How to Winterize a House with Home Assistant

Start with the failure modes that can damage the house

Winterizing a house happens in layers. Insulation, weather sealing, heat, and sensible plumbing come first. Home Assistant comes after those basics, not instead of them. A dashboard cannot thaw a frozen pipe, and an automation cannot make a neglected furnace safe. What it can do is turn a slow, quiet failure into an early decision while there is still time to act.

That boundary keeps the system honest. Do not build a clever chain of service calls for every cold-weather scenario. Build a small monitoring system that answers four questions: Is the house getting too cold? Is the heat source producing a result? Has water appeared where it should not be? Is the house still powered and reporting?

The winter automations worth keeping are deliberately boring. They use sustained thresholds, more than one notification path, and a health signal for sensors that stop being trustworthy. A 30-second temperature wobble should not wake anyone. A sensor that has been unavailable all afternoon should not silently remain in a safety plan.

Map the risk zones before building alerts: the coldest basement area near water lines, the garage wall shared with conditioned space, an exterior utility closet, the mechanical room, a freezer, and each place with a leak sensor. For ordinary indoor risk zones, my default is the SONOFF SNZB-02D from SONOFF with code JPK (Amazon fallback). The guide to Zigbee temperature and humidity sensors covers placement and alternatives. Sensor count matters less than knowing which failure each sensor is meant to catch.

Build thresholds around time, not panic

A threshold without a duration is a noise generator. Garage doors open. A delivery person opens a basement exterior door. A dehumidifier changes a nearby reading. If every dip below a number becomes an alert, the alerts that matter become background noise.

For ordinary interior monitoring, set an early warning at 45 F after 20 to 30 minutes below that mark. That leaves time to check the trend, inspect the thermostat, and decide whether someone needs to visit. Exposed plumbing deserves a separate critical threshold closer to freezing, also sustained rather than instantaneous. A finished basement with buried plumbing does not need the same trigger as a thin exterior wall with a pipe passing through it.

Every low-temperature alert needs a recovery message. If Home Assistant warns that the basement has been below its warning threshold, send a clear follow-up only after it has stayed safely above it again. That closes the incident instead of leaving a stale critical notification with no context.

This single-location pattern is intentionally narrow. It sends a durable alert only after the condition lasts, and it does not pretend to repair a heating problem.

Before adding these automations, create the Home Assistant toggle helper input_boolean.basement_freeze_warning_active and use it as the alert state. The warning automation sets it after it sends the warning; the recovery automation can then close only an alert that was actually opened.

alias: Winter freeze warning
mode: single
triggers:
  - trigger: numeric_state
    entity_id: sensor.basement_temperature
    below: 45
    for: "00:25:00"
conditions: []
actions:
  - action: notify.mobile_app_phone
    data:
      title: "Freeze warning"
      message: "Basement temperature has stayed below 45 F for 25 minutes. Check heat and exposed plumbing."
  - action: input_boolean.turn_on
    target:
      entity_id: input_boolean.basement_freeze_warning_active

A recovery notice should wait for a comfortably higher reading, not merely the warning threshold. That 47 F threshold adds hysteresis, so small fluctuations around 45 F do not make the alert state bounce between warning and recovered. The helper condition also prevents a normal rise above 47 F from producing a false recovery when no warning fired.

alias: Winter freeze warning recovered
mode: single
triggers:
  - trigger: numeric_state
    entity_id: sensor.basement_temperature
    above: 47
    for: "00:10:00"
conditions:
  - condition: state
    entity_id: input_boolean.basement_freeze_warning_active
    state: "on"
actions:
  - action: notify.mobile_app_phone
    data:
      title: "Freeze warning recovered"
      message: "Basement temperature has recovered and stayed above 47 F for 10 minutes. The freeze warning condition has cleared."
  - action: input_boolean.turn_off
    target:
      entity_id: input_boolean.basement_freeze_warning_active

Replace sensor.basement_temperature and notify.mobile_app_phone with your own entity and mobile-app notification service IDs. Keep a separate, lower critical alert instead of piling conditional branches onto this one. Separate alerts are easier to test, acknowledge, and tune after the first cold snap.

Watch the furnace as a system, not a single entity

A thermostat can report that it is calling for heat while the house continues to cool. That distinction matters. A thermostat state of heating proves only that the thermostat asked for heat, not that the furnace is healthy.

Treat furnace failure as a combination of demand and outcome. When the thermostat has been calling for heat for a sustained period, the room is materially below its set point, and the relevant indoor sensors are still falling or not recovering, send an alert. A furnace fault entity is useful additional evidence when the equipment exposes one. It cannot be the only signal because many systems publish no fault state and some faults do not present cleanly.

Alert text should state what Home Assistant actually knows. “Heat demand has been active for 45 minutes and the basement temperature is still falling” is more actionable than “furnace failed.” The first points toward the next decision without claiming a diagnosis. Fuel, a tripped breaker, a blower, the thermostat, or a heat source overwhelmed by unusual weather can produce the same pattern.

Keep thermostat control local, especially when the house is empty. Local thermostat control with Z-Wave and Home Assistant explains why local control is a resilience measure rather than a luxury. For a compatible low-voltage HVAC system, shop the current Resideo T6 Pro Z-Wave TH6320ZW2007/U, not the discontinued 2003 model. The winter monitoring layer still needs to work when an internet service is having a bad day.

Use a notification ladder, not one dramatic push notification. The first sustained warning goes to the normal household channel. A critical combination, such as low temperature plus prolonged heat demand, escalates to people who can respond. If nobody can respond safely, the playbook needs a real-world fallback: a neighbor, property manager, or qualified HVAC contact. Home Assistant routes the signal. It does not solve the access problem.

Give the garage and basement their own rules

The garage is not simply a colder room. It is a different risk zone. Its sensor can drop rapidly after a door opens, and an attached garage may be cold by design while the adjacent living space is fine. It should not use the same threshold or escalation timing as a conditioned room.

The useful question is whether plumbing, appliances, or a shared wall make that cold meaningful. If there is a water line, hose connection, or plumbing in a cabinet against the garage wall, put a sensor near that risk, not in the most convenient spot on the ceiling. Wait to alert until the door has had time to close and the temperature has had time to settle. A garage reading below its chosen threshold for 30 minutes is useful. A brief dip while unloading groceries is not.

Basements have the opposite problem. They change slowly, so a slow decline can look harmless until it has been happening for hours. Put at least one sensor near the coldest plumbing exposure rather than only beside the thermostat or electrical panel. In a large basement, two inexpensive sensors in distinct zones are more useful than one premium sensor in the middle. Their difference often tells the story faster than either reading alone.

Use outside weather as context, not as the trigger that decides whether a pipe is at risk. An exterior sensor or weather integration explains why a threshold deserves attention and flags a forecast that warrants a manual check. The Home Assistant weather-station guide covers the broader monitoring side. The interior sensor near the pipe remains the signal that drives an urgent response.

Plan for the leak that appears after the thaw

The most expensive part of a freeze event may arrive after the temperature recovers. A pipe can crack, stay dry while frozen, then leak when water starts moving again. Freeze monitoring and water monitoring belong in the same winter plan.

Place leak sensors where water will first reveal itself, not where a pipe is easiest to see: below the water heater, beneath utility sinks, near laundry supply hoses, under a basement appliance, and at low points where a slow leak will collect. The SONOFF SNZB-05P from SONOFF with code JPK (Amazon fallback) is useful where its optional detection cable can cover a longer run. The water-leak sensor guide for Home Assistant covers placement and alternatives.

A leak notification should be immediate, unlike a temperature warning. There is little value in waiting 20 minutes for a confirmed puddle. But an alert is only half a response. For homes with a compatible quarter-turn main shutoff and enough clearance, the Zooz ZAC36 Titan adds a local Z-Wave actuator and wired leak-probe fallback. Decide in advance which conditions may close it automatically and which only notify. The Home Assistant water shutoff valve guide lays out that decision in more detail.

Do not close the water automatically because one temperature sensor reports a low number. A bad battery or open garage door should not leave the house without water. Automatic closure is easier to justify when a leak sensor is wet, especially in a vacant property, and the household understands the tradeoff. Test the notification and valve-control path before winter, with people present and a manual override available.

Treat freezer loss and utility loss as related failures

A freezer sensor is cheap insurance, but it needs a different rule set from a pipe sensor. Freezers warm during defrost cycles and whenever the door opens. Alert on a high temperature that remains elevated long enough to rule out those normal events, then escalate if it continues upward. A sudden rise paired with a power alert deserves faster attention than temperature alone. The local fridge and freezer monitoring guide covers practical placement and alerting details.

Utility outages expose a weakness in otherwise good automations: a cloud push notification may not leave the house when the internet equipment has no power. A UPS for the modem, router, Home Assistant host, and network coordinator buys time for the system to report what happened and shut down cleanly. The APC BR1500MS2 is the pure-sine model used in my NUT with Home Assistant and graceful shutdown guide, but size the UPS for the actual load and verify NUT compatibility before buying. The goal is not to run the house indefinitely. It is to preserve observation long enough to distinguish a brief outage from a failed sensor and keep the automation host from corrupting storage.

A safety sensor that is silent is not a safety sensor

Every winter dashboard needs a health view, even if it is only a short list. Check battery level, last-reported time, and state for every temperature, leak, power, and freezer sensor. A device that is unavailable, unknown, or stale needs its own maintenance alert. Otherwise the sensor meant to warn about a freeze can fail in the quiet weeks before one.

Set the stale interval from how the device actually reports. A sensor that normally sends a reading every few minutes can be called stale sooner than a battery device designed to check in hourly. Do not suppress stale warnings because the latest temperature looks comfortable. Replace batteries before a forecasted cold period, not after a low-battery warning becomes an outage.

Test the whole chain before the forecast does it for you

Test each alert as a workflow, not merely as YAML that loads. Use a safe test value or helper to cross the warning threshold and wait through the full sustained delay. Verify that the intended notification reaches its destination, then verify that the recovery message arrives after the value returns to normal. Repeat the test for the critical threshold with a different title and escalation path.

Then simulate bad information. Make a noncritical test entity unavailable, confirm that the stale-sensor alert fires, and restore it. Check a low-battery indicator if the device exposes one. For a leak-and-shutoff plan, confirm the notification path and manual valve control while someone is home. Do not turn a real house into a test fixture by closing its water main or disabling heat without a deliberate, supervised plan.

Write down the human response as well. Who receives the first message? Who can enter the house? Where is the manual water shutoff? What is the HVAC service contact? Which neighbor can check a vacant property? The automation is complete only when its handoff is clear.

Frequently Asked Questions

What temperature should trigger a freeze alert?

Use a sustained 45 F warning for ordinary interior risk zones, then set a lower critical point for the pipe location and sensor reliability. Duration matters as much as the number. Twenty to 30 minutes filters ordinary disturbances without turning a real downward trend into background noise.

Can Home Assistant tell me that my furnace failed?

It can identify a credible failure pattern, not make a diagnosis. Alert when there is prolonged heat demand, a room is below set point, and indoor temperature is still falling or failing to recover. Add a furnace fault signal when available, but do not depend on it alone.

Should I automatically close the water valve during a freeze warning?

Usually no. A low temperature is an early warning, and false positives are costly. Automatic shutoff is more defensible after a confirmed wet leak sensor, especially for a vacant home, with a manual override and a tested notification path.

How often should I test winter automations?

Run the full alert and recovery test before the cold season, after material automation changes, and after replacing sensors or network hardware. Check stale-sensor and battery alerts monthly during winter. Testing finds a broken notification route before it becomes a frozen-pipe problem.

The Bottom Line

Winter protection in Home Assistant is not a collection of dramatic automations. It is reliable early warning: sustained low-temperature thresholds, evidence that heat is not recovering the house, immediate leak detection after a thaw, freezer and outage context, and sensor-health checks that expose blind spots. Build the response plan around those signals, test it before winter, and let the system buy time rather than promise certainty.

Hero photo: National Institute of Standards and Technology, a US federal government public-domain work, cropped from the original.