Intelligent Fault Diagnosis Methods Of Hydro-generating Unit And Vibration Trend Prediction Research

With the rapid development of hydropower, hydropower has attracted more and more attention of the world due to its character of flexibility of the output adjustment in the electrical industry and enormous potential resources. Hydro-generating unit capacity become lager and the proportion of the power system have shown a trend of rapid growth. Hydro-generating unit have always been regarded as the key equipment in the hydropower enterprises,the health status and safety operation of the unit has a close relationship with hydropower enterprise security and economic benefits,unexpected fault will cause serious security accident and may lead to huge economic losses. Therefore, security, reliability and stability of hydro-generating units not only the goals of hydropower enterprises pursuit, but also the hotspot of many scholars researched.Hydro-generating unit fault diagnosis and vibration trend prediction can provide technical support to grasp the status of the unit,and they also can reduce the accident rate and help reduce economic loss. On the basis of the current research of this area, this paper base on the principle of vibration,applied fuzzy integral fusion of multiple classifiers method for hydro-generating units fault diagnosis and General Regression Neural Network(GRNN).optimized by intelligent algorithm for hydro-generating units vibration trend prediction, the main work includes the following two parts:The first part is based on fuzzy integral hydro-generating unit fault diagnosis.This paper extract the vibration fault feature of hydro-generating unit by wavelet packet energy algorithm.A fault diagnosis model based on fuzzy integral combination of multiple classifiers is introduced,a weighted method improved Naive Bayesian classifier, applied Mahalanobis distance classifier and BP neural network as a sub-classification,use the fuzzy measure represented the interaction of sub-classifications,through the fuzzy integral comprehensive diagnostic results of the three sub-classifier to get the final diagnostic results. The simulation experiments shows that the fusion diagnostic model not only has higher diagnostic accuracy,but also can improve the diagnostic model generalization.The second part is a study on vibration trend prediction by GRNN which optimized by intelligent algorithm.This section applied the chaos theory of time series prediction to the vibration trend prediction of hydro-generating unit, the phase space reconstruction,embedding dimensions and time lags problems are discussed. A vibration trend prediction model base on GRNN which distribution parameters (spread) optimized by fruit fly optimization algorithm(FOA) is prosed. The mean square error of predicted results compared the BP shows that this model has higher stability and accuracy,can reflect signal trend accurately.This model proposed a new method of hydro-generating unit vibration trend prediction and provide a strong technical support to grasp the changes state of the hydro-generating unit.