Hybrid Neuro-Fuzzy System for Optimal Generation scheduling in Electrical Power systems

Abstract

The real time controls at the central energy management center in a power system, continuously track the load changes and endeavor to match the total power demand with total generation in such a manner that the operating cost is least. Conventional optimization techniques are cumbersome for such complex optimization tasks and are not suitable for on-line use due to increased computational burden. This paper proposes a neuro-fuzzy power dispatch method where the uncertainty involved with power demand is modeled as a fuzzy variable. Then Levenberg-Marquardt neural network (LMNN) is used to evaluate the optimal generation schedules. This model trains almost hundred times faster that the popular BP neural network. The proposed method (Hybrid Neuro Fuzzy System) has been tested on two test systems with six and thirteen generating units and found to be suitable for on-line economic dispatch.