Design And Implementation Of Wind Turbine Fault And Early Warning Recognition System Based On Condition Monitoring

Wind energy,as a renewable and non-polluting energy source,has attracted a lot of attention.In recent years efforts have been devoted to increasing the efficiency of wind energy utilization and power output when the problem of energy shortage is becoming more serious.In addition,wind farms are normally located in remote areas where the condition is relatively harsh and the wind turbines are tens of meters from the ground,which contribute to the high maintenance cost of wind farms.Based on the real-time monitoring data in combination with the analysis of historical operating data of wind turbines,the paper presented a computational method for wind turbine fault detection and early warning trigger for unit status through data mining and neural network analysis,which could be used in identifying faults of frequent occurrence and high reparation cost and early warning and intervening of output power and temperature of the turbines.In this paper,the k-Nearest Neighbor algorithm is used for model training on data related to wind turbine blade icing failure.In addition,the variance within the time window is used to characterize data fluctuations caused by blade structure changes,and TF-IDF is used to amplify the "waves" to achieve the purpose of determining whether the current wind turbine blade has broken.Based on the time series data of wind speed,the generated power of the wind turbine is predicted,and a regression model for real-time prediction of the generated power of the unit based on the wind speed is obtained.At the same time,the Convolutional Neural Network model is used to predict the mechanical state of the wind turbine,the generator winding temperature is predicted by the characteristics of the unit,and real-time warning is provided.The system has been approved to use and received positive feedback from the maintenance personnel of the wind farms.With the help of this system,the maintenance personnel was able to resolve the faults faster and prevent them from happening more effectively.