Research On Smart Grid Fault Diagnosis Method Based On Multi-source Information Fusion

With the continuous expansion of the scale of the smart grid,the regional interconnection has gradually become closer.At the same time,it also makes the power grid structure more complicated,and the uncertain factors affecting the stable operation increase.When a complex fault occurs,the control center will receive a large amount of alarm information uploaded by each acquisition unit.If the dispatcher cannot make timely and accurate judgments,it may cause a safety accident.The switching value information of protection and circuit breaker action has high redundancy,but it is uncertainty,which makes it difficult for traditional diagnosis methods to obtain accurate results when the alarm information is wrong.The voltage and current information contains a large number of fault features and has high accuracy,but due to the incomplete layout of the acquisition device,its redundancy is poor.Therefore,this thesis studies the smart grid fault diagnosis method based on multi-source information fusion by using the switch quantity alarm information and the electrical quantity information of fault current and voltage.Firstly,the characteristics of the multi-source information system based on the fault diagnosis of the power grid are analyzed,and the protection,circuit breaker action and alarm information are classified,and the uncertainty of the alarm information,relay protection configuration and action principle are analyzed.A smart grid fault diagnosis framework based on multi-source information fusion is designed.Secondly,the diagnosis method of switch quantity information based on fuzzy Petri net is researched.Because most of the diagnosis methods do not fully utilize the action sequence constraint property between the protection and the circuit breaker,and when the complex power grid and it’s structure is changed,the model matrix dimension is too large and the adaptability is poor.To this end,considering the timing constraint attribute between protection and circuit breaker action,the circuit breaker action is divided by timing constraint reasoning.Then the method of hierarchical modeling is introduced to improve the adaptability of the model,the TFPN model of component main protection,near-backup protection and far-backup protection subnets is established.Improved model adaptability through hierarchical modeling.And the simulation model was built in MATLAB/simulink,the validity of the model were verified,and the degree of suspected component failure based on the switch information was obtained,which constituted the evidence source of switch information.Then,the fault diagnosis method of electrical quantity information based on wavelet analysis is studied.Through multi-scale wavelet decomposition,the fault voltage and fault current obtained by building the system model in PSCAD are analyzed,and the characteristic mutation value of voltage amplitude,voltage energy characteristic mutation value,current amplitude characteristic mutation value,and current energy characteristic mutation value of each component are extracted.Then convert it into the electrical quantity failure degree of the probability interval,which constitutes the electrical quantity information evidence source of voltage and current.Finally,the smart grid fault diagnosis method based on multi-source information fusion is studied.The thesis analyzed the fusion algorithm of D-S evidence theory,and proposed a D-S evidence theory based on virtual alliance to improve the weighted fusion rules for the problem that the fusion result is contrary to common sense when the evidence is strongly conflicted.Considering the normalization of fusion results,a decision-making strategy is proposed by condition the non-propositional fusion results M(-).Then,the fault retrospective reasoning is carried out on the faulty components,and the accuracy of protection,circuit breaker action and alarm information is analyzed.Using switch quantity information and electrical quantity information to improve D-S evidence theory multi-source information fusion diagnosis.And compared with the classic fusion algorithm before improvement,the simulation results showed that the method in this thesis can avoid the occurrence of misjudgment and omission of judgment,which improved the fault tolerance and diagnostic accuracy of the model.The thesis has 36 figures,22 tables,and 87 references.