Cooperative co-evolutionary dynamic multiobjective optimization for burden surface profile decision in blast furnace ironmaking

Burden surface profile (BSP) is crucial for energy utilization improvement and stable operation in the blast furnace (BF) ironmaking process. The BSP decision can be expressed as a multiobjective optimization problem. However, the production process has nonlinear dynamic characteristics, leading to certain defects in conventional stationary optimization. To address the aforementioned issue, this study presents a novel dynamic multiobjective decision scheme to identify the optimal BSP. In the proposed scheme, a dynamic multiobjective optimization problem with constraints is first constructed. Then, to get rid of the dilemma of deriving parameter dependence using the physics domain knowledge, we employ kernel extreme learning machine (K-ELM) to establish the multiobjective optimization models. Subsequently, aiming to solve this dynamic multiobjective problem (DMOP), a novel dynamic optimization framework based on environment sensitivities (CCE-D-NSGA-II) is developed, which embeds the parameter decomposition, change response strategy, and optimal solution selection, for adapting to the nonstationary characteristics of BF ironmaking process. Finally, experiments using the actual production data are carried out to verify the effectiveness and feasibility of the proposed scheme. The results demonstrate that the proposed scheme achieves a set of Pareto optimal solutions for the BSP optimization problem and outperforms the selected baseline methods in both convergence and population diversity.