The problem EcoGrid solves

Every campus, hospital, school, and public space in India is managing its environment the same way it did 30 years ago, with people walking around and checking things manually.

The result:

Dustbins overflow mid week in canteens and corridors because nobody knew they were full

Enclosed labs, wards, and seminar halls accumulate CO₂ and pollutants for hours before anyone notices

Water pipes run through the night and electricity stays on in empty labs, discovered only when the bill arrives

These aren't failures of effort. They are failures of visibility.

EcoGrid Insights solves this by giving institutions a real-time nervous system for their environment:

Smart Waste: Ultrasonic sensors monitor bin fill levels and fire alerts before overflow happens. Staff are dispatched proactively, not reactively.

Air Quality: AQI is monitored continuously in enclosed spaces. When levels cross danger thresholds, air purifiers activate automatically, no human command needed.

Water & Electricity: Flow meters and load sensors track consumption per zone, flagging anomalies the same night they happen.

Every data point feeds an AI prediction engine that learns patterns over time, so EcoGrid doesn't just respond to problems, it prevents them.The platform works across colleges, schools, hospitals, bus stands, railway stations, and corporate offices, any space where people gather and resources are consumed invisibly.

Challenges we ran into

1. MQ135 sensor instability on ESP8266
   The air quality sensor was the most frustrating component to stabilise. On the ESP8266, the single analog pin (A0) is shared across the board and picks up significant electrical noise, causing wildly inconsistent readings, values would jump 150 points between consecutive reads. The purifier relay was switching on and off every 10 seconds near the threshold, which is both functionally useless and damaging to the hardware.

The fix required three layers:

Averaged sampling: taking 20 readings, sorting them, and discarding the top and bottom outliers before averaging the rest

Hysteresis band: the purifier turns ON at AQI 420 but only turns OFF when AQI drops below 340, preventing relay chatter near the boundary

30-second warm-up delay: the MQ135 filament needs thermal stabilisation before readings are valid; skipping this produced garbage data on cold starts

2. Firebase Firestore rate limiting
   Writing all three bins and air data every 10 seconds hit Firestore's free-tier write limits during extended testing. The fix was tracking state locally and only writing to Firestore when values actually changed beyond a meaningful threshold, reducing writes by ~70% without losing data fidelity.

3. SSL time sync on ESP8266
   Firebase authentication kept failing with SSL errors until we added configTime() with NTP servers before initialising Firebase, a non-obvious fix that cost significant debugging time.