CMAC neural network based network reconfiguration for loss minimization in distribution networks

Network reconfiguration of distribution systems is an operation in configuration management that determines the switching states for a minimum loss condition. Neural networks have the capability to map the perplexed and extremely nonlinear relationship between loads and system topology. This paper is intended to propose a method for network reconfiguration based on a cerebellar model articulation controller (CMAC) neural network. The trained CMAC network determines the relationship between the various load patterns and the according topology with minimum losses. The proposed method is tested to a 16-bus test system. Test results indicate that the developed method can provide accurate and fast configuration predication for minimum losses.