Specificatie van het rekenmodel

Authoritative for the terugverdientijd-engine. Supersedes v1. v2 changes: Regime-A saldering annual-balance model (v1 treated all export as fully salderable — wrong for overproducers) · intra-month capture factor (v1 monthly sums overstated small-battery utilization) · scenarios renamed conservative/base/favourable with per-variable direction table (fixes monotonicity ambiguity) · trading uplift defined net (gross − revenue share − cycling cost − subscription), cap clarified · year-by-year degradation simulation replaces flat 0.90 multiplier · economic sizing replaces median-export matching (guardrail retained) · payback-vs-warranty honesty flag · headline = first-year simple payback + separate 10-yr cumulative net benefit (inflation removed from headline) · roof-orientation input · EV/heat-pump behavioural qualifiers (fixes double counting) · golden tests hardened + G11–G15.

0. Principles (unchanged from v1)

Calculate, never presume · every constant has source + date + VERIFY flag; CI blocks production recommendations while any required constant is unverified · versioned assumptionSet + engine version on every result and PDF · ranges with named drivers, not false precision.

1. Inputs (v1 table plus/amended)

• roofOrientation: zuid / oost-west / noord / weet_niet (default oost-west). Adjusts pvFactor profile shape and effective yield (zuid ×1.0, oost-west ×0.85 yield but flatter daily curve → selfUse +0.03, noord ×0.6 with warning; VERIFY factors vs PVGIS).
• hasEV + evChargingPattern: overdag_thuis / avond / gemengd / weet_niet. Self-use bonus (+0.05) ONLY when overdag_thuis.
• hasHeatPump + type hybride/all_electric: increases winter consFactor weighting; NO summer self-use bonus (v1's flat +0.03 removed — heat pumps barely run when PV peaks).
• hasElectricBoiler + boilerSolarControlled ja/nee/weet_niet: self-use bonus only if ja.
• Consumption defaults, contract, export compensation, feed-in structure, phase, budget, goal, ownership: as v1. "Weet ik niet" first-class everywhere; every default disclosed with its effect.

2. Scenarios (replaces v1 L/B/H)

Three named scenarios; each variable set in the battery-favourability direction, guaranteeing payback(conservative) ≥ payback(base) ≥ payback(favourable):

3. Energy & value model

3.1 Monthly profiles

pvFactor[12] per orientation (zuid peaked, oost-west flatter; VERIFY PVGIS NL), consFactor[12] per household config (heat-pump variant winter-weighted). Σ=1 each.

3.2 Direct self-use & evening demand

s0 by profile (thuis_overdag .40 / gemengd .30 / werkend_buitenshuis .22) + qualified bonuses per §1 (max cumulative bonus +0.10). eveningShare e as v1.

3.3 Monthly flows (per battery b, per year y with capacity(y) from degradation curve)

pvM       = pvProduction × orientationYield × pvFactor[m]
consM     = annualConsumption × consFactor[m]
exportM   = max(0, pvM × (1 − s0))
deficitM  = max(0, consM × e − pvM × 0.05)
theoreticalChargeM = min(exportM, usable(y) × daysM, maxChargeKw × window × daysM)
chargeM   = theoreticalChargeM × captureFactor          # §3.4
dischargeM= min(chargeM × rte, deficitM)
shiftedM  = dischargeM ; chargeActualM = dischargeM / rte

3.4 Capture factor

Accounts for intra-month solar volatility (monthly sums assume surplus arrives conveniently; real surplus clusters on sunny days a small battery can't fully absorb). Scenario-set scalar in v1 of the engine (see §2); flagged for calibration against representative-day profiles in engine v2. Methodology page discloses this explicitly.

3.5 Regime A — saldering annual balance (until 31-12-2026) [v2 correction]

importBefore = max(0, annualConsumption − pvProduction × s0_effective)   # grid imports pre-battery
exportBefore = Σ exportM
salderable   = min(exportBefore, importBefore)
nonSalderable= exportBefore − salderable

Battery charging is allocated first from nonSalderable export (marginal value of shifting = importPrice − exportComp, genuinely positive), then from salderable export (marginal value ≈ efficiency loss only, negative). Overproducing households (export > import) can therefore see positive pre-2027 value; balanced households see ~none. The engine computes this allocation; copy explains both cases honestly.

3.6 Regime B — post-2027 (per year y)

valueYr(y) = Σm [ shiftedM×importPriceCt − chargeActualM×exportCompCt + chargeActualM×feedInChargeCt ] − emsSubscriptionCtYr − netTradingUplift excluded (shown separately)

3.7 Payback & cumulative benefit

• Headline: paybackSimple = netCost / valueYr(y=1, base scenario)
• Cumulative: simulate y=1..10 with capacity(y); tenYearNetBenefit = Σ valueYr(y) − netCost; cumulative payback year = first y where Σ ≥ netCost
• netCost per offer prijsType; quote_required → [min,max] band widens the range
• Warranty flag [v2]: effectiveWarrantyYears = min(garantieJaren, garantieThroughputKwh / avgAnnualThroughput, garantieCycli / avgAnnualCycles). If paybackBase > effectiveWarrantyYears → mandatory result banner: "De geschatte terugverdientijd overschrijdt de garantieperiode." Ratio payback/effectiveWarranty ≤0.7 strong / 0.7–1.0 borderline / >1.0 weak feeds recommendation copy.

4. Dynamic trading uplift [v2 definition]

netTradingUpliftYr = grossTradingEurPerUsableKwhYr × usableKwh × (1 − aggregatorRevenueSharePct)
                     − incrementalCycleDegradationCt − emsDeltaCt

Defaults (VERIFY, aggregator publications are interested sources — label confidence low): gross 0/20/35 €/kWh-usable/yr (conservative/base/favourable), revenue share 30%, cap on NET uplift = €40 × usableKwh per year total. Displayed always as separate line, never in payback. R4 (below) uses net, favourable-capped-at-base uplift and requires battery×aggregator×contract compatibility all verified.

5. Recommendation decision tree (v2)

Primary scenario = Regime B base. Regime A shown transparently (incl. §3.5 overproducer case).

R0 huur/no-fixed-install → plug-in track
R1 pv==0 & goal!=backup:
     wants smart-charging/trading path & compatible setup → DYNAMISCH_CONTRACT_EERST(trading-motivatie, low confidence)
     else → NIET_KOPEN
R2 goal==backup → backup-capable shortlist, payback labeled secondary → KOOP_NU(backup)
R3 paybackBase ≤ 7 AND payback ≤ 0.7×effectiveWarranty → KOOP_NU / KOOP_PLUG_IN per track
R4 7 < paybackBase ≤ 10:
     not on dynamisch AND netTradingUplift (§4) would bring ≤7 AND compatibility verified → DYNAMISCH_CONTRACT_EERST
     else → WACHT_randgeval: both paths shown, email CTA "Mail mij een seintje op 1 januari 2027 met herberekende tarieven"
R5 10 < paybackBase ≤ 12 → WACHT (same email CTA)
R6 paybackBase > 12 OR payback > effectiveWarranty×1.3 OR valueYr ≤ 0 (G13) → NIET_KOPEN(in jouw situatie) — alternatives explained, no battery CTA

6. Battery shortlist — economic ranking [v2]

For every compatible published+verified battery (hard filters: phase, track, budget): run the full engine → rank by tenYearNetBenefit (tiebreak: paybackBase, then warranty throughput). Guardrails: exclude usable > 2× median Apr–Sep daily surplus ("te groot gekocht" flag) and G12 diminishing-utilization check. Shortlist 2–4. Offer fallback chain per PRD §6 so a missing affiliate link never breaks the flow.

7. Outputs, confidence, reproducibility

As v1 §6–7 plus: scenario table uses conservative/base/favourable columns; warranty flag; trading line separate; Regime A/B both shown; result stores full intermediate vector (charged, shifted, residual export, salderable split, valueYr per year) — required for golden tests and PDFs.

8. Golden tests

Frozen per case after assumption verification: annual charged/shifted/residual export, salderable split, valueYr(1), tenYearNetBenefit, paybackBase, range, recommendation, confidence, estimated-input list, shortlist slugs (or empty), top sensitivity drivers. G1–G10 as v1 (re-derived under v2 model), plus:

• G11 saldering cap: import 2000, export 4000 → assert nonSalderable=2000 and positive Regime-A value from that tranche only
• G12 oversize: same household, 5 vs 15 kWh → assert per-kWh utilization declines and 15 kWh not shortlisted when tenYearNetBenefit lower
• G13 negative value: favourable-to-export world (exportComp high, import low, subscription) → valueYr ≤ 0 → "geen terugverdientijd" state, R6, no nonsense numbers
• G14 quote band: installed min/max only → payback shown as band, widened range asserted
• G15 stale/unverified battery: best-ranked but unverified → excluded from shortlist, admin warning logged

9. Engineering constraints

As v1 §9, plus: engine simulates years 1–10 (still trivially fast); intermediate vector exposed for tests; captureFactor and all §2 scenario values live in the single assumption config; dedup key for click logs is a daily-rotating HMAC (no raw IP), retention 30 days.