| 基于贝叶斯推断的多层软测量建模在丁醇发酵中的应用 |
| |
| 引用本文： | 朱湘临,顾雯炜,王 博.基于贝叶斯推断的多层软测量建模在丁醇发酵中的应用[J].教育技术导刊,2020,19(4):183-188. |
| |
| 作者姓名： | 朱湘临 顾雯炜 王 博 |
| |
| 作者单位： | 江苏大学 电气信息工程学院，江苏 镇江 212000 |
| |
| 基金项目： | 江苏省自然科学基金项目（BK20140568，BK20151345）；吉林省重大科技攻关专项（20140203005SF）；江苏高校优势学科建设工程资助项目（PAPD） |
| |
| 摘 要： | 针对丁醇生产过程中发酵产物品质参量难以实时测量，现有测量方法精度不高、测量结果受不确定因素影响较大的问题，提出一种基于贝叶斯推断和支持向量回归（Support vector machine regression，SVR）的多层软测量建模方法。首先应用贝叶斯推断计算后验概率、筛选偏置数据，并对偏置数据校准，建立第一层SVR模型；然后利用贝叶斯推断进行二次校准，建立第二层SVR模型，对第一层SVR模型输出进行修正，得到最终预测结果，克服干扰和偏差引起的模型不准确问题。将基于贝叶斯推断的多层支持向量回归（Bi-SVR）预测模型应用于丁醇发酵过程，仿真及实验结果表明，相较于传统SVR预测模型，系统在低干扰的情况下预测精度提高了4.52%，在高干扰时预测精度提高了5.37%。
|
| 关 键 词： | 微生物发酵 贝叶斯推断 支持向量机回归 软测量 |
| 收稿时间： | 2019-06-04 |
Application of Multilayer Soft Measurement Modeling Based on Bayesian Inference in Butanol Fermentation |
| |
| ZHU Xiang-lin,GU Wen-wei,WANG Bo.Application of Multilayer Soft Measurement Modeling Based on Bayesian Inference in Butanol Fermentation[J].Introduction of Educational Technology,2020,19(4):183-188. |
| |
| Authors: | ZHU Xiang-lin GU Wen-wei WANG Bo |
| |
| Institution: | School of Electrical and Information Engineering， Jiangsu University， Zhenjiang 212000， China |
| |
| Abstract: | In order to to solve the problem that the quality parameters of fermentation products in butanol production process are difficult to be measured in real time， the accuracy of existing measurement methods is not high， and the measurement results are greatly affected by uncertainties， a multi-layer soft sensor modeling method based on Bayesian inference and support vector machine regression （SVR） is proposed. Firstly， Bayesian inference is used to calculate posterior probability to screen bias data， then the bias data is calibrated to establish the first-level SVR model. Secondly， Bayesian inference is used to calibrate the second-level SVR model， and the output of the first-level SVR model is modified to obtain the final prediction results， which overcomes the inaccuracy caused by interference and bias. When the multi-layer support vector regression （Bi-SVR） prediction model based on Bayesian inference is applied to butanol fermentation process， the simulation and experimental results show that the prediction accuracy of the system is improved by 4.52% under low disturbance and 5.37% under high disturbance， compared with the traditional SVR prediction model. |
| |
| Keywords: | microbial fermentation Bayesian inference support vector machine regression soft measurement |
|
| 点击此处可从《教育技术导刊》浏览原始摘要信息 |
| 点击此处可从《教育技术导刊》下载免费的PDF全文 |
