Modelling, Simulation, and On-line Detection of Rotor Fault in Hydrogenerators
MetadataShow full item record
- Institutt for elkraftteknikk 
Hydropower generator is a critical component in hydropower stations and is therefore subject to a preventive maintenance strategy to avoid failures of the generator's components. Fault detection techniques can be used to assess the technical condition of the generator either through on-line or off-line detection methods. Off-line measurements are carried out periodically during scheduled inspections and revisions, and require stop of the generator to perform the detection method. Available on-line detection methods require often installation of specific measurement equipment inside the generator. Therefore, non-invasive methods for continuous monitoring, with no need for installation of new measurement equipment, would be useful to continuously get information about the generator condition. By detecting a fault in early stages, problems can be improved at a lower cost and with a shorter downtime. The aim of condition monitoring is to offer the plant owners more information about the generators and provide a more effective production. This master thesis investigates fault detection of hydro generator rotor winding turn-to-turn short circuits. In this report, a new method for fault detection and continuous monitoring has been developed and tested through modeling of Kalvedalen hydropower generator in the simulation software ANSYS Maxwell. The modeling of Kalvedalen generator is based on technical drawings and data from Eidsiva Energi. Various fault severity cases are modeled in the simulation software, to test the new proposed fault detection method. The simulations are carried out at no-load and full-load operation. First an initial study was carried out, were RMS values of terminal voltages are used for fault detection. The model was simulated for the six fault cases and the induced voltage waveforms were plotted. It was observed that the short circuit faults reduce the output voltage. However, the changes in induced voltage are considered as too small to detect a fault at an early stage, although the method could be used to detect faults were large part of the windings are shorted. The main proposed method utilizes the frequency spectrum of induced voltage and stator current waveforms. By studying the induced voltage and stator current frequency spectrum, it was discovered that fault related signatures appear when short circuits are introduced. For a healthy generator, the frequency spectrum consist of harmonic order with considerable amplitudes at the electrical frequency and its odd multiples. When a fault is introduced, fault related harmonics appear as side-band harmonics on both sides of these original healthy frequencies. It was further discovered that the fault harmonics follow a certain pattern, which can be calculated based electrical frequency and the generator pole pair number. The fault severity could also be found by utilizing the amplitudes of the side-band harmonics, as they increase with fault severity. Based on modeling and simulations in ANSYS Maxwell, it was concluded that induced voltage and stator current could be used for detection of rotor winding short circuits and to determine the fault severity.