Industrial Steam Systems Optimization under Uncertainty Using Data-Driven Adaptive Robust Optimization

Steam system, which is an important component of utility system of the industrial process, provides power and heat to the process. Operational optimization methods can improve the efficiency of the steam system and increase the economic benefits for industrial plants. Because of the uncertainty in device efficiency, traditional deterministic optimization methods could lead to suboptimal or even infeasible optimization decisions of steam systems. This paper proposes a data-driven adaptive robust optimization approach to deal with the operational optimization under uncertainty for industrial steam systems. Uncertain parameters of the steam system model are derived from the historical process data based on steam turbine models. A robust kernel density estimation method is employed to construct the uncertainty sets. The data-driven uncertainty sets are incorporated into a two-stage adaptive robust mixed-integer linear programming (MILP) framework for steam systems operational optimization to minimize the total operating cost. By applying the affine decision rule, the proposed multi-level optimization model is transformed into a single-level MILP problem. An industrial case study of the steam system from an ethylene plant is presented to demonstrate the effectiveness of the proposed method.