In response to the issue of data quality loss in the actual collection, transmission, and storage of massive electricity data, this paper proposes a method for reconstructing electricity data based on CP decomposition and kernel norm regularization joint constraints, utilizing the temporal and spatial correlations of large-scale substation user electricity data. Firstly, a tensor of electricity consumption data for the substation area is constructed based on some potential similar characteristics of the electricity consumption behavior of large-scale substation users; Secondly, the F-norm and kernel norm are selected to regularize Gaussian noise and ideal complete electricity data, respectively, to establish tensor recovery models; Finally, the matrix factors obtained from CP decomposition are alternately updated column by column until convergence, thereby achieving high-precision completion of missing data. The effectiveness and accuracy of the algorithm were verified through simulation analysis and comparative experiments.