CO2-based recommendation system for predictive ventilation schemes in different real-life building applications using deep learning
Authors/Creators
- 1. EDP NEW
-
2.
Aristotle University of Thessaloniki
Description
The article presents a data-driven, low-intrusiveness system that uses historical CO₂ sensor data and deep learning to predict indoor CO₂ concentrations and provide ventilation recommendations. The system reduces the need for extensive IoT infrastructure by relying solely on CO₂ data. Six machine learning models (including LSTM, TCN, GRU, and S2S) are evaluated for their predictive accuracy across various room sizes and settings. The best-performing model achieved an R² of 0.97 and an RMSE of 50.46 ppm in large rooms. Real-world pilot studies in Spain, Portugal, Denmark, and Greece validated the system’s effectiveness in maintaining healthy indoor air quality through proactive window-opening guidance. The model proves to be scalable, cost-effective, and suitable for residential, institutional, and office environments.
Files
1-s2.0-S2352484725001507-main.pdf
Files
(10.5 MB)
| Name | Size | Download all |
|---|---|---|
|
md5:b3f438628d4fde46d2b92dbea93b0851
|
10.5 MB | Preview Download |
Additional details
Funding
Dates
- Available
-
2025