Diagnosis of Multiple Open-Circuit Switch Faults Based on Long Short-Term Memory Network for DFIG-Based Wind Turbine Systems

Abstract

Effective fault diagnosis for power electronic converters is an essential and mandatory operation to reduce failures and unscheduled shutdowns. This paper presents a data-driven fault diagnosis method, using long short-term memory (LSTM) network, to detect multiple open-circuit switch faults of the back-to-back converter in doubly fed induction generator (DFIG)-based wind turbine systems. Twelve sensor signals of the back-to-back converter for different fault types are measured. Wind speed fluctuation and sensor bias faults are considered as interference factors. The method has been evaluated in a grid-connected DFIG simulink model. Simulation results have demonstrated that, compared with least squares support vector machine (LS-SVM), back-propagation artificial neural network (BPANN) and recurrent neural network (RNN), the proposed method excavates the deep information of the fault signal with the highest diagnosis accuracy and strongest robustness with short time delay. Experimental tests undertaken on a hardware-in-the-loop testing platform further validate the effectiveness of the LSTM-based network.