Particle Filter-Based Estimation of Instantaneous Frequency for the Diagnosis of Electrical Asymmetries in Induction Machines

Abstract

Fault diagnosis of induction machines operating under variable load conditions is still an unsolved matter. Under those regimes, the application of conventional diagnostic techniques is not suitable, since they are adapted to the analysis of stationary quantities. In this context, modern transient-based methodologies become very appropriate. This paper improves a technique, based on the application of Wigner Ville distribution as time frequency decomposition tool, using a particle filtering method as feature extraction procedure, to diagnose and quantify electrical asymmetries in induction machines, such as wound- rotor induction generators used in wind farms. The combination of both tools allows tracking several variable frequency harmon- ics simultaneously and computing their energy with high accu- racy, yielding magnitudes and values similar to those obtained by the application of the fast Fourier transform in stationary operation. The experimental results show the validity of the approach for rapid speed variations, independently of any speed sensor.