On April 3, 2026, Baosteel officially launched its AI-driven smart blast furnace system globally, with initial deployments across Blast Furnaces No. 4, 3, and 2 at its Baoshan base. This milestone marks the entry of the steel industry into an era of AI-driven ironmaking. Traditionally, blast furnace operations have been regarded as a “black box” in industrial production due to their extreme high-temperature and high-pressure conditions, complex multi-variable coupling, and heavy reliance on human experience.
2026年4月3日，宝钢股份在全球范围内首发上线AI智慧高炉系统，并在宝山基地4号、3号、2号高炉先后投入应用。这一动作标志着钢铁重工业正式进入AI驱动的智能化炼铁阶段。此前，高炉炼铁因其内部环境的高温高压、多变量耦合及对人工经验的高度依赖，长期被视作工业生产中的“黑箱”。
In the complex and continuous chemical process of blast furnace ironmaking, real-time perception and precise control of internal conditions have long been core challenges for the industry. Baosteel’s AI smart blast furnace project, jointly developed with Baosight Software and Huawei, establishes a closed-loop system integrating sensing, decision-making, and execution.
在高炉炼铁这一复杂且连续的化学反应过程中，炉内工况的实时感知与精准调控是行业核心难题。宝钢股份的AI智慧高炉项目，通过自主研发并联合宝信软件、华为公司共同打造，构建起一套感知、决策、执行的闭环控制系统。
At its core, the system represents a shift from traditional experience-driven operations to a dual-driven model based on data and AI. In practice, the system has demonstrated prediction accuracy and control adoption rates exceeding 90%, significantly improving thermal control precision and the stability of hot metal quality. According to public disclosures, each blast furnace can achieve annual cost reductions exceeding tens of millions of RMB, while also contributing to greener and lower-carbon production.
这一系统的核心在于将传统的“经验驱动”转向“数据+AI”双轮驱动。其核心模型在实际运行中表现出超过90%的预测命中率与控制采纳率，有效提升了炉热控制精度与铁水质量稳定性。根据公开信息，该系统在实现生产指标提升的同时，单座高炉每年可实现降本超千万元人民币，并同步推进绿色低碳生产。
Looking across the industry, similar efforts in digital transformation reveal common patterns in the evolution of intelligent ironmaking. For instance, in the first half of 2025, Nanjing Iron & Steel (NISCO) launched its “Integrated Ironmaking Smart Center,” which connects five major processes—including sintering, pelletizing, and blast furnaces—across 25 systems. By leveraging big data to build a digital mirror of the furnace body, it aims to make the smelting process both visible and analyzable.
观察行业内其他企业的智能化实践，可以清晰地看到炼铁工序数智化演进的共性。例如，2025年上半年，南钢（南京钢铁）投入运行的“铁区一体化智慧中心”通过整合烧结、球团、高炉等5大工序及25套系统，利用大数据建立炉身镜像，旨在实现冶炼过程的可视、可判。
While both approaches seek to break the “black box” of ironmaking, their focus differs. NISCO emphasizes system-level integration to achieve a holistic mapping of the entire ironmaking process, whereas Baosteel’s AI smart blast furnace focuses more on applying large AI models to enable real-time, fine-grained operational decision-making at the level of individual furnaces.
对比两者的路径，虽然都是为了打破炼铁“黑箱”，但侧重点存在细微差异。南钢的模式侧重于通过全局系统整合实现“铁区一体化”的逻辑映射；而宝钢股份的AI智慧高炉则更聚焦于利用AI大模型技术，在单座高炉的精细化操控层面实现模型的实时决策。
From a broader industry perspective, steelmakers worldwide have been advancing digitalization in recent years. However, most initiatives remain at the level of data integration and process visualization. The application of AI to core metallurgical units—enabling real-time decision-making in production—remains at an early stage globally.
从更广泛的行业视角来看，全球钢铁企业近年来持续推进数字化转型，但多数仍停留在数据整合与过程可视化阶段。真正将AI引入核心冶炼单元，实现生产过程实时决策的探索，仍处于早期阶段。
At the same time, the evolution of AI is increasingly moving from virtual environments into real-world industrial systems. Industries such as steelmaking, characterized by continuous processes and high operational constraints, are becoming critical testing grounds for the practical value of AI.
与此同时，AI技术的发展也正在从虚拟世界走向现实工业场景。对于钢铁这样的连续流程工业而言，其复杂性与高约束环境，使其成为检验AI实际价值的重要试验场。
The emergence of this achievement is not accidental but the result of long-term technological accumulation. Since 2020, China Baowu has promoted smart manufacturing under its “Four All” strategy and has been among the first to announce a clear roadmap toward carbon peaking and carbon neutrality, linking low-carbon metallurgy with technological innovation. In 2024, Baosteel designated the year as its “AI Year,” during which it established foundational capabilities across computing power, models, talent, and organizational structure, alongside the creation of a centralized data platform.
宝钢股份此次成果的出现并非偶然，而是企业长期技术积累的产物。自2020年以来，中国宝武提出以“四个一律”推进智慧制造，并率先公布“双碳”时间表，将低碳冶金与技术创新绑定。2024年被宝钢股份定义为“AI元年”，公司在此期间完成了构建算力、模型、人才、组织四大底座的基础建设，并组建了公司大数据中心。
This strategic evolution reflects a shift from high-level smart manufacturing goals toward embedding algorithmic capabilities directly into specific production processes.
这种基于战略演进的动作，体现了从宏观智能制造目标向具体微观工序算法能力的下沉。
April 2026: Scaled deployment of AI-driven blast furnaces
2020年底：提出智慧制造战略，布局物联网与大数据
Despite its demonstrated benefits in cost reduction, efficiency improvement, and low-carbon production, the broader adoption of AI-driven blast furnace systems faces significant practical challenges.
尽管AI智慧高炉在降本增效与低碳减排方面取得了显著成效，但这种模式的推广仍面临复杂的现实门槛。
First, data quality and accumulation remain critical. Stable blast furnace operations depend heavily on large volumes of well-labeled historical data, requiring steelmakers not only to invest in hardware but also to develop strong data infrastructure and operational capabilities. Second, the generalization ability of large AI models across different production lines and varying raw material conditions still requires further validation.
首先是数据的清洗与沉淀能力，高炉运行的稳定性高度依赖于历史海量数据的高质量标注，这要求钢企不仅拥有硬件，更需具备强大的大数据中心运营能力。其次，大模型在不同产线、不同原料结构条件下的泛化能力仍需时间验证。
At present, the deployment of smart ironmaking systems across four blast furnaces at the Baoshan base is still ongoing. While Baowu’s approach provides a valuable reference for the industry, differences in raw material structures, equipment conditions, and organizational setups mean that such solutions cannot be easily replicated. Balancing the investment in AI systems with the timeline for realizing cost benefits remains a key challenge for the steel industry’s digital transformation.
目前，宝山基地4座高炉的智慧炼铁建设仍处于推进计划中。对于行业而言，AI智慧高炉的“宝武方案”提供了示范效应，但各家企业面临的原料结构差异、设备老旧程度以及管理架构的不同，决定了AI炼铁方案无法进行简单的“平移式”复制。如何平衡AI系统的开发投入与实际降本产出的周期，依然是摆在钢铁行业智能化转型面前的一道经营考题。
In the longer term, the significance of such initiatives may lie less in immediate success or failure, and more in their role in validating how AI can be integrated into complex industrial systems. For the steel industry, the challenge of translating intricate physical processes into computable and optimizable models remains an area worthy of continued attention.
从更长周期来看，这一探索的意义或许不在于短期的成功或失败，而在于其对工业与AI结合路径的验证。对于钢铁行业而言，如何将复杂的物理过程转化为可计算、可优化的模型体系，仍是一个值得持续关注的方向。
📅 2026年04月09日 写于Paris
✍️ 转载请注明：Steel Press