Prototype · TRL 3–4

Pilot readiness & proposal evidence

FlexCasa prototype mapping to COSMIC OC2 Challenge 21 (Energy Literacy Online Game). Source: Technical Guidelines pp. 14–15.

Proposal evidence summary

Current TRL
3–4
Target TRL (month 10)
6–7

Evidence available today:

  • Vercel-ready working prototype (all routes functional)
  • Pre/post knowledge quiz with normalized learning gain
  • Analytics dashboard with A/B framing comparison
  • AI recommendation accept / modify / reject tracking
  • System mirror with labelled simplified estimates
  • Reviewer demo mode — instant 7-day instrumented session

Path to TRL 6–7:

Pilot dissemination with COSMIC partners, structured user testing (n≥200 closed beta), multilingual content (PT/EN + 2 languages), behavioural validation (TPB instrument, CSF co-design), PVGIS/ERSE-calibrated simulation, and predefined reach target with impact reporting in the COSMIC evaluation framework.

Open dashboard

Data realism — prototype vs funded pilot

TRL 3–4 prototype: Uses deterministic synthetic profiles for weather, solar generation level, electricity price level, and grid stress. Each of the 7 campaign days is scripted with fixed condition enums (sunny/cloudy/rainy/cold; solar none/low/medium/high; etc.). Scoring applies simplified but directionally correct logic (solar hours reduce cost/CO₂; evening peak penalises grid support; cold days penalise delayed heating).

TRL 6–7 funded pilot: Will replace and calibrate these profiles with PVGIS hourly solar data for representative EU locations, weather inputs (e.g. Open-Meteo) for outdoor temperature and COP behaviour, ERSE-published tariff structures for time-of-use pricing, and partner-reviewed heat-pump assumptions validated with the Daikin pilot. System mirror scaling will be documented against published grid patterns, not score-derived linear extrapolation alone.

We do not claim the prototype uses live external APIs. All system-level numbers are labelled as simplified estimates in the UI.

CH21 fit

  • Free, open-access web-based game — deployed on Vercel, no install.
  • Teaches energy flexibility, renewable production, system resilience, AI-enabled optimisation (AI assistant mechanic).
  • Trade-offs between comfort, cost, emissions, grid stability — four live meters.
  • Embedded assessments (pre/post quiz), engagement telemetry, A/B mechanic variants.
  • Multilingual-ready (EN + PT labels); full i18n in funded build.
  • Aligned with Daikin heat-pump pilot via heat pump as core flexible load.

Behavioural measurement design

LayerFrameworkPrototype
CapabilityCOM-BPre/post knowledge quiz (8 items, 2 per module)
MotivationTPB + BCT feedbackA/B framing (money / climate / comfort)
OpportunityCOM-BTimeline shows solar windows & peak stress
AI acceptanceRevealed preferenceAccept / modify / reject tracking

Scientific & energy assumptions

Defensible in prototype

  • ToU peak/off-peak price structure
  • Solar curve shape (midday peak, scaled by day profile)
  • Heat pump preheat vs evening peak trade-off
  • Evening grid-stress pattern (duck curve)

Synthetic (labelled in UI)

  • Day profiles — not live PVGIS or weather API
  • Flat CO₂ factor 0.25 kg/kWh
  • System mirror linear scaling from game scores
  • AI recommendation is rule-based, not ML

Path to TRL 6–7 (10 months)

  1. M1–2: Co-design with CSF + Daikin; PVGIS/ERSE integration; TPB instrument
  2. M3–5: Full art, daily-challenge mode, PT/EN/FR/NL
  3. M6: Closed beta n≥200; instrument validation
  4. M7: Public launch; dissemination (10k player target)
  5. M8–10: Impact analysis; COSMIC framework data handover

Validation metrics (TG p. 15)

Unique players
Completion & retention rates
Average playtime
Measurable knowledge gains
Stated behavioural intention
Predefined reach target

Do not overclaim

  • Prototype does not prove population-level behaviour change — only instrument design.
  • Learning gain from demo sessions is illustrative until validated with real users.
  • System mirror figures are not validated against ENTSO-E or DSO data.
  • Reach and efficacy targets belong to the funded project, not this TRL 3–4 slice.