Transformer Health Condition Assessment Based On Gray Cloud Model

Transformer is the core equipment of power transmission and conversion in the whole power system.Its operational reliability is directly related to the safety and stability of power system.Implementing condition-based maintenance on transformer can improve maintenance efficiency and equipment reliability,lower costs of operation and maintenance and make maintenance more rational and scientific.The key to achieving condition-based maintenance is that transformer state is scientifically and effectively assessed.Thus,this thesis focuses on the research of transformer health state assessment from three aspects that include the establishment of evaluation system,the calculation of index weights and the selection of state assessment methods.In view of complexity of indicator that characterize state of transformer,this thesis selects 24 evaluation indicators that can be quantitatively calculated,easily obtained and representative.Common fault types of transformer and their causes are explained in detail.According to relevant evaluation guidelines and the feature of correlation between faults and indicators,a hierarchical condition index system of transformer is built.Meanwhile,condition classification and maintenance strategies of transformer are briefly introduced.In terms of weight calculation,calculation processes and principles of some arithmetic that include analytic hierarchy process method,entropy weight method,standard deviation method,CRITIC method and combination weighting method based on game theory are described.To overcome the defect of relying heavily on subjective opinions of decision makers,the method of calculating index weight based on the support and confidence of association rule is proposed.Finally,The case verifies the effectiveness and rationality of the proposed weight calculation method.In terms of the selection of state assessment methods,a hierarchical evaluation method for condition of transformer based on weight-varying gray cloud model is proposed.Firstly,in allusion to the defect that traditional gray clustering whitening weight function cannot effectively reflect ambiguity and randomness for state evaluation level information of transformer,cloud model is introduced into gray clustering whitening weight function to construct a gray cloud model.A method of constructing index cloud model is proposed.Cloud correlation is calculated by using index cloud model instead of index value to better reflect uncertainty for state information of transformer.Then,index weight is calculated by using the support and confidence of association rule,and gray cloud clustering is used to obtain fault evaluation scores of transformer fault layer.According to the rule of “50%correlation degree”,the fault layer condition is determined.The overall condition of transformer is obtained by weight-varying fusion.To acquire the final evaluation result,the condition of fault layer and the the overall condition of transformer are considered comprehensively.Finally,the case studies show that the proposed method can not only obtain objective and accurate evaluation results,but also find out causes that cause this condition.Therefore,this method can provide more substantial help for condition-based maintenance.