Basak (2025) Smart Irrigation Control Using IOT Sensors and Machine Learning for Optimized Water Management

Identification

• Journal: International Journal for Research in Applied Science and Engineering Technology
• Year: 2025
• Date: 2025-12-26
• Authors: Upangshu Basak
• DOI: 10.22214/ijraset.2025.76574

Research Groups

• Heritage Institute of Technology, Kolkata
• Institute of Engineering and Management, Kolkata

Short Summary

This project developed an intelligent IoT-based irrigation system integrating real-time sensor data with a K-Nearest Neighbors (KNN) machine learning model to automate water management, achieving approximately 66% accuracy in predicting irrigation needs and demonstrating water conservation.

Objective

• To develop an intelligent, automated irrigation system that leverages IoT-based sensor data and machine learning algorithms to optimize water usage and improve agricultural efficiency.
• To design and implement a hardware prototype that collects real-time environmental data such as soil moisture, temperature, and humidity using IoT sensors.
• To build and train a K-Nearest Neighbors (KNN) machine learning model that predicts whether irrigation should be ON or OFF based on sensor inputs.
• To integrate the ML model with the microcontroller-based hardware to enable automatic control of the water pump, with an optional manual override mode.
• To evaluate the performance of the ML model using appropriate accuracy metrics.
• To analyze the potential benefits of the system in terms of water conservation, cost efficiency, and scalability for precision agriculture.

Study Configuration

• Spatial Scale: Prototype system designed for small-scale to large-scale agricultural applications.
• Temporal Scale: Real-time data collection from sensors at regular intervals, with the machine learning model trained on historical environmental data.

Methodology and Data

• Models used: K-Nearest Neighbors (KNN) classifier.
• Data sources:
  • Real-time sensor data from a capacitive soil moisture sensor and a DHT11 temperature and humidity sensor.
  • A dataset named "TARP.csv" comprising 100,000 records, generated through a combination of simulated and sensor-driven methods, containing soil moisture, temperature, humidity, and a binary pump status (ON/OFF) as the target variable.

Main Results

• The K-Nearest Neighbors (KNN) machine learning model achieved an overall prediction accuracy of approximately 65.9%.
• The model demonstrated better performance in predicting the "Pump ON" state with a precision of 67%, recall of 73%, and F1-score of 70%.
• The prototype successfully demonstrated automated irrigation behavior that intelligently responds to varying soil conditions, conserving water and eliminating the need for manual monitoring.
• The system provides rapid, real-time predictions for pump control via an ESP32 microcontroller.

Contributions

• Integration of IoT hardware (ESP32 microcontroller, soil moisture, temperature, and humidity sensors) with a machine learning model (KNN) to create an intelligent, automated irrigation system.
• Development of a low-cost, scalable, and user-friendly practical framework for smart irrigation suitable for diverse farm sizes.
• Demonstration of effective water conservation through automated, intelligent irrigation decisions based on real-time environmental data.
• Prioritization of "Pump ON" prediction reliability to mitigate the risks of under-irrigation, aligning with critical agricultural needs.

Funding

No specific funding projects, programs, or reference codes were mentioned in the paper. The research was supported by academic infrastructure and resources from the affiliated institutions.

Citation

@article{Basak2025Smart,
  author = {Basak, Upangshu},
  title = {Smart Irrigation Control Using IOT Sensors and Machine Learning for Optimized Water Management},
  journal = {International Journal for Research in Applied Science and Engineering Technology},
  year = {2025},
  doi = {10.22214/ijraset.2025.76574},
  url = {https://doi.org/10.22214/ijraset.2025.76574}
}