SMART IRRIGATION SYSTEM USING ARTIFICIAL NEURAL NETWORKS FOR SCHEDULING IRRIGATION IN AGRICULTURAL SETTINGS

ABSTRACT

This project employs Artificial Neural Networks (ANN) to train predictive models for optimizing crop watering schedules based on meteorological data and soil type. Focusing on three soil classifications - Nitisol, Lixisol, and Acrisol, within Etsako West Local Government Area, Edo State, Nigeria - meteorological factors including average temperature, precipitation, humidity, rainy days, and average sunlight serve as input features, while soil moisture readings act as target variables. Employing historical data from meteorological sources and literature, the ANN models are trained and subsequently deployed on Arduino microcontrollers to enable real-time soil moisture content prediction. The system incorporates logic to trigger crop watering when predicted soil moisture levels fall below predefined thresholds, ensuring optimal irrigation practices. Furthermore, the project evaluates the predictive accuracy of the ANN models by comparing the coefficient of determination (R2 ) between observed soil moisture readings and their corresponding ANN predictions for Nitisol (R2 = 0.85), Lixisol (R2 = 0.80), and Acrisol (R2 = 0.81). This integration of ANN-based predictive modelling with Arduino-based automation offers a practical and efficient solution for precision agriculture, contributing to enhanced crop yield and water resource management in agricultural settings.