青岛能源所实现生物乙醇发酵从“人工补料”到“AI智控”的智能制造新突破

Core Viewpoint - The article discusses the development of an intelligent feedback control system for bioethanol fermentation, which integrates online Raman spectroscopy and deep learning to enhance the precision and real-time control of carbon source concentration during fermentation, addressing the limitations of traditional methods [1][7]. Group 1: System Innovation - The research team developed a spectral-time concatenation convolutional neural network (STC-CNN) model that utilizes non-destructive data streams from high-frequency online Raman spectroscopy to overcome issues such as prediction lag and lack of labeled data [3]. - Key innovations of the system include the introduction of a time series concatenation mechanism to capture dynamic changes in fermentation, a pseudo-label data augmentation strategy that expands the training sample size by 100 times, and the integration of a Kalman filter to enhance robustness under complex conditions [3][4]. Group 2: Performance Metrics - In practical applications, the system demonstrated significant advantages, achieving over 95% correlation between model predictions and offline HPLC measurements, with the maximum prediction deviation reduced from 8.3 g/L to 2.63 g/L [4]. - When controlling glucose concentration at 30 g/L, ethanol yield increased to 140.68 g/L, an 11.9% improvement over traditional batch fermentation, while glycerol concentration decreased to 6.72 g/L, resulting in a 64.6% increase in the ethanol/glycerol ratio [4][6]. Group 3: Broader Implications - The system not only addresses the precision control challenges in bioethanol production but also provides critical support for the intelligent transformation of the fermentation industry, reducing reliance on manual experience and enhancing process stability and production efficiency [7]. - The STC-CNN architecture and Raman feedback control system have been validated across various fermentation scenarios, indicating broad applicability in food, biopharmaceuticals, and green energy sectors [7].