The fault location problem has been tackled mainly through impedance-based techniques, the travelling wave principle and more recently by machine learning algorithms. These techniques require both current and voltage measurements. In the case of impedance-based methods they can provide multiples solutions. In the case of the travelling wave approach it usually requires high sampling and synchronized frequency measurements together with sophisticated identification algorithms. Machine learning techniques require training data and re-tuning for different grid topologies. In this work we propose a new fault location method based on the fault's transient intermediate frequency response of the system immediately after a fault occurs.

The transient response immediately after the occurrence of a fault is characterized by the travelling wave phenomenon together with intermediate frequencies of oscillation in the range of 5 to 500 kHz. These intermediate frequencies of oscillations are associated with the natural response of the cable/line system to the fault event. Their frequencies of oscillation are dependent on the faulted section and the fault location within that section. The proposed fault location methodology aims to leverage on that dependency, by firstly identifying these intermediate frequencies for different fault location scenarios for a given network. This process is performed offline using a linear time invariant (LTI) representation of the network. To compute this LTI representation, as part of this work an impedance representation in the modal domain is established for cable/line sections, which is able to capture the frequency-dependence and distributed nature of its electrical parameters. The offline methodology identifies these intermediate frequencies for different fault location scenarios, and then proceeds to fit the fault location dependence of each intermediate frequency using a polynomial regression.

An online methodology is also proposed to perform the fault location in real time by solving the polynomial regressions computed during the offline methodology using measurements of the intermediate frequencies present in the frequency spectrum of transient signals. The fault location is thus solved by using voltage or current measurements of the fault’s transient response at different locations in the network, together with simple signal processing techniques such as the Fast Fourier Transform. The full method is tested with an EMT simulation in PSCAD, using the detailed frequency dependent model for underground cables, together with realistic load models in a low voltage distribution network test system.