A Stacking Photovoltaic Power Prediction Model Based on MICE and BiGRU-Attention for Plum Rain Season

Considering the issues of insufficient data collection and large randomness in power generation during the plum rain season, this paper fully considers the lack of data in the plum rain season itself. Directly deleting missing samples would lead to waste of information. Firstly, the Multiple Imputation by Chained Equations (MICE) method based on Random Forest (RF) is selected to fill in missing values by using machine learning. At the same time, a Stacking ensemble prediction model based on BiGRU-Attention is adopted so that the algorithm can focus on important parts of plum rain data to achieve differential learning of the data. Finally, the control variable method is used in experiments to compare and analyze the MICE method and the improved Stacking algorithm. It is proved that both methods can improve the efficiency and accuracy of photovoltaic power prediction.