欢迎访问过程工程学报, 今天是

过程工程学报 ›› 2023, Vol. 23 ›› Issue (9): 1351-1358.DOI: 10.12034/j.issn.1009-606X.222382CSTR: 32067.14.jproeng.222382

• 研究论文 • 上一篇    下一篇

基于SSAE-FCM的燃料乙醇分批发酵关键时间节点自动识别

田晓俊1， 王梦1， 刘小辰1， 郑淏月1， 林海龙1*， 刘劲松1， 杨萌2， 温广瑞2   

  1. 1. 国投生物科技投资有限公司，北京 100034 2. 西安交通大学机械工程学院，陕西 西安 710049
  • 收稿日期:2022-10-14 修回日期:2023-02-16 出版日期:2023-09-28 发布日期:2023-09-27
  • 通讯作者: 林海龙 linhailong@sdic.com.cn
  • 基金资助:
    2021年度智能制造示范工厂项目

Automatic identification method of batch time node of fuel ethanol fermentation based on SSAE-FCM

Xiaojun TIAN1,  Meng WANG1,  Xiaochen LIU1,  Haoyue ZHENG1,  Hailong LIN1*,  Jinsong LIU1,  Meng YANG2,  Guangrui WEN2   

  1. 1. SDIC Biotechnology Investment Co., Ltd., Beijing 100034, China 2. School of Mechanical Engineering, Xi'an Jiaotong University, Xi'an, Shaanxi 710049, China
  • Received:2022-10-14 Revised:2023-02-16 Online:2023-09-28 Published:2023-09-27

RichHTML

PDF

摘要： 生物工程日趋复杂化和智能化，其生产过程也从实验室向工业化规模快速发展，给生物工程过程控制优化带来了新的挑战。本工作以燃料乙醇分批发酵这一复杂过程为研究对象，提出了一种基于堆叠稀疏自编码器(SSAE)和模糊C均值聚类(FCM)相结合的燃料乙醇分批发酵关键时间节点自动识别方法。通过SSAE由低到高逐层提取发酵过程原始数据中更能反映数据本质属性的各级高层特征，并将其作为FCM算法的输入数据，最终构建燃料乙醇分批发酵关键时间节点自动识别模型。为评估模型性能，以国投生物燃料乙醇发酵过程为应用对象，结果表明，本工作所提出的方法具有可操作性。同时，与基于动力学模型和过程多参数相关性分析方法对比，本工作所提方法具有更优的识别性能。

关键词: 燃料乙醇, 发酵过程, 堆叠稀疏自编码器, 模糊C均值聚类, 自动识别

Abstract: The control and optimization of the fermentation process are the key technologies in batch production of fuel ethanol, it has been attracting continuous attention in the control and optimization of a biological fermentation plant. Although many detection methods are proposed in the literature, the increasing demand for green renewable energy and the increasing reliability of biological fermentation plant also brings a new challenge to the control and optimization of the fermentation process. On the one hand, fuel ethanol has developed from laboratory scale to industrial scale production, but the traditional detection methods are also useful. On the other hand, the plant becomes extremely complex and intelligent which made the control optimization process more difficult. As a response, an automatic identification method for key time nodes of batch fermentation of fuel ethanol based on stacked sparse autoencoder (SSAE) and fuzzy C-means clustering (FCM) is proposed. Firstly, SSAE is employed to extract the high-level features of the original data (boolean data, process data, and energy consumption data) layer by layer from low to high levels. Then, the high-level features that reflect the essential attributes of the data are taken as the input data of FCM clustering operation, and an automatic identification model of batch time nodes in fuel ethanol fermentation based on FCM is established. Finally, the batch process data of fuel ethanol fermentation from SDIC Bioenergy Company is used as training samples to validate the SSAE-FCM. And the two indicators of recognition accuracy (RV) and recognition speed (RN) are employed to characterize the effect of the detection model. In contrast, the control of the fuel ethanol fermentation process based on kinetic models and multiparameter correlation analysis method are introduced to carry on optimization results under the same condition. By comparison, the results show that the method proposed in this work has better identification performance, which satisfies the requirements of batch process control of ethanol fermentation.

Key words: fuel-ethanol, fermentation process, stacked sparse autoencoder, fuzzy C-means clustering, Automatic identification