An Improved IoT Based Hybrid Predictive Load Forecasting Model for a Greenhouse Integrated With Demand Side Management

Greenhouse farming enhances agricultural productivity but remains highly energy-intensive, requiring advanced energy management strategies to ensure sustainability. Traditional load forecasting and demand-side management (DSM) methods often fall short in adapting to the dynamic and highly variable environmental conditions within greenhouses. This study hypothesizes that combining hybrid machine learning models with IoT-based DSM can optimize energy consumption while maintaining critical microclimatic conditions for crop growth. A hybrid predictive model integrating Extreme Gradient Boosting (XGBoost) for static feature learning and Long Short-Term Memory (LSTM) for sequential pattern recognition is proposed, coupled with a priority-based DSM framework deployed on a Raspberry Pi IoT platform. The model was trained and tested using one year of real-world greenhouse data, achieving a Mean Absolute Percentage Error (MAPE) of 5.3%, reducing grid energy dependency by 7.1%, and lowering the average electricity cost by 17.3%. These results demonstrate that the proposed system offers a scalable, economically viable, and sustainable solution for intelligent energy management in greenhouse environments, significantly advancing the integration of AI and DSM technologies in controlled agriculture. A laboratory-scale prototype was developed for physical verification using an EcoSense 500W solar microgrid, Raspberry Pi 3B+, relay modules, current sensors (SCT-013, ADS1115 ADC), and load simulators. Furthermore, this work contributes directly to Sustainable Development Goals (SDGs) 2, 7, and 12 by promoting food security, increasing renewable energy utilization, and enhancing responsible resource consumption in greenhouse farming.