Power System Reliability Assessment Based On Bayesian Theory

Reliability of the power grid plays an important role in the maintenance and replacement of the power system.Therefore,it is necessary to study the reliability of the power grid.The current major concerns of power reliability including analyze historical and test data of components,establish system model based on the power system structure,and construct a system model when components are correlated.Although the reliability analysis of the power system focuses on different objects and the adopted indicators may not be the same.The analyzing method,in essence,are identically the same: First,a model should be established to evaluate the life property of a product.Then,based on the acquired data,the uncertainty of the key parameters in the model could be approximately evaluated for the purpose of life prediction.However,traditional parameter estimation method may lose their effects with the consideration of the condition of small samples.In this regard,if some prior information is acquired,the way of pooling style of prior information should be highlighted to improve the accuracy of the parameters evaluation.From the above analysis,main topics of this thesis could be summarized as follows:First,the reliability for relay elements is explored based on hierarchical-Bayesian.As a key component in Grid,the failure of relay element will influence the stability of the Grid and even lead to major break down.Besides,as the development of the relay element,failure data is extremely hard to be acquired.Traditional parameter evaluation method may lose their effects if the failure data is not full enough.In order to solve this problem,a method is proposed to apply the thought of hierarchical-Bayesian into the traditional Bayesian method.Simulation and experimental data will be both provided to validate the effect of the proposed method.Second,different pooling patterns of prior information are studied to evaluate the parameters in reliability model.Although the field data is hard to be acquired,expertise experience and the historical data of similar component do exist.For such scenario,the modelling of the prior information for traditional Bayesian should be focused.In this part,different pooling styles of prior knowledge are studied.Accordingly,the accuracies of the parameter evaluations are demonstrated.The optimal pooling style of the prior information is explored based on simulation studies.Third,parameter evaluations in the relative model are studied.Considering the relativity of the component in the system and the small amount of the failure data,the accuracy of the parameter evaluation based on Bayesian method is discussed.Optimal sampling method is explored based on simulation studies.