A survey of 'modern techniques' for distribution feeder reconfiguration for loss minimization

Reducing energy losses in power distribution systems is important. Besides reduced energy losses, the economic benefits resulting from loss minimizati on include released generation, transmission and substation capacity, and d eferral or elimination of system expansion. As well, reducing losses also l eads to reduced feeder voltage drop and consequently improved voltage regul ation. One method of minimizing losses is to use existing switches to recon figure the distribution network to achieve minimum losses. However, the pro blem is a mixed-integer, nonlinear programming problem that presents a heav y computational burden for even a moderately-sized distribution system, and researchers have turned to "modern" optimization techniques to solve the p roblem. "Modern" optimization techniques include techniques based on artifi cial neural networks, fuzzy systems, expert systems, simulated annealing, g enetic algorithms and evolution strategies. These techniques represent comp utation paradigms based on a physical or biological metaphor, and are rapid ly gaining popularity among power system researchers as a result of theoret ical and application successes in many disciplines. In this paper, a brief overview of the theory of these techniques is presented, and then their use is examined for the problem of loss minimization through reconfiguration i n power distribution systems.