Calculation chain

This page is the map of the engine: what is calculated, in what order and where each step is explained. Everything is done hour by hour over a year of 8,760 hours.

Overview

Weather data (GHI, DNI, DHI, T, wind)
        │
        ▼
1. Sun position
2. Irradiance on the module plane (transposition)  ← shading (objects, rows, horizon)
3. Optical losses: angle of incidence (IAM) + soiling
4. Rear-side gain (if the module is bifacial)
5. Cell temperature (Faiman)
6. Module power: single-diode model (irradiance + temperature)
7. Spectral effect
8. Electrical losses (mismatch, cables, connections, availability)
        │   ← one string
        ▼
9. Sum of strings → MPPT input → inverter (efficiency, clipping)
        │
        ▼
10. Balance with consumption and battery  →  self-consumption, export, import
        │
        ▼
11. Economics: savings, cash flows, NPV, IRR, payback, LCOE
12. Loss waterfall, P50/P90/P99 uncertainty, report

Order of the corrections

Engine principles

• One string at a time. Each string is simulated with its orientation, its module model and its shading; then they are aggregated per MPPT input and per inverter.
• Open, proven models. Published models are used (Perez, Hay-Davies, Faiman, Martin-Ruiz, First Solar, single diode with the PVsyst parameterisation), with the pvlib library as the implementation reference.
• No hidden adjustments. Everything you can change is in Advanced settings, and what was applied is recorded in Annex B of the report.
• Traceability. The loss waterfall splits the difference between the nominal energy and the delivered energy across each of these stages.

Conventions

• Time: the data are in UTC; consumption, tariffs and battery use local time.
• Azimuth: 0° = north, 90° = east, 180° = south, 270° = west.
• Tilt: angle relative to the horizontal.
• Power in W or kW, energy in kWh; irradiation in kWh/m².