Onderzoek: rendement van plug-inbatterijen

Roadmap item: A5 · Researched: 29 July 2026 · Researcher: Claude (not an independent verification — see verification-checklist.md; a second reviewer must confirm before any value is marked verified)

Question: can the five plug-in records be made engine-eligible by sourcing a product-specific round-trip efficiency, per option A of the v1 scope memo Part 4?

Answer: no — not from manufacturer documentation. Plug-in manufacturers do not publish round-trip efficiency at all. What they publish instead is systematically more optimistic and on the wrong basis.

---

1. What the engine needs, and why the bar is where it is

buildShortlist excludes any battery whose roundTripEfficiency is null, with reason no_rte. The rule exists because fixed and plug-in systems have materially different efficiency ranges, so a borrowed or cross-track figure would silently distort the ranking. Since the v1 scope change the rule is stricter still: simulateBatteryScenario throws rather than substituting a default.

A usable value must be a round-trip (AC in → storage → AC out) efficiency, rated rather than maximum, and identified as AC-AC rather than DC-DC.

2. Findings per product

Coverage is 3 of 5 at an official source. Zendure and EcoFlow rate-limited every route tried today; both must be re-attempted before this file is treated as complete.

3. The pattern, and why it matters more than the individual gaps

Across every source reached, plug-in manufacturers publish one of three things, none of which is a round-trip efficiency:

1. Maximum one-way conversion efficiency. Marstek's ">93.5%" is battery→AC only, and is a maximum rather than a rating. Squaring it to approximate a round trip (≈87.4%) would be an inference from a best-case number, in the battery-favourable direction, on an undocumented basis. That is precisely the error the no_rte rule exists to prevent.
2. Cycle life to a retention floor. "6000 cycles to 70%+ capacity" (HomeWizard, EcoFlow) is a degradation figure. It says nothing about efficiency.
3. A derived "usable for AC loads" figure. EcoFlow's blog states ~1.7 kWh of a 1.92 kWh pack. That silently blends usable capacity with conversion loss — two separate engine inputs. Adopting 1.7 kWh as usable capacity would double-count the conversion loss the RTE already applies.

This is the same failure mode as the competitor calculator. thuisbatterijoverzicht.nl assumes 90% efficiency for every product including plug-ins (see design-reference.md §5). Our engine refuses to, and the reason is now documented rather than assumed: the number simply is not published, so anyone using one has invented it.

Separately confirmed: the Marstek brochure still self-contradicts on capacity. The cover claims "5.12kWh of Usable Capacity" while the specification table gives Capacity 5.12 kWh with DoD 90%, implying 4.608 kWh usable. The earlier decision to reject the value stands. Ev's email to Marstek support remains the open route.

4. Recommendation — option A cannot close as specified

The scope memo's option A assumed the values existed and needed finding. They do not exist. Three routes remain, in order of preference:

A′ — an independent measured plug-in efficiency, applied per track. The objection encoded in the engine is to a cross-track fallback, not to a track-level value: fixed and plug-in ranges differ, but plug-ins resemble each other. A measured figure for plug-in AC-coupled systems from a disinterested source, applied to the plug-in track with a visible disclosure and an explicit confidence downgrade, would be defensible where a manufacturer maximum is not.

Best candidate source: the CE Delft / Vereniging Eigen Huis field study (Jan 2026, 13 households, Oct 2024–Oct 2025) already cited in the assumption dossier §3 for self-use. It measured real installed systems; whether it reports round-trip efficiency needs checking. Other candidates: TNO, an accredited test institute, or a Dutch consumer-testing body.

This route needs founder approval and must be put to the 0.6 reviewer explicitly, because it loosens a rule the engine currently enforces strictly.

A″ — ask the five manufacturers directly, as with Marstek. Slow, low hit-rate, but a "we do not publish this" answer is itself publishable and reinforces the positioning.

B — disable KOOP_PLUG_IN in v1, with designed copy on the plug-in tab explaining that the calculator does not rank plug-ins because their manufacturers do not publish the efficiency figure the calculation needs. Plug-ins stay visible and comparable on published specs in /vergelijken; they are simply not ranked.

Recommended: pursue A′ this week, hold B as the 11 August fallback. B is not a defeat — an honest "we will not rank these because the data does not exist, and here is what the manufacturers do and do not publish" is a stronger trust signal than a ranking built on a squared maximum. It is also directly publishable as editorial content, and no competitor can run it.

4a. Independent-source routes — both now closed (29 July)

Option A′ depended on finding an independent measured efficiency. Two candidate sources were chased to conclusion. Neither works.

CE Delft / Vereniging Eigen Huis field study — wrong product set

The study (Jan 2026, 13 households, Oct 2024–Oct 2025) measures real installed systems, but Table 1 lists every tested unit: Bliq 10/20/25 kWh · Sessy 5/10/10/15 kWh · Enphase IQ 3.5/7 kWh · Pylontech 20 kWh · Alpha-ESS 3.8/8.2 kWh. All thirteen are fixed-install. No plug-in system is in the study, so it cannot anchor a plug-in efficiency.

It remains highly valuable for other purposes and should be cited elsewhere:

• Its headline conclusion — "Thuisbatterijen kunnen het wegvallen van de salderingsregeling niet volledig compenseren" — is independent, disinterested confirmation of our positioning, from a study commissioned by a consumer organisation. This belongs on /hoe-we-rekenen.
• Measured self-consumption uplift: ~30% average; one setup 35% → 68% with a 5 kWh battery.
• Table 2 gives an installed price band of €1,400–2,500 incl. VAT for 2.7 kWh plug-in systems ("Systeem met stopcontact aansluiting"), a useful independent check on our price bands.

HTW Berlin Stromspeicher-Inspektion — right method, wrong products

Found via the BYD datasheet's "High efficiency certified by HTW Berlin" line. The Stromspeicher-Inspektion is the methodological gold standard for this question: it publishes measured Wirkungsgrad im Ladebetrieb and im Entladebetrieb, distinguishes AC-coupled from DC-coupled systems, and states its test standard. Exactly the boundary discipline we need.

It still does not help us:

1. The product set barely overlaps ours. The 2025 edition tested 22 systems from 17 manufacturers — Kostal, Fronius, RCT Power, Fox ESS, SAX Power, GoodWe, Fenecon, Energy Depot. Of our twenty catalogue records, effectively none appear.
2. Most results are anonymised (E1, K1, L1, M1 …), so even where a brand participates the figure often cannot be attributed to a product. sonnen appears only as "System C".
3. The one plug-in reference is not an efficiency measurement. Anker's marketing cites HTW for the Solarbank E1600, but the citation is to HTW's Stecker-Solar-Simulator — a self-consumption simulation tool — for a self-consumption rate (40% → 96%), not a measured round-trip efficiency. It is also a different model from our Solarbank 3 E2700 Pro.

Conclusion — option B is now the evidence-based answer, not the fallback

Three independent routes have been exhausted: manufacturer documentation on the plug-in track (nothing published), manufacturer documentation on the fixed track (only BYD, with no stated boundary — see fixed-rte-research.md), and independent measurement (wrong product sets in both candidate sources).

Recommend option B for v1: disable KOOP_PLUG_IN, with designed copy on the plug-in tab explaining that we do not rank plug-in batteries because their manufacturers do not publish the efficiency figure the calculation needs — and saying exactly what they publish instead. Plug-ins stay fully visible and comparable on published specs in /vergelijken.

This is a stronger position than it looks. It is defensible to a technical reviewer, it is publishable editorial no competitor can run, and it is the honest reading of three dead ends rather than a concession made under time pressure.

One route stays open but is not launch-critical: A″ — ask the five manufacturers directly, as with Marstek. A "we do not publish this" reply is itself publishable.

5. Open work

• Re-attempt Zendure and EcoFlow official specification pages (rate-limited today).
• Check the CE Delft/VEH study for a measured round-trip efficiency — closed: all 13 tested systems are fixed-install (§4a).
• Check HTW Berlin Stromspeicher-Inspektion — closed: right method, non-overlapping product set, anonymised results (§4a).
• Founder decision on option B by 11 August.
• Marstek reply on usable vs nominal capacity (Ev, outstanding).
• Founder decision on A′ vs B by 11 August, per the scope memo.

Sources

• MARSTEK VENUS E 3.0 EU brochure (PDF)
• HomeWizard Plug-In Battery
• Anker SOLIX Solarbank 3 E2700 Pro specifications
• EcoFlow STREAM Ultra review (EcoFlow's own blog — interested source)
• Zendure SolarFlow 800 Pro — official page unreachable today; retailer listings only