Earth Fault Distance Computation Methods Based on Transients in Power Distribution Systems

The most common fault type in MV distribution network is single line to earth fault. The initial transients of earth faults are important especially for unearthed and compensated neutral networks. The earth fault transient signals consist of many different frequency components, which result from charging and discharging of the network capacitances. The transient components provide valuable information for fault location purposes. The charging component has higher amplitude and lower frequency than the discharge component and hence is more suitable to be used for fault location purposes. In this thesis, we discuss algorithms to locate an earth fault in unearthed or a compensated neutral MV networks using the information of the measured transient signal. The networks considered are assumed to be radially operated and they are modeled using Electromagnetic Transient Program-Alternative Transient Program (EMTP-ATP). Five types of fault location algorithms have been developed which are called general model (GM) algorithm, exact model (EM) algorithm, continuous wavelet transform (CWT) based method, multiple regression analysis (MRA) based method and artificial neural network (ANN). GM algorithm is developed based on a simplified model of symmetrical components while EM algorithm is developed with exact "pi"-model of symmetrical components. Both algorithms utilize the frequency of charging transient to estimate the fault distance. CWT based algorithm requires both voltage and current of transient signals to estimate the fault path inductance. MRA and NN algorithms were developed using the transient signal measured from the secondary side of the MV/LV distribution transformer. In addition, an algorithm to find a correct path towards the position of fault in network which has many branches is presented. The results from intensive simulations and experiments in actual distribution networks are also presented in this thesis. The results are analyzed using signal processing techniques. The algorithms apply continuous wavelet transform (CWT) to locate the dominant charge transient frequency and extract the specific coefficient corresponding to the charge transient frequency. In this thesis, the properties of Hilbert transformation (HT) are used to estimate the damping attenuation of the transient signal. Finally the performance of the proposed fault location algorithms is evaluated and the results are compared. Based on the simulation results, it is found that the proposed algorithms work at a reasonable level of accuracy. The results from real experiment data show that both CWT and GM algorithms have a comparable result.