Heliosolve is currently in closed beta. If you'd like to take part, contact us →

Glossary

What do P50, P90 and P99 mean for a solar project?

By Ivan and David, co-founders of Heliosolve · · Last reviewed

These are exceedance probabilities for annual production. P50 is the energy met or exceeded with 50% probability, P90 with 90% and P99 with 99%. The higher the number, the more conservative the figure and the lower the energy.

Who uses which?

PercentileQuestion it answersWho looks at it
P50What does a typical year produce?Developer, equity investor
P90What does a 1-in-10 bad year produce?Lender, to size debt
P99Does the project survive a very bad year?Lender, as a stress test

Where does the gap come from?

From total uncertainty: each year’s weather, the resource dataset, the simulation model, soiling and availability. Under a normal distribution, P90 = P50 · (1 − 1.282 · σ) and P99 = P50 · (1 − 2.326 · σ). Since σ differs from project to project, subtracting a flat percentage doesn’t hold up.

There are two common ways to compute it:

  • Analytical: combine the uncertainties in quadrature and apply a distribution shape. It’s correct when the sources are independent.
  • Monte Carlo: simulate many combinations of errors. It separates what averages out over the years from what doesn’t, and it lets sources be correlated.

Two nuances

  • Horizon matters. A one-year P90 isn’t the same as the P90 of a 10-year average. See 1-year vs. 10-year P90.
  • P90 drives debt. Each year’s cash flow in the P90 case sets the DSCR, and with it how much debt the project supports.

Limits

  • The gap is project-specific. How far P90 and P99 sit below P50 depends on each project’s uncertainty: there is no percentage that carries over from one plant to another.
  • Uncertainties are assumptions. The starting profiles aren’t measurements of your site and need to be justified project by project.
  • Ten weather years bound the analysis. With few years, the result depends on which years are included.
  • Validated against measured data at one plant. The energy chain was checked against the NIST Ground Array (−1.1% against 371 MWh measured): one plant, not a fleet. No track record with lenders yet.

→ Full methodology: P50 vs P90: the number your debt is sized against.

Talk to an engineer

How the engine is validated, with its scope and limits