Optimization of Coal Blending for Coke Production : Using Linear and Mixed-Integer Programming

Sammanfattning

Coal blending is an important decision problem in coke production, where different coal types must be combined to satisfy technical, operational and economic requirements. The aim of this thesis is to develop and evaluate a mixed-integer linear programming (MILP) model for coal blending in coke production. The model is used to identify cost-efficient coal blends while satisfying quality, structural and availability constraints. The study is conducted as a quantitative case study in collaboration with SSAB Luleå. The model is implemented in Python using the Gurobi Optimizer and is applied in both a single-period and a multi-period setting. Due to industrial confidentiality, the results are presented using anonymized demonstration data. The results show that the model can identify feasible coal blends that satisfy the defined technical and operational constraints. The multi-period demonstration also shows that the model can handle varying demand, stock levels, reserved stock, safety stock and planned deliveries over time. As material availability changes, the selected blend composition and cost index also change. The study concludes that MILP is a suitable approach for modeling coal blending decisions under multiple simultaneous constraints. The model provides a transparent decision-support framework for evaluating feasible coal blends and analyzing how changes in input conditions affect blend composition and cost structure.