Kanwal Saima, Abdullah Muhammad, Kumar Sahil, Arshad Saqib, Shahroz Muhammad, Zhang Dawei, Kumar Dileep
Engineering Research Centre of Optical Instrument and Systems, Ministry of Education and Shanghai Key Lab of Modern Optical System, University of Shanghai for Science and Technology, No. 516 Jun Gong Road, Shanghai 200093, China.
Department of Biomedical Engineering, School of Health Science and Engineering, University of Shanghai for Science and Technology, No. 516 Jun Gong Road, Shanghai 200093, China.
Sensors (Basel). 2024 Dec 8;24(23):7842. doi: 10.3390/s24237842.
Smart fish farming faces critical challenges in achieving comprehensive automation, real-time decision-making, and adaptability to diverse environmental conditions and multi-species aquaculture. This study presents a novel Internet of Things (IoT)-driven intelligent decision-making system that dynamically monitors and optimizes water quality parameters to enhance fish survival rates across various regions and species setups. The system integrates advanced sensors connected to an ESP32 microcontroller, continuously monitoring key water parameters such as pH, temperature, and turbidity which are increasingly affected by climate-induced variability. A custom-built dataset comprising 43,459 records, covering ten distinct fish species across diverse pond environments, was meticulously curated. The data were stored as a comma-separated values (CSV) file on the IoT cloud platform ThingSpeak and synchronized with Firebase, enabling seamless remote access, control, and real-time updates. Advanced machine learning techniques, with feature transformation and balancing, were applied to preprocess the dataset, which includes water quality metrics and species-specific parameters. Multiple algorithms were trained and evaluated, with the Decision Tree classifier emerging as the optimal model, achieving remarkable performance metrics: 99.8% accuracy, precision, recall, and F1-score, a 99.6% Matthews Correlation Coefficient (MCC), and the highest Area Under the Curve (AUC) score for multi-class classification. Our framework's capability to manage complex, multi-species fishpond environments was validated across diverse setups, showcasing its potential to transform fish farming practices by ensuring sustainable climate-adaptive management through real-time water quality optimization. This study marks a significant step forward in climate-smart aquaculture, contributing to enhanced fish health, survival, and yield while mitigating the risks posed by climate change on aquatic ecosystems.
智能养鱼在实现全面自动化、实时决策以及适应多样环境条件和多品种养殖方面面临着严峻挑战。本研究提出了一种新型的物联网（IoT）驱动的智能决策系统，该系统可动态监测和优化水质参数，以提高不同地区和品种养殖环境下的鱼类存活率。该系统集成了连接到ESP32微控制器的先进传感器，持续监测诸如pH值、温度和浊度等关键水质参数，这些参数正日益受到气候引起的变化的影响。精心整理了一个包含43459条记录的自定义数据集，涵盖了不同池塘环境中的十种不同鱼类品种。数据以逗号分隔值（CSV）文件的形式存储在物联网云平台ThingSpeak上，并与Firebase同步，实现无缝远程访问、控制和实时更新。应用了具有特征转换和平衡功能的先进机器学习技术对数据集进行预处理，该数据集包括水质指标和特定品种参数。对多种算法进行了训练和评估，决策树分类器成为最优模型，取得了卓越的性能指标：准确率、精确率、召回率和F1分数均为99.8%，马修斯相关系数（MCC）为99.6%，多类分类的曲线下面积（AUC）分数最高。我们的框架管理复杂多品种鱼塘环境的能力在不同设置中得到了验证，通过实时水质优化确保可持续的气候适应性管理，展示了其改变养鱼实践的潜力。本研究标志着气候智能型水产养殖向前迈出了重要一步，有助于提高鱼类健康、存活率和产量，同时减轻气候变化对水生生态系统造成的风险。
