| 以下仅列出文章摘要、提纲简介，如需获取全文阅读权限，或原创定制、长期合作，请随时联系。 微信QQ:312050216 |
 |
 |
| 基于数据驱动的食用油精炼工序关键参数优化
摘 要
食用油精炼工序是食用油生产过程中必不可少的过程。该工序通过酸碱中和的方式来去除毛油中的磷脂等杂质，得到符合一定质量标准的成品油。食用油精炼工序的参数调节过程是保证油品质量及最终得油率高低的重要环节，通过调节磷酸添加量、碱添加量及轻向压力等关键工艺参数达到适合该批次毛油的生产设定值；生产过程中当出现化验值反馈油品不合格时需要根据观测到的反馈状态做相应调整，其调节过程的好坏直接关系整个精炼厂的经济和技术指标，对参数进行优化控制是提高精炼工序最直接有效的方法。本文针对食用油精炼工序人工经验无法高效调节的的问题，提出一种基于数据驱动的食用油精炼工序关键参数优化的方法，主要包括以下内容：
（1）大数据工艺约束规则挖掘，首先深入分析工艺过程，找出影响食用油质量的DCS数据跟化验数据，将DCS数据和化验数据基于时间对齐拼接成原始拼接表，然后针对无效值和缺失值处理，采用滤波算法来剔除异常值，去除冗余和矛盾数据，精确计算各参数之间的时间延迟关系，形成大数据决策表。采用粗糙集算法对决策表进行规则挖掘，得到磷酸添加量、碱添加量及轻向压力等工艺约束，为遗传算法寻优提供工艺约束规则。
（2）建立Xgboost得油率预测模型，然后根据历史数据找出不同油品下的最高得油率参数区间，作为遗传算法寻优约束区间，采用遗传算法从寻优区间中寻找一组最优参数组合，以Xgboost得油率预测模型的输出结果的高低最为评价指标，最后，基于专家经验跟规则挖掘算法获取现场调节规则对遗传算法输出结果进行约束校正。
根据现场工业数据验证，本文提出的基于数据驱动的关键参数优化方法，能够满足实际生产需要，证实优化出来的最优参数组合确实能够提高得油率。
关键词：得油率预测模型，食用油精炼工序，数据驱动
ABSTRACT
The refining process of edible oil is an essential process in the production of edible oil.In this process, phospholipids and other impurities in crude oil are removed byacid-base neutralization to obtain product oil meeting certain quality standards. The parameter adjustment process of edible oil refining process is an important link to ensure the oil quality and final oil yield. By adjusting the key process parameters such as phosphoric acid addition amount, alkali addition amount and light pressure,the set value suitable for the production of this batch of crude oil can be reached. In the production process, when the test value feedback oil is unqualified,the corresponding adjustment should be made according to the observed feedback state, and the adjustment should be carried out The quality of the refining process is directly related to the economic and technical indicators of the whole refinery. The most direct and effective way to improve the refining process is to optimize the parameters.In this paper, a data-driven optimization method for key parameters of edible oil refining process is proposed to solve the problem that manual experience can not be adjusted efficiently:
(1) In the mining of big data process constraint rules, firstly,the process is deeply analyzed to find out the DCS Data and test data that affect the quality of edible oil, and the DCS Data and test data are spliced into the original splicing table based on time alignment. Then, for the invalid value and missing value processing,the filtering algorithm is used to eliminate the abnormalvalue, remove the redundant and contradictory data, and accurately calculate the time delay between the parameters Late relationship to form big data decision table. Rough set algorithm was used to mine the rules of decision table,and the process constraints such as phosphoric acid addition,alkali addition and light pressure were obtained, which provided process constraint rules for genetic algorithm optimization.
(2) This paper establishes the xgboost oil yield prediction model,and then finds out the highest oil yield parameter interval underdifferent oil products according to the historical data, which is used as the optimization constraint interval of genetic algorithm. The genetic algorithm is used to find a group of optimal parametercombination from the optimization interval,and the output result of xgboost oil yield prediction model is taken as the evaluation index. Finally, based on the expert experience and rule mining algorithm,this paper proposes a new algorithm The regulation rules are obtained to correct the constraints of the output of genetic algorithm.
According to the field industrial data,the optimization method of key parameters based on data driving in this paper can meet the needs of actual production, and it is proved that the optimized parameter combination can improve the oil yield.
Keywords：oil yield prediction model, edible oil refining process, data driven2
|
