No PVsyst report is bankable by itself. Your lender or investor decides what it accepts. Prepare the project files, the weather data source, a reason for each loss, P50 and P90 results with their uncertainty inputs, and an independent review. IEA PVPS says yield results should come with an exceedance probability.
A PVsyst PDF is only the visible output of a model. A reviewer also needs the files and the evidence behind it. SurgePV sells solar design software and publishes this guide. PVsyst is made by PVsyst SA, a separate company.
Submission checklist
- Ask the lender or investor what they need, and who may rely on the report.
- Freeze the design revision you will report.
- Record where every input came from: weather, equipment, layout and operations.
- Give each material loss a basis: measurement, datasheet, contract, study or labeled assumption.
- Report P50 and P90 with the uncertainty inputs behind them.
- Export the project and variant with referenced files.
- Reproduce the reported result from the exported package.
- Write down every open difference and who will close it.
The sections below explain each step.
Agree on what the report is for
A tender estimate, a design comparison, an owner forecast and a financing review need different proof. Before you model, record:
- The project stage, design revision and equipment status.
- Who will use the report and who accepts it.
- The energy boundary, reporting periods and probability cases you must show.
- Whether the lender wants an independent assessment, a model rerun or extra measurements.
- Who may rely on the report, and who owns later revisions.
IEA PVPS Task 13 explains why this matters. Its executive summary on yield-assessment uncertainty says yield assessments feed the investment case, with cost data, to give LCOE (levelized cost of electricity), IRR (internal rate of return) and NPV (net present value). It says the results should come with an exceedance probability. It names the weather data as the most important input: database reliability, year-to-year variation and long-term trends.
Name the energy boundary and the period
Two models can both be correct and still differ, because they measure energy at different points. Say whether a number is array output, inverter output or energy injected into the grid.
PVsyst’s grid-system variable list separates EArray, the effective energy at the array output, EOutInv, the available energy at the inverter output, and E_Grid, the energy injected into the grid. Wiring, transformer, availability and grid-limit losses have their own variables.
| Record | What to write down |
|---|---|
| Output variable | The exact PVsyst variable and unit, not just “generation” |
| Location | Inverter terminals or the defined injection point |
| Included effects | Auxiliary power, transformer losses, export limits |
| Time basis | A specific weather year, a representative year or a long-term case |
| Design basis | Capacity, equipment and layout revision |
Example. One model shows 10,000 MWh at the inverter output. Another shows 9,800 MWh at grid injection. The gap is 200 MWh, which is 200 / 10,000 = 2% of the inverter figure. That is not a software error. It is the loss between the two points. Check that path first.
If your revenue is metered at a contract point, match that point to the model output. Do not subtract a downstream loss twice.
Keep an input register
A reviewer should be able to see where each input came from.
| Input | Keep this record | Reviewer’s question |
|---|---|---|
| Weather | Provider, product, version, coordinates, period, processing | What does this file represent, and how was it checked for this site? |
| Site measurements | Instruments, timestamps, quality checks, calibration | Are gaps and corrections shown? |
| Components | Module and inverter files with datasheets or test records | Do they match the equipment you will buy? |
| Layout | Drawing revision, module count, tilt, pitch, terrain, obstacles | Is the model the released design? |
| Electrical design | Strings, inverters, sub-array assignments | Does the wiring match the design and the shading model? |
| Operations | Cleaning, outages, export limits | Which items are contract, measured or still assumed? |
These are good records to keep. They are not a rule every reviewer shares. Ask your assessor to justify the weather source and any extra evidence the lender wants.
Give each loss a basis
For every material loss, write down where the number came from. Use measured data, equipment documents, a contract, a study, or a clear label that says “assumption”. Keep the date and the operating conditions. Do not tune losses to reach a chosen annual total.
PVsyst’s loss-diagram page says each loss is a percentage of the energy before it, so percentages do not add up. For example, 100 MWh reduced by 2% and then by 3% is 100 × 0.98 × 0.97 = 95.06 MWh. The total drop is 4.94%, not 5%.
PVsyst starts you on default values and tells you to change them for your system after the first run (array and system losses). These are the defaults it publishes. They are PVsyst’s starting points, not lender limits.
| Loss | PVsyst default or guidance | PVsyst page |
|---|---|---|
| DC wiring | 1.5% at STC (standard test conditions); the loss diagram usually shows about 60% of that, which is 0.9% | Ohmic wiring |
| Module mismatch | 0% for 1 module per string, 0.5% for 2, 1% for 3, 2% for more than 3 (thin film: 0%, 0.25%, 0.5%, 1%) | Mismatch |
| String voltage mismatch | 0% for 1 string per MPPT (an inverter input that tracks maximum power), 0.05% for 2, 0.1% for 3 to 10, 0.15% for more than 10 | Mismatch |
| LID (light-induced degradation) | No default. PVsyst says it is typically 1% to 3% or more for p-type crystalline modules. If you set it, the suggested value is 2% | LID loss |
| Unavailability | No annual value can be set. You enter downtime periods, and the energy loss follows from the weather in those periods | Unavailability |
IEA PVPS also gives an example of how to report losses. Its table lists each step from horizontal irradiation to grid injection, with an uncertainty for each step. It is a format example, not a target. This is the table from the executive summary, shortened to four columns.
| Step | Uncertainty % | Gain or loss % | PR % |
|---|---|---|---|
| Irradiation on module plane | 2.5 | 16.0 | |
| Horizon shading | 0.5 | -0.2 | 100.0 |
| Row shading | 2.0 | -1.7 | 98.3 |
| Object shading | 3.0 | 0.0 | 98.3 |
| Soiling | 0.5 | -0.5 | 97.9 |
| Reflection losses | 0.5 | -2.7 | 95.2 |
| Spectral losses | 0.5 | -1.0 | 94.3 |
| Irradiation-dependent losses | 0.8 | -1.5 | 92.9 |
| Temperature-dependent losses | 1.0 | -2.5 | 90.5 |
| Mismatch losses | 0.5 | -0.8 | 89.8 |
| DC cable losses | 0.5 | -0.8 | 89.1 |
| Inverter losses | 1.5 | -1.2 | 88.0 |
| Inverter power limitation | 0.5 | -0.1 | 88.0 |
| Additional consumption | 0.5 | -0.1 | 87.9 |
| AC cable losses, low voltage | 0.5 | -0.4 | 87.5 |
| Transformer, medium voltage | 0.5 | -0.9 | 86.7 |
| AC cable losses, medium voltage | 0.5 | -0.1 | 86.6 |
| Total | 6.5 | 86.6 |
PR is the performance ratio. The table also lists a global horizontal irradiation row (uncertainty 4.0%) and a high-voltage transformer row (0.0%), which are left out here. IEA PVPS shows this as an example of good practice.
Questions a reviewer will ask about your losses:
- Does the soiling number fit the site and the cleaning plan?
- Do the shading settings match the real layout?
- Is each wiring or transformer loss placed at the right energy boundary?
- Are availability, curtailment and clipping separate, and counted once?
- Does the degradation match the reporting period?
- Are the open assumptions visible, with an owner?
A zero or a rounded value is not wrong in itself. It is a problem when nobody can say why it was chosen.
P50 and P90
PVsyst’s P50-P90 page says the tool assumes annual yields follow a normal (Gaussian) distribution centered on your simulation result. You supply the uncertainties, and the tool adds them in quadrature to get one standard deviation (sigma). P50 is the median. P90 is the 10th percentile, which sits 1.28 sigma below the mean. The tool is not available for stand-alone and pumping systems.
PVsyst also says:
- Year-to-year variability of annual irradiation is typically 3 to 4% (RMS).
- Weather is the largest contributor to sigma. The PV module model is the main one after that.
- A specific weather year cannot be taken as the long-term P50.
- The tool has no general method for the bias of the weather file itself. It suggests running several weather providers to see the spread.
- Long-term degradation is not compatible with the P90 idea.
- P50-P90 values are meaningful for annual yield only, not for months or days.
Example. Take an assumed annual yield of 100 GWh. With an assumed sigma of 5 GWh, P90 = 100 − 1.28 × 5 = 93.6 GWh. With a sigma of 3 GWh, P90 = 100 − 1.28 × 3 = 96.16 GWh. With a sigma of 4 GWh, P90 = 100 − 1.28 × 4 = 94.88 GWh. A wider spread lowers P90. These are teaching numbers, not a project result or a lender rule.
IEA PVPS tested how far skilled assessors agree. Seven specialists got the same inputs for two sites. For Bolzano, Italy, their P50 ran from 1,095 to 1,406 kWh/kWp and their P90 from 997 to 1,274 kWh/kWp. For Alice Springs, Australia, P50 ran from 1,757 to 1,985 kWh/kWp and P90 from 1,631 to 1,819 kWh/kWp. The IEA authors say the gaps came mainly from the modeler’s choices, such as the weather database, site adaptation and degradation assumptions. That is why you document each choice.
Label scenario runs, such as an outage or an export cap, separately from probability cases. They answer different questions.
Export files that reproduce the result
Send more than the PDF. PVsyst’s export tool lets you export a full project or chosen variants. It asks whether to include referenced files. If you say no, it copies only the project, and the reviewer may miss inputs.
- Note the software version, project, variant and design revision.
- Export the agreed scope, with referenced files where allowed.
- Keep the input register, source documents and any adjustments you calculated outside PVsyst.
- Open the export in a clean workspace.
- Check that all files load and that the agreed output matches.
- Write down warnings and differences, and who will resolve them.
Do not overwrite earlier component or weather files. Keep a revision record so a reviewer can see which input changed a result.
Compare your model with the reviewer’s
Agree in writing what the independent review covers. It may include source-data checks, a rerun, separate modeling and sensitivity work. Then record each comparison:
| Field | What to write |
|---|---|
| Cases | Variant names and versions on both sides |
| Common basis | Weather, equipment, layout, energy boundary, period |
| Difference | The changed input or method, and the reason |
| Result | Output change in kWh and in %, with the base value stated |
| Status | Accepted, queried or open, with an owner |
For a sensitivity run, change one named assumption and keep the rest fixed. Keep it apart from your main forecast.
Mistakes that make a report hard to review
These come from the PVsyst and IEA PVPS sources above.
- Showing P50 only, with no exceedance probability.
- Using a specific weather year as if it were the long-term P50.
- Giving P90 with no uncertainty inputs behind it.
- Leaving PVsyst default losses unchanged.
- Adding loss percentages instead of multiplying them.
- Reporting energy at one point and billing at another.
- Sending the PDF without the project files.
- Using a P90 value for months or days.
Buying PVsyst report and simulation services
Start with the decision the model must support. State the project stage, reporting period, probability cases and who accepts the work. Ask for the PVsyst project and variant files, the software version, the input register, the loss diagram, the assumptions log, sensitivity cases and a revision record.
Heaven Designs lists PVsyst forecasts and shading analysis in its solar pre-design service. Confirm in writing the modeling scope, data sources, uncertainty method, files you receive, reviewer qualifications and who may rely on the result. A preliminary forecast does not replace detailed engineering, site checks, an independent energy-yield assessment or lender acceptance.
Disclosure: SurgePV is promoting Heaven Designs through a commercial relationship. Apply the same source-data review, reproducibility test, sensitivity review, and independent QA process to every provider.
Review the Heaven Designs solar pre-design service, visit the Heaven Designs website, and request a PVsyst report sample before seeking a quote.
For how to pick and manage a provider, use the solar design services buyer’s guide. For an introduction to the tool, read what PVsyst software is. To see how P50, P75 and P90 appear in sales conversations, read P50, P75 and P90 in sales. For the performance ratio, see the solar performance ratio guide.
See yield and financial outputs in SurgePV
Bring a project to a guided demo. We will show how SurgePV handles energy-yield and financial modeling in its generation and financial tool.
Book a Free DemoGuided demo · Bring a real project
The generation and financial tool page lists what that part of SurgePV covers.
Frequently asked questions
What is a bankable PVsyst report?
There is no universal bankable-report standard. A lender or investor decides what it accepts. Ask for their requirements first. Then prepare the project files, traceable weather data, a stated basis for each loss, P50 and P90 results with their uncertainty inputs, and an independent review.
Do lenders use P50 or P90 for solar project finance?
Ask your lender which case it needs and over what period. P50 is the value the annual yield has a 50% chance of beating. P90 is the value it has a 90% chance of beating. IEA PVPS says yield results should come with an exceedance probability. PVsyst puts P90 1.28 standard deviations below the mean.
Which meteo data sources do lenders accept in PVsyst?
Agree on the weather source with your assessor and reviewer. Record the provider, period and checks. IEA PVPS describes site adaptation, which joins a short on-site measurement campaign, typically around one year, to a long satellite record. A specific weather year cannot stand in for the long-term P50, according to PVsyst.
What loss assumptions raise red flags in a PVsyst report?
Watch for losses with no stated basis, PVsyst default values left unchanged, losses counted twice, and loss percentages that were added instead of multiplied. PVsyst says its defaults should be adapted to each project. A round or zero value is fine if you can show why.
What is the role of the independent engineer in solar financing?
The scope in the appointment sets the checks. Ask whether it covers source-data checks, rerunning your variants, uncertainty review and a written assessment, and who may rely on the result. IEA PVPS found that seven skilled specialists, given the same inputs, did not reach the same yield.
How long does bankability review of a PVsyst report take?
We could not find a sourced review time, so none is given here. Agree the schedule, reviewer availability and number of revision rounds in the appointment. A complete package makes review possible but does not promise first-pass acceptance.
Where this fits
This article is part of SurgePV's Solar Design hub, which works through the topic from first principles to the decisions a project team actually has to make.
