Contents I. Introduction
In a complex power system, natural calamities and unexpected events are all possible causes of faults. To ensure a reliable power supply, the location of fault section is anticipated to grasp in the early stage. Conventionally, this task can be completed by retrieving the statuses of relays and circuit breakers (CBs) from supervisory control and data-acquisition (SCADA) systems. However, once CBs fail to operate or face the multiple faults along with the high stresses to interpret the voluminous alarms, the work may become tremendously difficult. Several reasoning-based methods, such as expert systems [1]–[3]; artificial neural networks (ANNs) [4]–[7]; cause–effect networks [8]; Petri nets [9], [10]; and Bayesian networks [11] have been applied. Yet, not only are the knowledge acquisition and the database establishment burdensome, but the confirmation of knowledge sufficiency is also laborious to accomplish [12]. Recently, metaheuristic algorithms, such as genetic algorithms (GAs) [13]; evolutionary programming [14], [15]; tabu search [16]; immune algorithms [17]; ant systems [18]; and particle swarm optimization [19] have been successively proposed. They are further followed by the development of hybrid methods, by which various methods can be integrated to find a feasible solution with higher efficiency [20]–[27].
