Scholarlink

<p>Global climate change and water scarcity pose severe challenges to sustainable agricultural development in desert regions. Aquaponics systems, as an efficient ecological agricultural model, demonstrate tremendous potential for optimizing resource utilization. However, they face complex water resource management challenges under extreme climatic conditions. Traditional linear control models cannot effectively capture the complex nonlinear relationships among environmental variables, while existing nonlinear models exhibit suboptimal performance in prediction accuracy and system robustness. To address this technological bottleneck, this study developed an intelligent water resource prediction framework based on ElasticNet Multi-Kernel Hybrid Support Vector Machine (EN-MH-SVM). The framework integrates ElasticNet feature selection with multi-kernel SVM algorithms, effectively capturing complex nonlinear dependencies among variables under extreme environmental conditions through a combination strategy of radial basis function, polynomial kernel, and sigmoid kernel. The research conducted comprehensive validation using over 10,000 sample data points, covering three typical scenarios: steady-state (1,000 samples), dynamic (6,378 samples), and complex random environment (2,632 samples). Model performance was evaluated through systematic comparison with traditional PLC control, single-kernel SVM, random forest, deep neural networks, and genetic algorithm and particle swarm optimization models. Experimental results demonstrate that the EN-MH-SVM model achieved prediction accuracy exceeding 92% under all testing conditions, significantly outperforming all comparative models. The model exhibited excellent performance in water resource consumption optimization and system robustness, particularly demonstrating outstanding adaptability under dynamic and complex environmental conditions. The research findings provide an effective technical solution for intelligent agricultural water resource management in desert regions and offer new research directions for development in time series modeling, data augmentation, and renewable energy integration.</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

Intelligent water resource optimization in desert greenhouse aquaponics: an ElasticNet multi-kernel SVM approach for sustainable agriculture

  • 期刊论文

  • 发表日期：

    2025年9月12日

查看全文
  • Wei Wang,
  • Jiukai Liu,
  • Tao Wen,
  • Sina Dang,
  • Jue Qu

摘要

Global climate change and water scarcity pose severe challenges to sustainable agricultural development in desert regions. Aquaponics systems, as an efficient ecological agricultural model, demonstrate tremendous potential for optimizing resource utilization. However, they face complex water resource management challenges under extreme climatic conditions. Traditional linear control models cannot effectively capture the complex nonlinear relationships among environmental variables, while existing nonlinear models exhibit suboptimal performance in prediction accuracy and system robustness. To address this technological bottleneck, this study developed an intelligent water resource prediction framework based on ElasticNet Multi-Kernel Hybrid Support Vector Machine (EN-MH-SVM). The framework integrates ElasticNet feature selection with multi-kernel SVM algorithms, effectively capturing complex nonlinear dependencies among variables under extreme environmental conditions through a combination strategy of radial basis function, polynomial kernel, and sigmoid kernel. The research conducted comprehensive validation using over 10,000 sample data points, covering three typical scenarios: steady-state (1,000 samples), dynamic (6,378 samples), and complex random environment (2,632 samples). Model performance was evaluated through systematic comparison with traditional PLC control, single-kernel SVM, random forest, deep neural networks, and genetic algorithm and particle swarm optimization models. Experimental results demonstrate that the EN-MH-SVM model achieved prediction accuracy exceeding 92% under all testing conditions, significantly outperforming all comparative models. The model exhibited excellent performance in water resource consumption optimization and system robustness, particularly demonstrating outstanding adaptability under dynamic and complex environmental conditions. The research findings provide an effective technical solution for intelligent agricultural water resource management in desert regions and offer new research directions for development in time series modeling, data augmentation, and renewable energy integration.