In modern power system, the quality of data influences the safety and stability of the power system directly. In paper, GSA-based data-mining technique was analyzed and applicated into bad data detection of the power system. Shortcomings of the technique were found by a lot of simulations. Toovercomethe shortcomings, a new judgment was presented to estimate the optimal number of the cluster: elbow judgment based on GSA, and it also was applicated in bad data detection of the power system. Elbow judgment based on GSA is a kind of techniques what can estimate the optimal cluster number well. In bad data detection of power system, BP neural network is used first, then the testing results are clustered, and then elbow judgment based on GSA is used to estimate the optimal cluster number, at last good and bad data will be separated distinctly .In paper, the real-time data measurements of a part of Jiangsu power network was simulated. The results revealed that in kinds of bad data conditions the bad data could be detected by the new technique efficiently and correctly. |