Yield Assessment

A yield assessment exists to answer one question: how much energy will actually reach the meter, and how confident can anyone be in that number. Everything else in the document is the working that supports it. Ours are built so that a lender’s engineer can check the arithmetic rather than take it on trust.

Modelling the array

Models are built in PVsyst, calibrated against local ground-station data where it exists rather than against satellite data alone.

• Bifacial and tracker geometry: rear-side irradiance modelled against the actual ground albedo of the site, with backtracking behaviour simulated rather than assumed from the tracker datasheet.
• Near-shading in three dimensions: the scene reconstructed from the real layout, so row-to-row and structure shading is calculated instead of applied as a flat loss factor.
• Horizon and terrain: digital elevation data used for the far horizon, which matters more on the hilly inland sites than on the coastal ones.

Losses at sub-hourly resolution

Hourly averaged meteorological data flattens the peaks. That flattening hides two things that cost real energy, and it usually hides them in the direction that flatters the project.

• Inverter clipping: irradiance spikes inside the hour are truncated by the inverter but invisible in an hourly average. On a high DC-to-AC ratio this is not a rounding error.
• Low-light behaviour: module efficiency at low irradiance modelled from the manufacturer’s curve rather than from the nameplate figure.
• Thermal time constants: cell temperature follows irradiance with a lag. Simulating that lag changes the loss estimate on clear days with moving cloud.

Degradation across the asset life

• PID and LID: quantified from factory and field data for the specific module, not carried across from a generic assumption.
• Non-linear loss: degradation is steeper in year one and flattens afterwards. A flat annual percentage applied across twenty-five years is wrong at both ends.
• Temperature and cycling coupling: how local heat cycles interact with the degradation curve, which is where a warm inland site and a coastal one stop being comparable.

P50, P90 and what a lender does with them

Non-recourse finance is not sized on the central case. It is sized on the downside case, which is why the derivation of that downside matters more than the headline number.

• P50: the median expectation. Useful for the equity case and almost nothing else.
• P90 and P99: the exceedance probabilities the debt is actually sized against, with the assumptions behind each stated rather than buried in the model.
• Inter-annual variance: calibrated from long satellite and reanalysis records for the site. A P90 is only as good as the variance estimate underneath it.

Where the numbers come from

Every input is attributed. Satellite and reanalysis irradiation, ground-station records where available, module and inverter data from the manufacturer, and the European Commission’s PVGIS dataset as an independent cross-check. Where an input could not be verified, it is labelled an assumption and listed separately, so the reader can see which conclusions depend on it.

What this is not

A yield assessment from us is not an independent engineer’s certificate. We do not issue certified opinions for lenders until professional indemnity cover is in place in the relevant jurisdiction. Where a financing requires one, the model and its documentation are built so that an independent engineer can review them rather than rebuild them from scratch.

Related work

Before a model goes to a lender or a grant panel, Pre-Submission Technical Review reads it the way their reviewer will. If the constraint is the connection rather than the resource, see Grid Connection Strategy and Project Development Support. For a connection point already held, the fixed-scope Grid Position Review answers it in two to three working days.