This thesis investigates the effects of winding faults on the doubly fed induction generator (DFIG) operation. A DFIG is a wound rotor induction machine that is most commonly used in present day wind power applications. In combination with a back-toback converter in the rotor circuit the DFIG is capable of generating power at constant frequency while operating at variable speed. The aim of the research in this thesis is to develop analytical and experimental tools that would enable the work on establishing of DFIG condition monitoring techniques. DFIG winding faults are examined using the machine current signature analysis (MCSA) method. This method is based on exploring the machine current frequency spectra, and defining and monitoring the fault specific changes in the spectra harmonic content. For this purpose a detailed analytical machine model is developed in this work. The model is capable of representing the effects that various winding faults have on the current spectra and is based on the generalized harmonic analysis. Higher order m.m.f. ~ ·harmonics are taken into account in the model calculations. Combining the model with a connection matrix enables the machine current frequency analysis for various healthy and faulty DFIG winding configurations. The model equations are solved in a time stepping algorithm based on Euler's method. For the purpose of experimental work a DFIG laboratory test rig was built. The test rig is designed to enable the simulation of various DFIG operating regimes and winding faults. The test machine stator winding terminals are taken out to a specially constructed terminal box, where by simple reconnecting of the terminals different winding configurations are achieved. The effect that open-circuit and short-circuit faults have on the DFIG current spectra harmonic content, when compared to healthy operating conditions, was analyzed in both the analytical model and on the test rig. The comparison of the model results against the experimental results shows that they are in very good agreement, thus verifying that the developed model is capable of predicting the real life operating conditions that are present in the simulated system. Some fault specific harmonic components are defined from model predictions and corresponding test rig experiments for each of the winding fault scenarios and DFIG operating conditions analyzed. !tis shown that the frequency of the fault specific harmonic components is dependant on the DFIG operating speed.