Fault Diagnosis Of Converter In 3MW Wind Power System

Currently, wind energy is the fastest growing renewable energy, there are a lot of power electronic devices in the converter of large wind power system, which has many types of failure, relying on past experience solely to judge obviously does not meet the requirement complex system. The study of intelligent fault diagnosis system which can simulate human brain function is the goal of the fault diagnosis experts and scholars and has great significance.This thesis describes the present situation and prospects of wind power technology, the working principle and function of converter in wind power systems and the fault diagnosis methods of the power electronic devices in converter. Taking the open-circuit fault of the diode and insulated gate bipolar transistor (IGBT) for example, Wavelet analysis method and artificial neural networks are applied to fault diagnosis. And for the special operating mode in converter of wind power system, the frequency division control is applied to fault diagnosis method.Firstly, simulation models of wind power system with each component are built and the fault conditions in converter are analyzed and classified. Wind power converter system includes two parts which are rectifier and inverter. The output voltage of rectifier which is chosen and normalized input the BP neural network and the RBF network to diagnose. The results show that the diagnosis efficiency of the RBF network is better than the BP network in the fault diagnosis of single converter.Secondly, there are many categories when rectifier converter and inverter converter part fails at the same time. So the representative types of the fault are selected and the wavelet analysis is applied to extract the energy value of the faults frequency, RBF network is the fault classifier respectively. The simulation results show the feasibility of wavelet analysis.Finally, for the particularity of the converter's operation mode in wind power system, that is the frequency fluctuates with the wind speed, the sub-frequency band control is applied. Using different neural network at different frequencies to output the correct diagnosis results through a decision, which make the fault diagnosis result accurately and achieve accurate fault diagnosis results when the frequency change.