Fault location based on wavelet energy spectrum and neural network for ±800 kV UHVDC transmission line

The inherent frequency of fault traveling wave is mathematically associated with fault distance and its transient energy containing rich information about fault distance is concentrated around this frequency. Because of its fitting capability for non-linear function, an ANN(Artificial Neural Network) model of HVDC line is built to locate its faults. Based on the equidistant characteristic of wavelet transform, the transient energy spectrum of line voltage modulus at one end is extracted in seven scales, which are used as the samples to train and test the ANN model. The proposed method takes the inherent frequency band, instead of point, to extract fault information, which is easier and more reliable. Results of digital test show faults at any line position and with any transition resistance can be accurately located.