摘要
针对氧化铝生料浆配料过程中存在成分信息不完备,工况变化规律难统计等诸多不确定性的特点,提出一种从建模和优化计算两个层次上弱化不确定性的实时优化方法,采用神经网络、物料平衡建立生料浆质量主规律预测模型,再利用灰色理论对模型的偏差数据进行信息挖掘,补偿不确定因素的影响;然后,基于质量预测模型,提出一种专家分级推理机制实现不确定环境下的配比优化,提高入槽生料浆质量,实现了生料浆质量指标的优化控制,其中碱比[N/R]的合格率达到了99%,钙比[C/S]和铝硅比A/S分别达到了96%和94%,并简化了生产工艺,有效降低了能耗,为存在信息不确定性的长流程工业过程的优化控制提供了新思路。
Based on uncertainties resulting from the incomplete information of raw materials＇ ingredients and the ruleless variations of operating condition in the process of aluminium production, the paper introduces a real-time optimization method weakening the effect of uncertainty by modeling and optimizing algorithm. Firstly, the master-rule prediction model for raw slurry quality is built by adopting neural networks and material balance principle, and information mining for deviation data of model is made by using the gray theory to compensate effect of uncertainty. Then, based on the prediction model, the paper proposes a kind of expert hierarchical reasoning mechanism realizing proportioning optimization under uncertainty so as to improve the raw slurry quality in tanks and realize optimization control of charge pulp quality index among which yield rate of alkali ratio reaches 99%, calcium ratio and aluminium-silicon ratio reaches 96% and 94%. It simplifies the process and reduces the energy consumption effectively.It also provides new thoughts for optimization control of long-flow industrial process with existing information uncertainty.
出处
《有色冶金设计与研究》
2008年第5期1-6,10,共7页
Nonferrous Metals Engineering & Research
基金
国家自然科学基金项目(60874069
60804037)
国家自然科学基金重点项目(60634020)
博士点基金项目(0050533016)资助
关键词
氧化铝
配料
不确定
实时优化
专家推理
遗传算法
alumina
blending
uncertainty
real-time optimization
expert reasoning
genetic algorithm
