Scholarlink

<p>Atmospheric Water Harvesting (AWH) has emerged as a promising approach for addressing freshwater scarcity in arid and water-stressed regions. This study presents a machine learning-based framework for predicting and optimizing water production in a desiccant-driven AWH system using experimental data. Gaussian Process Regression (GPR) was employed to model four performance indicators, including Water Production (WP), Cumulative Water Production (CWP), and their logarithmic transformations. To improve predictive performance, Differential Evolution (DE) and Genetic Algorithm (GA) were integrated for hyperparameter optimization. The results demonstrated that log-transformed models significantly improved prediction accuracy, achieving an R² value of 0.981 compared with 0.649 and 0.664 for the standard WP and CWP models, respectively. Humidity, outdoor temperature, and thermal regulation parameters were identified as the dominant factors affecting water production efficiency. The optimized hybrid DE-GA framework provided stable and reliable predictions across varying environmental and operational conditions. The findings demonstrate the potential of integrating machine learning and evolutionary optimization for improving the performance of AWH systems. This framework can support the development of decentralized and energy-efficient water supply solutions, particularly in regions facing increasing water stress and climate-related challenges.</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

Optimized Desiccant-Driven Atmospheric Water Harvesting: Machine Learning and Evolutionary Algorithms for Sustainable Water Resource Management

  • 期刊论文

  • 发表日期：

    2026年6月25日

查看全文
  • Seyed Amir Ahghar,
  • Elshan Soltani,
  • AmirHamzeh Farajollahi

摘要

Atmospheric Water Harvesting (AWH) has emerged as a promising approach for addressing freshwater scarcity in arid and water-stressed regions. This study presents a machine learning-based framework for predicting and optimizing water production in a desiccant-driven AWH system using experimental data. Gaussian Process Regression (GPR) was employed to model four performance indicators, including Water Production (WP), Cumulative Water Production (CWP), and their logarithmic transformations. To improve predictive performance, Differential Evolution (DE) and Genetic Algorithm (GA) were integrated for hyperparameter optimization. The results demonstrated that log-transformed models significantly improved prediction accuracy, achieving an R² value of 0.981 compared with 0.649 and 0.664 for the standard WP and CWP models, respectively. Humidity, outdoor temperature, and thermal regulation parameters were identified as the dominant factors affecting water production efficiency. The optimized hybrid DE-GA framework provided stable and reliable predictions across varying environmental and operational conditions. The findings demonstrate the potential of integrating machine learning and evolutionary optimization for improving the performance of AWH systems. This framework can support the development of decentralized and energy-efficient water supply solutions, particularly in regions facing increasing water stress and climate-related challenges.