Fault diagnosis for wind turbine systems using a neural network estimator

Faults in dynamic systems are caused basically by malfunction of actuators, sensors or other components in the system. In this work a neural network (NN) estimator is used for diagnosing the wind turbines(WT) sensors faults. Radial basis function (RBF) is the type of NN used here. This is because of the ability to approximate a nonlinear input into a linear output. The RBF is trained using sample data collected during a fault free operating condition. The benchmark model has three sensor faults simulated. The proposed method after being applied to the benchmark model was effective as the residual signals were all sensitive to the three sensor faults. The three sensor faults were also isolated as would be in the simulation results.