Bad Data Identification In Power System Based On The Gsa Data Mining

The quality of operation data is becoming increasingly important in a modern power system. When bad data intrudeds into the power system, the operating personnel will make wrong dicisions according to the bad data, thus threatens the safe and stable operation of the power system. The paper provides a new method for solving the bad dada identification problem: GSA-based (gap statistic algorithm) data mining technique.The GSA technique is new technique based on the cluster analysis theory and statistic analysis. The advantage of this technique is its ability to find the appropriate number of clusters of a data set automatically for unsupervised learning. In the paper, the GSA technique was embedded into an artificial neural network. The raw measurements were first inputed into a well-trained neural network, then the outputs of the neural network were grouped by the k-means cluster analysis, finally the optimal number of clusters was chosen by caculating the gap values vesus the cluster number. Thus the bad data groups and the normal data groups were separated and the bad data would be well identified.In the paper, the proposed method has been tested through a simple power netwok model. In the test, the voltage, power and reactive power were selected as inputs for testing in several bad data scenario. The results of the test validated that the proposed method was an automatic, simple, flexible and effective tool for bad data identification of the power system.