⚙️ Smart HVAC Controller: Privacy-First Optimization

Build a modular HVAC system with Raspberry Pico W, MQTT, and Home Assistant to optimize furnace and AC efficiency while keeping your data private. This Homatica project by Claudio Cabete uses servo-controlled gas modulation and real-time sensor data for a smarter, greener home.

🔍 Why I Built This Smart HVAC Controller

Traditional thermostats react slowly and waste energy, while many smart thermostats rely on cloud services, risking your privacy. My goal was a privacy-first, locally hosted HVAC controller that anticipates thermal behavior, modulates fuel usage, and adapts dynamically—all without third-party data sharing. This project embodies Homatica’s mission: modular, resilient, and customizable home automation built with open-source tools like Home Assistant.

🧰 Hardware Setup

• 3 × Raspberry Pi Pico W: All run identical MicroPython firmware for interchangeability.
• Main Pico W Controller:
  • 8-channel relay board for switching.
  • 2 temperature sensors for precise monitoring.
  • 1 RC servo motor for gas flow modulation.
• 2 Pico Ws: Each with a temperature sensor for distributed zone monitoring.
• Raspberry Pi 4 Hub: Hosts:
  • Docker, MariaDB, Home Assistant.
  • FastAPI, Nginx, MQTT broker, Portainer.

All components communicate via MQTT, with the Pi 4 orchestrating logic and logging telemetry to MariaDB.

🧠 How the HVAC Controller Works

• Dynamic Scheduling: Uses a FastAPI schedule endpoint to set HVAC modes and setpoints, with override and fallback logic.
• Stateful Control: Tracks states to prevent erratic switching, enforcing minimum run/off times.
• Gas PWM Modulation: Servo adjusts gas flow based on air in vs. out delta for efficiency.
• Circulation Fan Logic: Manages post-run cooling, hysteresis, and cycle-based fallbacks.
• Stale Data Protection: Validates timestamped thermostat data to ensure safe operation.

🔧 Technical Highlights

• Runtime-tunable hysteresis and dead band for precise control.
• Debounce and timestamp guards to avoid erratic behavior.
• Modular YAML and MQTT-based configuration for scalability.
• Robust MQTT reconnect logic and availability publishing.
• Fallbacks for expired overrides and missing schedules.

Sample MicroPython Code

import machine
import time
from umqtt.simple import MQTTClient

# Servo control for gas modulation
servo = machine.PWM(machine.Pin(16))
servo.freq(50)

def set_servo_duty(duty):
    servo.duty_u16(int(duty * 65535 / 100))

# MQTT setup
client = MQTTClient("pico_w", "mqtt_broker_ip")
client.connect()

def on_message(topic, msg):
    if topic == b"hvac/ctrl/outputs":
        duty = float(msg.decode().split("gas_pwm:")[1])
        set_servo_duty(duty)

client.set_callback(on_message)
client.subscribe("hvac/ctrl/outputs")
    

🧾 Sample Logs

Active schedule selected:
  eid=0
  mode=hold
  sch_type=AC
  sch_hvac_mode=AC
  setpoint=24.7
  fan_on=0
  fan_off=1800
        

HEAT_TEMP_TRIGGER: 18
AC_TEMP_TRIGGER: 26
heat_max_setpoint: 20
ac_max_setpoint: 25
dead_band: 0.7
ac_min_run_secs: 900
ac_min_off_secs: 300
ac_fan_delay_secs: 60
furnace_fan_delay_secs: 90
hvac_mode: AC
sch_mode: AC
settemp: 24.7
setadjtemp: 24.7
fan_cycle_on_secs: 0
fan_cycle_off_secs: 1800
udv_hvac_mode: AC

🌦️ Weather:
  temperature: 72°F
  humidity: 97%
  condition: rainy
  forecast:
    temperature: 77°F
    templow: 72°F
    humidity: 95%
    wind_speed: 13.17 km/h
    uv_index: 5.3
    precipitation: 0.47
    datetime: 2025-09-25T16:00:00+00:00

🌡️ Sensors:
  HVAC Central Unit CTLR:
    In Air Temp: 22.81°C
    Internal Temp: 29.86°C
    Out Air Temp: 11.63°C

  Portable Thermostat:
    Ambient Temp: 22.94°C
    Internal Temp: 29.86°C

  BedRoom Thermostat:
    Ambient Temp: 25.69°C
    Internal Temp: 33.60°C
        

Mode: AC
Setpoint: 24.7
AvgTemp: 24.8
TempDiff: 0.10
ac_state_change_lap: 575.31
ac_state: 1
        

Time since last fan state change: 514.78s

Air Temp Delta:
  temp_diff_air: 11.19
  fan_diff_on: 6.0
  fan_diff_off: 3.0

Thermostat Temp Delta:
  temp_diff_thermostats: 2.75
  thermostat_diff_on: 4.5
  thermostat_diff_off: 3.7

During AC Active:
  temp_diff_air: 11.19
        

registers:
  air_in_out_delta: True
  thermostats_delta: False
  fan_cycle: False
  ac_fan_delay: True
  furnace_fan_delay: False
  post_ac_fan: False
  post_furnace_fan: False

circfan_state: 1
fan_reason: air_in_out_delta, ac_fan_delay
        

Topic: hvac/ctrl/outputs
Payload:
  device_id: 7
  settemp: 24.7
  setadjtemp: 24.7
  udv_hvac_mode: AC
  fan_cycle_on_secs: 0
  fan_cycle_off_secs: 1800
  sch_mode: hold
  gas_pwm: 11.4
  hvac_mode: AC
  avg_temp: 24.8
  data:
    - ioid: 0, io_v: on
    - ioid: 1, io_v: off
    - ioid: 2, io_v: on
    - ioid: 16, io_v: on, io_freq: 50, io_duty: 11.4
        

📈 Next Steps for the HVAC Controller

I’m enhancing the override logic for smoother manual interventions and developing animated diagrams to visualize system states. A full tutorial with schematics and MicroPython code is in progress—stay tuned on my blog!

🤖 Built with AI Collaboration

This HVAC controller was crafted with AI assistance, from optimizing MicroPython code to refining control logic and documenting the build. AI helped me explore edge cases and streamline the system, but the core design—every wire, code line, and debug session—came from my passion for DIY automation, sparked by building an FM transmitter at age 12.