Intelligent Fault Diagnosis of Wind Turbines Using Adaptive Fuzzy Threshold

Wind turbines are exposed to a variety of faults some of which can cause irreparable economic losses. Therefore, identifying the faults in a short time, ensures the correct operation of the system and prevents the mentioned losses. In this paper, using a dynamic model for wind turbines which includes mechanical and electrical parts with appropriate details, an intelligent fault detection and isolation system is designed utilizing recurrent neural networks. The proposed system can identify the occurred faults in pitch sensors and pitch actuators. Then, in order to consider the robustness of the system, it is suggested to use an adaptive fuzzy threshold in decision making block. Simulation results for the fixed threshold, robust thresholds, and the proposed adaptive fuzzy threshold validate that the suggested adaptive threshold reduces the detection time. In addition, the number of false alarms, and the number of missed ones are reduced by using the intelligent fault detection system.