How smart heat pump control works in 2026

The heat pump is probably the most expensive appliance in your home. In a Swedish villa it accounts for 50–70 % of the electricity bill. And in most homes it runs on a single rule: "keep it warm inside." That works, but it's incredibly expensive — because the electricity price varies enormously throughout the day.

Smart heat pump control is the art of breaking that rule in the right way. In this guide we cover how it works, what you need, and the pitfalls to avoid. It's written for 2026 — when Swedish households have hourly spot contracts more than ever, SMHI's forecasts are better than they've been, and AI control has finally become a mainstream product.

What "smart control" is — and isn't

Smart heat pump control means shifting when heat is produced without changing how much heat is produced overall. The result: the same comfort at significantly lower cost.

It's not the same thing as:

• Lowering the indoor temperature. That trades comfort for money — it's not optimisation. Smart control doesn't reduce your comfort.
• Turning the pump off at night. The classic "night setback" is often more expensive, because the pump works harder when it has to catch up in the morning — exactly when the price peaks.
• A timer schedule. Running the pump "between 02 and 06" is a simplification. Real smart control is dynamic and adapts to the spot price every day.

The four ingredients

To do smart control for real you need four things. If any one of them is missing, you're working with approximations.

1. The hourly spot price

Nord Pool publishes tomorrow's hourly prices at 13:00 every day. Those are the prices your electricity contract follows if you have hourly spot, which most modern Swedish contracts are today (Tibber, Greenely, Cheap Energy, etc.). Smart control needs:

• The current hour: what does electricity cost right now?
• The rest of the day: when is it cheapest and most expensive?
• Tomorrow, after 13:00: what's coming?

Without these three you can't plan. Just reacting to "is it expensive right now?" often leads to sub-optimal decisions.

2. The weather forecast

SMHI's forecasts are free and update around the clock. For a heat pump the following matters:

• Outdoor temperature (hourly, 48 hours out). When is it coldest? When does it turn mild?
• Sun and cloud cover. South-facing windows can warm a house noticeably even in winter.
• Wind. Houses with poor airtightness lose more heat in wind.

A smart controller that knows it'll be +8 °C tomorrow night can let the house cool slightly tonight without risk — it'll be cheap and warm again tomorrow. Without the forecast that's impossible.

3. The home's thermal inertia

This is the hard ingredient. How fast does your home lose heat when the pump pauses? And how fast does it recover when the pump starts again?

Inertia depends on:

• House construction (stone/brick = high, lightweight = low)
• Insulation quality
• Heating system (underfloor = very high, radiators = lower)
• Outdoor temperature (colder outside = faster losses)

In practice a modern AI can learn your unique inertia in 5–10 days. It doesn't need a building spec — it learns by observing how the temperature moves as the pump works at different intensities.

4. A comfort floor

This is the user's part. You tell the system: "I never accept it going below 20.5 °C indoors." That's the system's hard limit — regardless of how expensive electricity gets, regardless of what the forecast says, the system never goes below that number.

Without a comfort floor smart control becomes uncomfortable. With a comfort floor it becomes invisible — you only notice that the bill is lower.

Three examples of how it thinks

Here are three everyday scenarios — how a good AI controller actually reasons.

Example 1: Normal winter weekday, clear weather

The spot price peaks 07:00–09:00 (people wake up, industry starts) and 17:00–20:00 (people come home, cook). It's cheapest at night, 03:00–06:00, and mid-day 12:00–14:00.

The AI does the following:

1. Pre-heats the home overnight (03:00–06:00). The pump works a bit extra and lifts the indoor temperature from 21.0 °C to 21.5 °C.
2. Pauses during the morning peak (07:00–09:00). The temperature slowly drifts back toward 21.0 °C — but never below the comfort floor.
3. Mild operation midday (10:00–16:00). The pump follows the heat demand normally, since the price is neither high nor low.
4. Pauses during the evening peak (17:00–20:00). Again, the temp drifts down slowly. Daytime sun warmth on the south side helped.
5. Returns to normal operation overnight (21:00→).

Result: 8 hours paused, 4 hours pre-heating. Total energy roughly the same — but cost 20–25 % lower.

Example 2: Cold outside, expensive all day

Worst case: it's −15 °C outside and spot prices are high all day (a classic January cold snap). What does the AI do?

It stops trying to save. Safety and comfort first — otherwise the house freezes. A real AI recognises that the conditions for optimisation aren't there today and runs normally. Better to have a day with 0 % savings than a day with the house at 18 °C.

Example 3: Mild and cheap in the morning, cold and expensive in the afternoon

SMHI says it'll be +5 °C until 14:00, then the wind shifts and it drops to −8 °C by evening. The price is cheap in the morning and expensive in the afternoon.

The AI pre-heats aggressively in the morning. The house is heated to 22.0 °C while electricity is cheap and the pump is efficient (warmer outside = better COP). When the cold arrives in the afternoon the house coasts on stored heat for hours. The pump only takes short peaks.

That kind of thinking — combining weather, price and inertia — is what separates rule-based scheduling from real AI.

What you need yourself

To run smart control at home you need:

1. A controllable heat pump. In practice nearly every modern pump (Qvantum, Mitsubishi, Nibe, IVT, Bosch, etc.). Either directly via the manufacturer's cloud API or via Home Assistant.
2. An hourly spot contract. Without hourly spot there's nothing to shift — you pay the same price all day regardless. Tibber, Greenely, Cheap Energy, Bixia and several others offer this.
3. A power meter (recommended). Shelly EM, HomeWizard P1 or Tibber Pulse. Without a meter the prediction is weaker — not impossible, but you make life harder for yourself.
4. A control service. You can build your own (Home Assistant + a bunch of automations), but it gets complex fast. Or you use a ready-made service like Therilly.

Common pitfalls

• "I switch it off at night — that's enough, right?" No. The heat pump works hardest and most expensively right after a shutdown, not during normal operation. Smart control is dynamic, not schedule- based.
• "My pump already has an 'eco' mode." That's usually a fixed-time schedule. It doesn't know the spot price, weather or dynamics. Often worse than nothing during winters with volatile pricing.
• "I need a new pump." No — in 95 % of cases just adding AI control on top of the pump you already have is enough.
• "AI = magic." No. The AI here is a model that learns your home and optimises against price and weather. It doesn't invent physics. A poorly insulated house will save less than a well-insulated one — but both save something.

How big is the saving?

Realistic numbers for Swedish villas:

Home type	Typical saving
Modern villa, underfloor, well insulated	25–35 %
Standard 70s villa, radiators	15–25 %
Poorly insulated, high comfort target	10–15 %
Home without hourly spot contract	0 % (no basis)

If you pay SEK 25,000/year for the heat pump, smart control can drop it to SEK 18,500–21,250. Payback on an SEK 89/month service is typically under 2 months.

Summary

Smart heat pump control in 2026 isn't a future technology — it's a mature product category that works today, for Swedish homes, with your existing equipment. The ingredients are: good price data, good weather data, a model of your home, and a clear comfort floor. If one of them is missing, you get approximations; if all four are there, your bill drops 15–30 % without you noticing.

Want to read more? See our guide on saving 30 % on the heat pump with hourly spot pricing.