ThermoStat

An ESP32-based smart thermostat built with MicroPython — no cloud, no app, no subscription. The ESP32 is the entire stack: thermostat brain, IR transmitter, web server, and database.

Controls a Daikin mini-split AC and a backup Pelonis oil heater via IR, reads room temperature from a DHT22 sensor, and hosts a single-page web UI served directly from the microcontroller.

---

What makes this interesting

Most "smart thermostat" projects wire a microcontroller to a relay and call it done. This one handles the logic that commercial thermostats get wrong:

Reverse-engineered IR protocols. The Daikin ARC433 remote sends a stateful 19-byte frame — every command includes the full AC state, not a diff. This project builds and transmits that frame directly from MicroPython, giving full control over mode, fan speed, and setpoint without the Daikin's unreliable built-in auto mode.

Adaptive learning. The schedule supports pre-conditioning — start heating N minutes before you arrive home. The adaptive optimizer measures actual heating rate each day (how many minutes per degree °F) and adjusts the lead time automatically. It converges on the right value without manual tuning.

Served entirely from the ESP32. The web app is a single-page React-style UI with a live dashboard, Gantt-style schedule editor, and canvas history chart — all served from the ESP32's flash over WiFi. The HTML/JS is split into 2 KB chunks to work around MicroPython's memory limits.

No USB needed after initial flash. OTA updates use MicroPython's built-in WebREPL. remote.py deploys all changed files and reboots in one command. fetch_logs.py pulls logs locally for analysis.

---

Features

Thermostat logic

• Own heat/cool decision engine — avoids the Daikin's unreliable built-in auto mode
• Dead-band control, season guard, dynamic fan speed, heat boost, 20-min short-cycle lockout
• Heartbeat resend every 20 min to keep Daikin in sync; boot resend on restart

Schedule

• Weekly schedule: away windows (protect against extremes) and home windows (target temp)
• Optional pre-conditioning: starts heating N minutes before you arrive home
• Adaptive learning: measures actual heating rate daily and auto-adjusts lead time

Outdoor temperature

• Polls Open-Meteo (free, no key) every 30 min for outdoor °F
• Used by season guard, heat boost scaling, backup heater thresholds, and history logging

Web interface (served from ESP32 on port 80)

• Dashboard: live sensor readings, target temp ±1°F (1-hour override), AC state, away/override cards
• AC Control: full manual IR command — mode, fan speed, setpoint, Pelonis on/off
• Schedule: Calendar Gantt view + list editor; add/edit/delete windows with pre-con timing
• History: canvas chart of indoor/outdoor/humidity over 6h–7d; heating/cooling background bands; system log viewer

IR protocols

• Daikin ARC433: stateful 19-byte frame — power, mode (heat/cool/fan/dry), fan speed, setpoint
• Pelonis PSHO06MR6ASB: NEC protocol, power toggle only (repeat=1 mandatory)

Deployment

• WebREPL on port 8266: wireless OTA upload, no USB needed after initial flash
• remote.py: deploy all files and reboot in one command
• fetch_logs.py: archive system.log and history.csv locally
• Watchdog timer (90s): reboots if thermostat thread stalls

Hardware

Component	Detail
Microcontroller	ESP32 running MicroPython
Temperature/humidity	DHT22 on GPIO 26
IR transmitter	GPIO 23 → NPN transistor → 22Ω → 2× 940nm IR LEDs in parallel
Primary HVAC	Daikin mini-split (ARC433 protocol)
Backup heat	Pelonis oil radiator (NEC protocol)

Setup

1. Flash MicroPython to your ESP32
2. Copy secrets.py.example → secrets.py and fill in your WiFi credentials and lat/lon
3. Edit remote.py and fetch_logs.py to set your device's IP address
4. Upload all device files:

   python3 remote.py deploy

Project structure

Device files (MicroPython):
  main.py          — entry point; thermostat + web server threads
  boot.py          — WiFi connect, WebREPL start
  ac_mgr.py        — smart thermostat brain and IR dispatcher
  ir_daikin.py     — Daikin ARC433 frame builder and IR transmitter
  ir_pelonis.py    — Pelonis NEC IR transmitter
  sched_mgr.py     — schedule storage and evaluation
  learn_mgr.py     — adaptive pre-conditioning optimizer
  outdoor_mgr.py   — Open-Meteo outdoor temp fetcher with 30-min cache
  data_mgr.py      — 5-min history logger (history.csv)
  log_mgr.py       — rolling system.log capture
  state_mgr.py     — target temp persistence (state.json)
  time_mgr.py      — NTP sync + US Eastern DST handling
  wifi_mgr.py      — WiFi connect/reconnect
  web_ui.py        — dynamic JS var block injected into web page
  _web_p1.html     — web app HTML + CSS (static, served in 2 KB chunks)
  _web_p2.html     — web app JavaScript (static, served in 2 KB chunks)

PC-side tools:
  remote.py        — OTA deploy via WebREPL; reboot
  fetch_logs.py    — pull system.log and history.csv to local archive
  ir_angle_test.py — IR LED pointing calibration tool
  webrepl_cli.py   — MicroPython WebREPL client (upstream project)

Development:
  .dev/            — capture and decode IR signals, OTA helpers
  docs/            — full project documentation and changelog

API

The ESP32 exposes a REST API on port 80:

Endpoint	Method	Description
/api/data	GET	Live sensor + state JSON
/api/schedule	GET/POST	Read or save schedule
/api/ac	POST	Send manual IR command (sets 1-hour override)
/api/command?cmd=set&val=<F>	GET	Set target temp + 1-hour override
/api/command?cmd=override_clear	GET	Resume schedule immediately
/api/history	GET	Stream history.csv
/api/log	GET	Stream system.log
/api/log/clear	POST	Delete system.log
/api/reboot	POST	Reboot ESP32