Loss tree template for utility-scale PV yield reports

Anders NakamuraRadiological Engineering · 23 yrs

The basis column is the right insight and it generalises well beyond PV — the objection is never to judgement, it is to judgement presented as measurement. One refinement worth adding to the four categories: split engineering judgement into judgement from comparable operating assets and judgement without a reference. Those are treated very differently in due diligence. A soiling assumption backed by two years of measured data from a plant 40 km away in similar conditions is close to a measurement in the reviewer's eyes; the same number with no reference is a discussion.

On ordering, I would move availability and curtailment out of the sequential chain and present them separately, because they are contractual and grid assumptions rather than energy-conversion losses. Keeping them in the multiplicative chain makes them look like physics, and reviewers then have to mentally extract them to test the contract. A chain that runs from irradiance through to AC energy at the meter, followed by a separate block for availability, curtailment and degradation, has consistently drawn fewer queries for me than a single flat list.

Linnea WeberFusion Engineering · 18 yrs

Verified expertise in Fusion Engineering

Using something close to this structure and the line that causes most argument is mismatch, because it is doing at least three different jobs in most reports — module-to-module electrical mismatch, string-level mismatch from unequal shading, and mismatch from non-uniform soiling or degradation. Reporting it as one number invites a reviewer to challenge the whole figure, and you cannot defend it because the components have different bases. Splitting it into electrical mismatch with a manufacturer basis and shading-related mismatch with a modelling basis removes that entire exchange.

The other line worth splitting is inverter power limitation, separating clipping from efficiency-curve effects. Clipping is a direct consequence of the DC to AC ratio, which is a design decision the lender's adviser will want to test against the yield outcome, and burying it inside a general inverter loss makes that test impossible without asking you for the breakdown. Presenting clipping on its own, with the DC to AC ratio stated next to it, turns a query round into a line item.

Leila NakamuraGrid Storage Engineering · 23 yrs

A format suggestion that made ours materially more useful: add an uncertainty column alongside the basis column, even where it is a rough band. It forces the same discipline for uncertainty as the basis column does for the value, and it means the P90 derivation traces back to the individual lines rather than appearing as a single combined figure at the end. Reviewers who want to test your P90 currently have to reconstruct that; giving it to them directly removes a round.

It also surfaces where the effort should go on the next project. When the uncertainty column is filled in honestly, it is usually obvious that two or three lines dominate the total uncertainty and the rest are noise. That tells you where a measurement campaign is worth paying for and where it is not, which is a much better argument to put to a developer than a general request for more site data.