The European flat steel industry has come under immense competitive pressure as a result of globalization. To retain its market position, costs must be further reduced and unique selling points, such as quality, further expanded. A contribution to these requirements can be provided by an advanced production scheduling in a flexible flow production, without the need for an investment in capital-intensive plant components. Within the EU-funded RFCS project DynReAct, a scheduling approach is under development to generate optimized production plans for each individual coil at each production step considering real-time plant information. This concept offers immediate reactions to critical situations like insufficient plant performances or coils that are off the quality specifications. The optimal routing and sequencing will be estimated using real-time-capable plant performance models derived from machine learning on large historical data, which will be incorporated in multi-objective optimization methods. This approach enables the associated complexity to be mastered and other factors, such as quality, energy, or maintenance, to be taken into account in scheduling in addition to the classic logistical targets. In this work, the concept of multi-objective production scheduling is presented by the necessary foundations for the implementation in a practical use case in the tin plate industry.