Reducing Carbon Footprint in Smart Buildings with Data Science
summary
A deployed system called Optimus Power uses historical sensor and weather data to build a thermal model and trains an agent with thermal model + reinforcement learning to control HVAC under user rules, then runs a four week live test across seven Portuguese bank offices and reports 20–35% energy savings. Real challenges are heterogeneous sensor APIs, poor data quality (one indoor sensor was installed outside), and occupant overrides during tests.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 1:12Smart Buildings
- 2:56Optimization
- 6:02Live Testing
- 6:57Energy Savings
- 7:33Challenges
- 8:00Integration
- 9:34Funding
- 10:34Conclusion
related talks
31mPredicting Optimal Temperature in The Transmission System
26mDriving towards a cleaner future through data
48mUrban Data & Decisions
29mOne Size Doesn't Fit All: How data can support tailored approaches in public health
28mHow to make your CEO believer in Data Science
31mKeeping up with the renewable energy transition
44mKEYNOTE: ClimateTech and the Future of the Planet 🌳
07mLightning Talk: Data Science - What Can Go Wrong with Carbon Footprint Calculations