Simulation Physics & Data Reference

The physics chain, in order

A project (site, array geometry, components, tariff, dispatch) is handed to the simulation engine, which runs one ordered pipeline and returns energy, a per-stage loss cascade, and financials. The order is canonical and every loss term is applied at a fixed stage, so the loss diagram (Sankey) always reads top-to-bottom in the same sequence:

1. Transposition (GHI → POA). Horizontal irradiance is projected onto the tilted plane of the modules. This is a gain (typically +20 to +35 % at a good tilt), shown as the first, positive line of the cascade.

2. Shading and stow. Far/horizon shading, near (inter-row) shading, the share of near shading attributable to vegetation, a tracker's backtracking and rotation limit, and any tracker wind-stow are peeled off the POA before the cells ever see it.

3. Module physics. Angle-of-incidence (IAM), soiling, snow and spectral mismatch trim the irradiance that actually reaches the cells; STC conversion (η_STC) turns what is left into DC power; then the low-light (irradiance-level) and cell-temperature terms apply — all through the single-diode I–V model.

4. DC. The electrical (bypass-diode) effect of partial shading, module quality, year-1 LID, module-to-module mismatch, then DC wiring (ohmic) losses on the string and in the combiner box.

5. Inverter. Any module-level DC optimisers first (their own partial-load curve plus power and current clipping), then MPPT tracking, the inverter's partial-load efficiency curve, the five inverter-limit clamps (power/voltage min/max, night drain) and — on a non-unity cosφ setpoint — the reactive-power clipping increment.

6. AC / grid. AC wiring, transformers (MV then HV), the MV line, auxiliaries, availability and grid-side (weak-grid) losses down to the meter, less any battery-backup credit; beyond year 1, module ageing closes the chain.

The inverter's partial-load curve is the efficiency profile of its .OND file when the inverter came from one. Otherwise it is built from the inverters' datasheet efficiency, weighted by their AC power — the CEC figure first when the project's Electrical code on Diagram is NEC (US), the Euro figure first under every other code — and it is a typical 96 % curve when the datasheets give neither. With no inverter in the electrical design, the stage is a flat conversion loss that never clips, so the yield can read a few percent high: on a grid-connected system the Warnings card on Results says so, as it does when the results use the generic starter inverter that Create initial design places, which has a rating but no datasheet efficiency (Electrical Design & the One-Line Diagram).

A new loss term takes its own place in the chain without moving the existing ones, so every entry keeps its position relative to the others from one release to the next. Every entry also carries a provenance tag (where the number came from — PVGIS-attested, computed from weather, integrated from your bill-of-materials, or a literature default), which the loss diagram surfaces so a reviewer can audit each line. See Charts, Heatmaps & Tables for the loss diagram and Reporting & Export for the audited PDF export.

Weather and irradiance input

Everything downstream depends on the irradiance and temperature series the engine is fed. You choose it with Weather source, in the Weather & models card of the Configuration section on Production (/production).

Data sources

Two more sources feed the chain even on a PVGIS run:

• Open-Meteo archive — the site's hourly weather history: rainfall, snowfall, temperature, humidity, irradiance and 10 m wind from the Copernicus ERA5-Land (0.1°, from 1950) and ERA5 (0.25°, from 1940) reanalyses, and PM10 / PM2.5 particulates from the CAMS air-quality archive, which starts in spring 2022. The hosted service reads them through pvSpline's own instance of the open-source Open-Meteo server. Every year from 2014 to last year is fetched. The soiling (Kimber 2006, from rainfall and particulates) and snow (Marion 2013, from snowfall and temperature) losses average the last three complete years, and a year without particulate data uses Kimber's constant deposition rate. Both are worked out for each array zone at its own tilt — a tracker at its Rotation limit, or at a representative 45° when the module sets none — and weighted by the modules placed on each zone (equally when that is unknown); buildings, obstacles and vegetation do not count. Snow slides off a near-flat array at no less than the rate the Marion model gives a 5° slope, a stand-in for the melting the model leaves out; at 5° and above the published model applies unchanged. The same record gives the spectral correction its precipitable water and the snow-covered-ground albedo its snow depth, and its coldest daylight hour is the design cold temperature a batch checks the strings it plans against when the Diagram tab sets no site minimum (Deciding and applying). When you point Wind source at it, it also supplies the wind channel that drives tracker wind-stow and the thermal model's wind term. When present, soiling and snow losses become computed rather than flat constants.

• Enedis — the French metering operator, source of your consumption curve (30-minute load), not irradiance. It matters here only because its timestamp convention must be aligned with production before self-consumption is computed (see The Consumption Tab).

When PVGIS radiation database is left on Automatic, PVGIS picks the best radiation database for the location; you can force PVGIS-SARAH3 (satellite, Europe/Africa) or PVGIS-ERA5 (global reanalysis) to compare sources or to satisfy an independent engineer who standardised on one. The choice also sets the irradiance-uncertainty sigma used in the P50/P90 calculation (SARAH3 is treated as more accurate than ERA5).

How the series reaches the engine

You do not have to fetch anything by hand. The app requests the PVGIS and Open-Meteo series for you — when the array geometry changes, and again before a run that would otherwise have no weather — and writes the result into the project. A row of badges under the weather source says which series the project currently embeds (PVGIS, Open-Meteo, Enedis, weather file), and a project that carries them simulates offline, reproducibly, on any engine. Use the refresh control in the same card only when you want a newer fetch, or to retry one that was refused.

A project saved to your account keeps them as well: the series, an imported weather file included, are stored beside the save and come back when you open it, so it runs without fetching anything again. The measured Enedis series is the exception: once your client's authorisation for the delivery point is no longer in force, it is removed from every saved version, and only a new authorisation lets you download it again on Consumption (When your client's authorisation expires). A save that could not store its series reopens without them: its PVGIS and Open-Meteo series are then fetched again before the next run, while an imported weather file has to be applied again. An exported project file keeps them all (The portable project file).

A fetch that only partly succeeds keeps what it got: if the Open-Meteo part fails, the PVGIS series still lands, the project keeps any Open-Meteo series it already carried, and no other weather service is asked in its place. The badges show what the project ends up with; with no Open-Meteo series at all, soiling and snow use the percentages in the Loss cascade card.

Timestamp conventions

Two consequences are worth holding on to. First, every chart, table and export you read is built on start-of-interval stamps, so you never have to apply a shift yourself — doing so double-counts it. Second, the conversion happens at the moment the series is interpolated, which is also where the interpolation kernel and output resolution are chosen. Those two settings — Interpolation method (Linear, Cubic spline, PCHIP, Akima, Catmull-Rom, or Engine default) and Time resolution (1 to 60 minutes; 5 minutes is the recommended value and the engine default) — live in the Simulation band on Production (/production). They are saved with the project, so every run of it, single or batch, uses them.

One PVGIS detail shows up when you compare raw downloads: satellite (SARAH) rows are stamped at the satellite's scan minute rather than on the hour, and the engine treats those as instantaneous samples at that instant instead of hourly means. Reanalysis (ERA5) rows are stamped on the hour and treated as hourly means. You never configure this; it only explains why a raw PVGIS CSV may not show round hours.

Transposition: GHI → POA

Transposition converts horizontal irradiance (global GHI, diffuse DHI, direct DNI) into the plane-of-array irradiance that actually strikes a tilted module. The model matters most on diffuse-heavy hours (overcast, dawn/dusk, horizon brightening), where the choice moves annual POA by roughly 1–3 % on tilted geometry. Pick the model with Transposition model in the Weather & models card on Production (/production).

GHI decomposition (diffuse split)

When the weather source gives GHI but no diffuse (some TMY3 files, GHI-only satellite series, custom CSV), the engine must split GHI into direct and diffuse before transposing. PVGIS and EPW already carry diffuse, so on those paths no decomposition runs at all. This is not a setting you pick: the engine decomposes with Erbs, and falls back to it whenever a richer model's inputs are missing. The two models are described here because the difference shows up when you reconcile a GHI-only source against another tool.

Solar position

Sun zenith and azimuth are computed with the NREL Solar Position Algorithm (SPA, Reda & Andreas 2004), accurate to ±0.0003°, including atmospheric refraction. You do not configure this — it is used everywhere the chain needs a sun angle (AOI, tracker rotation, horizon shading). The azimuth convention is compass bearing from true north (N=0, E=90, S=180, W=270).

Module physics

Once POA reaches the module, a short sequence of physical effects converts it to DC power, each computed hour by hour. You set the module's parameters once, in the panel editor on Catalogue (/catalogue) — Add part… for a new module, Edit for one already in the catalogue, which opens the Add a panel reference form (The Component Catalogue). Leaving a numeric field empty keeps the engine's own value for it, so you only fill in what the datasheet actually states.

STC conversion and the single-diode I–V model

At standard test conditions the module converts POA to DC at its efficiency η_STC. Off-STC (real irradiance and temperature), the engine solves the single-diode equivalent circuit rather than a linear approximation, because the I–V curve shape drives low-light and voltage-rail behaviour. Pick the model with IV model.

The single-diode solution also produces the cascade's Irradiance level (low light) line: the I–V curve's own non-linearity at irradiances below STC, which PVsyst reports as “IL”. It only appears when the extraction succeeded — the linear-γ fallback cannot produce it — so an empty low-light line is a hint that a datasheet field is missing.

When a module's substrings and bypass diodes are described, the engine also models partial shading exactly: a partly-shaded substring is clamped by its bypass diode, so a 10 % shaded substring on a 60-cell module can delete ∼28 % of that module's power — an effect a naive irradiance-average misses. Set Bypass diodes (3 on virtually every crystalline module, 4 on 96-cell ones) and, optionally, Bypass diode Vf (0.45 V Schottky, 0.7 V silicon). Two or more diodes is what enables per-substring partial-shading modelling and lets a 3-D shading bake be consumed; left unspecified the engine assumes 3 rather than modelling a module with no bypass diodes at all.

Angle-of-incidence (IAM)

At oblique incidence the glass cover reflects part of the beam (Fresnel loss). The incidence-angle modifier IAM(θ) is the fraction transmitted. Choose the model with IAM model.

A read-only preview under the field plots the active IAM curve from 0° to 90°, so you can see the effect of the choice before saving. On the .PAN profile setting there is no curve to draw — its points are import-only — and the preview is replaced by a line saying so.

Cell temperature

Hotter cells produce less power (the power temperature coefficient γ_Pmpp is negative, typically -0.40 %/°C for c-Si). Two thermal models are available:

• Sandia / NOCT — T_cell=T_amb+(NOCT-20)/800· G_POA, driven by the module's Nominal Operating Cell Temperature (typically 45 °C for c-Si).

• Faiman (IEC 61853-2) — the PVsyst-canonical form, T_cell=T_amb+G_POA/(U_c+U_v· v_wind). The convective coefficient U_c follows how the array is actually mounted: 29 W/m²/K free-standing (rack, ground, carport), 20 for a semi-integrated array with a ventilated air gap or duct behind it, 15 for a fully integrated one with an insulated back — the hottest regime — and 32 for floating (on-water) arrays, whose cooler, breezier microclimate is folded into an effective coefficient.

Which regime applies is a property of the polygon, not of the module: set BIPV (Not integrated / Integrated (insulated back) / Semi-integrated (ventilated gap)) and On water on Map (/map). Either one overrides the module's NOCT, which is measured on an open rack and therefore does not describe an integrated or floating array. A polygon left on Not integrated and off water inherits the module's own Mounting setting instead.

The wind term is always on: the engine takes the mounting's wind-coupling coefficient U_v and applies it to the hourly wind series, so back-side airflow cools a free-standing or floating array as the wind picks up. There is nothing to enable. Both building-integrated regimes stay wind-naive on their own merits (U_v=0 — there is no free airflow behind them for the wind to scale), and an hour whose series carries no wind reading simply falls back to the no-wind result. Which wind series is used is the one choice you do make: Wind source (Field reference).

At sub-hourly time steps a lumped-capacitance transient model adds realistic thermal lag through cloud passage; it collapses exactly to the steady-state Faiman result at the hourly step.

The cascade's Temperature line is the year's net thermal effect: the energy the cells gain in hours colder than 25 °C is set against what they lose in the hotter ones, so on a site where cold hours win the line is a credit.

Spectral correction

Real sunlight's spectrum shifts with air mass and water vapour, and each cell technology responds differently. The engine applies a small technology-dependent spectral correction (air-mass polynomial, plus a precipitable-water term when the weather source carries humidity). Multi-junction (HCPV) cells are handled by a separate concentrator utilization factor instead, to avoid double-counting. You select the technology with Cell technology in the panel editor — it also drives the LID and degradation defaults. Where the source gives temperature and relative humidity but no water-vapour column, the engine estimates the column from those two; a pvSpline JSON weather series carries the measured value directly.

Bifacial, albedo and snow

Bifacial modules gain rear-side energy from ground-reflected light. The rear irradiance scales with the ground albedo (reflectance) and a self-shadow factor that depends on Ground clearance. Set Bifaciality (0–1) on the module (0 = monofacial, disables the gain; 0.65–0.85 for commercial bifacial) and Ground clearance (flush roof ≈0.05 m, ground-mount ≈1–1.5 m, canopy ≈2 m+). Ground albedo is set per polygon on Map (/map; see the material presets in Field reference); it also feeds the front-side ground-reflected term. When Open-Meteo snow data is present, the engine swaps in fresh-snow albedo (0.85) on snow-covered hours for both faces. The rear-side gain shows in the cascade as a negative loss (a credit), and the same multiplier carries all the way through the downstream DC, inverter and AC losses to the meter.

Concentrator optics (CPV / LCPV)

pvSpline models two concentrator regimes, each with its own group of fields in the panel editor. Both collect the direct beam only, so a beam-bearing transposition (one of the Perez settings) is more meaningful than the Hay-Davies path.

• High-concentration PV (HCPV). Selecting multi-junction as the Cell technology reveals the concentrator fields (pre-filled with generic ∼500× triple-junction defaults) and locks Tracker type to Two-axis, which a multi-junction cell needs; the editor says so in a note. A DNI-driven model replaces the flat-plate chain: the cell sees DNI× C×η_opt within the acceptance cone, a spectral utilization factor (air-mass polynomial) replaces the flat-plate spectral term, and cell temperature follows T_cell=T_amb+DNI/U.

• Low-concentration PV (LCPV). Tick LCPV concentrator optics to put trough / V-trough / CPC / Fresnel optics in front of ordinary c-Si cells — a regime most tools exclude because of non-uniform flux, which pvSpline models with a flux-uniformity penalty instead. Each optics type carries calibrated defaults; leave a field blank to inherit the per-type default. A concentrator admits only ∼1/C_geo of the diffuse sky, and the single-diode parameters (extracted at 1 sun) are extrapolated only up to Max valid concentration, beyond which the effective concentration is clipped and the P90 band widens.

Shading

Shading is split into distinct terms so the loss diagram tells you which kind is costing you energy:

• Far / horizon shading. The terrain or building skyline blocking the direct beam. Estimated from the PVGIS horizon profile, or measured from the skyline sampled in the 3D view (/earth, The 3D Earth Site View), crossed with the hourly sun position. PVGIS silently bakes the horizon into its G(i), so this cascade entry makes the hidden kWh visible. When the irradiance carries no horizon of its own — an imported weather file, for instance — Far horizon without a 3-D sweep on Production decides what an array the 3-D view has not swept gets: Auto and On shade its beam with the far horizon alone (the terrain model published with the 3-D tiles when there is one, else the PVGIS horizon fetched for the array), Off applies no far shading without a sweep. A run on PVGIS G(i) never consults it.

• Near / inter-row (mutual) shading. On regularly-spaced rows, the row in front blocks part of the beam and masks part of the sky diffuse at low sun. Which geometry applies comes from the polygon's Field setting on Map (/map) — Rows (sheds), Brise-soleil, Agrivoltaics (elevated) or Agrivoltaics (vertical bifacial) — or, for an East-West array, from the East-West dome switch and its Dome GCR. Modelled with an infinite-row 2-D geometry from the ground coverage ratio and profile angle. Two cascade lines separate the beam and the diffuse/ground portions.

• Near shading by vegetation. When a 3-D bake attributes part of the near shading to a tree canopy rather than to terrain, rows or buildings, that share is peeled out onto its own line (attenuation through the canopy, not an opaque block). The two beam lines together are what one line used to be, so the cascade still telescopes. Projects with no canopy-aware bake simply do not carry it.

• Electrical / mismatch effect of shading. A partially-shaded module loses extra power to string current-limiting and bypass activation, beyond the pure irradiance deficit. Surfaced as its own line when the polygon's field geometry sets an electrical-effect fraction.

• String-level mismatch. Beyond shading, series-connected modules that do not share an identical MPP force a common string current, pulling weaker modules off their own maximum. The engine applies it as one flat percentage. With the row's Auto toggle on (the default) and modules placed, the run takes PVsyst's default tier for your system: module mismatch 0 % below 10 modules, 1 % from 10 and 2 % from 100 (0.5 and 1 % for CdTe, CIGS or amorphous-Si modules), plus 0.10 % for 2 to 5 strings per MPPT input or 0.15 % for 6 or more. With Auto off, the Mismatch percentage stored in the project applies (1 % unless you type another).

Tracking

Tracking is a property of the module: set Tracker type in the panel editor on Catalogue (/catalogue). The six choices map onto three physical classes: fixed, single-axis (Single-axis, horizontal N–S or E–W, Single-axis, vertical, Single-axis, tilted) and Two-axis. Several refinements apply to single-axis trackers, and their fields appear only once you have picked one.

A zone on Map (/map) takes that tracking when the module is assigned to it, and its card then shows a read-only Tracker line (The Map & Site Design Tab). The run treats the four single-axis choices alike, and a vertical or tilted axis is not modelled as such: a single-axis zone turns about a horizontal axis along its Azimuth, 180° — the default — for a north–south axis (0° is the same axis) and 90° or 270° for an east–west one. On a PVGIS run the same axis chooses PVGIS's irradiance: a zone whose axis runs within 45° of east–west gets that of a horizontal east–west tracker, and every other single-axis zone that of a horizontal north–south one. On a tracker the simulation sets the modules' tilt hour by hour rather than using the zone's Tilt, and on a two-axis tracker their azimuth too. The weather-derived soiling and snow losses do not use the Tilt either: they count a tracker at its Rotation limit, or at 45° when the module sets none (Weather and irradiance input).

• Backtracking. Set Backtracking GCR (0–1) — the ground coverage ratio (module length / row pitch, typically 0.30–0.50). A non-zero GCR activates Anderson-Mikofski 2020 backtracking: near sunrise/sunset the tracker rotates back from the ideal angle to stop shading the next row. Leave at 0 for widely-spaced arrays or when GCR is unknown (residual ground-shade bias then shows up). What backtracking costs is the cascade's Tracker backtracking / rotation limit line (below).

• Rotation limit. Rotation limit is the tracker's mechanical Φ min/max (±45 to ±60° typical). It clamps every strategy — leave it blank for no limit (the ideal angle is already within ±90°). What the limit costs shows on the same cascade line.

• Wind-stow. Tick Enable wind stow to drive the tracker to a safe Wind-stow angle (0° = flat) whenever the hourly wind exceeds the Wind-stow threshold (12 m/s PVsyst default). The two fields appear only once the box is ticked. The sacrificed POA appears as a dedicated cascade line, and a day×hour heatmap shows exactly when trackers stowed. It reads the hourly wind chosen with Wind source on Production: PVGIS WS10m or Open-Meteo — on an imported weather file, the file's own wind or Open-Meteo.

PVGIS computes a tracker's irradiance for ideal tracking, with neither backtracking nor a mechanical stop. Both leave the modules flatter than the sun asks for — backtracking near sunrise and sunset, the rotation limit whenever the ideal angle lies beyond it — so they catch less of the direct beam, and the run charges that difference on a cascade line of its own, Tracker backtracking / rotation limit. On each hour where the real rotation departs from the ideal one, the default Hay-Davies model transposes the same horizontal irradiance twice — once onto the ideal plane PVGIS computed for, once onto the plane the tracker really presents — and scales the PVGIS plane-of-array value by the ratio of the two, so any bias of the model itself cancels and only the change of geometry is charged; the Perez models, which transpose onto the real plane anyway, report the same difference. Like the wind stow, the line is energy already taken off the plane-of-array total, shown rather than deducted twice, and the stow starts from the backtracked irradiance, so the two lines never count the same energy. For a horizontal north–south tracker field at a GCR of 0.4 expect about 1–2 % of the plane-of-array irradiance; an ideal tracker (GCR 0, no rotation limit) and a fixed array show no such line.

Sub-hourly time step

Physics time step (min) — in the Simulation band of the Configuration card on Production (/production) — controls how finely the physical chain integrates. Hourly (60 min) — default is the ordinary path. A finer step (30, 15, 10, 5, 1 min) synthesises an energy-conserving sub-hourly irradiance profile (clear-sky shaped) and engages the transient thermal model, capturing intra-hour clipping and thermal lag that an hourly average smooths over. 5 min is the recommended sub-hourly setting. Expect under 0.1 % on annual yield at stable sites, and 0.1–0.3 % on steep-angle or high-latitude arrays, from the sharper sunrise and sunset ramps.

The loss-diagram footer records which basis the run used, because the difference matters to a reviewer:

• Hourly — no intra-hour modelling at all.

• Synthesised clear-sky — the hourly irradiance redistributed across the sub-hour grid on a clear-sky shape. The ramps sharpen; no cloud-edge variance is invented.

• Native sub-hourly file — the sub-hour series came from an imported weather file whose records are finer than hourly. This is the only basis that is measured rather than modelled.

Two conventions keep sub-hourly results aligned with everything else. The synthesised and native sub-hour samples are instantaneous values at their own timestamp, while the hourly series they come from are interval means labelled at the start of their hour — the same start-of-interval rule as every cached series in the app (Weather and irradiance input). And the physics step is independent of the interpolation resolution that shapes the irradiance and load curves themselves, which is set per project beside it.

The loss cascade and provenance

Every loss is a line in one unified cascade, ordered POA → meter. The loss diagram (Sankey) on Results (/results) draws the whole of it, line by line; the Loss cascade card at the top of Production (/production) is its editable part — the post-processing stages, from soiling to LID, between the energy entering post-processing and the AC at the meter. Each line carries three things a reviewer reads:

• its value — a percentage of the energy remaining at that point, and a monthly series when the term is weather-driven;

• its position, which is canonical and stable from release to release; and

• its provenance — PVGIS-attested, PVGIS typical-share (marked with *), a real PVGIS horizon profile or a skyline sampled in the 3-D view, computed from Open-Meteo (Kimber soiling / Marion snow), integrated from your electrical bill-of-materials (I²R), a per-hour inverter-curve or optimiser-curve integration, an integrated outage or weak-grid schedule, the local physics chain, or a user constant.

What the lines are, in order

Not every line appears on every project — a term whose energy is zero is skipped, and whole families (concentrator, optimiser, transformer, weak-grid) appear only when the project has that equipment. The order below is the one the loss diagram always reads; the names here are short ones, and the diagram's own wording for each line is in The loss cascade, line by line:

1. Transposition gain (a credit), then the four transposition-stage lines that report energy already removed from POA: Far / horizon shading, Near shading, Tracker backtracking / rotation limit and Tracker wind stow.

2. Optical: on an LCPV project the concentrator optics come first — Concentration gain (LCPV) as a credit, then LCPV optical loss, LCPV acceptance loss and LCPV diffuse rejection; then Row shading (beam), Near shading (vegetation, 3-D bake), Row shading (diffuse), Angle of incidence, Soiling, Snow, Spectral mismatch.

3. Conversion: Module STC conversion, which converts the flux the cells actually see (on an LCPV project, the concentrated one). On an HCPV project the concentrator stages instead follow the conversion: Concentrator optics, Spectral utilization (multi-junction), Acceptance angle / tracking error.

4. DC module: Irradiance level (low light), Temperature (the net thermal effect, a credit where cold hours win), Bifacial rear-side gain (a credit), Row shading (electrical), Module quality, Light-induced degradation (LID), Module mismatch.

5. Wiring and conversion: DC ohmic wiring, LCPV flux uniformity on a low-concentration project, Combiner box ohmic, then — with module-level optimisers — Optimizer partial load, Optimizer power clipping and Optimizer current clipping.

6. Inverter: Inverter conversion, plus the six operating-envelope lines it decomposes into — Inverter MPPT operation, below start power, over nominal power, below MPPT voltage, above MPPT voltage, Inverter night consumption — and Power factor (cos φ) clipping on a non-unity setpoint.

7. AC and grid: AC ohmic wiring, MV transformer, HV transformer, MV line ohmic, Auxiliaries (fans, monitoring), Availability (maintenance / outage), Weak grid (trips + resync) and, when a battery covered outages, Battery backup recovery as a credit.

8. Beyond year 1: Module aging (degradation), which is not a year-1 stage at all — it compounds over the project horizon and therefore carries no bar.

The two MPPT-voltage lines, Inverter below MPPT voltage and Inverter above MPPT voltage, come from the strings wired on Map: each MPPT input is solved hour by hour with the strings routed to it, each at its own length, so a string too short or too long for the window shows up there (The Map & Site Design Tab). A batch scenario that changes the array works them out on the string layout planned for it (Deciding and applying).

You edit the user-adjustable losses in the Loss cascade card itself — one row per loss, with a Value column you type in and a Source column naming the provenance actually in use. Five rows carry an Auto toggle that hands the value to a computed source when data is available: Soiling and Snow (Auto (weather), from the weather profile), Mismatch (Auto, PVsyst's default tier for the module count, the strings per MPPT input and the technology — Shading), AC ohmic losses (Auto (BOM), integrated over the sized bill-of-materials) and LID (year 1) (Auto (tech), from the inherited cell technology). Two further rows are display-only engine figures rather than inputs — DC ohmic losses (info) and Combiner ohmic losses (info) — because those come from the electrical design (Electrical Design & the One-Line Diagram), not from a percentage. Reset all to defaults restores the percentages and the degradation parameters together. See Charts, Heatmaps & Tables for the diagram and Loss cascade — user-adjustable losses for the fields.

Probabilistic yield (P50 / P90)

Every run is framed as a distribution — there is nothing to switch on. The Probabilistic production card on Production (/production) carries one control: a four-segment Exceedance percentile selector (P50, P75, P90, P99), each segment showing its own year-1 figure, and P90 selected by default. The segment you pick is the percentile the financial table is computed on, and the one the run's year-1 energy is shown at: the energy tiles on Production name it in their caption (Annual production · P90), and the energy charts and the loss diagram on Results are scaled to it. The spreadsheet exports' energy rows and the PDF report's cover stay at P50 (Probabilistic production (P50/P75/P90/P99)). The ? in the card head opens the methodology note.

Annual yield in operating year t is treated as a normal variable Y(t)∼ Normal(µ(t),σ(t)·µ(t)); the exceedance P_X is the value the plant beats in X % of years (P90 = beaten 90 % of the time). The methodology follows IEC 61724-3:2016 and IEA PVPS Task 13.

The spread is the quadrature sum of the relative uncertainty terms below. The two ageing terms are zero in year 1 and grow with plant age; the last two contribute only where the project carries that exposure:

Interannual variability, soiling, availability and weak grid are treated as random year-to-year terms and average down as 1/√(N) over a lifetime; irradiance bias, model residual and the concentrator term are systematic and do not.

The mean losses that anchor the distribution (LID, soiling, snow, mismatch, AC ohmic, availability, bifacial gain) are pinned from the same loss cascade the Sankey and cashflow use, and both are shown at the percentile you picked: the card's figure for that segment equals the cashflow's year-1 net AC and the Sankey's net-meter output. On the default P90 that is the P90 segment, which the Annual production · P90 tile on Production also shows; the P50 segment is the deterministic year, the figure the spreadsheet exports' Net AC at meter (pvSpline loss cascade) row carries (Reporting & Export). Where a datasheet warranty floor binds, the percentile curve and the cashflow follow the same floored decay path.

Field reference

Every control that shapes the physics chain, grouped by the card it lives in. Ranges and effects are as the application enforces them. In the panel editor an empty numeric field means “keep the engine's value”, not zero — fill in what the datasheet states and leave the rest alone.

Production — Weather & models card (/production)

Production — Simulation card (/production)

There is no “local physical chain” switch and no wind-sensitive-cell-temperature switch: the engine behind the web app always runs the local chain, and always couples wind into the thermal model (The physics chain, in order, Module physics).

Production — lifetime degradation row (/production)

The cascade's last row is lifetime degradation, and it is driven from the warranty rather than from a rate you type: you state the datasheet line, the engine derives the implied annual rate and shows it read-only beside it, with a green / amber / red verdict on whether the pair sits in a plausible corridor.

Production — Probabilistic card (/production)

Panel editor — electrical specifications (STC)

Panel editor — mounting, tracking & IV model

Panel editor — concentrator (HCPV)

Visible only when Cell technology is multi-junction; selecting it also locks the tracker to two-axis and pre-fills the optics with the usual ∼500× literature values. A DNI-driven model replaces the flat-plate chain.

Panel editor — concentrator (LCPV)

Hidden behind the LCPV concentrator optics checkbox (off by default), which unlocks the fields below. A blank numeric field inherits the calibrated per-type default.

Map — mounting and ground (per polygon)

Four polygon properties on Map (/map) feed the physics chain directly, and none of them lives on the module.

Pick a material preset or enter a custom value:

A carport deck takes neither BIPV nor On water: whatever integration, on-water, field or dome setting its zone still carries, the run treats it as a free-standing deck with its module's own Mounting, and a butterfly deck as two opposed faces of equal area at the deck tilt with no row-to-row shading, so a bifacial module keeps its rear-side gain (The Map & Site Design Tab).

The zone card also shows, read-only, the tracking the zone took from its module — Tracker: Single-axis or Two-axis, nothing on a fixed zone. It changes only when a module is assigned to the zone again (Tracking).

Loss cascade — user-adjustable losses

Each loss below is a row of the Loss cascade card on Production (/production): a percentage you type in the Value column, with the provenance of the figure actually in use named in the Source column. Five rows carry an Auto toggle that defers to the computed source when the data is there.

Schedules & grid card — reactive power

The Schedules & grid card on Production holds three editors as tabs, reached either directly or from a Schedule… button on the matching cascade row: the availability schedule, the grid-unavailability schedule, and the reactive-power setpoint below.