DeJoule: AI Powered Building Management System for Real-Time Hospital Cooling Optimization
India
Environment
Implementing Organisation
Smart Joules Pvt. Ltd.
India, Kerala
Implementing Point of Contact
Listin Mathew
Contributor of the Impact Story
International Energy Agency
Year of implementation
2026
Problem statement
Commercial buildings in India, especially hospitals, consume a lot of energy, with cooling systems using nearly 30% of a building’s electricity. Hospitals operate 24/7 with constantly changing occupancy and activity levels, so cooling needs vary throughout the day and across seasons. At Kerala’s leading 600-bed super specialty hospital, the 430 TR chiller plant with pumps and cooling towers was managed using fixed schedules and manual adjustments. Operators had to guess when to turn equipment on or off and at what speeds, often running air conditioning at full power even when rooms were empty or the weather was cool. This “one size fits all” approach led to energy waste, slow response to sudden changes in cooling demand, constant operator intervention, and extra wear on equipment. The system consumed 0.87 kW per TR, well above efficient benchmarks below 0.80 kW/TR, showing a clear need for a solution that could match cooling output to real-time demand while maintaining precise climate control for patient care.
Submission Overview
Smart Joules is a company that develops and deploys AI-powered building automation systems. They own and operate DeJoule, an AI-driven energy management platform that uses reinforcement learning and deep learning algorithms to optimize cooling systems in commercial buildings. The solution has been commercially deployed in more than 39 buildings across India.
AI Technology Used
Key Outcomes
Hospitals in India often operate cooling systems on fixed schedules and manual controls, running pumps, cooling towers, and air conditioning at constant speeds regardless of occupancy or weather—resulting in energy use around 0.87 kW/TR, above best-in-class benchmarks below 0.80 kW/TR. DeJoule applies reinforcement learning and deep learning to continuously analyze operational data, occupancy trends, weather patterns, and equipment performance, automatically optimizing chiller sequencing and system speeds in real time. The platform anticipates demand spikes, adjusts for humidity and seasonal shifts, and adapts as equipment ages, ensuring cooling is delivered efficiently and reliably. Deployments have reduced overall energy consumption by up to 15%, improving efficiency from 0.87 to 0.74 kW/TR, saving hundreds of kilowatt-hours daily while lowering annual CO₂ emissions. With continuous monitoring and automated alerts to prevent disruptions, the solution sustains performance through monsoons, peak summers, and equipment wear, and has been commercially deployed across more than 39 buildings in India, delivering an average 7% energy improvement across healthcare, commercial, and industrial facilities.
Impact Metrics
Improvement in chiller plant energy efficiency
Baseline Value
0.87 kW per Ton of Refrigeration (TR)
Post-Implementation
After implementation, energy consumption per Ton of Refrigeration (TR) was reduced to 0.74 kW/TR.
Daily energy savings achieved through AI-optimized cooling
Baseline Value
NA
Post-Implementation
Post-deployment, the system reduced daily energy use by 525 kWh.
Annual greenhouse gas emissions avoided through optimized cooling operations
Baseline Value
NA
Post-Implementation
The optimized cooling operations reduced greenhouse gas emissions to approximately 192 metric tonnes of CO₂ annually.
Implementation Context
Super-specialty hospital in Kerala and commercially deployed across over 39 buildings nationwide, including hospitals, commercial, and industrial facilities.
Key Partnerships
Replicability & Adaptation
1. DeJoule's design allows it to work alongside existing building management systems rather than replacing them, reducing adoption costs and disruption. 2. A layered approach makes it suitable for hospitals, commercial buildings and industrial facilities nationwide.
* The data presented is self-reported by the respective organisations. Readers should consult the original sources for further details.