| 15 | 0 | 25 |
| 下载次数 | 被引频次 | 阅读次数 |
合成生物学技术极大提升了菌株改造效率，然而传统发酵优化方法依赖于专家经验且需要反复试错，开发效率低，难以满足大量高产菌种对应发酵工艺优化的需求。基于数据和模型驱动的智能化发酵优化方法可大幅提升发酵工艺优化的效率，然而缺乏成熟可靠的智能化模型严重限制了发酵工艺优化智能化的发展。针对以上问题，本研究提出一种基于卷积神经网络(convolutional neural network,CNN)、双向长短期记忆网络(bidirectional long short-term memory,BiLSTM)、注意力机制(attention mechanism,AM)多模型整合的发酵过程多时间步预测方法(命名为CNNBiLSTM-AM)并应用于发酵过程预测。首先，对提出的整合模型进行超参数优化，提高模型的预测精度；其次，通过消融实验和与各分模型的对比，对整合模型进行验证；最后，分别利用青霉素发酵数据集和左旋多巴发酵数据集进行模型应用。结果表明，本文模型在10个时间步长下，预测的决定系数大于0.9，能够利用已获得的发酵过程数据成功预测产物浓度多时间步后的变化。
Abstract：Synthetic biology technology has greatly improved the efficiency of strain modification. However,traditional fermentation optimization methods rely on expert experience and require repeated trial and error,which has low development efficiency and cannot meet the demand for optimization of fermentation processes corresponding to a large number of highyield strains. Data-and model-driven intelligent fermentation optimization methods can greatly improve the efficiency of fermentation process optimization. However,the lack of mature and reliable intelligent models severely limits the development of intelligent fermentation process optimization. In response to the above issues,this study proposes a multi-time step prediction method for fermentation processes based on the integration of convolutional neural network(CNN),bidirectional long short-term memory(BiLSTM) and attention mechanism(AM),named CNN-BiLSTM-AM model and applies it to fermentation process prediction. First,the proposed integrated model was optimized for hyperparameters to improve the model's predictive accuracy. Second,the integrated model was validated through ablation experiments and comparisons with each sub-model. Finally,the model was applied using penicillin fermentation data and levodopa fermentation data. The study found that at ten time steps,the determination coefficient was greater than 0.9,and the changes in product concentration at multiple future time steps have been successfully predicted using the obtained fermentation process data.
[1]夏建业，刘晶，庄英萍.人工智能时代发酵优化与放大技术的机遇与挑战[J].生物工程学报，2022,38(11):4180-4199.
[2]CHAI W Y,TEO K T K,TAN M K,et al. Fermentation process control and optimization[J]. Chemical engineering&technology,2022,45(10):1731-1747.
[3]GEORGIEV T,RATKOV A. Mathematical modelling of fed-batch fermentation processes for amino acid production[J]. Mathematics and computers in simulation,1997,44(3):271-285.
[4]WARNES M R,GLASSEY J,MONTAGUE G A,et al.On data-based modelling techniques for fermentation processes[J]. Process biochemistry,1996,31(2):147-155.
[5]DU S,LI T,YANG Y,et al. An LSTM based encoderdecoder model for MultiStep traffic flow prediction[C]//IEEE. 2019 International Joint Conference on Neural Networks(IJCNN). New York:IEEE,2019:1-8.
[6]GRAHOVAC J,JOKIC A,DOCID J,et al. Modelling and prediction of bioethanol production from intermediates and byproduct of sugar beet processing using neural networks[J]. Renewable energy,2016,85:953-958.
[7]DEL RIO-CHANONA E A ,MANIRAFASHA E ,ZHANG D,et al. Dynamic modeling and optimization of cyanobacterial C-phycocyanin production process by artificial neural network[J]. Algal research,2016,13:7-15.
[8]KE W ,HUANG D ,YANG F ,et al. Soft sensor development and applications based on LSTM in deep neural networks[C]//IEEE. 2017 IEEE Symposium Series on Computational Intelligence(SSCI). Honolulu:IEEE Computer Society,2017:1-6.
[9]CHANDIA R,GOYAL S,GUPTA R. Evaluation of deep learning models for multi-step ahead time series prediction[J]. IEEE Access,2021,9:83105-83123.
[10]JI C ,MA F ,WANG J ,et al. Profitability related industrial-scale batch processes monitoring via deep learning based soft sensor development[J]. Computers&chemical engineering,2023,170:108125.
[11]PETSAGKOURAKIS P,SANDOVAL I O,BRADFORD E,et al. Reinforcement learning for batch bioprocess optimization[J]. Computers&chemical engineering ,2020,133:106649.
[12]LU W,LI J,WANG J,et al. A CNN-BiLSTM-AM method for stock price prediction[J]. Neural computing and applications,2021,33(10):4741-4753.
[13]GOLDRICK S ,DURAN-VILLALOBOS C A ,JANKAUSKAS K,et al. Modern day monitoring and control challenges outlined on an industrial-scale benchmark fermentation process[J]. Computers&chemical engineering,2019,130:106471.
[14]BIROL G，ÜNDEY C,CINAR A. A modular simulation package for fed-batch fermentation:penicillin production[J]. Computers&chemical engineering,2002,26(11):1553-1565.
[15]XUA B ,YANG G. Interpretability research of deep learning:a literature survey[J]. Information fusion ,2024,115:102721.
[16]HUANG J C,GUO Q,LI X H,et al. A comprehensive review on the application of neural network model in microbial fermentation[J]. Bioresource technology ,2024,416:131801.
[17]ZHANG Q,YANG L T,CHEN Z,et al. A survey on deep learning for big data[J]. Information fusion,2018,42:146-157.
[18]HU X,CHU L,PEI J,et al. Model complexity of deep learning:a survey[J]. Knowledge and information systems,2021,63:2585-2619.
基本信息:
DOI：10.13364/j.issn.1672-6510.20240203
中图分类号:TQ920.6;TP183
引用信息:
[1]韩炫州,韩雪莉,徐超,等.基于CNN-BiLSTM-AM的发酵过程多时间步预测方法及其应用研究[J].天津科技大学学报,2026,41(01):20-28+60.DOI:10.13364/j.issn.1672-6510.20240203.
基金信息:
天津市重点研发计划项目（25ZXWCSY00190）; 山西省重点研发项目（202202140601018）; 合成生物学海河实验室项目（22HHSWSS00013）
2026-02-25
2026-02-25