An expert system and dynamic programming hybrid for unit commitment

The authors propose a hybrid expert system/dynamic programming program for the solution of the unit commitment problem. The three concerns of schedule feasibility, program flexibility, and optimality are not satisfactorily achieved by either purely heuristic or numerical methods. A program which couples an efficient numerical optimization program with an expert system which can encompass the heuristic knowledge of an experienced scheduler is shown to be a possible solution to this problem. The significant features of the hybrid expert system thus developed are: (1) a user interface expert consultant, to make possible the use of the program by nonexpert users through its ability to guide the user through important decision-making processes in the problem setup period; (2) a knowledge base, containing expert knowledge of both schedulers and mathematical programmers, represented by if/then rule statements; (3) coordination of the mathematical programming algorithm input-output with expert knowledge base (rules) and data (facts); (4) a flexible program structure that is easily adaptable to changes in network configuration and system operation policy; and (5) the answering of user queries about any aspect of the problem setup phase.<
>