Methods And Research Of Power Grid Fault Diagnosis With Multi-Source Information

With the rapid development of our country’s economy, the demand on the safe operation of power grid and the reliability of power supply from the whole community have become particularly important. However, the power system may be interrupted inevitably by man-made operation mistakes, natural force and other factors. As an important subject of self-healing of power system, the power system fault diagnosis has been one major research direction of smart grid.Traditional methods of power grid fault diagnosis are mostly based on the information of protections and circuit breakers, and accurate diagnosis results can be obtained when the information is complete. When fault occurs, a lot of alarm information produced by devices would enter the control center within a short time, which may cause obstruction, distortion or frame-loss of information easily. So there is a lot of uncertainty in the information of protections and circuit breakers. As the scale and complexity of power grid become increasingly larger, more and more automatic monitoring devices are applicated to the power system. The fault recorder works as a black box which can analyse power grid accident through recording waveforms of voltage and current before and after the failure time, and it contains lots of fault information. To improve the accuracy of diagnosis results, it is necessary to bring a new information source (recorded data) into Power System fault diagnosis. To solve these problems, this paper carries out fault diagnosis of power grid with fuzzy petri nets and technology of wavelet analysis. The detailed research work is shown as follows:(1) This paper introduces information sources used for power grid fault diagnosis and analysis methods of power network topology. A serial-structure framework of mixing fault diagnosis is set up according to the difficulty level of obtaining information.(2) The reliability assessment model of protection system configured on lines is improved by the Markov state space method. A method is proposed to calculate credibility degrees of protections configured on lines. The credibility degrees of protections configured on buses and transformers are corrected by statistical data of right protective actions. The credibility degrees of protections can supply accurate initial values for fuzzy reasoning and diagnosis.(3) Fuzzy petri nets will be used in power system fault diagnosis to solve the question that the action information of circuit breakers and protections has the uncertainty. First of all. associated trees of suspicious fault components are formed by the breadth-first search method. Circuit breakers are classified in accordance with the priority order of action. Secondly, a method based on topology and protection knowledge mapping into petri nets is proposed to construct models of fault diagnosis, which is applicable to the line, single busebar and transformer. At the same time, the models of fault diagnosis based on the logic between protections and circuit breakers are applicable to the double-buse and3/2Wiring. Finally, we can identify fault components by fuzzy reasoning and evaluate the motion behavior of relay protections and circuit breakers.(4) The recorded fault data can be analyzed with the wavelet technology. Firstly, we can determine the fault occouring time by detecting current mutation point from the generation to the removal of the failure. Secondly, fault characteristics can be constructed by analyzing the wavelet energy variation of voltage, current before and after the failure time. Finally, the fault components can be found out.From the above, this paper makes a thorough research to the power systerm fault diagnosis, handles the problem of the uncertain information with fuzzy petri nets, initially realizes fault diagnosis with electrical quantity information. Simulation results show that methods in this paper are effective.