Fault location in Transmission System based on Transient Recovery Voltage using Stockwell transform and Artificial Neural Networks

Abstract

Fault location creates specific technical challenges for power system protection, providing reliability in Transmission Systems (TS). This paper presents a new approach for locating faults in multiterminal buses. The method is based on the application of the Stockwell transform (ST) on Transient Recovery Voltage (TRV). TRV appears between the poles of a Circuit Breaker while isolating a fault due to the change of impedance, and it contains a wide-band frequency information. ST facilitates high accuracy in frequency, and therefore, it is possible to compute the energy content in frequency dimension associated with the system. Dominant frequency and its associated energy value are obtained from energy analysis, which are closely related with the fault distance. Thus, a general-purpose algorithm, as Artificial Neural Network (ANN), is used to accomplish this location. This ANN was optimised using a hyper-parameter algorithm that select optimal configuration of the network. Hence, this new approach applies the precision of ST frequency analysis to a well-known and easily measurable TRV magnitude. Robustness of the features used in the proposed approach has been studied using a sensitivity analysis. Finally, the method has been tested on an TS IEEE 14 bus model with less than 1 second execution time.