Abstract
In this paper, we present the design of a Very low frequency (VLF) generator made of power electronics converter coupled to a mechanical system. This device is used to detect and locate faults in cables. This VLF test is an AC type test at 0.1 Hz of a cosine-rectangular waveform performed on an underground cable. Simulations were carried out for the following cable faults: open-circuit, short-circuit, resistance fault and spark gap. For each fault type, simulations are performed for different locations in order to collect databases in a neural network relating the distance and the corresponding voltage from the fault location. This allows to register the range of voltages variation in each fault, which is useful for its identification. In addition, these databases are used to determine the fault location using the Spline interpolation method. The tests were performed on a 20 km cable length. The obtained results show the high performance and efficiency of the investigated methods in terms of cable fault identification precision and localization.