Technical due diligence is the filter that separates solid investments from projects carrying hidden risk. Yet even reports prepared by well-known consultancies contain systematic errors that can affect billion-euro decisions.
Based on our experience reviewing more than 40 utility-scale projects, these are the five mistakes we encounter most often.
1. Solar resource based on a single data source
The most widespread error is building the simulation on a single meteorological data source — usually PVGIS or a commercial satellite product — with no cross-validation.
The risk: individual sources carry systematic biases that can compound to 3–5% in the solar resource estimate. On a 100 MW project, that translates into a difference of millions of euros across the project’s lifetime cash flow.
Correct practice: combine at least two independent sources (ERA5 + satellite) and verify against nearby ground station data where available.
2. Linear degradation models
Most reports assume linear module degradation of 0.5–0.7% per year. That model is wrong for two reasons:
- Degradation is steeper in the first years of operation
- Modules of different technologies (monoSi, biSi, CdTe) follow different degradation curves
A linear degradation model can overestimate production in years 15–25 of the project by as much as 5–8% — precisely the period that matters for equity payback.
3. Soiling losses underestimated or generic
The soiling coefficients typically applied (1–2%) are generic values that ignore:
- The site’s rainfall regime
- Local pollution and particle type
- The planned operational cleaning frequency
At arid sites in southern Spain or North Africa, soiling losses without washing can exceed 8–12%.
4. Shading analysis without modeling real obstacles
The shading calculation is frequently the weakest part of the report. Typical errors include:
- Using the theoretical horizon instead of the measured one
- Failing to model surrounding vegetation and nearby buildings
- Ignoring row-to-row shading in tracker configurations
5. P90 calculated as P50 × a conservative factor
The most serious conceptual error: some reports calculate P90 by simply applying an 8–10% discount to P50. This is methodologically incorrect.
A real P90 requires a Monte Carlo analysis that propagates every source of uncertainty simultaneously: solar resource uncertainty, model uncertainty, loss uncertainty, and so on. Without that analysis, the reported P90 has no statistical backing and banks may reject it.
Every one of these mistakes comes down to the same thing: a model whose assumptions nobody can trace or reproduce. Heliosolve runs native Monte Carlo P50/P90 in the browser and exports a report your lender can audit — validated against NREL and Sandia.