A two-stage robust optimization approach for oxygen flexible distribution under uncertainty in iron and steel plants

TL;DR

This paper introduces a two-stage robust optimization model for oxygen distribution in steel plants, effectively managing demand uncertainties to improve operational efficiency and profitability.

Contribution

It develops a novel two-stage robust optimization framework with a Gaussian process-based demand forecasting method for flexible oxygen distribution under uncertainty.

Findings

Model effectively handles demand fluctuations in steel industry.

Forecasting approach improves demand prediction accuracy.

Computational results validate model's practical applicability.

Abstract

Oxygen optimal distribution is one of the most important energy management problems in the modern iron and steel industry. Normally, the supply of the energy generation system is determined by the energy demand of manufacturing processes. However, the balance between supply and demand fluctuates frequently due to the uncertainty arising in manufacturing processes. In this paper, we developed an oxygen optimal distribution model considering uncertain demands and proposed a two-stage robust optimization (TSRO) with a budget-based uncertainty set that protects the initial distribution decisions with low conservatism. The main goal of the TSRO model is to make wait-and-see decisions maximizing production profits and make here-and-now decisions minimizing operational stability and surplus/shortage penalty. To represent the uncertainty set of energy demands, we developed a Gaussian process (GP)-based time series model to forecast the energy demands of continuous processes and a capacity-constrained scheduling model to generate multi-scenario energy demands of discrete processes. We carried out extensive computational studies on TSRO and its components using well-synthetic instances from historical data. The results of model validation and analysis are promising and demonstrate our approach is adapted to solve industrial cases under uncertainty.