Generic Buffer: A data-driven flexibility modeling approach for predictive control in industrial steam networks

• A Generic Buffer accurately represents the intrinsic flexibility of a steam network. • Predictive control based on the Generic Buffer results in cost and emissions savings. • Our controller is tested in a high-fidelity Modelica emulator (ClaRa and TransiEnt). • Industrial steam network flexibility seems an attractive substitute for batteries. • Achieved weekly reductions of up to 13780 tCO 2 emissions and 1860 kEUR costs. Steam networks offer a practical solution to increase flexibility in energy-intensive industries, thanks to their potential for electrification and integration of various energy sources. However, effectively using this flexibility requires models that accurately represent system constraints and dynamics for control. While first-principle models provide accuracy, their complexity can hinder practical control applications. This paper proposes a data-driven model using Generic Buffer to model the intrinsic flexibility and dynamics of a steam network in order to efficiently leverage the available flexibility through model predictive control. The simulations show significant weekly reductions in CO 2 emissions (between 8630 and 13780 tCO 2 ) with an emissions-oriented controller, and in costs (between 29 and 1860 kEUR) with a cost-oriented controller, compared to a pressure-regulating proportional-integral (PI) controller.