P99 is the annual energy yield a solar plant is expected to exceed with 99% probability, derived from the uncertainty distribution in an energy yield assessment. Under a normal distribution it equals P50 minus about 2.326 times the combined uncertainty. Lenders and analysts use it mainly for stress testing and worst-case debt service checks.
What Is P99?
Formal definition
The P99 annual energy yield is the value Q where the cumulative probability function P(annual energy ≥ Q) = 0.99. In plain English: a solar plant modeled at P99 is expected to produce at least that much energy in 99 out of every 100 years, an extreme downside case, not a typical one.
Engineering definition
Like P50 and P90, P99 is derived from the same statistical distribution of simulated annual energy. It applies a larger confidence factor to the same mean (P50) and uncertainty band (σ) that drive every other percentile in an Energy Yield Assessment.
Industry definition
P99 is the most conservative percentile commonly reported in a bankable EYA. It exists to answer a narrow question: “how bad could a genuinely bad year get, and can the project still service its obligations?”
Why it’s requested
Commercial lenders almost always size debt against P90, because P99 is so conservative that using it as a routine sizing benchmark would leave real financing capacity on the table. P99 shows up instead where the counterparty is unusually risk-averse, government-backed loan guarantee schemes, multilateral development bank co-financing, or pension and infrastructure funds where capital preservation outweighs upside.
How P99 Is Calculated
P99 uses the same formula structure as every other confidence-level yield: start from the mean simulated annual energy (P50), then subtract a multiple of the combined uncertainty (σ). The multiple, the z-score, is what changes between percentiles:
- P50 = μ (no adjustment; z = 0)
- P90 = μ − 1.28 × σ
- P99 = μ − 2.33 × σ
The larger z-score for P99 reflects that it sits further into the tail of the distribution, it’s a rarer, more extreme outcome than P90.
Worked example
Using the same site data an EYA would carry through all four percentiles:
- Mean simulated yield (P50), μ = 250 GWh/yr.
- Combined uncertainty σ (inter-annual weather + modeling, RSS-combined) = 6% of mean ≈ 15 GWh.
- P90 = 250 − 1.28 × 15 = 231 GWh.
- P99 = 250 − 2.33 × 15 = 215 GWh.
The P90 - P99 gap (≈16 GWh in this example) grows on sites with higher weather variability, and shrinks on low-variability sites such as deserts with long, stable irradiance records.
P99 vs. P50 vs. P90
| Metric | Confidence | Typical Use |
|---|---|---|
| P50 | 50% | Equity IRR, expected revenue |
| P90 | 90% | Senior debt sizing, EPC energy guarantees |
| P99 | 99% | Government-backed loans, pension-fund stress tests |
Both P90 and P99 are read off the same underlying distribution, so a project with a wide P50 - P90 spread (a highly variable site) will also show a wide P50 - P99 spread, the two move together, just at different confidence levels.
Related Reading
P99 is one line in a larger yield-modeling exercise, it’s calculated alongside P90 and P50 from the same underlying distribution, so it’s rarely useful to read in isolation.
For EPCs and developers scoping out how deep that modeling and reporting work needs to go for a given project, energy yield assessments, permitting drawings, post-design reporting, the complete guide to solar design services in India covers where this kind of engineering scope typically sits alongside financing due diligence.
Key Takeaways
- P99 = 99th-percentile annual yield.
- Used for stress testing, government loans.
- P99 = P50 − 2.33 × σ.
- Pension funds and conservative lenders use.
- Standard EYA reports P50, P75, P90, P99.
Frequently Asked Questions
3 commonly searched questions about P99.
Who uses P99 instead of P90?
How does P99 differ from P90?
Is P99 used for routine debt sizing?
Need engineering-backed solar designs?
Heaven Designs prepares solar drawings, structural calculations, electrical engineering documents, and project-specific permit plan sets.
Akash Hirpara