6-2_A-Model-for-Efficient-Shading-Evaluation-in-Large-Scale-1

07 May 2023 A Model for Eficient Shading Evaluation in Large-Scale PV Plants Based in Hemicube Geometry Anja Neubert, Javier Lopez-Lorente, Mike Hamer, Tony Mercer DNV © 07 MAY 2023 PV, DNV assurance fast and acurate! •But not so easy for complex terain or for arbitrary near shading obstacles •Ray tracing •For beam shading, plot a line from each point on module to every solar position and lop over al the obstacles to evaluate whether or not that beam path is obscured •Great for acuracy but an intensive computation / computational eficiencies likely to be complex •… and, for difuse, ned to acumulate shading through weighted colections of paths over al of the sky 3 Is there an accurate enoughmethod for complex terain and arbitrary shading obstacles that’s faster, or simpler? DNV © 07 MAY 2023 Hemicube finite element aproach •Established method in computer graphics for solving radiosity; rendering scenes with difusely reflected light https:/en.wikipedia.org/wiki/Radiosity_(computer_graphics) •3D surface polygons amalgamated into patches, then compute the light emited and colected across al combinations of patch pairs. An eficient solution entails regularised patches and view factor computations htps:/ww.doc.ic.ac.uk/~dfg/graphics/graphics2010/GraphicsHandout07.pdf •SolarFarmer re-purposes just the first step; the so-caled hemicube step… 4 DNV © 07 MAY 2023 Hemicubes for beam shading Pre-compute step •A virtual unit hemicube placed over each point on the module •Grid each of the 5 faces, M * N pixels •Evaluate beam shading for each pixel considering al shading obstacles; black = shaded •Amenable to optimised graphics CPU/GPU instructions •Store al these hemicube surfaces, for a fixed aray they won’t change 5 Run-time •For each solar position, just lokup the apropriate pixel in each hemicube; easy because the hemicube is regular •Fast. Especialy useful when you want to consider a lot of solar angles / smal time steps DNV © 07 MAY 2023 Hemicube silhoueting with multiple tilt angles •To re-enforce how the hemicube works… •… visualise how the shading from the row in front spreads up the hemicube side as the tilt angle increases •Al shading obstacles handled in a single method; row-to-row, near obstacles 1 per module through 9 per cel •For a fixed tilt layout on simple terain, so we can easily compare (the back row) with infinite sheds •Al methods similar at times of day without shade •1 per module assumption under estimates (row-to-row) shading towards each end of the day –underestimates shade •5/submodule, 1/cel, 9/cel: 0.01% yield diference •Compute time relative to the 1/module case for 5, 20 compute time »proportional to N hemicubes e.g. *4 increase from 5 per submodule to 1 per cell 8 SolarFarmer defaults to 5 hemicubes per submodule DNV © 07 MAY 2023 Curent status •Core method proven striving for continuous validation •11 SolarFarmer methodology what tok hours now takes minutes •Private Preview … while we kep working on the cluster scaling and queuing 9 DNV © 07 MAY 2023 Future Work •More validation … continuously •Colaborations? •Comparisons with other 3D methods? •More compute optimisations… explore more accuracy / compute time trade-offs •Bifacial •3D front side but currently using view factors for the backside -needs hemicubes on the ground and on the backside •Multi-angle hemicube interpolation for trackers 10 DNV © 07 MAY 2023 DNV’s Solar Energy Assessment Validation using SolarFarmer Madison Ghiz, here, today @ 10:5! DNV solar experts share the latest Solar Energy Validation Methodology, Webinar, 18th May 1pm ET htps:/ww.dnv.com/power-renewables/events 11 Uncertainty Considerations in Bifacial Photovoltaic Systems with High Albedo Seasonality, Javier Lorente Lopez, PVSC Light and Shade –A Hemicube Aproach for Eficient Shading Calculations in Utility-Scale PV Plants, Anja Neubert EU-PVSEC See more at … DNV © 07 MAY 2023 www.dnv.com Thank you 12 tony.mercer@dnv.com renewables.suport@dnv.com DNV © 07 MAY 202313