Power Dispatch With Guaranteed Convergence

This paper formulates and solves the economic power dispatch (ED) problem with practical operation constraints using potential games. Each generator operates as an independent player in a self-optimizing manner with marginal contribution utility functions to minimize the total generation cost. The proposed distributed formulation converts inequality constraints into feasible action sets, incorporates equality constraints by penalty functions, and extends to practical cases that exhibit non-convex or non-smooth objective functions.

Two learning algorithms with guaranteed convergence to Nash equilibria and/or optima are applied to solve the proposed formulation. How generators react as best responses to others is analyzed to capture the reasoning of operations. As a numerical example, the solutions obtained using the proposed ED method in a benchmark system are analyzed. Examples are provided to emphasize how priority for renewable sources are incorporated.