基于迁移代价的连续碳酸化分解过程序列化最优操作方法

As one of the key processes for alumina production,the continuous carbonation process of sodium aluminate is a typical long, complex and unstable process with multi time-delay and large inertia. Good gradient distribution of precipitation rate indicates the quality and quantity of the product. But the process fluctuation resulted from fickle feed condition arouses low precipitation rate and poor quality, which brings difficulty to successive processes. Based on the proposed idea of ordered migration for fluctuation restraining, a method of optimal operation sequence based on minimum migration cost for carbonation decomposition is proposed. To preestimate the dynamic performance of the process, the interacting continuous stirred-tank reactor model with time-correlation analysis matrix is built. A high efficiency carbonation pattern set is extracted from sparse distributed value space to collect gradient distributions of efficient precipitation rates. By synthesizing the ICSTR dynamic model, the high efficiency carbonation pattern set and time sequence analysis, the synthetic cost of process migration is computed. The operation sequence of minimum cost is discovered by proper optimization algorithm with variable-dimension solutions to implement optimal precipitation rates and smooth operating transition. The proposed operation theory and method adapted to long process stabilizing operation is of wide application and generalization benefit for energy saving, emission reduction, production enhancing and quality improving in other complex processes.

连续碳酸化分解过程是氧化铝生产中的关键工序，是典型的长流程大惯性生产过程，具有多重大滞后，工况难以稳定。良好的分解率梯度分布是碳分过程产量、质量的保证。但是原料条件多变导致工况波动引起分解率下降和产品质量的恶化，给氧化铝后续生产流程节能增效带来严重困难。本课题根据提出的工况有序迁移思想，构建了基于迁移代价的碳分过程最优序列化操作方法。主要包括建立关联连续搅拌反应釜多时效动态模型，实现过程动态性能估算；提出价值稀疏分布的高效分解率梯度提取方法，获取高效碳分模式集；研究基于动态模型、高效模式集和时间序列分析的工况迁移代价综合评价方法；设计针对非定维解的操作序列优化方法，获取迁移代价最小的操作序列，从而实现碳分过程的分解率梯度优化和工况状态平稳过渡，形成基于最小迁移代价的工况迁移操作理论与方法。工况波动问题在长流程复杂工业过程中具有普遍性，本课题对工业生产节能减排提产增效具有广泛的推广应用价值。

连续碳酸化分解过程是氧化铝生产中的关键工序，是典型的长流程大惯性生产过程，具有多重大滞后，工况难以稳定。本课题根据提出的工况有序迁移思想，构建了基于迁移代价的碳分过程序列操作优化方法。首先建立了基于时效关联分析的关联连续搅拌反应釜动态模型，提出基于离散状态转移算法的多重时滞辨识辨识方法；针对调整过程的工况空间性能退化敏感区域进行风险分析，获得性能退化风险评价指标，从而获得操作序列的最小迁移代价目标函数；提出采用机理参数辨识、数据驱动和灵敏度自校正的集成模型进行末槽出口成分预测；提出多属性评价的高效操作序列提取方法，获取高效碳分模式集；提出基于链式个体的非定长操作序列表示方法和功能性多种群的混合差分进化算法实现非定长操作序列智能优化,同时提出了基于知识协调的操作序列快速调整策略；并在生产企业开发形成了相应的软硬件系统，应用验证表明本课题成果能够保证氢氧化铝产品的质量稳定性，末槽分解率水平有所提高。工况波动问题在长流程复杂工业过程中具有普遍性，生产过程的稳定运行和优化操作是绿色高效生产面临的重要科学问题，本课题相关成果不仅为氧化铝生产的增产节约和节能降耗提供支撑，而且对不同工业过程、各类复杂系统的精细化分析、建模、评估、控制和优化研究具有借鉴作用。