The wide application of digital technology in power systems generates a large amount of data. The analysis and mining of these data can generate great value. It is important to ensure and improve data quality, as the basis of further data mining works. This paper studies characteristics of household electricity data and uses order dependencies (ODs) to define data criterion on them. This paper then provides a repairing algorithm for electricity data based on dynamic programming. Further, the article optimizes the time complexity of the algorithm so that the algorithm can be repaired in a suitable time on a large data set. Finally, we conduct experiments to compare our approach against common signal processing methods in terms of multiple dimensions such as algorithm running time and repair quality. Experiments show that our algorithm has a significant improvement over the electricity data compared to these approaches.