An evolutionary algorithm for a non-standard scheduling problem

The paper describes an evolutionary computation approach to solving a real-world scheduling problem for a group of production lines. The problem is non-standard in that it requires scheduling of process interruptions rather than processes themselves, and the schedule cost is determined by energy consumption on the machines. A scheduling system was developed to cope with instances of the problem on a daily basis. It consists of a heuristic procedure to generate initial schedules and an evolutionary algorithm to iteratively improve the schedules. Experimental evaluation of the system on numerous problem instances shows that, through proper scheduling, energy costs on the production lines can be substantially reduced.