Energy-efficient scheduling of a single resource with production requirements: a joint simulation and scenario-based mathematical programming approach

Motivated by paint oven on-off scheduling in an automotive plant, this study investigates the control of a single energy-intensive resource that can operate in on, off, warm-up, and multiple idle modes. The problem is to determine the energy mode of the resource over a finite planning horizon so as to minimise the average total cost of energy consumption and product waiting. We propose a data-driven on-off scheduling procedure that collects real-time upstream vehicle locations at the beginning of each shift, uses a digital twin to generate arrival scenarios conditional on these locations, and then determines the resource schedule. Two solution approaches are developed and used in parallel. The first is a scenario-based mathematical programming formulation strengthened by a scenario reduction technique based on clustering, and the second is a problem-specific genetic algorithm that encodes the resource schedule directly. Numerical experiments show that incorporating advance arrival information into multi-mode scheduling can reduce the normalised energy cost per vehicle by up to 41% relative to benchmark policies, with energy measured in equivalent on-mode operating minutes. The results highlight the value of digital-twin-based arrival information for energy-intensive manufacturing systems.