HeatSchedule — Energy-aware heating for EPFL rooms
A LauzHack 2022 build — a Python framework that schedules room heating from real occupancy data, cutting energy use without compromising comfort.
Built in 24 hours at LauzHack 2022 with Martin Zwifl and Thomas Rimbot: a framework for scheduling room heating in EPFL buildings based on actual occupancy, designed to reduce energy use while preserving comfort.
The problem
University buildings heat empty rooms by default. EPFL has detailed room-reservation data per building, but no system uses it to drive HVAC schedules — heating runs on fixed timetables that ignore when rooms are actually booked. The wasted energy compounds across hundreds of rooms over a winter.
Approach
A Python framework with three parts:
- Occupancy signal — pulls room reservation schedules directly from EPFL’s booking system. Rooms with no reservations in a given time window are flagged as candidates for heating cutbacks.
- Heating curve parameterisation — each room has a configurable thermal model (warm-up time, set-point, decay rate) so the scheduler can pre-heat a room before its next reservation without over-shooting.
- Simulation framework — runs the schedule against a parameterised building model and reports projected energy savings vs. the baseline fixed schedule.
The whole thing is object-oriented so individual rooms, buildings, and heating policies can be swapped without rewriting the core scheduler.
Result & scope
A working 24-hour proof of concept that ingests EPFL reservation data and outputs heating schedules, plus a simulation environment to compare strategies. The natural next step is wiring the scheduler to a real HVAC controller — the abstraction layer is there, but live integration was out of scope for the hackathon.
Code: github.com/Thomas-debug-creator/LauzHack · Devpost: devpost.com/software/heatschedule