Multi-scenario data-driven robust optimisation for industrial steam power systems under uncertainty

Abstract

In actual industrial production, the deterministic optimisation of the steam power system cannot meet most production scenarios due to the influence of uncertain factors such as product demand and environmental conditions. This paper proposes an operational optimisation method for SPS under uncertainty by combining multi-scenario partition and data-driven adaptive robust optimisation algorithm. A hybrid equipment model was developed to modify the critical equipment models based on industrial data and process mechanisms. Considering that demand uncertainty varies with the different steam quality, the clustering method divides the entire time horizon into several periods, and the uncertainty set is constructed by variable robust kernel density estimation for each period. A multi-scenario data-driven robust optimisation model is developed by incorporating uncertainty sets into deterministic optimisation, and the counterpart model is obtained through the affine decision rules. Furthermore, the proposed framework is applied to the SPS of a coal chemical plant to verify the feasibility. The annual operating costs before and after optimisation are 125 million USD and 123 million USD, respectively, and the system's energy efficiency can be improved by more than 5%.