Key takeaways
- A shade report answers a geometry question — solar access, TOF and TSRF, per module or per roof plane. It does not predict kilowatt-hours.
- An energy model predicts kilowatt-hours from a weather file, a module and inverter model, and a stated loss stack. Change one input and the number moves.
- A typical meteorological year file is a synthesized typical year built from many years of records, not a forecast of the year the system will see.
- Two proposals with unstated loss assumptions are not comparable. Below, the same 8.4 kW array models about 12 percent apart on assumptions alone.
- P50 is the midpoint of a distribution, not a floor. Once a lender or a PPA offtaker is involved, someone will ask for P90.
A shade report proves geometry: how much of the available sunlight reaches each module after obstructions, expressed as solar access. An energy model predicts kilowatt-hours under a stated set of assumptions. Neither one is a measurement, and neither one is a promise. Read a solar production report for its assumptions before you read its headline number, because the assumptions are where two proposals for the same roof quietly stop being comparable.
What does a solar energy model actually compute?
An energy model computes a predicted kilowatt-hour figure by stepping an hourly weather file through a chain of physical models: irradiance on the plane of the array, shading, module temperature and its effect on output, DC array behavior, inverter conversion, and a stack of derate factors. It produces a prediction under assumptions. It does not measure anything, and it does not commit anyone to an outcome.
Each stage in that chain narrows the number, and each carries an assumption somebody chose — usually by accepting a default and never looking at it again.
What the model does not compute matters just as much. It does not know your customer’s future load, the year’s actual weather, whether the inverter will trip on a weak grid, or whether anyone will ever wash the array. It does not know construction tolerance: modules land where the crew puts them, not where the 3D model put them.
This is why our solar production reports are built as sales-support and financing documents with their inputs stated on the face of the report, rather than as a single number in a proposal. An EPC or engineering firm reviewing a bid needs the weather source, the loss stack, and the shading treatment before the kilowatt-hour figure means anything.
What does a shade report prove that a production estimate does not?
A shade report proves obstruction geometry: how much of the available solar resource reaches a given module or roof plane once trees, chimneys, parapets and neighboring structures are accounted for, usually as an annual and monthly percentage. It is frequently the specific artifact a utility incentive program or a financier requires, and a kilowatt-hour estimate will not substitute for it.
The vocabulary is standardized enough to check. Aurora Solar’s documentation defines Solar Access as “the incident solar energy given shading, divided by the incident solar energy if there were no shading,” TOF as the Tilt and Orientation Factor comparing actual tilt and azimuth against the optimum, and TSRF as the Total Solar Resource Fraction combining both. When a rebate program sets a threshold, it sets it against one of those three, per plane or per module — not against annual kilowatt-hours.
An energy model can absorb a bad shading input and still return a confident-looking number; a shade report puts the geometry on the page where a reviewer can argue with it. Which makes the report only as good as its obstruction data. Heights and offsets estimated from the driveway produce solar access figures that look precise and are not. That problem belongs upstream: we cover what a site survey has to capture separately, along with the design considerations before drafting that decide whether the geometry is worth modeling at all.
One distinction worth holding on to: an annual solar access average can hide a plane that is unusable in December. Report solar access per roof plane, and per month where the array straddles orientations.
Why does the weather file move the answer?
Because the weather file is the model’s entire view of the sun. A typical meteorological year file contains one year of hourly data assembled to represent median conditions across a multi-year record. Two models of the same array, run on different weather sources, will disagree — and neither is wrong, because a TMY describes a climate rather than predicting a year.
NREL’s National Solar Radiation Database states it plainly: TMY files “contain one year of hourly data that best represents median weather conditions over a multiyear period.” The same page adds the caveat that gets skipped in sales conversations — “because TMYs represent typical rather than extreme conditions, they are not suited for designing systems to meet the worst-case conditions occurring at a location.”
Three consequences show up in real proposals. First, a TMY is synthesized: months are selected from different years and stitched together, so the file carries no meteorological continuity from one month to the next.
Second, outlier years are deliberately excluded. NREL’s P50/P90 analysis for solar energy systems notes that data representing months during outlier events “are excluded to ensure that the file represents the long-term climate,” and that TMY datasets “can overpredict the P50 value” relative to a multi-year simulation over the same period.
Third, source and vintage matter: a satellite-derived dataset at the project coordinates and a ground station eighteen miles away are different inputs, and new dataset releases move published values without anybody touching the design.
None of this makes TMY the wrong choice. It is the right input for comparing design options, and the wrong input for telling a customer what next year will look like.
Which losses should a credible model itemize?
A defensible model itemizes losses by category rather than applying one blended derate. The standard set is soiling, shading, snow, mismatch, DC and AC wiring, connections, light-induced degradation, nameplate tolerance, age, availability, and inverter conversion. Two proposals that do not state these are not comparable, however precise their figures look. NREL’s PVWatts Version 5 manual publishes the defaults much of the industry inherits without comment.
| Loss category | PVWatts V5 default |
|---|---|
| Soiling | 2% |
| Shading | 3% |
| Snow | 0% |
| Mismatch | 2% |
| Wiring | 2% |
| Connections | 0.5% |
| Light-induced degradation | 1.5% |
| Nameplate rating | 1% |
| Age | 0% |
| Availability | 3% |
| Total system losses | 14% |
Two things get misread. The total is not the sum — PVWatts combines the categories multiplicatively, which is why those percentages resolve to 14 percent rather than 15. And every value is a default, not a site condition: snow at 0 percent is wrong in Bend, soiling at 2 percent optimistic next to an unpaved county road.
PVsyst’s documentation describes its loss diagram as identifying “the main sources of losses” step by step, with incidence angle, ohmic wiring and array mismatch documented as separate stages. HelioScope presents the equivalent as a system loss chart alongside its annual production table.
The itemization is the point. Eight percent total losses may be defensible on a clean, unshaded commercial roof under a maintenance contract and indefensible on a low-slope retrofit under a maple, and you cannot tell which from the total. Same discipline that governs solar engineering plans: state the assumption on the document.
Worked example: two proposals, one roof, two loss stacks
Two firms model the same 8.4 kW array — 21 modules at 400 W, a south plane and a west plane, one mature tree off the southwest corner. Both use the same TMY dataset, module and inverter. Pre-shading, pre-loss modeled DC energy at the array plane is 12,700 kWh in both cases.
| Assumption | Proposal A | Proposal B |
|---|---|---|
| Shading treatment | Per-plane solar access, weighted to 96% | Single array-wide figure, 99% |
| Soiling | 2% | 1% |
| Mismatch | 2% | 1% |
| Wiring | 2% | 1.5% |
| Light-induced degradation | 1.5% | Not itemized |
| Nameplate tolerance | 1% | Not itemized |
| Availability | 3% | Not itemized |
| Total system losses (multiplicative) | 11% | 3.5% |
| Modeled year-one output | ~10,850 kWh | ~12,140 kWh |
Proposal B models roughly 1,280 kWh higher — about 12 percent — on the same roof, same equipment, same weather file. Nothing in B is arithmetically wrong. B declined to model three loss categories and reported shading as one array-wide number, so the west plane’s December problem never surfaces.
A salesperson holding both documents sees two kilowatt-hour figures and concludes one design is better. A reviewer who reads the assumption rows sees the designs are identical and the models are not. The fix is procedural: require the loss stack and per-plane solar access on every report you accept, including your own. On commercial solar PV designs where the model feeds a financing package, that becomes a condition of getting funded.
What do P50 and P90 mean, and when does the difference matter?
P50 is the production level a system is modeled to exceed half the time. P90 is the level it is modeled to exceed nine years in ten. Most residential sales proposals quote something close to a P50 without labeling it, which is fine for comparing options and inadequate the moment a lender, a tax equity investor, or a PPA offtaker is sizing debt against the project.
NREL’s P50/P90 analysis, cited above, puts it in plain arithmetic: a P50 value of 10,000 kWh means there is a 50 percent likelihood annual output will be greater than 10,000 kWh, while a P90 of the same figure means the system is likely to generate over that amount 90 percent of the time. The same paper notes that banks and investment firms use these values “to determine the risk associated with a project’s ability to service its debt obligations.”
The gap between the two is not a fudge factor. It is a calculated consequence of uncertainty — interannual variation in the solar resource, uncertainty in the irradiance dataset, and model uncertainty in how the array converts that resource to energy. A volatile site produces a wider gap than a stable one.
If a report does not say which exceedance level it represents, it is a P50 by convention and should be described that way, not as an expected annual yield.
And a P90 is not “the model minus ten percent.” It is derived from the distribution, which means it needs multi-year resource data rather than a single TMY run. That is where utility-scale PV system design work and bankable commercial packages diverge from sales-stage modeling: different data, different deliverable, different reviewer.
How do optimizers and microinverters change the shading penalty?
They change where the penalty lands, not whether it exists. In a conventional string, modules in series share one current path, so a shaded module drags the whole string toward its own operating point. With DC optimizers or microinverters each module is tracked independently, so the loss moves closer to the shaded modules’ own share of the array. A model that treats the two architectures identically will be wrong in a predictable direction.
On an unshaded roof the two architectures model within a point or two of each other. On a roof with a chimney at the ridge and a tree to the southwest, the modeled difference can run several percent — and the direction of the error depends on which tool made which assumption.
Two things to check on any report for a shaded roof. First, whether the model was run at module level at all, or whether an array-average shading derate was applied over a string-inverter topology. Second, whether the string layout in the model matches what the crew will build, since a string routed across two roof planes shades differently from one kept on a single plane.
Underneath it, the electrical design has to be real: the voltage window, temperature coefficient and MPPT arithmetic covered in our string sizing math decides whether the modeled configuration is buildable at all.
Module-level electronics are not a shading solution. They redistribute the loss. Trimming the tree still beats modeling it.
Aurora, Helioscope and PVsyst: what is each one for?
They solve overlapping problems at different resolutions. Aurora is built around remote site modeling and sales-stage shade analysis. Helioscope is built around fast commercial array layout with an integrated loss and production report. PVsyst is built around detailed simulation for bankable and technically demanding cases. The right choice follows the deliverable, not a ranking.
| Tool | Where it earns its place |
|---|---|
| Aurora | High-resolution 3D site models, per-module solar access and TSRF, shade reports for incentive and financing submissions, stringing and inverter configuration |
| Helioscope | Rapid commercial roof layout, preliminary shade analysis, itemized loss chart and annual production table for large arrays |
| PVsyst | Detailed loss diagram, bifacial gain modeling, near and far shading scenes, the format most often expected for bankable yield assessments |
Avila Solar Drafting works in all three, alongside AutoCAD. One operational detail changes how a Helioscope job is ordered: after 31 December 2023 we discontinued our internal Helioscope subscription, so Helioscope designs now run on the client’s own account — either by adding us as a user under your subscription, or by supplying credentials for us to work in your environment. Aurora and PVsyst work carries no such requirement.
Where a model travels with a permit or financing package, it needs company: an array layout that reconciles with the roof plan, and in many jurisdictions the stamped engineering deliverables a reviewer asks for before the package is complete.
Where does a production model stop being reliable?
A model is a prediction under stated assumptions, and it stops at the boundary of those assumptions. It cannot account for the actual weather of any given year, module degradation over the asset’s life, maintenance quality, grid availability, or how the system is finally built. Treating a modeled figure as a performance guarantee is a commercial risk, not a modeling one.
Four boundaries are worth naming to a customer before a signature.
Weather variance. A TMY run describes typical climate. Any single year can land above or below it, which is the entire reason the P50 and P90 vocabulary exists.
Degradation. Models apply a degradation rate as an assumption, uniformly, from a population effect. No individual array was consulted.
Operations. Soiling loss in the model is a fixed percentage. Soiling loss on the roof depends on whether anyone cleans it. Availability at 3 percent assumes a standard of response when an inverter faults.
As-built reality. Conduit runs where the structure allows, and the string map in the field is not always the string map in the model. This is where solar commissioning earns its keep — the first point at which anyone measures rather than models.
Naming them also keeps the model in its lane. A production report is not a permit document: the kilowatt-hour figure never reaches the plan checker, and a shade report has no conductor schedule, no placard sheet and no structural detail. That work belongs to solar plan sets, and the checks in our pre-submittal checklist apply to the drawing package, not to the model that sized it.
FAQ
What is a solar production report used for?
It supports the sale and the financing, not the permit. A production report gives a customer, a lender, or an incentive program a documented estimate of annual and monthly output, along with the shading geometry and loss assumptions behind it. Permit review uses a different document set.
How accurate are solar shade reports?
A shade report’s accuracy is set by its obstruction data, not by the software. Given correct heights, offsets and azimuths, remote shade analysis is precise enough that many rebate authorities accept it in place of on-site measurement, as HelioScope’s own documentation notes. Given estimated tree heights, it returns a confident number built on a guess.
Do I need PVsyst for a commercial project?
Not always. Helioscope or Aurora is usually sufficient for layout, sales support and preliminary shade analysis. PVsyst tends to be expected when a lender is reviewing the yield assessment, when bifacial gain has to be modeled, or when the loss diagram itself is a required deliverable.
What is the difference between P50 and P90 production numbers?
P50 is the level the model expects the system to exceed half the time. P90 is the level it expects to be exceeded in nine years out of ten. P90 is derived from the distribution of possible years, so it needs multi-year resource data rather than a single typical-year simulation. Sales proposals usually quote a P50 without labeling it.
Do you need my Helioscope account to run a report?
Yes. Avila Solar Drafting discontinued its internal Helioscope subscription after 31 December 2023, so Helioscope designs run on the client’s own subscription — add us as a user, or supply credentials for us to work in your environment. Aurora and PVsyst reports do not require a client-provided account.
Get the model and the assumptions in one document
Most proposals lose credibility in the second conversation, when a customer asks why two firms modeled the same roof twelve percent apart and nobody can point to the assumption that caused it.
Avila Solar Drafting produces Aurora, Helioscope and PVsyst reports with the weather source, the itemized loss stack and per-plane solar access stated on the face of the deliverable, backed by free AHJ and utility revisions, plus 3 months of free change orders from data submission.
Create a free account at my.avilasolar.com/signup and send us the site.
Working on something commercial, bankable, or with a shading condition you already know will be argued about? Call 971-410-0655 and we will scope it directly.
