A Resonant Fault Current Limiting Prediction Technique based on Auditory Machine Intelligence

Abstract

Faults are a major problem encountered by power system operators particularly single-line-to-ground faults. To mitigate such faults and assure enhanced services to consumers, power system operators need to deploy appropriate hard and soft-computing solutions. In this paper, we present a novel approach to fault mitigation based on a new type of artificial intelligence technique dedicated to time series prediction called Auditory Machine Intelligence (AMI). The actual fault mitigation approach uses a Resonant Fault Current Limiter (RFCL) to fine-tune inductances in circuit in order to estimate the clearance times for a fault. The fault mitigation approach is cast as a time series problem where the resonant inductances (L) and associated clearance times (tc) are re-sequenced in a temporal aggregated fashion; this approach is then applied to a double-circuit transmission line (Alaoji-Afam sub-transmission) of the Nigerian power network. The results using the proposed technique on a generated L-tc sequence are compared with that of the Group Method of Data Handling for time series (GMDH time-series) which is a state-of-the-art neural network; the results indicate that the both techniques are competitive but the AMI technique will outperform the GMDH time-series to the tune of 0.57% for a number of GMDH time-series and AMI equal simulation trials.