Energy-oriented scheduling based on Evolutionary Algorithms

Energy efficiency has become more and more critical for the success of manufacturing companies because of rising energy prices and increasing public perception of environmentally conscious operations. One way to increase energy efficiency in production is to explicitly consider energy consumption during short-term production planning. In many cases, final energy sources (FES) are not directly consumed by production resources and thus have to be transformed by conversion units into applied energy sources (AES), such as steam or pressure, so the relationship between AES and FES has to be considered. Therefore, we present an energy-oriented scheduling approach for a parallel machine environment. These parallel machines require production order and process time specific amounts for AES and the objective is to minimize the demand of FES. This minimization can be achieved by smoothing the cumulated demand of AES to avoid the frequent load alternations that are responsible for the inefficient operation of conversion units. Therefore, resource leveling is used as a surrogate objective for optimization. To solve the resource leveling problem for large problems, a Genetic Algorithm and two Memetic Algorithms are developed. The evaluation of the proposed Evolutionary Algorithms is based on small test instances and several real-world instances. These latter instances are based on an application case from the textile industry, and promising results concerning energy costs and carbon dioxide emissions are reported.