Fault Detection and Isolation in Wind Turbines: Type-3 Fuzzy Logic Systems and Adaptive Random Search Learning

Ensuring the reliability of wind energy conversion systems (WECSs) is a crucial task for maximizing energy capture from the wind. A detailed model incorporating mechanical and electrical components is essential for accurately diagnosing system errors and assessing their impact on subsystems. Additionally, a fault detection and isolation system is necessary to quickly identify recurring faults and prevent significant economic losses. This study introduces a fault detection and isolation system using dynamic model of WECS based on type-3 (T3) fuzzy logic systems (FLSs). The adaptive random search (ARS) is employed to optimize the T3-FLS parameters and structure for enhanced fault detection accuracy. T3-FLSs handle higher levels of uncertainty and variability compared to traditional FLSs and neural networks. This allows for more accurate fault detection in complex and dynamic systems. One T3-FLS model replicates the system’s normal operation, while another simulates faulty conditions. These T3-FLS models are run in parallel with the actual plant, allowing for comparison of their outputs with the real system’s outputs to pinpoint error timing and location. The ARS is utilized to train the T3-FLSs, eliminating the need for gradient expression calculations. The appropriate number of rules for the T3-FLS is determined using Akaike and final prediction error criteria. Simulation results demonstrate the system’s ability to rapidly detect and isolate errors with minimal false alarms. This research framework can be applied to identify errors in various system components effectively.