Fault Diagnosis Research On The Frequency Converter Of The Permanent Magnet Synchronous Wind Power Generator

When entering the twenty-first Century, in response to the global energy crisis, many countries including our China are generally accelerate the development of the new energy technology and industry. By the end of2009,the Copenhagen conference is made a clear further requirements for the industrial countries to cut down the greenhouse gas emissions, since then the new energy become a important form of energy. Among the new energy, wind power comes to the most representative and most potential new energy sources for its reserves enormous, simple in form conversion, can be large-scale mining and no pollution, and become the future economic development of science and technology. Now days, more and more wind-farms have been built in our county, but at the same time same faults happened for different reasons and the frequency converter fault must be the most common one. As we know different faults will affect both the wind-farm working and the electric energy quality, so how to diagnosis the fault become a hot topic.Wind power industry in Xin Jiang province development very quickly, and the study of the fault diagnosis on wind power generating converter of the Permanent magnet synchronous wind power generator will have a good representative meaning. In this paper, we analysis the working principle, topological structure and control strategy of the Permanent magnet synchronous wind power generator first. Then, on the basis of these we build a reasonable simulation model and a series of voltage and current curves have been got during the fault simulation of the frequency converter. Thirdly, the result of simulation data are processed and used as a sample data to training and testing the neural network, in order to make the neural network has the ability of fault diagnosis and orientation. Finally, consider factors such as the neural network training time and forecast result is not too ideal, we attempts to use the genetic algorithm to optimize the weights and the threshold parameter of neural networks, thereby reducing the neural network training time and improve the precision of forecasting.Through the theoretical analysis and simulation study we found that, different converter fault types of the direct-drive wind power generation corresponding different fault appearance, a simple sensory judgment could not distinguish of fault type or locat fault element position. With the aid of neural network, we can effectively complete the fault diagnosis task, but the diagnosis speed is relatively slow and prone to local optimum problem. Through the genetic algorithm optimization neural network weights and threshold operation can better improve the two disadvantages. In addition to being supported by the National Natural Science Fund project task, the completion of this paper to other related fields to complete a certain reference value.