Background: I wanted to move my Home Assistant setup beyond simple scheduling and into the realm of professional building science. My goal was to create a “Digital Twin” of my 2,440 sq ft home’s thermal properties to track exactly how well (or poorly) the envelope performs against climate norms.
The Setup:
- Hardware: 2-zone Gas Furnace (60,556 BTU/hr), Honeywell T6 Pro thermostats.
- Data Strategy: 10-minute sensor validation/self-healing, 15-minute Riemann Sum sampling, and external CSV persistence for multi-year historical analysis.
Key Technical Features:
- Self-Healing Data Pipeline: I built corruption detection logic that monitors my outdoor temperature proxy. If the sensor spikes to physical extremes (-50/150) or goes
unavailable, an automation triggers an auto-correction within 10 minutes to prevent “dirty data” from poisoning my long-term metrics. - True HDD Integration: Instead of using a 2-point proxy for setback calculations, I’m using a 15-minute sampling interval during “the coast” to calculate a true integrated average. This captures the non-linear thermal decay of the home far more accurately.
- Energy Use Intensity (EUI) & UA Estimates: I calculate the heat loss coefficient (UA) daily. This allows me to see the “Wind Penalty” on the building envelope and quantify the ~15% extra loss caused by my 2nd-floor cathedral ceilings.
- Climate Norms Comparison: Using a Python script, I cross-reference daily performance against 18 years of historical P10/P90 climate data for my specific coordinates.
- Database Independence: While HA handles the logic, I export daily, monthly, and yearly summaries to CSV via shell commands. This keeps my data portable and immune to HA database purges or corruption.
Current Dashboards: My UI is built on a “Snippet Library” system of Mushroom and ApexCharts.
- Runtime-per-HDD Control Chart: Monitors if the house is “drifting” from its efficiency baseline.
- Recovery Rate Trend: Tracks how efficiently the system recovers from a 4°F setback vs. the fuel saved during the setback.
What I’m looking for feedback on:
- Thermal Decay Modeling: I’m currently estimating furnace output. I’m looking for peer review on moving to a measured Delta-T (Supply/Return) model using an ESP32.
- Sampling Rates: Is a 15-minute Riemann sum sufficient for a residential thermal mass, or should I tighten the interval?
- Baseline Drifting: How are others handling the “seasonal drift” in UA estimates as ground temperature changes affect slab loss?
Repo/Files: GitHub - wkcollis1-eng/home-assistant-config: Home Assistant configuration for energy performance tracking and HVAC monitoring I’ve documented the math and entity relationships in the CLAUDE.md. Would love to hear from anyone else treating their home like a laboratory!