Missing Data Imputation for Photovoltaic Station Based on Spatiotemporal Features

Accurate and complete data is essential for data-driven applications. However, data loss is inevitable in photovoltaic (PV) station due to the vulnerability of transmission networks. Traditional imputation methods relying solely on the temporal features of individual inverters fail to achieve high accuracy under drastic environmental changes. This paper proposes a spatiotemporal feature-based data imputation method for PV stations, incorporating both the temporal features of single inverters and the spatial features among multiple inverters. This approach significantly improves imputation accuracy and maintains precision under rapidly changing conditions. Experimental results validate the effectiveness and superiority of the proposed method.