A 5 MWp solar plant with a P50 forecast of 8,400 MWh per year and a P90 forecast of 7,500 MWh carries a 900 MWh gap between the central estimate and the bankable floor. That gap is not a modeling error. It is the price of weather uncertainty, expressed as a probability. Misread it, and a project can sail through development only to breach debt-service coverage in its second year of operation.
P90 solar is the annual energy production level that a solar plant is expected to meet or exceed in 90% of years. It is the conservative number banks use to size debt, the floor investors use to stress cash flow, and the value developers often guarantee under a power purchase agreement. Understanding P90 is essential for anyone who reads solar proposals, yield reports, or term sheets.
Quick Answer
P90 solar means there is a 90% probability that the plant will produce at least that much energy in any given year. It is a conservative, bankable yield estimate typically 8–15% below the P50 median forecast, depending on weather variability and data quality.
In this guide:
- What P90 means as an exceedance probability and how it relates to P50, P75, and P99
- Why lenders, investors, and PPA buyers care about P90
- The formula and uncertainty inputs behind a P90 calculation
- A worked example you can replicate with a calculator
- Common misreadings of P90 and how to avoid them
- When P90 matters and when it does not
- How to tighten your P90 number with better data and design
What Does P90 Mean in Solar?
P90 is a probability statement, not a production target. The “P” stands for probability of exceedance. The number tells you how likely the plant is to hit or beat a specific annual energy value.
- P50 — Exceeded in 50% of years. This is the median, or most likely, annual output.
- P75 — Exceeded in 75% of years. A moderate risk-adjusted case.
- P90 — Exceeded in 90% of years. The conservative, bankable estimate.
- P99 — Exceeded in 99% of years. An extreme downside case used for stress tests.
If a project has a P90 of 7,500 MWh, you expect it to produce at least 7,500 MWh in 9 out of 10 years. In 1 out of 10 years, weather or system issues may push output below that level. The higher the P-number, the more conservative the estimate, because it must be easier to exceed.
These values come from a normal distribution fitted to the expected annual yields. P50 is the mean. P90 sits 1.282 standard deviations below the mean. P99 sits 2.326 standard deviations below. The spread between them is driven by total project uncertainty, not by the size of the system.
Pro Tip
Think of P90 as the yield that shows up even when the sun does not cooperate. It is the number you would use if you had to promise a minimum production level to a skeptical lender.
Why P90 Matters for Solar Projects
Solar production is not deterministic. A nuclear plant produces a predictable output as long as fuel and operators show up. A solar plant depends on clouds, aerosols, temperature, soiling, snow, and inverter availability. Long-term averages are stable, but single years can swing 10% or more from the mean.
That variability creates a problem for finance. A lender who sizes debt against P50 accepts a coin-flip chance that revenue falls short in any given year. That violates standard debt-service-coverage covenants. So lenders use P90, the production level that almost always arrives.
P90 shows up in three places:
- Debt sizing — Lenders calculate debt-service coverage ratio using P90 revenue. A higher P90 means more debt capacity.
- Power purchase agreements — Buyers may require a P90 production guarantee from the developer. Missing it triggers liquidated damages.
- Investor returns — Equity investors use P50 for upside, but they stress-test downside using P90 or P99.
The gap between P50 and P90 also signals project quality. A tight gap means the site is well understood, the data is strong, and the design is conservative. A wide gap means the project carries more uncertainty, which translates into higher interest rates or lower leverage.
How P90 Is Calculated
The math behind P90 is straightforward once you know the inputs. The formula is:
P90 = P50 − (1.282 × σ_total)
Where σ_total is the total combined uncertainty, expressed in the same units as P50.
The 1.282 multiplier comes from the standard normal distribution. It is the z-score that leaves 10% of outcomes below the P90 value and 90% above.
Combining Uncertainty Sources
Total uncertainty is not guessed. It is built from independent uncertainty sources combined with a root-sum-square method:
σ_total = √(σ_irradiance² + σ_model² + σ_interannual² + σ_local² + σ_module²)
The main uncertainty sources are:
| Uncertainty Source | Typical Range (% of P50) |
|---|---|
| Long-term irradiance data | 2.5 – 4.0% |
| PV simulation model | 3.0 – 5.0% |
| Inter-annual weather variability | 2.0 – 4.0% |
| Soiling, snow, and local effects | 0.5 – 2.0% |
| Module performance and binning | 0.5 – 1.5% |
For a typical commercial rooftop project, total uncertainty lands between 6 and 9%. For a utility-scale site with multi-year ground measurements, it can drop to 4.5 to 6%. For a small residential system with no on-site data, it can rise above 10%.
Worked Example
Take a project with:
- P50 = 10,000 MWh/year
- σ_irradiance = 3.5%
- σ_model = 4.0%
- σ_interannual = 3.0%
First, combine the uncertainties:
σ_total = √(3.5² + 4.0² + 3.0²) = √(12.25 + 16.0 + 9.0) = √37.25 = 6.10%
Convert to MWh:
σ_total = 10,000 × 0.061 = 610 MWh
Then calculate P90:
P90 = 10,000 − (1.282 × 610) = 10,000 − 782 = 9,218 MWh/year
So the project has a 90% probability of producing at least 9,218 MWh in any given year. That is 92.2% of P50.
Software like PVsyst, Solargis, and SurgePV’s design suite automate this calculation from multi-year irradiance datasets and system inputs.
P90 vs P50: The Real Difference
P50 and P90 answer different questions. P50 asks, “What is the most likely annual output?” P90 asks, “What output will almost certainly show up, even in a bad year?”
| Metric | Meaning | Typical Use | Relation to P50 |
|---|---|---|---|
| P50 | Median annual output | Equity planning, production targets | Baseline |
| P75 | Exceeded in 3 out of 4 years | Moderate risk case | ~5% below P50 |
| P90 | Exceeded in 9 out of 10 years | Debt sizing, PPA guarantees | 8–15% below P50 |
| P99 | Exceeded in 99 out of 100 years | Extreme stress tests | 12–20% below P50 |
The most common mistake is reading P90 as “90% of P50.” It is not. A project with 6% total uncertainty has a P90 about 7.7% below P50. A project with 10% total uncertainty has a P90 about 12.8% below P50. The gap widens as uncertainty grows.
Another common mistake is treating P50 as a promise. A P50 forecast means the plant will underperform in roughly half of all years. That is not a flaw in the model. It is the definition of a median.
When P90 Is (and Is Not) Useful
P90 is not always necessary. It adds value when money, contracts, or guarantees are on the line.
When P90 matters
- Commercial and industrial projects with financing or tax-equity structures
- Utility-scale projects seeking non-recourse project finance
- Power purchase agreements with production guarantees or liquidated damages
- Portfolio valuation for asset owners selling or refinancing operating plants
- Off-grid and mission-critical systems where low production has serious consequences
When P90 matters less
- Residential sales proposals, where a simple production range is enough
- Early-stage site screening, before measurements or detailed design exist
- Internal pre-feasibility studies, where P50 plus a sensitivity range is faster
For residential installers, presenting a single production number as a guarantee is risky. A better approach is to show a P50 central estimate with a plus-or-minus range based on historical weather variability. That sets realistic expectations without over-engineering the proposal.
How to Improve Your P90 Number
A better P90 does not mean a more productive plant. It means a more certain plant. You improve P90 by reducing uncertainty, not by adding panels.
Use better irradiance data
Long-term satellite datasets from Solargis, Meteonorm, NSRDB, or SolarAnywhere reduce irradiance uncertainty. For bankable reports, lenders prefer 15 to 25 years of hourly data. For a deeper look at irradiance components, see our GHI, DNI, DHI guide.
Run an on-site measurement campaign
A 12-month pyranometer campaign at the project site can cut GHI uncertainty from ±3.5% to ±2.0–2.5%. A 24-month campaign cuts it further. This is why most utility-scale projects require at least one year of on-site data before financial close.
Choose proven components
Module power tolerance, temperature coefficients, and degradation rates all feed into model uncertainty. Tier-1 modules with tight binning and verified performance reduce module uncertainty.
Minimize shading and soiling
Shading is non-linear. A small shadow on one string can disproportionately cut output. Use shadow analysis to validate row spacing, nearby obstacles, and seasonal angles. In dusty regions, account for realistic soiling losses and cleaning schedules.
Validate the simulation model
Cross-check the model against independent datasets or measured performance from similar nearby projects. A model bias of 2% can shift P90 by more than you expect.
Frequently Asked Questions
What does P90 mean in solar?
P90 is the annual energy production level that a solar plant is expected to meet or exceed in 90% of years. It is a conservative, bankable estimate used by lenders and investors to size debt and manage weather risk.
How is P90 different from P50?
P50 is the median expected output — the system produces more than P50 in roughly half of all years. P90 is the conservative output exceeded in 9 out of 10 years. P90 is typically 8 to 15% lower than P50, depending on site uncertainty.
Why do banks use P90 instead of P50?
Banks use P90 because it represents the revenue floor that almost always shows up. Debt service must be covered even in below-average sun years, so lenders size loans against the 90% probability case rather than the average case.
How do you calculate P90 solar production?
P90 = P50 minus 1.282 times the total uncertainty. Total uncertainty combines irradiance data uncertainty, simulation model uncertainty, and inter-annual weather variability using a root-sum-square method.
Is P90 the same as 90% of P50?
No. P90 is the production level exceeded with 90% probability, not 90% of the P50 value. The gap between P50 and P90 depends on the project’s total uncertainty, not a fixed percentage.
Do residential solar proposals need P90?
Formal P90 analysis is usually not required for residential systems. It becomes important for commercial, industrial, and utility-scale projects that need bankable yield reports for financing or power purchase agreements.
What is a good P90 to P50 ratio?
A typical well-modeled project lands between 0.88 and 0.94, meaning P90 is 6 to 12% below P50. Ratios above 0.92 usually indicate high-quality irradiance data and low weather variability.
How can I improve my project’s P90?
Reduce uncertainty. Use long-term satellite or ground-station irradiance data, run an on-site pyranometer campaign, choose proven components, minimize shading, and use a validated simulation model.
Model bankable P50/P90 yields in one platform
SurgePV runs hourly simulations and outputs probabilistic energy yield estimates lenders can trust. Build your design, stress-test the uncertainty, and export a report-ready P50/P90 analysis.
Book a DemoNo commitment required · 20 minutes · Live project walkthrough
Key Takeaways
- P90 is the annual solar production level exceeded in 90% of years, not 90% of P50.
- Lenders use P90 to size debt because it represents the revenue floor that almost always shows up.
- P90 = P50 − (1.282 × total uncertainty), where total uncertainty combines irradiance, model, inter-annual, local, and module sources.
- A typical P90 lands 8–15% below P50. A tight gap signals high data quality and low weather variability.
- P90 matters most for financed commercial, industrial, and utility-scale projects. It is usually overkill for residential proposals.
- You improve P90 by reducing uncertainty: better irradiance data, on-site measurements, proven components, and validated modeling.
For a deeper dive into the full P-value curve, read our P50 vs P90 energy yield explained guide.
