Research On Fault Diagnosis For Large-Scale Wind Turbines

Facing the crisis of global energy shortage, the wind energy due to its clean and renewable has been rapid development. Although the wind energy has rapid development, but produced a series of problems, such as the uncertainty affected of wind speed and ambient temperature, make the wind components often fault, the maintenance is high costs. For the pitch system higher failure rate, based on SCADA system data, establish online monitoring outlier identification model determining pitch systems state. On this basis, the establishment of the pitch system fault diagnosis model for accurate fault location.By analyzing the pitch system works, combined with the SCADA system related operating parameters. Using Relief algorithm for fault feature parameters extraction,using the similarity function to eliminate redundant data.on this basis, based on nonlinear state estimation(NSET) method to establish health model to cover all the normal pitch system operating conditions. using sliding window residuals mean the statistical analysis method, eliminate the influence of outlier disturbances and uncertainties in the process of plant operation.and finally, to validate the model's feasibility in the pitch system status monitoring outlier identification.Based on the results of online monitoring outlier identification model, use of pitch system fault diagnosis model achieve accurate positioning of the fault. Used BP neural network and support vector machine established fault diagnosis model of pitch system,compared the diagnostic accuracy, to verify the superiority of the BP neural network fault diagnosis for pitch system model.Online monitoring outlier identification technology,reducing the wind turbine failure rate, to ensure the safe operation of the wind turbine; Outlier fault diagnosis system, it can accurately located the fault reason,provide reliability technical guidance for the maintenance personnel, reduce the equipment repair time and maintenance costs.