Research On Wind Power Converter Main Circuit’s Monitoring System For Fault Diagnosis

Permanent magnet direct-drive wind generation has been widely applied with its various advantages, such as simple structure, reliable operating technology and low cost in maintenance. Although it eliminates the need for a gearbox which has a high failure rate in doubly-fed wind generation, the full power converter is also one of the high failure-rate components. It is significant that studying the fault diagnosis system for the safe and steady operation of wind generation. The current fault diagnosis system of wind generation still needs to overcome such problems as large human consumption, much environmental affection and the inaccuracy of diagnosis.In this paper, we study permanent magnet direct-drive wind generation, and the wind power converter main circuit topology is diode rectifier+boost chopper+voltage source PWM inverter. A set of fault diagnostic monitoring system of wind power converter is established.Firstly, in this paper, we establish a permanent magnet direct-drive wind power system model and select the three-phase output current of the wind power converter to extract the fault feature. The normal operation of the wind power generation system as well as25kinds of open-circuit fault condition is simulated in the Matlab/Simulink software, and then the current waveforms in the case of normal operation and under different fault can be obtained. I have decomposed and restructured the three-phase output current by using wavelet analysis method, extracted the phase energy value, combined the average current parameters with the tectonic fault feature vectors, as well as listed all the fault of feature vectors.Secondly, we select the fault feature vector constructed by using wavelet analysis and the average current parameters as input of BP neural network in this paper. According to fault feature vector and expectations characteristics, the structure of BP neural network can be determined. We learn and train the determined neural network, then test the fault phenomena that contain noise by using the trained neural network, and it verifies that the system has strong stability.Finally, we desgine a monitoring interface for fault diagnostic monitoring system by LabVIEW software.And then links the wind power model in matlab and monitoring interface by using LabVIEW Simulation Toolkit, The system can display wind power converter’s failure for simulation model in real time, and simulation is carried out.In this paper wavelet transform time-frequency localized features and neural network is the combination of self-learning function, applied to fault diagnosis system making the whole system has a strong approximation ability and fault tolerance.