AI-Based Energy Demand Forecasting for a Smart Farm

ABSTRACT: Modern agricultural systems increasingly struggle with inefficient resource utilization, rising energy costs, and the absence of predictive decision-support tools for farmers. This paper presents the design and implementation of the AI-driven Smart Farm Energy Forecasting System, a platform developed to address these challenges by integrating machine learning and deep learning based energy prediction with smart irrigation advisory modules. The system utilize key environmental parameters temperature, humidity, soil moisture, rainfall, and sunlight hours which are processed through a trained XGBoost and LSTM Hybrid model to generate real-time forecasts of farm-level energy demand.

The platform features a streamlit based web application supported for role-based user interaction enabling farmers to instantly obtain energy predictions and workable guidance. The methodology includes automated irrigation-type classification, where the system selects drip, sprinkler, or manual irrigation types based on environmental conditions, alongside rule-based recommendations. To improve interpretability, Explainable AI (XAI) components using SHAP values provide insight into feature contributions and model behavior.

Key functionalities demonstrate a fully functional system capable of real-time energy demand forecasting, environmental-aware irrigation recommendations, and adaptive scheduling that identifies optimal irrigation periods and watering durations. By combining predictive analytics and explainable modeling, the Smart Farm Energy Forecasting System establishes a scalable framework for energy-efficient, cost-effective, and sustainable smart farm management.

Keywords - Energy demand forecasting, XGBoost, LSTM, Explainable AI (XAI), Smart Scheduling