes a photovoltaic power prediction method based on variational mode decomposition(VMD).Firstly,the original photovoltaic power data is subjected to variational mode decomposition,which decomposes it into modal components with relatively stable frequencies.Secondly,the permutation entropy of different modal components is calculated,and the different components are further merged based on the permutation entropy.Under the condition of considering different influencing factors(temperature,radiation,etc.),the different frequency modal components are predicted by the bi-directional gated recurrent unit-attention(BiGRU-Attention)model.Finally,the predicted results of different frequency components are superimposed and reconstructed to obtain the final predicted value.Experimental tests were conducted on photovoltaic power data in a certain region of China,and the results showed that compared to the BiGRU model,the model proposed in this paper reduced the mean absolute percentage error(MAPE),mean square error(RMSE),and mean absolute error(MAE)by 11.25%,8.51%,and 11.92%,respectively,and significantly reduced its prediction error.
deep learningphotovoltaic power forecastVMDGRU networkattention mechanism
维普
万方数据
