Research And Study Of Fault Diagnosis System For Wind Turbine

Wind turbine which is installed in high place, due to its bad working conditions, itsfailure is not convenient to be repaired. According to the bad working conditions of windturbine, we can predict its faults effectively, timely, and accurately, avoid or reduce faults,and even to reduce the repair cost. As to the whole wind driven generator’s system, thegenerator is most easy to have failures, so we can begin the research of faults from thegenerator firstly. The generator’s structure is complex, and we are required to accuratelyjudge the working conditions, the quality of its performance, the changes of its parametersand the trend of its faults under the conditions that the equipment is not torn down ordestroyed and the equipment is in normal operation, thus we need a set of state detection andfaults diagnosis system which has a perfect function and a stable performance. Because thegenerator is rotary machine, rotary machine will causes vibration when it at the high-speedoperation. Through the real-time monitoring of the vibration signal, the system will be able toidentify the status from the vibration signal.Like a set of system suitable for faults diagnosis of wind turbine, firstly, the sensors isused without prejudice to the operation of the machine, sensors should be installed as close tothe source of vibration as possible, from various angles and direction to choose the measuringpoint and the position, to measure the common faults of generators (such as rotor’smisalignment, axial crack and oil film oscillation) by measure the vibration displacement, andto do the faults vibration signal simulation. Secondly, to analyze the vibration signal by usingthe MATLAB wavelet analysis methods, through this way we can complete noise separationby ignoring the useless noise signal and retaining the useful signals. Through the energyspectrum decomposition of the wavelet packet, work out the percentage of each band in totalenergy, and they can be the characteristic parameters in the faults diagnosis. Thirdly, take theenergy characteristics of the MATLAB wavelet packet decomposition of each band as theinput sample of BP neural network. Through the excitation function of hidden layer, conductneural network training, to get the fault type of output. By selecting test samples for networktesting to achieve the effective identification of test samples, this BP neural network trainingis simple and feasible to identify the type of faults. Finally, by doing Visual Basic program todesign the interface of the wind driven generator’s faults diagnosis and vibration testingsystem, which can be divided into two type systems, namely the user system and themanagement system. The user system can be used for vibration signal acquisition and analysis,and to extract the useful signal to judge the type of faults, and to output diagnosis results. Themanagement system can view and delete the Access database and print the final results. Vibration method is used for faults diagnosis system of the wind driven generatordeveloped in this dissertation, can simply predict the occurrence and development of itsfaults, and can reduce the economic losses effectively, and also has a certain of referential andguiding value for the mature faults diagnosis system of the wind driven generator in thefuture.