Fault Diagnosis of Pitch Actuator Using Improved Polytope Observer

We investigate the fault diagnosis of nonlinear wind-turbine systems with disturbance and noise, and design an improved polytope unknown input observer to diagnose faults in the pitch actuator. First, to decouple the interference term, we design a polytope unknown input observer based on a linear variable parameter model of the wind turbine. Next, considering the influence of noise, we use a Kalman filter algorithm to enhance robustness to noise. Then, we used the mean square method to design a threshold for evaluating the residual, which determines whether or not a fault occurs. Lastly, we verify the performance of the improved polytope hull-type unknown input observer by simulating the fault residual signal of the wind-turbine pitch-actuator system.