Transmission Line Fault Locator Using ANN

Ann can be used as a pattern recognition tools to estimate the fault location. The back propagation algorithm was successfully used for the training of the ANN. Different configurations were..investigated to reach the optimum number of layers and theoptimum number of neurons in each hidden layer. Itwas noted that increasing the number of neurons in the hidden layers to a specific number reduces the error and improve the network generalization. Increasing the number of neurons in the hidden layers beyond this specified number would increase the error and defect the generalization of neural network.CJ ANNs provide easy and fast estimation of the fault location on over head transmission lines, keeping its reach accuracy when faced with different fault conditions. This is an improvement in performance compared with traditional methods. It is found that the accuracy of the network estimation is mainly related to the network configuration as will as the training patterns. Neuro- Wavelet algorithm which use the Wavelet signal processing to process the simulated data (three phase voltages and currents from the relay end) in order to feed the neural network gave bad result than the preprocessing based Fourier transform algo~ithm. which indicat~ that F ~urier t~ansform is more reliable than Wavelet analysis in preprocessing the data for ANNs. A typical 220 k V double circuit doubly fed transmission line loaded from one end was simulated with various types of loads, vanous fault location, fault type, fault inception angle, and vanous fault resistance (time varying) to feed four different networks one for each fault type. The four different networks were tested using the data generated by the• distributed parameter and frequency dependent transmission line model. The error in estimating the fault location was within the acceptable range. The networks have a great capability of generalization for the unseen patterns. Due to the simplicity of the proposed algorithm, it can be used not only for fault location (off-line application), but also in the field of distance protection (on-line application). The tests of the proposed algorithm (off-line application) confirmed that the algorithm convergence is fast enough to satisfy the requirements of fast and reliable operation. Due to the fact that the majority of line faults is single line-toground fault concentration on the correct fault location for this type of fault was taken into consideration. The traveling wave theory of transmission lines together with Wavelet transform were used to get the fault distance, the transient signals (three phase voltages of the two ends of the line) are firstly decomposed into their modal components.