Open-switch Fault Diagnosis Of Inverter In DFIG-based Wind Power Systems Based On Wavelet Transform And SVM

With the installed capacity of wind farms increasing year by year, the design and manufacture of wind turbine is no longer a problem. After running for some time, the maintenance and repair has become an unavoidable problem when the wind power equipment broke down. In the present, the double-fed wind turbines are most widely used among wind power generation equipment. Compared with the generator and other mechanical components, the inverter is one of the relatively poor stable links in a wind power system. The power devices of inverter must withstand the changes in temperature as well as the conditions, which easily leads to failure during its working period. Once any switch of the inverter breaks down, the entire wind turbine might suffer serious damage or lead to downtime. So the fault diagnosis of the inverter is significant for the stability improvement of the wind power system.This paper uses Matlab/Simulink software to build a double-fed wind system according to the model built by the DFIG-based Wind Power System in laboratory to simulate open-switch faults of the inverter. Simulation results show that the wind power system operates stably in normal conditions and simulates open-switch faults well in fault conditions, which lays the foundation for the subsequent fault diagnosis.Secondly, since FFT can only reflect the information of signals globally in frequency domain instead of the local anomaly information in time domain. This paper proposes a wavelet transform method on open-fault current waveform feature extraction of rotor side inverter. The method uses wavelet transform to extract feature vectors which are the low-frequency energy value of three-phase currents of rotor side inverter in double-fed wind system. Simulation results show that the extracted current energy feature vectors present good separability, which are good for open-switch fault classification in the rotor side inverter.Finally, since the long training time and complex network design problem of neural network, a doubly-fed wind power rotor-side converter open-switch fault diagnosis based on wavelet and support vector machine is proposed. The method uses feature vectors extracted by wavelet transform to adopt the feature vectors by trained support vector machines for fault diagnosis. Simulation results show that the proposed method is not only reduce fault diagnosis time but also improves the classification accuracy.27maps,12tables,69references.