In this document

We will explain how to verify the PV simulation models. It can be achieved by breaking down the models into individual components, enabling a more detailed and precise analysis of the various calculations involved.

Overview

The verification of PV simulation models is critical for ensuring the accuracy and reliability of energy yield predictions. This verification process, applied to key simulation steps, ensures that the software correctly represents real-world conditions, including complex scenarios like partial shading and variable irradiance.

The verification of Solargis PV simulation software is presented here for two core models: Ray tracing and electrical simulation under shading conditions.

Validation of ray tracing model

The Solargis ray tracing model has been validated against ground measurement data from the NREL – SRRL BMS station. This validation aimed to assess the integration of backward ray-tracing and transposition models using both isotropic and anisotropic sky models for simulating Global Tilted Irradiance (GTI). The analysis includes various mounting configurations and compares the Root Mean Square Difference (RMSD) and Bias between measured and modeled values.

The validation setup

Location: The validation was conducted using data from the NREL – SRRL BMS station located at latitude 39.742, longitude -105.178, and an altitude of 1828 meters.
Time Period: The data covers the entire year of 2021 with a 1-minute time resolution.
Devices Used: Different pyranometers were used for measuring Global Horizontal Irradiance (GHI), Direct Normal Irradiance (DNI), and Global Tilted Irradiance (GTI) for various orientations and tracking systems.
Quality Check: Only data passing a quality check were considered valid for analysis.

Results analysis

The results are summarized in the following table, comparing RMSD and Bias for isotropic and anisotropic models across different mounting configurations and data aggregation intervals.

	RMSD [%]				Bias [%]
Data aggregation	1-min data		15-min data		1-min data		15-min data
Sky model	ISO	ANISO	ISO	ANISO	ISO	ANISO	ISO	ANISO
2-axis tracker	6.7	6.5	6.0	5.9	-0.7	-1.1	-0.4	-0.7
1-axis N-S tracker	5.2	5.0	4.4	4.2	-0.5	1.0	-0.2	0.6
Fixed tilt 40° South oriented	5.1	5.0	4.0	3.9	-0.7	-1.0	-0.4	-0.7
Fixed tilt 90° South oriented	10.5	9.6	10.1	9.1	-6.9	-6.0	-6.8	-5.6
Fixed tilt 90° East oriented	16.0	15.8	16.1	15.8	-11.2	-10.6	-11.1	-10.5
Fixed tilt 90° West oriented	13.9	14.9	13.4	14.5	-7.9	-9.5	-7.6	-9.0
Fixed tilt 90° North oriented	23.6	20.8	24.1	21.5	-14.1	-10.5	-14.3	-10.7

RMSD and Bias Comparison

Sky Model Performance

Sources of Discrepancies

2-axis and 1-axis Trackers: Both isotropic and anisotropic models show relatively low RMSD values (around 5-7%) and small biases, indicating good performance for tracking systems.
Fixed Tilt Orientations: Higher RMSD values are observed for fixed tilt orientations, especially for 90° tilts, with significant biases. This suggests that the models struggle more with fixed orientations, particularly those not facing south.
Data Aggregation: The RMSD and Bias values generally decrease slightly with 15-minute aggregation compared to 1-minute data, indicating some smoothing effect over time

The anisotropic model tends to perform slightly better than the isotropic model in most cases, as it can capture more complex sky conditions. However, the differences between the two models are generally small, suggesting that both models provide comparable results for many configurations.

Model Limitations: The choice of sky model and ray-tracing accuracy can impact results. Anisotropic models generally perform better due to their ability to capture more detailed sky conditions.
Measurement Accuracy: The precision of GTI measurements, particularly for certain orientations like north-facing surfaces, can affect validation outcomes.
Albedo Representation: The use of a simple local albedo and Lambertian assumption might not fully capture complex ground reflectivity conditions.
Scene Complexity: The lack of simulation for complex 3D scenes and local shading could contribute to discrepancies.

Figure 1: GTI - ground measurements versus Solargis ray tracing model simulation (40° tilt, south oriented, 1-minute data).

Figure 2: GTI - ground measurements versus Solargis ray tracing model simulation (40° tilt, south oriented, 15-minute data).

Bifacial pv module validation

As part of the comprehensive validation of the Solargis ray tracing model, the performance for bifacial PV modules was also assessed. This involved comparing the ray tracing method with established tools and validating against ground measurements, using the same setup and data as the general validation process.

Comparison with bifacial_radiance

To validate the implementation of the ray tracing algorithm for bifacial PV systems, a comparison was made with the Bifacial_radiance toolkit, a widely recognized tool in the field. The comparison analyzed the front and rear side GTI distributions to ensure accurate implementation of the ray tracing algorithm (Fig.3).

Overlap and Differences: The results showed a good overlap between the Solargis ray tracing method and Bifacial_radiance, indicating largely accurate implementation. However, moderate differences were observed in the rear GTI during sunrise and sunset times. These discrepancies could be attributed to the absence of specular reflection in the Solargis implemented algorithm.

Figure 3: Comparison of Solargis sky model versus Bifacial_irradiance results (Clear sky, 2 rows of PV modules).

Evaluation Metrics

Simulation Settings

Root Mean Square Difference (RMSD): Measures the difference between simulated and measured values, indicating overall accuracy.
Bias: Assesses any systematic deviation between simulated and measured values, helping to identify potential model biases.

Simulations were conducted with no losses to directly compare with measured data.

Simulation Configurations:

PV Module Settings: Simulations were performed for non-shaded, fixed tilt (tilted at 40°), south-oriented PV modules with a constant ground albedo set to 0.2.
Sky Models: Both Perez isotropic and Perez All-weather sky models were utilized to evaluate their effects on GTI simulations.

Identified discrepancies

Discrepancies: Sources of discrepancies included limitations in sky models, ray tracing algorithms, and the accuracy of GTI measurements. Future improvements aim to address these discrepancies by refining sky models and incorporating additional reflection phenomena.

Validation of simulation under partial shading conditions

The Solargis ray tracing model has also been validated under partial shading conditions, which is crucial for accurately simulating real-world scenarios where shading can significantly impact PV system performance.

Simulation approach

The validation involved using the Monte Carlo backward path tracing method, which allows for detailed simulation of complex 3D scenes, including partial shading effects. This approach enables the accurate modeling of how shading affects PV module performance by tracing rays from the cell to the light source and accounting for interactions with surrounding objects.

Comparison with established tools

To ensure the accuracy of the Solargis simulation tool under partial shading conditions, comparisons were made with other established simulation tools such as LTSPICE. While specific results from these comparisons are not detailed here, the process involves verifying that the Solargis simulation aligns with industry standards and other recognized software packages (Fig.4).

Figure 4: Comparing LTSPICE simulation results with Solargis IV and Power curves calculation. (10 modules in string).

Evaluation Metrics

Simulation Settings

Root Mean Square Difference (RMSD): Measures the difference between simulated and measured values, indicating overall accuracy.
Bias: Assesses any systematic deviation between simulated and measured values, helping to identify potential model biases.

Simulations were conducted using detailed 3D models that account for the spatial arrangement of PV modules and surrounding objects. The simulations considered various shading scenarios, including inter-row shading in PV arrays

Results and identified discrepancies

Accuracy Assessment: The validation process highlighted the importance of accurately modeling shading effects to ensure reliable PV system performance predictions. The Solargis tool demonstrated its capability to simulate complex shading scenarios, though discrepancies may arise from simplifications in the 3D scene representation or assumptions about shading conditions.
Discrepancies: Sources of discrepancies included limitations in representing complex 3D scenes and the accuracy of input data regarding shading conditions. Future improvements aim to enhance the tool's ability to handle detailed shading scenarios by refining 3D modeling capabilities and incorporating more precise shading data.

Conclusion

The Solargis ray tracing model validation demonstrates good performance for tracking systems but highlights challenges with fixed tilt orientations, especially those not facing south. The anisotropic model shows slightly better results than the isotropic model, reflecting its ability to handle more complex sky conditions. Further improvements could focus on enhancing model accuracy for fixed orientations and incorporating more detailed albedo and scene complexity representations.

The validation of the Solargis ray tracing method for bifacial PV modules demonstrates good alignment with established tools like Bifacial_radiance, though discrepancies exist, particularly during sunrise and sunset. Ground measurements further validate the model but highlight areas for improvement, especially for rear side GTI simulations. Enhancements to sky models and the inclusion of specular reflections are key areas for future development.

The validation of the Solargis ray tracing model under partial shading conditions demonstrates its effectiveness in simulating real-world PV system performance. While the model shows promise, ongoing improvements are necessary to better capture complex shading effects and enhance overall accuracy. Enhancements to 3D modeling and shading data integration are key areas for future development.

Further reading

“A new simplified version of the Perez diffuse irradiance model for tilted surfaces” by Perez, R., Seals R., Ineichen P., Stewart R., Menicucci D., 1987, Solar energy vol.39, Issue 3.
“Comparison of Bifacial Solar Irradiance Model Predictions with Field Validation.” by Silvana Ayala Pelaez, Chris Deline, Sara M. MacAlpine, Bill Marion, Joshua S. Stein, Raymond K. Kostuk, 2019, IEEE Journal of Photovoltaics, vol.9, Issue 1.
“All-weather model for sky luminance distribution-preliminary configuration and validation” by Perez. R., R. Seals, J. Michalsky, 1993, Solar Energy vol. 50, Issue 3.
“Implementation of Ray Tracing Rendering Technique for Improved Solar Radiation Modeling of Bifacial PV Modules“ by Dvonc L, Orosi P., Cebecauer, T. Schnierer B., 2022, WCPEC-8.
“ Self-shading analysis in PV simulation: Comparison of different software packages“ by Schnierer B., 2020, PVPMC Webinar.
“Comparison of ray tracing rendering technique with ground measurements for improved solar radiation modeling“ by L. Dvonc, P. Orosi, T. Cebecauer, B. Schnierer, 2022, PVPMC Salt Lake city.