Resonant Overvoltages in Offshore Wind Farms Analysis, modeling and measurement

Abstract

In this thesis, the resonant overvoltages at the terminal and inside wind turbine transformers are studied. The motivation of this study is the requirement for improved transformers for wind farms, especially Offshore Wind Farms (OWF). Compared to conventional medium voltage transformers, these transformers should operate in more onerous conditions; switching operation, harmonics and spikes from converters at LV side, loading cycles and operation under fault conditions. In addition, access for condition monitoring and maintenance is restricted. The main sources of the resonant overvoltages are energization of transformers with short cables and close-up earth faults. In the case of Wind Turbine Transformers (WTTs), high frequency spikes from power converters can also cause resonant overvoltages, which is out of scope of this thesis. The first goal is to study resonant overvoltages at the terminal of WTTs. Both energization of a typical offshore wind farm and close-up earth faults are modeled and simulated. The model used for WTT (in this part) is a black-box model based on swept frequency admittance measurements on the transformer terminals. The other important components to be modeled are cables, power converters, Vacuum Circuit Breakers (VCBs) and platform transformer. It is found that resonant overvoltages occur when the frequency of a sustained voltage oscillation at the transformer MV terminals matches with a resonant frequency in the WTT voltage transfer function. Furthermore, this research shows that the appropriate protection scheme against resonant overvoltages is Resistive-Capacitive (RC) filter. The simulation results for a 11/0.230 kV 300 kVA transformer shows that energizing the transformer with a 23 m cable may lead to resonant overvoltage at the LV terminal with a rate of rise of voltage (du/dt) equal to 93 p.u./μs if there is not any type of protection installed on the LV terminal. The installation of surge arrester or RC filter can decrease the du/dt to 27 or 11 p.u./μs, respectively. Therefore, RC filter is superior to surge arrester as a protection measure against resonant overvoltages. Surge arresters just limit the overvoltage amplitudes while the rate of rise of overvoltages are still high and this can lead to internal resonant overvoltages as reported by other researchers. It should be noted that the main duty of surge arrester is to protect the transformer against lightning overvoltages. The other goal of this thesis is the investigation of internal resonances. Since the cable lengths in wind turbines are pre-known, the quarter wave frequency of these cables can be obtained. This frequency is the dominant resonant frequency during energization. In this way, the effect of winding designs on the internal resonant overvoltages can be assessed. A 500 kVA 11/0.23 kV with three different windings, i.e. disc, layer and pancake, on three limbs is designed and manufactured for this purpose. Frequency response of the windings for input HV and LV impedances and transferred voltages are measured and compared. The amplitude of transferred voltages to LV terminal at dominant resonant frequencies is more critical for layer and pancake windings compared to disc winding. Meanwhile, the transferred voltages inside disc winding have many resonant frequencies in the range 100kHz-1MHz. To draw general conclusions about resonant frequencies of these windings, an analytical lumped-parameter model is developed and verified with measurements. It is observed that the calculations results from the model follow the trend of the measurement for input HV impedance in the case of Layer winding. The resonant frequency at 10kHz is not represented since flux linkage with other phases is not included in the model. This resonant frequency is not also represented for disc and pancake windings. The shortcoming of the model result for layer winding is that the resonant frequencies at 100 and 300 kHz for the measured HV winding is presented with a shift in the calculated model at 50 and 130 kHz. Except from 10 kHz, the calculation result for HV input disc impedance is in good agreement with measurement for disc winding. The shortcoming is the shift in the resonant frequencies and slight higher amplitude of the calculated results. For the pancake winding, the calculated results for HV impedance has the same decreasing trend and same amplitude range of the measurement results. Same as disc and layer windings, shift in the frequency response is the drawback. The double resonant frequencies at 40 and 50 kHz in the measurement results is only represented with one resonant frequency at 30 kHz in the calculation results. In addition, the resonant frequency at 600 kHz is not represented in the calculation results. In the case of induced voltage to LV, the model results for layer winding are in good agreement with measurements for f<1MHz. But, for f>1MHz, it only follows the decreasing trend of the measurements. The dominant resonant frequency in 1.6 MHz is not well represented. In the case of disc winding, the model is capable of good calculation of induce LV amplitude for resonant frequencies around 100 kHz. It however deteriorates the trend of induced LV voltage for higher frequencies. In the case of pancake winding, the amplitude of resonant frequencies around 1 MHz is considerably lower than measurements and it requires improvement. Base on the FRA measurements performed for transferred voltages inside three winding types, disc winding has critical resonant behavior in the range 65 kHz to 1.5 MHz. In this range, there are several resonant frequencies where Disc winding reaction is such as standing wave with several nodes. The internal resonant frequencies for layer and pancake windings are sporadic. Considering the repair and accessibility costs of offshore wind farms, the design and characteristics of wind turbine transformers should be improved. Pancake windings with their modular designs can be proposed as a proper solution.