DT Guardian
A simulated, offline-first edge controller for a 200 kVA distribution transformer. An IS 2026-7 thermal model drives MILP dispatch of battery storage, EV charging and pump loads, with a fail-safe fallback when the forecast or solver fails.
- my role
- co-developer · team of 2
- stack
- Python · MILP (PuLP / HiGHS)
- when
- oct 2026
cat problem.md
Schneider Electric Yuva Yodha Energy Tech Hackathon, Challenge 3 (grid reliability and renewable intermittency), prototype round. The project aims to turn a distribution transformer's thermal headroom into neighbourhood flexibility: EV and pump demand shifts toward renewable-rich hours, and critical loads stay on during supply gaps.
The prototype is a Python simulator with a thermal twin, a load forecaster, a rolling-horizon dispatcher, baselines and an experiment pipeline. It is an offline-first edge controller.
tree pipeline/
cat results.tsv
| metric | result | setup |
|---|---|---|
| essential energy not served | 121.2 → 56.7 kWh | simulated 6-hour, battery-stressed outage |
| change vs no control | −53% | same outage |
| change vs battery-only | −31% | same outage |
| load forecast coverage | 80.2% | 80% nominal |
| automated tests | 365 | repository test suite |
All results are simulated. Neighbourhood loads and control experiments are synthetic; the weather input is real Lucknow reanalysis data.
cat limits.md
- No field deployment and no utility integration. This is a prototype-round entry.
- The electrical model uses a voltage proxy. OpenDSS integration and the dashboard are unfinished.
- The outage result covers a simulated 6-hour, battery-stressed outage only.
Code is private until the competition ends.