Skip to Main Content
Detection and location of single pole-to-earth faults on the fourth-rail power supply system at London Underground Ltd. can enable early preventive measures before the faults escalate into more dangerous pole-to-pole faults. Single pole-to-earth faults on the infrastructure are difficult to detect using conventional methods because of the small fault current compared to the load current and the wide range of possible fault resistance values. A new fault detection and location method for single pole-to-earth faults using the 600 Hz harmonic ripple from the substation is presented. Circuit analysis has been carried out to determine fault detection and location information from the 600 Hz harmonic data. Because of the non-linear characteristics of the traction system, the relationship of the 600 Hz component to the fault location is not straightforward, and therefore neural networks have been used to attempt to capture this function. Both simulation and field test results show that the neural-network-based method provides a way to locate single pole-to-earth faults.