A Data-Driven Fault Location Algorithm Based on the Electromagnetic Time Reversal in Mismatched Media

The classical electromagnetic time reversal (EMTR) fault location method in power systems requires multiple simula-tions in the backward step, assuming different guessed fault loca-tions along the line. This process can be time consuming especially when a high location accuracy is desired, as it may require a significant number of guessed fault locations. To cope with this issue, the concept of EMTR in mismatched media has recently been introduced, allowing reducing the backward simulations to a single run and, thus, substantially improving the computation efficiency of the EMTR-based fault location technique. In this paper, we present a detailed study of the mismatched-media-based mirrored minimum energy property. This property has been applied in a few recent studies but never been theoretically studied and rigorously demonstrated. First, we infer a transfer function that relates the fault source to the voltage along the line resulting from back-injecting the time-reversed transients measured at a given observation point. We present a theorem according to which, at the fault switching frequency and its odd harmonics, the mirror-image point of the fault location with respect to the line center corresponds to a local minimum of the squared modulus of the transfer function. Then, it is proved that the mirrored minimum energy property is a corollary of this theorem. Based on these theoretical findings, we propose an algorithm that utilizes the reversed-time voltage energy as a fault location metric in the frequency domain. We further advance a data-driven strategy to maximize the computation efficiency of the fault location procedure. The applicability and robustness of the proposed frequency-domain fault location metric are numerically and experimentally validated.