Smart Classroom CO2 and Comfort Monitor with Ventilation Dashboard

A classroom of forty students exhales a surprising amount of CO2 — levels routinely climb past 1,000 ppm in a closed room, and the research literature associates that stuffy range with drowsiness and poorer concentration. Yet nobody in the room can feel CO2 rising; ventilation happens by habit or not at all. Commercial indoor-air monitors exist but are priced per room and opaque inside. This project builds the capability as an understandable student system: an MH-Z19-class NDIR CO2 sensor (true CO2 measurement, not an estimate from VOCs) plus a DHT22 for temperature and humidity, read by an ESP32 and streamed over Wi-Fi to a companion dashboard. The dashboard plots CO2, temperature and humidity trends, computes a simple comfort/ventilation recommendation from the data, and logs high-CO2 episodes with timestamps. The hardware demonstrates NDIR sensing and telemetry; the software demonstrates turning raw sensor data into a decision a teacher can act on.

1. The MH-Z19-class NDIR sensor measures CO2 via infrared absorption and reports ppm over UART; the DHT22 reports temperature and humidity alongside.
2. Every 30 seconds the ESP32 publishes the triplet over Wi-Fi (MQTT/HTTP) to the dashboard backend and updates the onboard OLED.
3. The dashboard renders live gauges — CO2 with green/amber/red bands around the 1,000 ppm stuffy-air guideline — plus rolling trend charts.
4. The recommendation engine watches CO2 level and its trend: sustained readings above threshold produce a "ventilate now" state; falling readings clear it.
5. Each threshold crossing opens a timestamped episode in the log with peak CO2 and duration; closing it records the recovery.
6. On the hardware, the RGB LED shifts green→amber→red with CO2 bands and the buzzer chirps on entering the red band.
7. The student runs the unit through a real occupied-room session and exports the day's chart and episode log as the report's measured evidence.

Tech stack:

• ESP32 development board (Wi-Fi telemetry)
• MH-Z19-class NDIR CO2 sensor (UART)
• DHT22 temperature/humidity sensor
• 0.96-inch OLED display + RGB LED + buzzer
• MQTT/HTTP telemetry
• Companion web dashboard (gauges, charts, recommendations)
• Arduino IDE (C/C++ firmware)

• [True NDIR CO2 sensing] An MH-Z19-class NDIR sensor measures actual CO2 (0–5,000 ppm range class) — not a VOC-derived estimate — with UART output to the ESP32.
• [Temperature and humidity] A DHT22 adds the other two comfort variables, so the dashboard reasons about the full indoor environment.
• [Live ventilation dashboard] Current CO2, temperature and humidity render as gauges with color bands, plus rolling trend charts for the school day.
• [Data-driven ventilation advice] The dashboard computes a recommendation — ventilate now / air is fine — from CO2 level and trend, not from a timer.
• [High-CO2 episode log] Every crossing of the stuffy-air threshold is timestamped and logged with peak value and duration, building an evidence record for the report.
• [Local alert] An RGB LED and buzzer on the unit signal rising CO2 in the room itself, so action does not depend on someone watching the dashboard.
• [Wi-Fi telemetry] Readings publish every 30 seconds over Wi-Fi (MQTT/HTTP) to the dashboard; the unit also shows live values on its small OLED.

• Working hardware prototype: ESP32, NDIR CO2 sensor, DHT22, OLED, LED, buzzer
• Companion dashboard web app (gauges, charts, episode log, ventilation advice)
• Complete firmware (sensing, telemetry, local alerting)
• Sensor warm-up and calibration notes (NDIR ABC calibration explained)
• Wiring diagram and MQTT/HTTP setup notes
• Project report PDF (background, NDIR sensing theory, comfort science basis, methodology, logged session)
• PPT presentation for final review
• Viva Q&A preparation document

• The NDIR sensor needs a warm-up period (a few minutes) after power-up before readings stabilize — the firmware flags warming-up state explicitly.
• CO2 accuracy of roughly ±50 ppm + 5% (sensor-class expectation) suits trend and threshold use, not laboratory-grade measurement; the report states this plainly.
• The ventilation advice is a simple threshold-plus-trend rule, not a building-HVAC control signal — it advises a human, it does not drive dampers or fans in this build.
• The dashboard runs on the local network; internet-remote access needs a hosted broker, listed as future scope — no cloud features are claimed.
• One unit covers one room; whole-building coverage would need multiple units, discussed as future scope.