Artificial intelligence algorithm for daily scheduling of thermal generators

Abstract
An artificial-intelligence-based algorithm is developed for scheduling thermal generators in a 24-hour schedule horizon. The primary requirement of meeting the daily system load demand and spinning reserve, the ramp rate characteristics and minimum up and down times of units, and the operational constraints including the crew constraints, station synchronisation intervals, station shut-down intervals and the allowable operating limits of units, are incorporated in the algorithm. The scheduling algorithm consists of an algorithm for the run-up-to-peak period, an algorithm for the period between the midday and midafternoon peaks and an algorithm for the end-of-day decommitment period. The algorithms are based on the heuristic-guided depth-first search technique. A ‘look-ahead’ and ‘constraint satisfaction’ technique is also developed in the algorithm for the scheduling period between the midday and midafternoon peaks. The combinatorial explosion in the search space of the scheduling problem is greatly reduced by the use of the heuristics derived, some of which are based on economic considerations. A method for further reducing the search space to a manageable size is also developed. In addition, a new and recursive algorithm for solving the power dispatch problem is developed. This algorithm is employed in the 24-hour scheduling algorithm in the process of evaluating the costs of the generated schedules. The application of the developed scheduling algorithm to schedule ten thermal generators, implemented in Prolog, is presented.

This publication has 1 reference indexed in Scilit: