Partial Discharges in Voids at Variable Voltage Frequency and Temperature Diagnostic Testing of Stator Mainwall Insulation

Abstract

The main purpose of this thesis is to improve diagnostic testing by PD detection at variable voltage frequency by developing and experimentally testing theoretical models proposed for measurement of PDs in voids in a mica/epoxy generator bar insulation. The focus of this thesis is to compare estimated and measured values, and from that determine which physical mechanisms best describe the voltage frequency and temperature dependence of observed PD activity. The main contribution to theoretical modelling was to propose an impedance abc-model and include the dielectric response. Two types of test object were examined: 1) laboratory made, square-shaped insulation sheets using mica/epoxy tape, made with cylindrical voids with different diameter and gap distance and 2) 50 cm long sections of service-aged generator bars. A frequency sweep procedure for PD detection was developed, based on the relevant condition assessment standard IEC 60034-27 to study the frequency dependence experimentally. PDs were recorded during applying an increasing voltage of 10 equal steps, at reducing voltage frequency, starting at 300 Hz to 0.1 Hz for laboratory objects, and starting at 50 Hz to 0.1 Hz for service-aged bars. The temperature was varied in the range of 20 °C to 155 °C. Experimental testing compared to the proposed model showed that the capacitive abcmodel needed to be expanded to include both conductive elements and dielectric response; the impedance abc-model, to explain a frequency and temperature dependence for the partial discharge inception voltage (PDIV) and increased residual charge relaxation at high temperatures. The impedance abc-model was also expanded by a time lag and limited discharge area to describe the apparent charge magnitude. The capacitive abc-model describes PDIV for most cases. The effect of a high conductivity and dielectric response can explain the behaviour at low frequencies according to the impedance abc-model. The maximum apparent charge is frequency-dependent, which can be explained by PD ignition at an increased voltage caused by a time lag. Measured apparent charge magnitude indicates that the void area only partly discharges during each PD. This was observed to cause a distribution of PD magnitudes, instead of a few large PDs. The total apparent charge per period is, however, close to the expected values based on the void geometry, and is frequency independent below 100 °C. The total apparent charge per period is frequency-dependent at 130 °C and 155 °C. This dependency was here modelled by an increased relaxation of residual charges, which enabled more PDs to occur during the same period. The relaxation was modelled by the material dielectric response. The impedance abc-model fits to the experimental results for both insulation systems containing cylindrical voids and generator bars. This includes the important fit to the lowered PDIV and the increased total apparent charge due to an increased dielectric response at high temperatures. A correlation was found between electrical PD measurements and results from microscopy investigations of cross-sections of the generator bars. A higher PD activity was correlated to a larger void content compared to bars with lower PD activity. This correlation to void geometry is in line with results from test objects with known void geometry. Measured material properties, estimated time lag, and estimated void conductivity were used to describe the voltage frequency and temperature dependence.