A neural networking based fault detection system for pitch and yaw control of a HAWT under different operating conditions

A fault detection system for the pitch and yaw control of a 1.5 MW horizontal axis wind turbine based on a data-driven technique is proposed. The impact of high ambient temperature and dust accumulation on the proposed fault detection technique was investigated. The system was modeled using the simulation programs FAST, TurbSim, and MATLAB under various operating scenarios and wind speeds. Pitch angle faults from −10° to 15°, blade imbalance faults from −3% to 7%, and nacelle-yaw angle faults from −10° to 20° were investigated. Furthermore, fault detection was considered under different operating temperatures. The Fast Fourier Transform (FFT) was applied to the Tower Top Deflection (TTD) data. Results show that the peaks of the TTD while applying faults to the blade are in the range of 5.09 to 21.42 Hz. YAW faults produce a single peak at 5.67 Hz while blade faults have two peaks. Moreover, the ambient temperature affects both the frequency and amplitude spectrum values. The TTD data were utilized to develop a Neural Network that can identify the errors in the pitch and yaw control systems. This network categorizes the faults into three categories with a best validation performance of 0.0086482.