@Călin Sas and @David Moser: this is the only real evidence I have seen showing direct impact of EPC quality on O&M: https://solargrade.io/articles/2025-solargrade-pv-health-report/

@Călin Sas I did not notice that David Penalva is here too! 😀

@Fredy Canizares thank you for sharing this! It sounds like a lot of DC/AC AC/DC conversions and wiring, so I wonder what are the main reasons to opt for such design as opposed to DC-coupled? Is it lower complexity and/or higher reliability? For sure it is lower yield and maybe higher costs? I assume curtailed energy can be stored even in an AC-coupled system? Thanks again!

@Fredy Canizares: These are brilliant insights, thank you very much for taking the time to explain!

To stop recycling 10+ year old contracts and move on to performance guarantees (assuming stakeholders are involved from the beginning of the project).

@Stefan Wippich , we also saw deviations of up to 5%, see Figure 3 here: onlinelibrary.wiley.com/doi/full/10.1002/pip.3615 These modules were purchased from the open market so they might not be representative of what is installed in utility scale. Large module procurement contracts typically include provisions for adjusting the rating following tests by a 3rd party lab.

@Joshua Hamsa, I really enjoyed this information, thank you. I posted about it 2 days ago on linkedin: https://www.linkedin.com/feed/update/urn:li:activity:7391636790425743360/ We need to come up with a standardized reporting approach. The IEC 61724 talks about predicted and expected energy. The problem is that the expected energy is supposed to be calculated by re-running PVsyst with actual weather data. We all know that PVsyst was not designed for operations. Because of that, you are forced to run a "proxy" model. After discussing with ~20 PVMAC stakeholders (software providers, owners, IPPs, O&M providers), we grouped their approaches in 3 categories, and this is where things can get confusing. We are working on a process and these "definitions" are not finalized but here they are to give you an overview: 1) Independent Proxy (IndProxy): This is a model (another software, or any pvlib model that you do not necessarily fit on data) that you simulate independent of PVsyst. You use the as-built information, derate assumptions, actual weather, etc and you create a model that has nothing to do with PVsyst: the information or most inputs could be the same, but it is not fitted on PVsyst. 2) Calibrated Proxy (CalProxy): This is a model (e.g., PVUSA, or PVWatts) that you fit on your PVsyst time-series to extract some coefficients. Then you re-run with actual weather to achieve weather adjustment. 3) FieldProxy: This is an independent model (e.g., based on ML) that is fitted on field time-series, and has no information about PVsyst and captures actual performance. Each approach has its pros and cons. For example, with a FieldProxy, you will not be able to diagnose degradation, because you fit a model on field data that already include degradation. Or with CalProxy, you might not do well with nonlinearities. On top of these, we need to be clear about what is included in the expected yield model. Where is the model bias (e.g., model error or hourly to subhourly bias), do we account for issues that are not due to plant underperformance (e.g., curtailment, grid unavailable etc)? This is easier to explain with a picture, so if you are interested, please check this presentation's first 3-5 slides:

@Stephen Lynch I totally agree with what you say. I would expect that performance engineers would not rely on PVsyst though, otherwise it is an apples-to-oranges comparison. This is where proxy models (or recalibrated models as you say) come in. Both models (PVsyst and recalibrated) are useful, but for different purposes and there are different metrics depending on purpose: 1) Financial purposes: use baseline energy performance index (BEPI) which is the ratio of predicted energy (e.g., PVsyst) to measured energy. 2) Performance purposes: use EPI which is the ratio of expected energy (e.g., whatever the recalibrated model is, please see my post above that breaks those in 3 categories) to measured energy. We have standards (e.g., IEC 61724), but they are subject to interpretation so this is why we need to work together to develop best practices and minimize confusion and calculation uncertainties.