Bad Data Detection Algorithm for Power System Based on K-Prototypes Clustering and Broad learning

In order to solve the problem of low identification rate of bad data in power grid state estimation, this paper proposes a bad data detection method based on K-prototypes clustering and broad learning. Firstly, the K-prototypes clustering algorithm is improved by increasing the adaptive part of the algorithm and selecting the best class. Secondly, the broad learning algorithm is used to compare the predicted quantity and the modified quantity measurement of state estimation, and the two algorithms are organically combined through the algorithm joint parameters to screen out bad data and improve the recognition rate and operation speed. Finally, through the simulation analysis of IEEE-30 bus case, the results show that compared with common detection methods, the proposed algorithm can effectively improve the identification rate and operation speed of bad data.