Surender K. Yellagoud and Purnachandra R. Talluri
Fault location, electric distribution lines, impedance-based algorithm, artificial intelligence-based networks
Artificial intelligence (AI)-based tools were successfully employed for automated fault location in medium voltage (MV) distribution networks by many researchers to date. This paper presents a comparative evaluation of some popular AI-based tools – feedforward neural networks (FFNN), radial basis function neural networks (RBFN) and adaptive network-based fuzzy inference system (ANFIS). The power system faults were triggered at various locations of a MAT- LAB simulated IEEE distribution network to generate training and testing fault data. ANFIS demonstrated a better fault location accuracy compared to other tools. A classical impedance-based algorithm is also presented in this work to surface its multi-estimation setback for multi-terminal distribution networks, which in contrast demonstrates the advantage of modern AI-based network models over classical electrical circuit-based network models. The detection capabilities of fault type classification of feedforward neural net-works were better than other tools. RBFN models exhibited faster learning than other tools. The automated fault location enables the substation maintenance crew to preliminarily estimate the locus of fault event sitting in the substation and then can undertake swift search for fault landmarks. Thus, automated fault location with higher accuracy reduces outage duration for affected customers and amount of power drainage into fault sinks, thereby contributing towards power system reliability and economy.
Important Links:
Go Back
