(595f) Coordinated Steel Plant Scheduling Solution

Saliba, S., ABB Corporate Research
Biondi, M., ABB Corporate Research
Xu, C., ABB
Sand, G., ABB Corporate Research
Harjunkoski, I., ABB Corporate Research

The steel industry has experienced dramatic changes during the last years, where both the steel price and raw-material prices have been volatile. In order to maximize the profit, it must be ensured that the operational costs are kept as low as possible and that the production can flexibly adapt also to rapidly changing conditions.

This raises the challenge for industrial solutions on plant-wide planning and scheduling problems. Subsystems of plants are often optimized locally without an overall coordination on plant level. Here, a new optimization method will be discussed that utilizes local scheduling algorithms and optimizes the overall schedule by coordinating coupling parameters. The performance is evaluated on a real steel plant coupling melt shops and hot rolling mills showing the coordination scheme's advantage to fully decentralized or centralized scheduling algorithms in terms of solution quality and computational effort.

The production constraints in the melt shop mainly result from metallurgical rules, whereas in the hot rolling mill the production constraints mainly arise from physical restrictions and the quest for small changes in product changeovers. Because of the complex and different production constraints in these two processes, we have developed scheduling algorithms separately for both plants with different optimization objectives for each of them. Both scheduling problems are large, complex and highly constrained. Therefore, we decompose the problems into smaller optimization problems, which are solved by formulating them as mixed integer linear programs (MILP) and put together again using integer programming.

The general coordination method is based on the idea to take up the uncoordinated local solutions, evaluate them according to plant-wide objective criteria. The evaluation results will be used by the coordinator to tune the parameters of the distributed schedulers, such that they are coordinated iteratively. The coordination iteration will be terminated when the coordination objective exceeds a certain value or the maximum iteration number is reached.

An industrial-size example illustrates the potential benefits and solution time for a typical model instance. Potential benefits do not only comprise lower production costs but also contribute to reduced energy demand and respectively lower emissions.