一种用于氧化铝生料浆优化调配的改进遗传算法

An improved genetic algorithm for raw slurry optimal arrangement in alumina process

为解决烧结法氧化铝生料浆调配过程中评判指标复杂、计算劳动强度大、难以获取最优调配组合,造成料浆成分波动大等问题,以物料平衡为基础,结合生产运行经验,建立综合考虑生料浆质量指标要求、当前满槽质量指标和调配后剩余各槽质量指标的生料浆调配过程优化模型;针对优化模型中存在的多目标、非线性和多约束特点,提出一种根据适应度函数自动调整交叉概率和变异概率的改进遗传算法,并结合惩罚策略求解最优调配方案。该方法与传统的枚举递归寻优算法相比,减小了时间复杂度,总能保证在1 min内快速求得最优调配方案。工业应用结果表明,该算法满足现场调配工艺要求,可大幅度减轻计算劳动强度,减少生料浆质量指标的波动,使熟料指标碱比、铝硅比的平均合格率分别提高0.34%和5.95%,为后续生产的稳定发挥了重要作用。

In order to solve the problem of alumina raw slurry arrangement as much composition fluctuation of raw mix slurry because of the difficulty of obtaining an optimal arrangement scheme caused by complicated judgement criteria and arrangement computational complexity, based on material balance principle and expert experience an optimal arrangement model was presented, which integrated the required composition of raw mix slurry, the current composition of raw slurry in each full trough, and the composition of raw slurry in the full trough left after mixing. Considering the existence of multi-objective function, nonlinearity, and multiple constraints, an improved genetic algorithm with penalty strategy was proposed, in which the crossover rate and mutation rate were changed according to fitness function, to reduce time complexity greatly compared to the traditional enumeration algorithm. The optimal arrangement scheme can be obtained efficiently in less than 1 min. The industrial application results show that the optimal arrangement scheme meets the technics requirements and reduces the composition fluctuation of raw mix slurry, resulting in the increase of eligibility rate of clinker that increased the eligibility rate of soda and alumina-silica by 0.34% and 5.95 %, respectively, such that contributes to stabilizing the subsequent process of alumina production.