Scholarlink

<p>Efficient water management in agriculture has become increasingly critical due to resource constraints and climate variability. This study introduces an innovative two-phase smart irrigation framework that leverages Internet of Things (IoT) technologies and Machine Learning (ML) to optimize water utilization and advance precision agriculture practices. In phase 1, we developed an IoT-based intelligent irrigation system that synthesizes data from multiple sensors monitoring soil moisture, temperature, rainfall, water flow, and air quality to facilitate real-time, crop-specific decision-making. By employing Edge Computing techniques with Arduino UNO and NodeMCU, this system successfully optimized water usage while reducing reliance on cloud services. Key innovations included the implementation of a Weighted Decision Index (WDI), dynamic threshold adjustments, and real-time alert mechanisms, resulting in a 25% reduction in water consumption, an 18% increase in crop yield, and a 35% enhancement in energy efficiency. Phase 2 involved the integration of ML models to predict irrigation needs based on a customized dataset. We trained and evaluated thirteen classifiers, with XGBoost demonstrated superior performance, achieving an accuracy of 99.96% and a Mean Squared Error (MSE) of 0.000433 following hyperparameter tuning via GridSearchCV and fivefold cross-validation. Statistical analysis using a t-test revealed significant performance disparities between XGBoost and other models, including Support Vector Classifier (SVC), Multi-Layer Perceptron (MLP), AdaBoost, logistic regression, and Gaussian Naïve Bayes (<i>p</i> &lt; 0.05), while ensemble methods such as LightGBM, random forest, and CatBoost exhibited comparable results. The proposed framework significantly surpasses existing methodologies, including SVR + k-means (MSE = 0.1), Random Forest (MSE = 0.551), K-Nearest Neighbors (<i>K</i> = 4, MSE = 0.521), and long short-term memory (LSTM) networks (MSE = 0.0072). By integrating the exceptional forecasting capabilities of XGBoost with IoT-enabled smart irrigation, this framework establishes a new benchmark in precision agriculture, providing a scalable, energy-efficient, and sustainable solution that effectively reduces water consumption while enhancing crop yield.</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

An intelligent framework for monitoring and irrigation prediction for precision agriculture

  • 期刊论文

  • 发表日期：

    2025年5月6日

查看全文
  • Rashmita Sahu,
  • Priyanka Tripathi

摘要

Efficient water management in agriculture has become increasingly critical due to resource constraints and climate variability. This study introduces an innovative two-phase smart irrigation framework that leverages Internet of Things (IoT) technologies and Machine Learning (ML) to optimize water utilization and advance precision agriculture practices. In phase 1, we developed an IoT-based intelligent irrigation system that synthesizes data from multiple sensors monitoring soil moisture, temperature, rainfall, water flow, and air quality to facilitate real-time, crop-specific decision-making. By employing Edge Computing techniques with Arduino UNO and NodeMCU, this system successfully optimized water usage while reducing reliance on cloud services. Key innovations included the implementation of a Weighted Decision Index (WDI), dynamic threshold adjustments, and real-time alert mechanisms, resulting in a 25% reduction in water consumption, an 18% increase in crop yield, and a 35% enhancement in energy efficiency. Phase 2 involved the integration of ML models to predict irrigation needs based on a customized dataset. We trained and evaluated thirteen classifiers, with XGBoost demonstrated superior performance, achieving an accuracy of 99.96% and a Mean Squared Error (MSE) of 0.000433 following hyperparameter tuning via GridSearchCV and fivefold cross-validation. Statistical analysis using a t-test revealed significant performance disparities between XGBoost and other models, including Support Vector Classifier (SVC), Multi-Layer Perceptron (MLP), AdaBoost, logistic regression, and Gaussian Naïve Bayes (p < 0.05), while ensemble methods such as LightGBM, random forest, and CatBoost exhibited comparable results. The proposed framework significantly surpasses existing methodologies, including SVR + k-means (MSE = 0.1), Random Forest (MSE = 0.551), K-Nearest Neighbors (K = 4, MSE = 0.521), and long short-term memory (LSTM) networks (MSE = 0.0072). By integrating the exceptional forecasting capabilities of XGBoost with IoT-enabled smart irrigation, this framework establishes a new benchmark in precision agriculture, providing a scalable, energy-efficient, and sustainable solution that effectively reduces water consumption while enhancing crop yield.