Solar Yield Calculation & Energy Simulation: P50, P90 and Bankable Reports Guide

The energy yield simulation is the foundation of every solar financial model. If your P50 estimate is 10% optimistic, the investor's IRR drops by 1–2 percentage points. For residential customers, it means proposals that overstate savings and create complaints 18 months after commissioning. Accurate yield calculation requires correct inputs, defensible loss assumptions, and an honest probability framework.

What Is a Solar Yield Calculation?

A solar yield calculation predicts annual energy output (kWh/year) under defined assumptions. It is the single number that feeds the financial model, grid connection application, and insurance documents for any solar project.

The difference between AC yield and DC yield matters here. DC yield is raw panel output before inverter conversion losses. AC yield is what the meter actually records — and what you invoice. Proposals should always state AC yield.

The core formula: Yield = System size (kWp) × Specific yield (kWh/kWp), where specific yield = Peak Sun Hours (PSH) × Performance Ratio (PR) × 365. PR for a well-designed system is typically 0.75–0.85. This formula gives you the ballpark; a full simulation gives you the number you can defend.

Simulation Inputs: The 5 Data Categories

Every simulation engine needs five categories of input. Getting any one wrong by a material amount will throw off the result.

1. Solar Resource Data

The most important input. Weather data quality drives simulation accuracy more than any other factor. Use TMY (Typical Meteorological Year) data from an appropriate source for your project region. Poor data means a yield estimate that can be 5–15% off before you've made a single design choice.

2. System Configuration

Module specifications (Pmax, temperature coefficient, bifaciality factor if applicable), inverter specifications (efficiency curve, MPPT range, clipping threshold), and string configuration. Use manufacturer datasheet values, not generic defaults from a simulation library that may be outdated.

3. Site Geometry

Tilt, azimuth, and shading. Tilt of 30–35° is typical for central Europe; south-facing is optimal in the northern hemisphere. Even 15° of azimuth deviation from south costs only 2–3% in most locations — but unmodeled shading can cost 8–15%. This is why Chapter 4 (Shading Analysis) feeds directly into this chapter.

4. Loss Assumptions

Soiling, temperature, wiring resistance, inverter availability, module degradation. Each loss is a decision: too optimistic and you'll overperform your proposal; too conservative and you lose deals. Use industry-standard ranges and document your reasoning.

5. Output Metric

Are you producing a P50 (median expected output) or P90 (exceeded in 9 years out of 10)? The answer depends on the audience. Residential proposals use P50. Bank financing requires P90. Using the wrong metric for the wrong audience is a commercial and liability risk.

Solar Resource Data: TMY, Meteonorm, and PVGIS

Weather data is the single most variable input. Poor data equals unreliable yield estimates. A 5% error in irradiance translates to roughly a 5% error in yield — and a 5% error in yield can swing a residential payback period by 6–9 months.

TMY (Typical Meteorological Year) is a synthetic year assembled from real measurements to represent the "typical" climate at a location. Based on 10–30 years of historical data, it smooths out anomalous years and gives a median expectation.

For residential and small commercial projects below 500 kWp, PVGIS data is generally acceptable. For projects requiring debt finance or insurance, lenders expect Meteonorm or Solargis. Historical irradiance varies by ±5–8% year-to-year in most European climates — this inter-annual variability is the primary driver of the gap between P50 and P90.

P50 vs P90: What They Mean and When You Need Each

This is the most commonly misunderstood concept in solar yield assessment. Get it wrong in a financing context and you create a liability problem.

P50 estimate is the median yield. Statistically, 50% of years will produce more and 50% will produce less. Use for: residential proposals, operational planning, internal business cases.

P90 estimate is the yield exceeded in 90% of years — only 1 year in 10 produces less than this. Use for: project finance applications, lender base case, insurance sum insured.

The P90 calculation uses a normal distribution: P90 = P50 × (1 − 1.28 × total uncertainty)

Total uncertainty (1σ, combined in quadrature) typically includes:

• Historical irradiance variability: ±5–8%
• Weather data source uncertainty: ±2–4%
• Simulation model uncertainty: ±2–3%
• Combined: typically ±7–10% (1σ)

Example: P50 = 100,000 kWh/year, combined 1σ uncertainty = 8%.
P90 = 100,000 × (1 − 1.28 × 0.08) = 100,000 × 0.898 = 89,800 kWh/year

The gap between P50 and P90 in this example is 10,200 kWh/year — about 10%. For a commercial project selling electricity at €0.30/kWh, that gap represents €3,060/year in revenue uncertainty. Banks lend against P90 so that even in a bad year, the system generates enough cash to cover debt service.

The Loss Tree: Where Energy Goes

Simulation tools produce a loss tree (or waterfall chart) showing each loss category. Understanding these lets you improve layouts and set defensible assumptions.

A Performance Ratio of 0.78–0.82 is typical for European residential systems in 2026. Below 0.75 warrants investigation. Above 0.85 is unusual and may indicate overly optimistic loss assumptions.

Running a PVsyst Simulation: Step-by-Step

1. Create new project. Enter the site location — PVsyst loads PVGIS or Meteonorm data automatically based on coordinates.
2. System configuration. Add modules from the database (verify manufacturer specs match actual purchase order), add inverter, configure string lengths.
3. Orientation. Enter tilt and azimuth for each array section. For complex multi-pitch roofs, create separate sub-arrays.
4. Near shading. Build the 3D scene with obstructions, or import a measured horizon profile. This step is where shadow analysis software data feeds in directly.
5. Simulation. Run — 10–30 seconds for a standard residential project.
6. Loss tree review. Check each loss category against the expected ranges in the table above. Anomalies indicate input errors.
7. Report generation. Export the PDF simulation report — required for any bankable assessment.

Common PVsyst errors that produce inflated yield estimates:

• Using "standard shading" instead of "module layout" shading — consistently underestimates partial-shade losses by 2–5%
• Forgetting to activate soiling losses (default is 0 in some versions)
• Using a database module without verifying it matches the actual purchased product
• Not checking inverter MPPT range against string Voc/Vmp at minimum and maximum temperatures

Validating Simulation Against Real Plant Data

For installers with an existing portfolio, benchmarking simulated vs. actual yield calibrates your loss assumptions. This is one of the most useful things a design team can do.

The benchmark metric is specific yield (kWh/kWp/year). Compare actual monitored output against the P50 simulation for the same period.

• Actual/simulated above 100%: your loss assumptions are conservative. Safe for sales, but overstates uncertainty for lenders.
• Actual/simulated below 95%: your shading, soiling, or availability assumptions are too optimistic.

Systems with well-modeled shading typically perform within 3–5% of P50. Systems with unmodeled shading underperform by 8–15%. If your portfolio consistently underperforms P50 by more than 5%, revisit the shading methodology before the next project cycle.

Bankable Energy Yield Reports

For projects requiring external finance, a bankable yield report must meet specific requirements. Lenders and their technical advisors review these documents closely, and a gap in methodology is enough to stall a financing process.

Minimum contents of a bankable yield report:

• Site description and coordinates
• Solar resource data source and justification for selection
• Simulation tool and version number
• System configuration with module and inverter data sheets referenced
• Shading analysis methodology
• Loss assumptions with explicit justification for each value
• P50 and P90 annual yield
• Monthly yield breakdown (12-month table)
• Uncertainty analysis methodology and combined σ calculation
• Simulation software output files (attached as appendix)

For projects below roughly 500 kWp, reputable solar design software like SurgePV outputs reports used for internal due diligence and smaller commercial transactions. For projects above 1 MWp, independent energy assessors (DNV, TÜV, WSP) are typically required by institutional lenders. The line is moving downward as solar software quality improves, but for any transaction involving bank debt on a large commercial system, factor in the cost of an independent technical advisor.

Frequently Asked Questions

What is a good specific yield for a European solar system?

Specific yield (kWh/kWp/year) varies by location and system quality. Benchmarks: Oslo 700–850, London/Berlin 850–1,000, Paris/Amsterdam 950–1,100, Madrid/Rome 1,300–1,600, Lisbon 1,500–1,750 kWh/kWp. Systems significantly below these ranges warrant investigation of shading, soiling, or underperforming inverters. Use our generation and financial tool to model specific yield for any European location.

Should I present P50 or P90 to customers?

For residential customers: present P50 as the expected output and explain it is the median estimate — half of years will produce more, half will produce less. For commercial customers financing with a bank: always include both P50 and P90. Presenting only P50 to a lender is technically incorrect and could create liability if the lender uses it as a conservative base case.

Why does my PVsyst result differ from the PVGIS estimate?

PVGIS provides a simplified yield estimate without detailed loss modeling. PVsyst with proper loss assumptions typically produces 5–12% lower yield than PVGIS default output. This is expected — PVGIS is a quick reference tool, not a design-grade simulation. Always use a full simulation for project documentation. A 10% discrepancy between PVGIS and a careful PVsyst simulation is normal; a 20% discrepancy suggests an input error in PVsyst.

About the Contributors

Author

Rainer Neumann

Content Head · SurgePV

Rainer Neumann is Content Head at SurgePV and a solar PV engineer with 10+ years of experience designing commercial and utility-scale systems across Europe and MENA. He has delivered 500+ installations, tested 15+ solar design software platforms firsthand, and specialises in shading analysis, string sizing, and international electrical code compliance.

View all posts