基于改进PGA优化VMD的ARIMA-LSTM混合模型的光伏功率预测OA

[Objective]The inherent volatility of large-scale solar power generation poses significant challenges to the energy management of the smart grid.Accurate prediction of solar power is crucial for ensuring the economic operation of the smart grid.A photovoltaic power prediction model is proposed which is based on the variational mode decomposition(VMD)optimized by the improved phototropic growth optimization algorithm(PGA),and integrates autoregressive integrated moving average(ARIMA)and long short-term memory neural network(LSTM).[Methods]The improved PGA is used to optimize the mode number and penalty factor of VMD,decomposing the original non-stationary power sequence into stationary sub-modes.Based on the fluctuation characteristics of each mode,ARIMA and LSTM prediction models are constructed,and the final prediction value is obtained by stacking and reconstructing.The comparative experiments are carried out among LSTM,ARIMA-LSTM,PGA-VMD-ARIMA-LSTM,and the improved PGA-VMD-ARIMA-LSTM using historical data from actual photovoltaic power stations as samples.Multiple evaluation indicators are selected to evaluate the proposed improved model.[Results]The MAE of the improvedPGA-ARIMA-LSTMis reduced to 0.5,the RMSE is 0.78,and the MAPE is as low as 1.37,achieving the highest accuracy,which is 18.9%higher than that of PGA-VMD-ARIMA-LSTM.[Conclusion]The ARIMA-LSTM model with VMD optimized by improved PGA outperforms traditional methods in terms of goodness of fit,prediction accuracy,and stability.By optimizing the parameters of traditional methods through improvement,the prediction performance is significantly enhanced.