Combination model prediction of photovoltaic power generation based on an improved bidirectional gated recurrent unit

ZM Chen and YX Pu, JOURNAL OF MECHANICAL SCIENCE AND TECHNOLOGY, 39, 4191-4203 (2025).

DOI: 10.1007/s12206-025-0640-4

Photovoltaic (PV) power generation forecasting refers to the prediction of PV power generation at a certain point or over a period of time in the future. Accurate power forecasting can promote sustainable development of PV energy and advance green development. Considering that the current single model ignores the complexity of power generation data when predicting PV power generation sequences, this study develops an improved model to enhance the prediction accuracy of the PV power generation output, which involves the use of complete ensemble empirical mode decomposition adaptive noise to decompose power generation sequence. Futhermore, this study constructs a combined model of the category boosting algorithm and the bidirectional gated recurrent unit networks to predict data of different frequencies. In addition, the multi-head attention mechanism and the grey wolf optimization algorithm are introduced to further improve the accuracy of the model. Finally, multiple experiments are conducted using data from a PV power station in Zhuhai. Experimental results show that the proposed model achieves reductions of 34.6 %, 34.2 %, and 57.2 % in root mean square error, mean absolute error, and mean square error, respectively, along with a 5.2 % improvement in R2 compared with the baseline model. Data from different regions and seasons are also selected for the experiments to prove the transferability of the proposed method.

Return to Publications page