Filter Based Fault Diagnosis of Wind Energy Conversion Systems Subject to Sensor Faults

Abstract

The operational reliability of wind energy conversion systems (WECSs) has attracted a lot of attention recently. This paper is concerned with sensor fault detection (FD) and isolation problems for variablespeed WECSs by using a novel filtering method. A physical model of WECS with typical sensor faults is first built. Due to the nonGaussianity of both wind speed and measurement noises in WECSs, an improved entropy optimization criterion is then established to design the filter for WECSs. Different from previous entropyfiltering results, the generalized density evolution equation (GDEE) is adopted to reveal the relationship among the estimation error, nonGaussian noises, and the filter gain. The sensors FD and isolation algorithms are then obtained by evaluating the decision rule based on the residual signals generated by the filter. Finally, simulation results show that the sensor faults in WECSs can be detected and isolated effectively by using the proposed method.