基于人工智能混合模型的乙二醇电渗析脱盐优化  

Optimization of ethylene glycol electrodialysis desalination based on artificial intelligence hybrid model

作　　者：李耀翔 范峥[1] 郝新宇 刘姝延 张叶 韩洁 LI Yaoxiang;FAN Zheng;HAO Xinyu;LIU Shuyan;ZHANG Ye;HAN Jie(College of Chemistry and Chemical Engineering,Xi’an Shiyou University,Xi’an 710065,Shaanxi,China;CNPC EastChina Design Institute Company Limited,Qingdao 266071,Shandong,China;Shaanxi Yanchang Petroleum(Group)Company Limited Refining and Chemical Company,Yan’an 727406,Shaanxi,China;College of Natural Resources and Environment,Northwest A&F University,Xianyang 712199,Shaanxi,China;Shenzhen Ruqinba Technology Group Company Limited,Shenzhen 518126,Guangdong,China;Shaanxi Mingze Yisheng Energy Technology Company Limited,Xianyang 712000,Shaanxi,China)

机构地区：[1]西安石油大学化学化工学院,陕西西安710065 [2]中石油华东设计院有限公司,山东青岛266071 [3]陕西延长石油(集团)有限责任公司炼化公司,陕西延安727406 [4]西北农林科技大学资源环境学院,陕西咸阳712199 [5]深圳如钦巴科技集团有限公司,广东深圳518126 [6]陕西铭泽易昇能源技术有限公司,陕西咸阳712000

出　　处：《化工进展》2024年第11期6049-6058,共10页Chemical Industry and Engineering Progress

基　　金：中国国家留学基金(201908610135);西安石油大学研究生创新与实践能力培养项目(YCS23213076);2021年西安石油大学研究生教育综合改革研究与实践项目(2021-X-YJG-004)。

摘　　要：利用小波神经网络和模拟退火-粒子群算法对电渗析法脱盐工艺进行参数优化。首先,通过单因素实验初步探讨了电渗析单位膜电压、操作时间、极板间距和极液浓度对脱盐效果的影响规律,然后利用小波神经网络模型对数据样本进行训练和预测,并对实验影响因素进行了Sobol灵敏度分析,最后将小波神经网络模型与模拟退火-粒子群算法相互耦合,得到了该体系下优化的电渗析条件及其对应的脱盐率。试凑法结果表明,4-10-8-1小波双隐层神经网络模型为适宜的预测模型;各因素对脱盐效果的作用程度由大到小依次为单位膜电压、操作时间、极液浓度和极板间距;当单位膜电压为0.42V/cm^(2)、运行时间为13.85h、极板间距为12.11cm、极液浓度为0.21mol/L时,预测的优化脱盐率可达97.13%,经统计检定值(t)检验该值与验证实验结果高度一致。该研究可为乙二醇电渗析法脱盐工艺的全面推广和深度应用提供准确可靠的理论支撑和数据来源。Utilizing wavelet neural network and simulated annealing-particle swarm optimization algorithm to optimize the parameters of electrodialysis desalination process.Firstly,a single factor experiment was conducted to preliminarily explore the influence of electrodialysis operation voltage,operation time,electrode plate spacing,and electrode liquid concentration on desalination efficiency.Then,a wavelet conducted on the experimental influencing factors. Finally, the wavelet neural network model was coupled with the simulated annealing-particle swarm optimization algorithm. The optimized electrodialysis conditions and corresponding desalination rates in this system were obtained. The trial-and-error method results indicated that the 4-10-8-1 wavelet double hidden layer neural network model was a suitable prediction model. The degree of influence of various factors on the desalination effect was in descending order: operating voltage, operating time, electrode liquid concentration, and electrode plate spacing. When the unit membrane voltage was 0.42V/cm^(2), the operating time was 13.85 hours, the electrode spacing was 12.11cm, and the electrode liquid concentration was 0.21mol/L, the predicted optimized desalination rate reached 97.13%. After t-test, this value was highly consistent with the validation experimental results. This study can provide accurate and reliable theoretical support and data sources for the comprehensive promotion and deep application of ethylene glycol electrodialysis desalination process.

关 键 词：乙二醇 电渗析 脱盐 神经网络 算法 粒子群优化 模拟退火 

分 类 号：TE646[石油与天然气工程—油气加工工程]