The Research On Fault Diagnosis Method On Photovoltaic Grid Connected Inverter

With the sustainable development of fossil energy and environmental problems become more and more serious,it has become one of the main hot spots to find and study new energy.Photovoltaic power generation has many advantages such as no pollution and wide distribution.In order to guarantee the normal operation of photovoltaic power generation,photovoltaic inverter reliability problem has become particularly prominent,once the fault will may cause damage or even a threat to the personal safety of equipment and equipment,thus realizing the rapid and accurate detection and fault location has become more practical and realistic significance.At present,the feature extraction and diagnosis methods used in the fault diagnosis of photovoltaic inverter include: wavelet analysis,Fourier transform,current root locus,fault tree,fault dictionary and so on.In this paper,a three-phase six pulse inverter circuit is selected as the research object.According to the characteristics of the output current waveform,a fault feature extraction method is proposed.Combined with three coding and intelligent algorithm to verify.(1)The fault of three-phase six pulse photovoltaic inverter circuit is simulated and the fault signal is extracted.By removing the IGBT trigger pulse to simulate the open circuit fault of the power transistor,the open circuit phase is not normally triggered,so as to realize the fault simulation of the inverter switching tube.22 kinds of fault waveforms are extracted from the AC output side,and the data of the 22 kinds of fault waveforms are extracted and processed by noise.(2)the method of extracting the characteristic value of the extremum ratio method.According to the waveform characteristics of A,B,C three-phase current under 22 kinds of faults,the feature extraction method of extreme value ratio is proposed based on the ratio of the maximum and minimum values of each phase current.Feature extraction by the extremum ratio method has the advantages of dimensionality reduction and automatic normalization.(3)A fault diagnosis method based on the combination of coding and extremum ratio is proposed.This method uses standard extremum ratio method table encoding obtained in the ideal case,considering the effect of external interference or current signal and transient adverse actors,the formula of the ratio of extremum method analysis,fault diagnosis in non ideal conditions.The method has the advantages of high speed,high accuracy,good anti noise ability and easy to understand.(4)A fault diagnosis method for photovoltaic inverter based on GA-BP is proposed.The genetic algorithm is used to optimize the BP neural network.At the same time,the extreme value ratio method is used as the fault characteristic input to diagnose the fault of the photovoltaic inverter.The simulation results show that the single BP network is compared with the GA-BP classifier,the GA-BP diagnosis results show that the proposed method has higher diagnostic accuracy,convergence speed and better robustness.In order to further improve the robustness of fault diagnosis,the extreme value ratio method combined with the average value of three-phase current is used as the composite extreme value ratio method,and the GA-BP algorithm is used to verify the robustness.The results show that the method of compound extreme value has better robustness.(5)A fault diagnosis method for photovoltaic inverter based on S_Kohonen is proposed.S_Kohonen neural network is used to realize the fault diagnosis of PV inverter,which is based on the ability of automatic clustering and the ability to identify the characteristics of the environment.The simulation results show that the method has the advantages of high accuracy and strong generalization ability.