Fault diagnosis of three-parallel voltage-source converter for a high-power wind turbine

A fault detection method for the switch devices of three-parallel power converters in a wind-turbine system is presented. The proposed method utilises the measured three-phase currents (i^s_a,i^s_b,i^s_c) that were already used for controlling the converters. Additional current and voltage sensors are thus not necessary. Consequently, this particular feature potentially cuts costs to apply the detection algorithm. The three-phase currents are transformed to a stationary reference frame (i^s_d,i^s_q). This reference frame possesses specific patterns in accordance to the conditions of the switch devices in the converter. Faulty switches can be detected by analysing the obtained pattern and pattern recognition is achieved by a neural network that has a learning capability. The proposed method is primarily able to diagnose an open fault switch in the converter. The reliability of the proposed fault detection method was proved by simulations and experiments using a 10 kW simulator.