Scholarlink

<p>Effective monitoring of microalgae cultivation is essential for optimizing energy utilization efficiency. This paper presents a deep neural network algorithm that combines a self-search mechanism with a group convolution strategy for predicting the concentration of the microalga <i>Chlorella vulgaris</i> FACHB-9 in real-time. The proposed algorithm achieves a prediction accuracy of 0.93 on sparse datasets, significantly outperforming traditional residual networks (0.72). The group convolution module demonstrates strong feature extraction capabilities in the datasets with uniform color characteristics, while the self-search architecture reduces the risk of underfitting in sparse datasets and thus lowers model training costs. The experimental prototype developed using this method showed excellent performance in field tests for online quantitative analysis. This study provides a cost-effective and feasible machine learning solution for online monitoring of energy microalgae cultivation in small dataset scenarios. These findings demonstrate the potential of the proposed approach for scalable, real-time monitoring in microalgae cultivation.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Online quantitative monitoring of microalgae concentration using a coupled neural network with self-search and group convolution

  • 期刊论文

  • 发表日期：

    2025年12月22日

查看全文
  • Yang Peng,
  • Shen Yao,
  • Huaichun Zhou,
  • Yang Chen,
  • Zhenyu Zheng,
  • Xun Gong,
  • Fanwei Meng

摘要

Effective monitoring of microalgae cultivation is essential for optimizing energy utilization efficiency. This paper presents a deep neural network algorithm that combines a self-search mechanism with a group convolution strategy for predicting the concentration of the microalga Chlorella vulgaris FACHB-9 in real-time. The proposed algorithm achieves a prediction accuracy of 0.93 on sparse datasets, significantly outperforming traditional residual networks (0.72). The group convolution module demonstrates strong feature extraction capabilities in the datasets with uniform color characteristics, while the self-search architecture reduces the risk of underfitting in sparse datasets and thus lowers model training costs. The experimental prototype developed using this method showed excellent performance in field tests for online quantitative analysis. This study provides a cost-effective and feasible machine learning solution for online monitoring of energy microalgae cultivation in small dataset scenarios. These findings demonstrate the potential of the proposed approach for scalable, real-time monitoring in microalgae cultivation.