Application Of Multivariate Statistical Process Control Based On Variable Weighting In Transformer State Evaluation

The premise to realize the condition based maintenance of power transformer is to accurately evaluate the operation status of transformer according to the equipment detection data information.Most of the methods focus on the study of complex model,ignoring the relevance and correlation of the data of transformer state detection.In view of this,this paper proposes a multivariate statistical process control method based on variable weighting to evaluate the transformer state.Firstly,this paper studies the dimensionality reduction method of data involved in multivariate statistical process control,the establishment of multivariate control chart,the evaluation system of control chart and the calculation method of evaluation index.According to the characteristics of the data during the operation of the transformer and the applicability of the relevant methods,the association rule method that does not depend on the correlation between variables is selected as the dimensionality reduction method of the data;based on the control chart of multivariate index weighted moving average(MEWMA),which includes the historical mean shift and high flexibility,the control method of multivariate statistical process is improved;use the Monte Carlo simulation method with large calculation amount but better applicability to calculate the average running chain length(ARL).Secondly,for process control problems with different degrees of importance of indicator variables,the weighted Mahalanobis distance is introduced to improve the multivariate statistical process control chart,and a multivariate statistical process control based on variable weighting that considers both the importance and the correlation of variables is constructed(VW-MSPC)figure.The ARL analysis of the overall shift of the sample mean and the individual shift of individual variables shows that the VW-MSPC chart is more sensitive to the shift of the mean of the variable with a higher weight.At the same time,compared with the traditional multivariate statistical control process,the VW-MSPC chart has Better denoising ability.Finally,according to the characteristics of the transformer condition assessment system and the multivariate statistical process control method,the two are organically combined,and the specific problems encountered by the VW-MSPC chart in the transformer condition assessment process are solved: based on the relational database theory and related Standards,combined with actual operating experience,select the appropriate fault type and state variable;use the support information of the association rules to classify the fault type and state variable,analyze the coupling relationship between the fault type and state variable,and establish the key index variable of the transformer System;through the confidence information of the association rules,the objective weight coefficients of the state variables are obtained;combining the attention threshold of each state variable and the data characteristics of the transformer during normal operation,a method for determining the multivariate significance level during the transformer state evaluation is given;the ARL of the VW-MSPC chart divides the operating state grade of the transformer,and further gives the control threshold of each operating state grade of the transformer;on the basis of the transformer operating state grade division system,the dynamic parameter optimization method of the VW-MSPC chart is proposed.Figure for dynamic Parameter optimization design.An example verifies the feasibility of this method.