This paper presents an adaptive neuro-fuzzy inference system (ANFIS) for vo
lt/var control in distribution systems. The control objectives are minimiza
tion of system losses without violating the voltage security of the system.
Power loss and voltage sensitivities obtained at the end of the base case
load flow are used to determine the dominant inputs to the fuzzy expert sys
tem. The rules are adapted using a neural network. Adaptability of rules co
ntributes to qualitative comprehension of the inference process by a human
operator. Application to a practical 30-bus system illustrates the economic
benefits and numerical results achievable through the fuzzy connectionist
scheme. (C) 1999 Elsevier Science S.A. All rights reserved.