Simplified Prediction-Based AI-IoT Model for Energy Management Scheme in Standalone PV Powered Greenhouse

• Published in: IEEE journal of emerging and selected topics in industrial electronics (Print) Vol. 6; no. 1; pp. 224 - 237
• Main Authors: Biswal, Soumya Ranjan, Choudhury, Tanmoy Roy, Santra, Subhendu Bikash, Panda, Babita, Mishra, Subhrajyoti, Padmanaban, Sanjeevikumar
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.01.2025; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Algorithms; Demand side management (DSM); Energy management; Green products; Greenhouses; Humidity; Internet of Things; Job shop scheduling; Load shifting; Machine learning; machine learning (ML); Photovoltaic cells; Predictive models; sensor-less system; Sensors; smart greenhouse; Soil moisture; Storage capacity; Sustainable development; Temperature dependence; Temperature distribution; Temperature sensors
• ISSN: 2687-9735; 2687-9743
• DOI: 10.1109/JESTIE.2024.3425670

Automated greenhouse is essential for sustainable development and food security. Photovoltaic (PV) power with physical sensors-based control using Internet of Things needs high initial investment and operational cost. This also needs significant installed storage capacity. In the proposed solution, the dependency on physical sensors like temperature, humidity, soil moisture sensors, etc., are eliminated due to the application of eXtreme Gradient Boosting-based machine learning (ML) algorithm. The training and testing of ML algorithm are performed with one-year physical data (approx. 50k @10 min interval) from greenhouse which provides accurate mapping (Temperature MAPE: 1.51%, R 2 : 0.9785 and Humidity MAPE: 1.68%, R 2 : 0.9867) between predicted and sensor data. Also, a novel priority-based demand side management scheme is implemented which includes load shifting which reduces the requirement of installed PV and storage capacity. A reduction of 63.27% storage capacity is possible with proposed control approach. ML algorithm is programmed using Python language and implemented in Raspberry Pi-3B+ SBC. For physical verification of the proposed control unit, a laboratory-based prototype is developed with PV emulator (1.5 kW), programmable electronic load box, and relay unit controlled through Arduino UNO, Raspberry Pi-3B+ SBC, ESP-32 Combo unit.