Research On Abnormal Condition Prediction Of Wind Turbines Based On Data

At present, the increase in the installed capacity of wind power and the phenomenon that wind is abandoned by wind farms seriously lead to the increase in operational and maintenance costs. Therefore, reduce maintenance and repair costs and power generation costs in the operation of the wind turbines are the urgent problem to be solved. Due to the state parameters of the wind farms are monitored in real time by the monitoring system, it is difficult to provide a basis to formulate plans for maintenance and repair of wind turbines. Therefore, the research on abnormal condition prediction of wind turbines is of great significance to plan for the wind turbines maintenance and repair scientifically. It can reduce the failure of wind turbines, guarantee the safe operation of the wind turbines, and increase economic benefits of wind farms.On the basis of the operation principle of doubly-fed wind turbines, the failure characteristics of the gear box and generator in wind turbines are analyzed, and the vibration characteristics of the gear box and generator are extracted by using wavelet packet theory. The operating conditions of wind turbines are divided by using k-means clustering algorithm. Based on this, according to the central-limit theorem, the criterions are established to evaluate whether the state characteristic parameters are normal or not. The operating condition identification model is established by using the support vector machine theory and the result of the sorted operating condition. The models are verified by model testing.Based on the operation data of wind turbines, the state parameters prediction models of the gear box and generator are established by three methods. By comparing with each model, the model that has better generalization ability is selected, and the models are verified by model testing.According to the prediction models and the sample data of the state parameters, the parameters of future changes are predicted. The parameters are devised into the appropriate operational conditions by the condition identification model. Based on the principles of the different operating condition, the prediction of the abnormal condition of wind turbines is realized by using the analytical hierarchy process and the fuzzy comprehensive evaluation method. The results can provide reference to the formulation of the maintenance and repair plan for wind turbines.