Metaheuristic Simulation-based Production Planning for Energy Efficiency: A Case Study

Abstract

Modern industrial production planning and control (PPC) systems are responsible for supporting planning decisions on how to optimally produce a given set of products while minimizing costs and retaining production constraints, such as delivery tardiness or offtimes. In recent years, more and more attention has also been paid on energy efficiency as part of production optimization, resulting in competing optimization targets. In order to solve such complex multi-objective scheduling problems in practice, metaheuristic methods are used because of their ability to deliver acceptable solutions in feasible time. In this paper, we demonstrate the application of a General Variable Neighborhood Search (GVNS) metaheuristic on a case study of flow shop scheduling in an industrial bakery in different scenarios and study the effect of different energy prices on the planning result. The case study features a simple production line with thermal processes for baking and freezing and also incorporates the energy supply system as well as a model of the thermal building hull. The metaheuristic is combined with a hybrid discrete/continuous simulation model to evaluate the energy efficiency of different production scenarios. The hybrid simulation enables to accurately capture material and energy flow within the production in an integrated and dynamicmanner. Overall, this simulation-based optimization method is intended to support energy-aware production scheduling in practical applications. Introduction Energy efficiency in industrial production has become an important topic in recent years because of the substantial potential for energy savings in the industrial sector [1]. Energy-aware Production Planning and Control (PPC) strategies can be used to influence energy demand and energy costs during operation, for example by shifting the production of energy-intensive products to the night hours, where energy is often cheaper. However, it is not sufficient to only consider energy as an optimization goal. Instead, energy efficiency must be seen as part of a multi-objective system of production targets together with production variables such as storage costs, throughput times or delivery delays. Such multi-objective problems with complex, sometimes time-dependent constraints are hard to solve for real-world problems. Modern solutions often rely on heuristic or metaheuristic methods [2]. For evaluating the fitness of solution candidates during metaheuristic search, simulation-based methods are gaining interest because they enable to capture complexity of real-world problems including difficult dynamic interactions without the limiting assumptions many other approaches have. However, with regard to energy optimization, interdisciplinary holistic simulation models are required which include dynamic interactions across engineering domains in order to get an accurate prediction of the overall energy demand, that not only includes production machinery, but also technical building services. For example, heating a production oven generates waste heat that is dissipated into the room and affects heating and cooling energy demand for the building. Similarly, the actual setup time for preheating the oven depends on different conditions, including which products have been produced before, and the setup time affects production throughput and scheduling. Incorporating energy considerations in production logistics simulations with their time-dependent interactions in an accurate manner requires advanced