Scholarlink

<p>The precise prediction and stable regulation of blast furnace molten iron temperature have long been constrained by its nonlinearity, significant time delay, and strong multiparameter coupling characteristics. This results in limited prediction accuracy, delayed control response, and compromised production stability with conventional methods. Therefore, this study proposes an intelligent prediction-control strategy deeply integrating TSMixer-CGC-PO-ADRC: employing a Time Series Mixer (TSMixer) to capture multidimensional temporal dependencies for precise temperature forecasting, utilizing novel Chaos-Genetic-Centroid Parrot Optimization (CGC-PO) to generate optimal process set points, and integrating Adaptive Disturbance Rejection Control (ADRC) for accurate tracking, thereby establishing a closed-loop “Prediction–Optimization–Control” system. First, an iron-water temperature prediction model based on the TSMixer is established. By employing a time-feature hybridization mechanism to capture multidimensional temporal dependencies, the 95% confidence interval of prediction results converges to ±5.54&#xa0;°C, demonstrating significantly superior performance compared with benchmark models. The core innovation lies in constructing the CGC-PO-ADRC collaborative control architecture: an enhanced CGC-PO algorithm synergistically optimizes multiple parameters (including air volume) to generate optimal set points; integrated with ADRC, it achieves precise tracking and dynamic disturbance rejection; this underpins a layered control strategy where “coke injection volume primarily targets temperature adjustment, air volume assists in temperature stabilization, and oxygen enrichment flow provides fine-tuning.” Building upon this intelligent prediction–regulation methodology for molten iron temperature, an industrial-grade application system was developed. Field results demonstrate that this strategy narrows temperature fluctuations to ±5&#xa0;°C, elevates target range hit rates from 68% to 92%, and increases gas flow stability by 16%, substantially enhancing both the operational stability and intelligent management of blast furnaces.</p> Graphical Abstract <p></p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Data-Driven Intelligent Prediction and Robust Control of Hot Metal Temperature in Blast Furnaces

  • 期刊论文

  • 发表日期：

    2026年4月28日

查看全文
  • Zhang Mengjie,
  • Liu Ran,
  • Liu Xiaojie,
  • Zhang Yujie,
  • Li Xin,
  • Li Hongwei

摘要

The precise prediction and stable regulation of blast furnace molten iron temperature have long been constrained by its nonlinearity, significant time delay, and strong multiparameter coupling characteristics. This results in limited prediction accuracy, delayed control response, and compromised production stability with conventional methods. Therefore, this study proposes an intelligent prediction-control strategy deeply integrating TSMixer-CGC-PO-ADRC: employing a Time Series Mixer (TSMixer) to capture multidimensional temporal dependencies for precise temperature forecasting, utilizing novel Chaos-Genetic-Centroid Parrot Optimization (CGC-PO) to generate optimal process set points, and integrating Adaptive Disturbance Rejection Control (ADRC) for accurate tracking, thereby establishing a closed-loop “Prediction–Optimization–Control” system. First, an iron-water temperature prediction model based on the TSMixer is established. By employing a time-feature hybridization mechanism to capture multidimensional temporal dependencies, the 95% confidence interval of prediction results converges to ±5.54 °C, demonstrating significantly superior performance compared with benchmark models. The core innovation lies in constructing the CGC-PO-ADRC collaborative control architecture: an enhanced CGC-PO algorithm synergistically optimizes multiple parameters (including air volume) to generate optimal set points; integrated with ADRC, it achieves precise tracking and dynamic disturbance rejection; this underpins a layered control strategy where “coke injection volume primarily targets temperature adjustment, air volume assists in temperature stabilization, and oxygen enrichment flow provides fine-tuning.” Building upon this intelligent prediction–regulation methodology for molten iron temperature, an industrial-grade application system was developed. Field results demonstrate that this strategy narrows temperature fluctuations to ±5 °C, elevates target range hit rates from 68% to 92%, and increases gas flow stability by 16%, substantially enhancing both the operational stability and intelligent management of blast furnaces.

Graphical Abstract