AbstractsMathematics

Demand-Side Management concepts have the potential to positively impact the financial as well as the environmental aspects of energy-intensive industries. More specifically, they allow reducing the energy cost for the industrial plants by dealing with energy-availability fluctuations. In this context, efficient frameworks for scheduling with energy awareness have been studied and showed potential to reduce the overall energy bill for energy-intensive industries, for instance stainless steel and paper plants. Those frameworks usually combine scheduling and energy optimization into one monolithic system. This work investigates the possibility of integrating the two systems by specific exchange of signals, while keeping the scheduling model separated from the energy-cost optimization model. Such integration means that the pre-existent schedulers and energy optimizers could be easily modified and reused without re-implementing the whole new system. Two industrial problems with different scheduling approaches are studied. The first problem is about pulp and paper production which uses the Resource Task Network (RTN) scheduling approach. The second one is about stainless steel production which is based on a bi-level heuristic implementation of an improved energy-aware scheduler. This work presents the decomposition methods that are available in literature and their application to the two industrial problems. Besides an improvement in the RTN approach for handling storages, this thesis describes a prototype implementation of the energy-aware RTN scheduler for paper and pulp production. Furthermore, this work investigates the performance of the application of different decomposition methods on different problem instances. The numerical case studies show that even though the decomposition decreases the solution quality compared to the monolithic system, it still gives good solutions within an acceptable duration with the advantage of having two separate pre-existent systems which are simply exchanging signals.