Haus Autonomie 2026: The Home as Energy Jurisdiction

Fission & Fusion: The Nuclear Frontier

Nuclear energy stands at a pivotal crossroad.

Capture the FlagAxel Quack

Baseload and the grid’s long game. The home is where the grid becomes personal.

I have spent the last years treating compute as a jurisdiction — local models, sealed control planes, inference you can unplug. In Germany in 2026 another jurisdiction is equally real: the household energy system. Photovoltaics, heat pumps, wallboxes, batteries, and now bidirectional cars are no longer separate purchases. They are a stack. Whoever owns the optimizer owns the stack.

Call it Haus Autonomie — not off-grid romanticism, but the right to route, store, and time your own kilowatt-hours under rules you understand.

This is a long read on purpose. The gap between a solar brochure and a working household is where §14a, EEBus, dynamic tariffs, V2H packages, household AI, and software lock-in actually live.

Why the home became a flag

Three legal-technical shocks arrived together. None of them is “smart home” fashion.

§14a EnWG — comfort as a negotiated dim

Since 1 January 2024, new steuerbare Verbrauchseinrichtungen (SteuVE) above 4.2 kW — typically wallbox, heat pump, and many storage systems — must be controllable by the distribution system operator in exchange for reduced network fees. The DSO may temporarily reduce power to protect the local grid. This is not arbitrary blackout theatre: the political bargain is dimming rights ↔ cheaper grid access.

What matters for autonomy:

• Without a clean control path, connection delays and missed discounts are common.
• A HEMS that understands EEBus LPC (limitation of power consumption) can throttle a wallbox steplessly toward 4.2 kW instead of hard-off. That is the difference between “the car still charges slowly” and “the heat pump just died mid-winter.”
• Related EEBus blocks: LPP (limit feed-in), MPC / MGCP (metering at device and at the grid connection point). BNetzA, VDE FNN, and ZVEH materials commonly recommend EEBus for this path.

Marketing slides say “up to ~60% network-fee relief.” Your actual module (1 / 2 / 3 under the §14a implementation) depends on DSO and whether you take a flat reduction, time-variable grid fees, or a more dynamic construct. Read the DSO letter, not the Instagram reel.

Dynamic tariffs — EPEX as household UI

Suppliers must offer time-variable products. A true dynamic tariff tracks the exchange (often hourly EPEX Spot) and generally needs an intelligentes Messsystem (iMSys) — smart meter + gateway — not only a digital Ferraris replacement.

Do not confuse four product shapes:

Only the last one makes a HEMS’s “shift the heat pump to 03:00” economically honest. Tibber-, aWATTar-, rabot-, 1KOMMA5° Dynamic Pulse–class products live here. Winter still exists: cheap summer hours do not cancel a dark January.

Price formation sits on EEX (forwards) and EPEX Spot (day-ahead / intraday). Retail never is the exchange; it is the exchange plus grid fees, levies, VAT, and the supplier’s book. Autonomy is knowing which layer you can actually optimize.

V2X left the pilot hangar

2026 is the first year German households can order commercial bidirectional packages rather than only join pilots:

• BMW + E.ON — first DE commercial V2G offer (early 2026); Neue Klasse / iX3-class; wallbox discount and “free km / ~€720-class” marketing under connection-hour rules
• VW Group / Elli — BiDi charger + Naturstrom V2G Flow + app; volume launch aimed Q4 2026; group claim of ~1 million MEB vehicles in Europe already prepared
• Nissan Energy + Valeo — AC bidirectional path (needs bidirectional onboard charger); Europe 2026 rollout story

“Prepared for bidirectional” is not customer-ready. You need matched OEM software, a listed EVSE, a tariff product, and usually a HEMS. More below.

The stack (without the brochure)

A working house is a pipeline of roles, not a single app:

1. Grid / DSO — connection, §14a signals, feed-in rules
2. Smart meter / SMGW — settlement and often the CLS path to a control box
3. HEMS / EMS — surplus routing, tariff windows, priorities, state-of-charge floors
4. Assets — PV inverter, stationary battery, heat pump, wallbox, optional EV
5. Cloud / VPP (optional) — aggregation, remote UX, flexibility markets

Germany’s HEMS market is crowded. The 2nd HEMS market overview 2025 (HS Ansbach / TUM / hems-finder): 41 vendors, 43 systems; ~70% can push PV surplus across BEV + heat pump + storage + heating rod; only ~38% yet do full dynamic-tariff optimization across that same set. Hardware is ahead of software. Buying a “smart” wallbox without a HEMS that can write schedules is buying a chart.

Naming trap: EMS / HEMS / “AI optimizer” are marketing synonyms. LEMS (IEC 63380) is not a cheaper HEMS — it is the standardized local manager toward chargers. Full-home optimization still needs HEMS-class scope (or a vertical that embeds both).

EEBus ≠ IEC 63380 (full stop). SPINE/SHIP entered IEC 63380-2/3, but 63380’s scope is EV-charging / LEMS–EVSE. Appliance interoperability still points at IEC 63510 / EN 50631. Cite the right standard per interface.

Open stacks exist: evcc is the power-user default for mixed brands (Pi/Docker, huge device list). Verticals (SMA Sunny Home Manager, E3/DC, SENEC, 1KOMMA5° Heartbeat) bury EMS in the package. Dedicated HEMS hardware+install often lands ~€500–3,000; bundles hide the line item.

Ownership vs rent (the flag that is not about kWh)

Two business models colonise the same roof.

Installment ownership (1KOMMA5°-class)

1KOMMA5° (Hamburg, 2021) sells integrated home energy: PV, battery, heat pump, wallbox, Heartbeat gateway, Heartbeat AI EMS, and Dynamic Pulse (EPEX-linked tariff; company: no markup on the hourly exchange price). Official 2026 facts: 120k+ homes, 300k+ energy systems, 3,000+ employees, 1 GW shiftable VPP (May 2026; 20 GW by 2030 ambition). June 2026: up to €1 billion installment program, terms up to 25 years, advertised 5.99%, ownership not rent. Heartbeat AI retrofit on third-party gear since Nov 2025 (~40% of DE market compatibility claim).

Heartbeat is not HomeKit. Inputs: weather, day-ahead + intraday prices, household history. Outputs: when to charge the car, run the heat pump, export. Fleet claim: millions of optimization decisions per day. Box in the building · AI layer · app.

Risks: 20–25 year debt can lose to cash TCO; deep software coupling; “0 ct/kWh” is a window, not your winter bill; sales-channel pressure in reviews. Company cites selected-customer evaluations for some price stats — model your roof and load.

Rent / lease (Enpal-class pattern)

Lower entry friction. Someone else owns the asset. Exit and residual value are the fine print. 1K5’s public line criticizes expensive rent; that is competitive rhetoric, not a TCO proof for your house.

The actual flag

Do you want the deed to the hardware, or a subscription to someone else’s balance sheet? Flag theory prefers deeds you can exit — only if you can also exit the software lock. A 25-year installment with a captive EMS is a residency you cannot easily leave.

Typical DE levers still worth listing (verify live): 0% VAT on residential PV (2023-era ruleset), BAFA/BEG heat-pump grants, KfW-class loans, EEG feed-in, smart meter / §14a. Subsidy “deadlines” are a sales tactic; regulatory volatility (heating law, rates) is a real demand risk.

V2H: the large battery that leaves

A long-range EV holds 60–100 kWh — often more than the wall box. V2L powers a kettle. V2H serves the building. V2G serves the market. V2X is the umbrella.

AC vs DC (the split that decides your wallet)

V2L Schuko on Hyundai/Kia/BYD is not whole-house V2H. Europe home V2H ≈ CCS DC wallbox + OEM-approved list.

Commercial economics (not free electrons)

Elli/BMW-class messaging: up to ~€720 first-year plug-in bonus if you stay connected 250 hours/month with session-length rules. That is behaviour-contingent. Model commute SoC floor vs parked hours. Literature/marketing bands for V2G value often sit **€200–800/year**; wallbox premium and install can eat years of that.

Net charges: the old killer was paying grid fees into the battery and again on the way out. 2025–2026 direction (EnWG storage privileges, BNetzA MiSpeL drafts) treats storage and EV batteries more like flexible resources and self-consumption assets. Directionally good. Not a billing calculator. Do not treat blog “full texts of §19 EEG” as law.

Warranty: older VW-class messaging cited discharge energy/hour caps. Read the PDF before you cycle the pack for a €60 bonus.

The car is a brilliant buffer when it is home. It is a terrible UPS at the office. Stationary storage and V2H are complements.

The role of AI (the optimizer as capital city)

The hardware stack is dumb without a dispatcher. In 2026 that dispatcher is sold as AI — sometimes honestly (forecast + mixed-integer-ish dispatch against prices), sometimes as a sticker on a 2018 rule engine. Autonomy means knowing which layer is actually learned, and who is allowed to change the objective function.

What “AI” is allowed to mean in a house

1KOMMA5°’s Heartbeat AI is the vertical exemplar: inputs are weather, day-ahead + intraday prices, and household history; outputs are when to charge the EV, run the heat pump, or export; the fleet is a VPP (company: millions of decisions per day, 1 GW shiftable). Heartbeat the box, Heartbeat AI the layer, the app the UI. That is not HomeKit. It is a specialized EMS with a market objective.

Open stacks (evcc, SMA-class managers) increasingly ingest the same price series. The difference is less “neural net vs not” than who sets the loss function: minimise your bill, maximise self-consumption, honour a VPP bid, or keep the living room at 21 °C no matter what EPEX did at 18:00.

Why rules break and models show up

A German house in 2026 has conflicting objectives that a static priority list cannot resolve:

• PV surplus wants the battery and the car and the heat pump.
• Dynamic tariffs want grid charge at 03:00 even if the sun will arrive at 10:00.
• §14a may cap the wallbox while you promised a full pack by 07:30.
• V2H wants to empty the car into the evening peak unless tomorrow is a 400 km drive.
• Negative residual prices want you to import; EEG export wants you to spill — at different hours.

That is a constrained sequential decision under weather and price uncertainty. Forecasts (irradiance, temperature, occupancy) plus a receding-horizon scheduler are the adult form of “AI.” Chatbots that explain the bill are not.

Edge vs cloud (same fight as compute)

The sovereignty instinct that puts inference on a sealed lab applies here: comfort constraints and SoC floors should be local and auditable; price forecasts can be rented. If the cloud dies, the house should still heat and the car should still leave with a commute reserve — dumb rules as fallback, not a bricked gateway.

VPP: your house as someone else’s model feature

A virtual power plant is AI at fleet scale. Your battery becomes a pixel in a larger bid. That can pay (tariff, plug-in bonus, implicit cheaper Dynamic Pulse). It also means:

• The objective may include the aggregator’s book, not only your bill.
• High-resolution load is training data for someone else’s flexibility product.
• Retrofit EMS (Heartbeat-class connectors on third-party inverters) extends that model over gear you already own — convenient, and a deeper lock.

Ask three questions before you enrol: What is optimised (my €, their MW, both)? What is logged, where, for how long? What happens if I revoke cloud access — do local schedules survive?

What AI is not

• A substitute for a smart meter or a legal §14a path.
• A reason to skip SoC floors and warranty caps.
• Explainability theatre: if you cannot see why the heat pump ran at 02:00, you do not have autonomy — you have a vibe. Demand a decision log (price, forecast, constraint), not a weekly “you saved 12%” card.
• General-purpose home agents with fuse-box tools. A coding agent with Modbus to the inverter is a confused deputy. Energy AI should be a narrow optimizer with a kill switch, not an LLM with root on the HAN.

Design rules for an autonomous house

1. Local constraints, rented forecasts — comfort and mobility reserves never depend on a vendor API staying up.
2. One objective you can name — bill vs autarky vs VPP; mixed objectives need explicit weights.
3. Human override that actually sticks (holiday, sick day, heat wave).
4. Decision log retained in the house, not only in an app.
5. No auto-approve for firmware or “new optimisation modes” that change discharge depth.
6. Treat the EMS like production software: pin versions, read changelogs, assume the model will drift.

AI does not create Haus Autonomie. It implements it — or implements someone else’s VPP using your basement as a free parameter.

Privacy is an energy topic

HEMS and VPP apps see high-resolution load. That is when you cook, charge, sleep, and whether you are on holiday. Cloud optimizers sell convenience; local-first stacks (evcc on a Pi, edge HEMS with optional cloud) sell opacity.

Same instinct as sealed compute: prefer edge decisions, audit exports, revoke aggregator access when the discount is not worth the mirror. EEBus certification maturing through 2026 is the interoperability hope. Until then, “works with my heat pump” is a Meister sentence, not a checkbox.

Economics in one decision frame

HEMS ROI collapses if you only have a small PV and no flexible loads. It compounds once you have two or more of: heat pump, wallbox, ≥5–10 kWh storage, dynamic tariff. Guides cite +20–35 percentage points self-consumption and ~15–25% bill cuts — profile-dependent marketing bands, not guarantees.

A practical autonomy checklist

1. Measure your Jahresverbrauch and winter heat load — not national averages.
2. List protocols on every device before buying a HEMS (SG Ready / EEBus / Modbus / OCPP).
3. Design the §14a path with the electrician (EEBus LPC preferred over dumb contactor drama).
4. Pair a dynamic tariff only if the HEMS can write schedules, not just draw charts. Need iMSys for honest hourly settlement.
5. If chasing V2H/V2G, buy from one OEM package list, not three hopeful PDFs.
6. Set SoC floors for mobility before you optimise for arbitrage.
7. Read battery warranties for discharge energy/hour caps.
8. Decide cloud vs local before the gateway phones home forever.
9. Get an independent second quote; read exit, tariff coupling, and ownership-transfer clauses.
10. Re-check BNetzA/DSO rules at commission date — not at brochure date.

Final thoughts

Haus Autonomie is not autarky. You will still touch the grid, the DSO, and the weather. It is the insistence that the home is a jurisdiction of flexibility you administer — not a dumb endpoint for someone else’s virtual power plant.

Nuclear and interconnectors decide the civilisation-scale mix. HEMS, §14a, dynamic tariffs, and bidirectional driveways decide whether your evening is priced by yesterday’s sun or by a peak you could have shifted.

Plant the energy flag deliberately. The optimizer is the capital city. The contract is the constitution. The car is a battery that commutes.

The website and the information contained therein are not intended to be a source of advice or credit analysis with respect to the material presented, and the information and/or documents contained on this website do not constitute investment advice. Energy and tax rules change; verify with your DSO, installer, and current statute before investment.

Addendum: Market figures (HEMS survey shares, VPP GW, package bonuses, 1KOMMA5° facts) are compiled from 2025–2026 research notes and company disclosures; re-check primary sources before decisions. Editorial assembly from a Markdown vault with AI assistance.