CONDITION MONITORING ARTEFACTS FOR DETECTING WINDING FAULTS IN WIND TURBINE DFIGs S. Djurovic S. Williamson The University of Manchester The University of Manchester Sinisa.Durovic@manchester.ac.uk Steve.Williamson@manchester.ac.uk P.J. Tavner W. Yang University of Durham University of Durham Peter.Tavner@durham.ac.uk Wenxiang.Yang@durham.ac.uk Abstract: Monitoring the condition of doubly-fed induction generators (DFIG) is growing in importance for Wind Turbines. This study presents the results of a comparison of DFIG steady state stator line current and instantaneous power when used as a means for generator condition monitoring, based on an examination of their frequency spectrum. For the purpose of this work, a detailed analytical model that makes it possible to simulate DFIG operation under a range of supply and winding balanced/ unbalanced operating conditions, was developed. Additionally, a purpose-designed DFIG test rig was built to facilitate the experimental validation of model results. The faulty machine current and power spectra are compared using experimental and model results. Keywords: Wind Turbine, DFIG, condition monitoring, current, instantaneous power, frequency spectrum. 1 Introduction At present doubly-fed induction machines are the dominant technology used for generators in variable speed wind turbines with ratings in excess of 1MW [1,2,3]. Monitoring the condition of such machines is an important aid to maintenance and improved reliability. Stator line currents and instantaneous power are commonly used as the basis of detecting electrical faults in induction machines [4,5,6]. The condition monitoring technique that is usually used in conjunction with these quantities is based on examining the frequency content of their steady state spectra and detecting the presence of tell-tale harmonic components or unexpected changes that may be linked to the presence of a fault within the machine. The aim of such an analysis is, if possible, to identify clearly defined fault-specific current and/or instantaneous power harmonic components that have the potential to constitute reliable fault indicators. The practical advantage of this approach is that it is non-invasive. The machine windings are used, in effect, as search coils, and only a small number of additional transducers is needed to establish an efficient data acquisition system. In order to monitor the condition of electrical apparatus it is essential to characterize the monitored signal so that the most likely faults can be unambiguously identified and appropriate action taken. The authors have developed a time-stepped coupled-circuit model expressly for this purpose. In this paper we compare current and power spectra for a DFIG operating with a stator open-circuit fault, using simulation data obtained from the time-stepped model. We also further substantiate the analytical findings with corresponding data measured experimentally on a specially- constructed test-rig. 2 DFIG Modeling Machine modeling in this work is based on coupled-circuit theory and takes into account the precise distribution of the winding conductors, as well as the instantaneous values of the line voltages for both stator and rotor. As such the analytical model is capable of representing winding unbalance and/or excitation unbalance on either the stator or rotor. Additionally, model considerations account for higher order air gap fields unlike most other conventionally used methods [8]. The set of model voltage and flux linkage matrix equations is defined as: