Research On Power System Bad Data Detection And Identification Algorithm Based On Data Mining

With the continuous improvement of the automation level of power system,the requirement of real-time data quality on power grid operation is getting higher and higher.So it is very necessary to detect and identify bad data.The traditional methods of bad data detection are basically algorithms before the estimation,which need to be repeated for many times to estimate the state.It may cause large amount of calculation and "residual submergence" or"residual pollution".So the ideal state is that bad data can be identified and detected before estimation.In this paper,bad data detection and identification based on association rule mining can solve these problems to a certain extent.The main contents of this paper are as follows.1)The traditional methods of detecting and identifying bad data is researched and then their limitations and shortcomings are analyzed.According to the superiority of bad data detection before state estimation,the feasibility of applying association rule mining algorithm to this problem is proposed.2)The classical algorithm of association rules——Apriori algorithm is researched and then their shortcomings are analyzed.On the basis of traditional Apriori algorithm,an improved multi-dimensional association rule algorithm is proposed.The two algorithms are compared to verify the effectiveness of the improved algorithm.3)The concept of temporal association rules is introduced and one of them——cycle association rule is focused on.The basic idea and algorithm realization of cycle association rule are elaborated and the improved algoritlhm based on the traditional algorithm is realized.which can reduce unnecessary cycle support count and the time cost.4)The improved association rule algorithm is applied to the detection and identification of bad data before estimation.Based on the historical data samples under the premise of unknowing the topology of the system,the association rules mining is used to obtain the corresponding forecasting for each time.Then,the method of setting bad data artificially is used to analyze the experiment data.The correlation between the time attribute and the state value is used to verify the correctness of the current state value,which verifies the feasibility and validity of the proposed method.