Research On The Detection And Diagnosis Method For Photovoltaic Power Station Equipment Fault

With the depletion of fossil fuel andincreasingly serious environmentalpollution,the research for renewable energy has been a hot topic in our society.Amongof the renewable energy, solar photovoltaic (PV) energy is no polluted, sustainable,good flexibility and high reliability.Because of these merits, it has been widelyconcerned by people.In order to guarantee the normal operation of PV plant, reducingserious accidents and revenue loss caused by equipment faults，the detection anddiagnosis methods of the faults are studied in this paper. The main contents are asfollows:1. This paper studies the types and topological structures of PV system, and thenanalyzes the main faults in the PV plant. The research objects of this paper arethe faultsof photovoltaic array and the open-circuit faults of power tube in the PVinverter.Furthermore, the reason of the faults generated is studied in depth.2.On the fault detection of the PV array set, thefaults of PVmodulesare equivalentto shadows, and the effect of shadows on the output characteristics of PV array issimulated.Aiming at power difference between normal and faultedPV modules, a PVarray fault detection method is proposed to detect the branch current and modulesvoltage simultaneously.The method realizes the location of faulted PV modules byanalyzing the power difference. The accuracy and cost of the detection method is alsostudied. Finally, the feasibility of detection method is proved by the model of8X3PVarray built in Matlab/Simulink.3. On the fault diagnosis of the power tube set,a fault diagnosismodel based onwavelet transform and neural network is proposed in this paper. Firstly, themathematical model and control strategy of inverter is studied and the currentcharacteristics of the faults is analyzed. And then,based on the simulation model of thephotovoltaic system,22kinds of fault combination are validated by simulation.Byintroducing the concept of waveform parameters, it constructs the compositefeaturecombining with energy feature and wave parameters. The fault diagnosis model based on probabilistic neural network is established. Due to the currentmutationdecliningthe diagnostic rate, this paper adopts the sumof three-phase currentfrequency energy as normalized base value of wavelet energy, effectively reducingthesensitive of the fault features for current mutation.Finally, the simulation result showsthat: the diagnosis model based on the composite feature performancebetter than the onebased on wavelet energy feature.