On incorporating variable consumption functions within energy-efficient parallel machine scheduling

Abstract: The increase in non-renewable energy consumption and CO2 emissions, especially in the manufacturing sector, is moving radical shifts in energy supply policies and production models. Renewable energy integration and regulated pricing policies require new and effective scheduling strategies, as highlighted by the emerging field of energy-efficient scheduling. In this paper, we aim to contribute to this field by addressing a scheduling problem where a set of jobs must be allocated to a set of machines over a discrete finite horizon and variable energy consumptions are required for job execution. Energy can be obtained by a renewable source or through transactions on the market. The goal is to minimize the total energy costs from the grid while scheduling all the jobs within the time horizon and adhering to an energy limit per time period. We introduce a novel time-indexed Mixed Integer Linear Programming (MILP) formulation capable of handling variable energy consumption functions, surpassing traditional models that assume constant energy usage of jobs. We then develop a matheuristic algorithm based on an Iterated Local Search (ILS) framework that exploits the MILP formulation for large neighborhood searches. We tested more than 200 instances with up to 200 jobs, 35 machines, and 120 time slots. The results show a good performance of both our methods and highlight the advantage of using the ILS when jobs are characterized by variable consumption functions.