Artificial intelligence-based machine-learning system for thermal generator scheduling

SHAPES, an artificial intelligence-based, machine-learning thermal-generator scheduling system, for the run-up-to-peak period, has been developed as an extension of earlier work on a heuristic-guided depth-first scheduling algorithm. SHAPES incorporates new machine-learning algorithms, capable of automatically acquiring heuristic-search guidance, alleviating the need for heuristics to be manually provided to the original scheduling algorithm. Further enhancements have also been introduced in the new system, through the use of best-first search to explore the problem space instead of depth-first search. The paper reports on the development of SHAPES, and application studies which have been conducted to determine the effectiveness of its learning subsystem in improving search efficiency, as well as the performance of the new system in relation to the original scheduling algorithm