The comparison indicates that the MGA-ELMAN model without error correction exhibits a certain model bias, with overall predictions being lower than the actual values. The cascaded error correction model proposed in this paper can mitigate the model bias inherent in the preliminary prediction model, significantly enhancing the accuracy of the predictions.

Conclusion: Accurate prediction of PV output enables the power dispatch department to timely adjust the scheduling plan, significantly enhancing the acceptance of PV grid connection and the stability and security of the power system. In this paper, an error-corrected PV output prediction model using a serial MGA-Elman neural network is established, based on the PV output data from actual PV power stations. The effectiveness of the serial error-corrected prediction model is verified through comparison with the MGA-Elman error-corrected prediction model, thereby providing a reference for predicting PV output.