基于多参数灵敏度分析与遗传优化的铁水质量无模型自适应控制

铁水硅含量(化学热)和铁水温度(物理热)是高炉炼铁过程最重要的铁水质量指标, 其建模与控制对于整个高炉炼铁过程的运行优化意义重大. 针对高炉炼铁过程极复杂动态特性以及铁水质量难以进行常规机理建模与控制的难题, 基于直接数据驱动控制思想, 提出一种基于多参数灵敏度分析与大规模变异遗传参数优化的高炉铁水质量无模型自适应控制方法. 首先, 基于紧格式动态线性化(Compact form dynamic linearization, CFDL)无模型自适应控制(Model free adaptive control, MFAC)技术确定铁水质量的多变量数据驱动控制器结构; 然后, 针对CFDL-MFAC众多可调参数对控制器性能影响大, 同时对众多参数整体优化非常耗时且效果不理想的问题, 基于多参数灵敏度分析(Multi-parameter sensitivity analysis, MPSA)技术, 提出基于大规模变异与精英局部搜索遗传优化的CFDL-MFAC控制器参数整定方法; 最后, 将参数整定后的CFDL-MFAC控制器应用到高炉炼铁过程多元铁水质量控制, 并与基于递推子空间辨识的数据驱动预测控制进行比较研究, 验证所提控制方法的有效性和先进性.

Model Free Adaptive Control of Molten Iron Quality Based on Multi-parameter Sensitivity Analysis and GA Optimization

The silicon content (Chemical heat) and the molten iron temperature (Physical heat) are two important molten iron quality indices of blast furnace ironmaking process, whose modeling and control is of great importance to the operation and optimization of the whole blast furnace ironmaking process. Considering the extremely complicated dynamic characteristics and the puzzle of conventional mechanism modeling and control of the blast furnace ironmaking process, a multi-parameter sensitivity analysis (MPSA) and large-scale mutation genetic parameter optimization based blast furnace molten iron quality model free adaptive control method is proposed based on direct data-driven control theory. First, a multivariable data-driven controller structure for molten iron quality is determined based on compact form dynamic linearization (CFDL) based model free adaptive control (MFAC) technique. Then, considering that it is very time-consuming and less effective to adjust all the CFDL-MFAC adjustable parameters, which have a high influence on the controller performance, a CFDL-MFAC controller parameter tuning method based on large-scale mutation and elite local search genetic optimization is proposed. Finally, applied the parameter-tuned CFDL-MFAC controller into the control of multivariate molten iron quality in the blast furnace ironmaking process and compare it to data-driven predictive control based on recursive subspace identification to verify the effectiveness and advancement of the proposed control method.