Validations

Summary

• First IE evaluation of PlantPredict’s site-level 3D scene capability — covering terrain-aware backtracking (TABT), the proprietary 3D transposition model, and 3D shade-scene modeling with electrical shading response.
• ICF rated PlantPredict’s 3D scene approach as “reflecting higher-fidelity approaches than is typical in solar industry-accepted PV performance modeling.”
• All 18 major algorithm categories were rated as “meeting expectations” and consistent with industry-accepted PV performance modeling — including solar position, air mass, clear-sky, GHI decomposition, transposition (incl. 3D), rear-side irradiance (bifacial), tracking and backtracking, irradiance optimization, wind stow, far / near / 3D shading, IAM, spectral correction, module cell temperature, single-diode I-V characterization, DC and AC wiring, inverter efficiency, and transformer efficiency.
• ICF performed an independent energy-yield benchmarking study comparing PlantPredict V12 against PVsyst v8.0.15 across 15 sample U.S. projects spanning Hawaii to Maine, 1.1 MW to 235 MW DC, cSi / HJT-cSi / CdTe module technologies, and both fixed-tilt and single-axis-tracker (SAT) configurations.
• The PlantPredict-versus-PVsyst comparison showed a mean energy-yield difference of just −0.16 %, median of 0.01 %, standard deviation of 0.73 %, and 95 % confidence interval of −0.57 % to +0.24 % — statistically indistinguishable from zero, confirming no systematic portfolio-level bias between the two tools.
• The −0.16 % mean delta is comparable in magnitude to the 0.13 % delta reported in the 2015 Passow study, extending PlantPredict’s validation lineage through Logic Version 12.
• PlantPredict’s modern stateless REST API and Python SDK were endorsed as well-suited for automation, large-scale scenario analysis, and integration into custom analytical workflows.
• The platform’s shared libraries, role-based governance, status management, and documentation tools were validated as supporting consistent, collaborative, and well-governed energy modeling across teams.

Summary

• Reviewed PlantPredict’s implementations of irradiance transposition, near and far shading, module thermal, single-diode, and inverter performance models — concluding that the underlying algorithms are appropriately selected and substantiated.
• Benchmarked PlantPredict’s outputs against PVsyst (the de facto industry-standard tool at the time) and against operational performance data from First Solar’s utility-scale portfolio.
• Confirmed PlantPredict’s alignment with PVsyst for comparable input assumptions and project configurations.
• Concluded that PlantPredict is suitable for bankable, utility-scale PV energy assessments.

Summary

• Reviewed PlantPredict’s user interface, platform hierarchy, model algorithms, reporting and integration capabilities, and revision management practices, finding the platform’s design “straightforward and well documented” and its hierarchy “logical.”
• Independently benchmarked PlantPredict against PVsyst across four geographically diverse theoretical projects spanning CSi and CdTe module technologies and both fixed-tilt and single-axis tracker configurations.
• Confirmed that PlantPredict’s algorithms are appropriately implemented and substantiated for utility-scale PV generation estimates.
• Concluded that PlantPredict produces PV generation estimates with uncertainty comparable to PVsyst and other bankable modeling platforms.
• Endorsed PlantPredict’s resource-center documentation of critical algorithms and source citations as supporting transparency for users and reviewers.

Summary

• Evaluated PlantPredict’s algorithms, validation methodology, and modeling outputs against the requirements of bankable utility-scale PV project finance.
• Reviewed the platform’s suitability for both engineering decision-making and lender / investor due-diligence processes.
• Confirmed PlantPredict’s suitability as a bankable energy assessment tool for utility-scale solar projects.

Summary

• Performed a comprehensive review of PlantPredict’s modeling approach, input/output structure, and underlying algorithms.
• Benchmarked PlantPredict against PVsyst, the de facto industry-standard PV simulation tool at the time.
• Concluded that PlantPredict provides modeling accuracy equivalent to other energy-prediction modeling tools currently used in the industry.
• Verified that PlantPredict’s predictions match operational performance data within the uncertainty expected of bankable simulation tools.

Summary

• PlantPredict was selected as one of seven industry-leading commercial PV modeling tools evaluated, alongside 3E SynaptiQ, PVsyst, RatedPower, SAM, SolarFarmer, and Solargis Evaluate.
• Predictions were submitted directly by the PlantPredict team, ensuring the platform was exercised exactly as intended by its developers.
• PlantPredict demonstrated sub-hourly modeling capability across both lab- and utility-scale benchmarks.
• Across all seven tools, annual energy-yield deviations from the mean were within ±2.5 % at the lab scale and ±6.0 % at the utility scale — placing PlantPredict squarely within the range of the industry’s most established modeling software.
• The paper publishes four side-by-side feature-comparison tables (POA transposition, module temperature, shading, and performance models) intended as a resource to help users select the most suitable software for their application.

Summary

• PlantPredict was one of eight bifacial PV modeling tools included in the validation, spanning commercial, freeware, and open-source categories.
• The study used consistent input parameters and on-site meteorological data (spectrally flat Class A sensors) across all tools to enable a true apples-to-apples comparison.
• State-of-the-art bifacial performance models — including PlantPredict’s — were shown to add only ~0.5 % uncertainty to the overall PV modeling chain.
• For fixed-tilt configurations, 2-D view-factor simulations (PlantPredict’s bifacial approach) agreed with measured monthly bifacial gain to within ±1 %.
• For horizontal single-axis tracker (HSAT) configurations, 2-D view-factor simulations agreed with measured bifacial gain to within approximately ±2 %.
• PlantPredict completed full-year annual simulations in under 10 seconds, comparable to the fastest commercial tools in the study.

Summary

• Strong agreement with PVsyst: across 51 equivalent simulations, the mean energy-yield difference between PlantPredict and PVsyst was just 0.13 % ± 0.52 % (1-σ).
• Strong agreement with real operating plants: actual-versus-expected analyses across 20 utility-scale systems totaling nearly 1 GW_DC of CdTe modules showed PlantPredict under-predicted energy by only 0.41 % ± 2.01 % (1-σ).
• PlantPredict’s Hayes advanced thermal model reduced hourly module temperature RMSE by 43 % compared to standard thermal modeling.
• The paper established the empirical foundation for PlantPredict’s accuracy claims, including documented benchmarks of transposition models (Perez vs. Hay–Davies) and hourly vs. monthly spectral correction.