In this document
We will discuss how solar software provides expected uncertainties in solar irradiance and yield simulations, thereby delivering more reliable results and ensuring the effective design, operation, and financial success of solar energy projects.
Overview
Addressing uncertainties through robust modeling and validation with real-world data is essential for enhancing the reliability of PV output calculations and enabling informed decision-making in solar energy projects.
Among the primary sources of uncertainty is the solar irradiance data used as a critical input for simulations. Irradiance data uncertainties stem from inherent limitations in models, including the incomplete representation of atmospheric conditions, geographic variability, and other environmental factors. By combining regional and site-specific analyses, it is possible to achieve a more comprehensive understanding of satellite model uncertainty.
Estimating the uncertainty of satellite-based irradiance data requires extensive validation using ground measurements and careful analysis of contributing factors that lead to deviations. While ground-based sensors are highly accurate, they are not without challenges. Issues such as calibration errors, sensor degradation, and environmental interference (e.g., shading, soiling, or adverse weather conditions) can introduce additional uncertainty in solar resource assessments.
In PV output simulations, further uncertainties arise from the assumptions and simplifications embedded in modeling tools, such as those related to PV system components and their performance under varying conditions.
Uncertainty of satellite-based irradiance
Semi-empirical irradiance models are subject to a range of uncertainties coming from a variety of factors, ranging from each site’s climatic, geographic, and environmental conditions to the limitations of current satellite technology covering the area and capturing the inputs for the model.
Some of the key factors that influence solar irradiance model uncertainty:
Cloud persistence and variability can significantly affect solar irradiance estimates. Especially with rapidly changing cloud formations, the irradiance estimates are less accurate. High albedo surfaces, such as snow-covered or desert areas, reflect more sunlight and may cause overestimations or underestimations of irradiance in satellite models.
Terrain variability, such as mountains and valleys, affects how solar radiation is distributed across a landscape, which is difficult to capture with coarse satellite resolution. The distance to water bodies introduces uncertainties due to localized effects like fog or maritime clouds.
Anthropogenic pollution, including smog and industrial emissions, adds variability to solar irradiance measurements by increasing aerosol content in the atmosphere. Extreme weather events, such as wildfires, hurricanes, or volcanic eruptions, drastically alter atmospheric conditions, often leading to spikes in uncertainty.
The spatial resolution of satellite data is critical for accurately modeling solar irradiance. Lower resolution can lead to errors in capturing small-scale variability. Pixel distortion, caused by sensor limitations on capturing Earth's curved surface, can also introduce inaccuracies, especially at the edges of images.
Estimation of uncertainty for Solargis irradiance model
High-quality satellite models are characterized by consistency across both space and time, making validation at multiple sites within a specific geography a reliable indicator of model accuracy for comparable regions. By validating model outcomes against reference values for a sufficient number of samples and thoroughly analyzing all factors influencing the results, it becomes possible to estimate expected uncertainty. In essence, if validation sites within a particular geography show consistent bias and root mean square deviation (RMSD) within a certain range, it can be reasonably assumed that the model will perform similarly in regions with comparable geography, even where no validation sites are available.
Solar model uncertainty is typically expressed as a percentage variation around the mean irradiance value (e.g., ±5%). Two types of model uncertainty analysis are commonly conducted:
This approach provides a broad approximation of model performance by categorizing regions into lower and higher uncertainty ranges based on validation results and environmental factors. For annual Solargis GHI values, the expected P90 uncertainty for areas outside validation sites ranges between ±4% and ±8%.
Lower uncertainty regions (~±4%): These include areas with stable atmospheric conditions and reliable ground measurements, such as most of Europe, North America below 50°N, South Africa, Chile, Brazil, Australia, Japan, Morocco, the Mediterranean, the Arabian Peninsula (excluding the Gulf region), and other regions with good availability of high-quality data.
Higher uncertainty regions (~±8%): These encompass latitudes higher than 50°N and 50°S, high mountain regions with snow and ice coverage, reflective deserts, urbanized or industrial areas, regions with high and variable aerosols (e.g., India, West Africa, the Gulf region, and parts of China), coastal zones (within ~15 km from water), and humid tropical climates (e.g., equatorial regions in Africa, America, the Pacific, and Southeast Asia). Limited or absent high-quality ground measurements further increase uncertainty in these areas.
Analyzing site-specific model performance is a complex process requiring in-depth expertise in the model’s internal algorithms, inputs, and environmental dependencies. To estimate site-specific uncertainty accurately, factors such as climate, geography, environmental conditions, and satellite technology must all be considered. This detailed approach provides a refined uncertainty estimate tailored to the unique characteristics of the site in question.
In the case of Solargis irradiance uncertainty values, the uncertainty of the validation instruments used to calculate uncertainties is already included in the given value of model’s uncertainty.
Uncertainty of measurements
ISO 9060:2018 classifies pyranometers into three categories:
Class A (Secondary Standard)
Class B (First Class)
Class C (Second Class).
For solar power plant development, Class A pyranometers are recommended due to their superior accuracy and reliability. Similarly, Class A pyrheliometers are preferred for precise DNI (Direct Normal Irradiance) measurements.
In contrast, Rotating Shadowband Radiometers (RSRs) are usually classified as Class B or C devices. While they are less suitable for high-precision applications, RSRs can be a cost-effective alternative for preliminary solar resource assessments.
Using state-of-the-art instruments alone does not guarantee low uncertainty. Measurements are inherently subject to error, and accurate results require accompanying information on associated uncertainty, as well as robust quality control and correction techniques applied to raw data.
Estimating the long-term uncertainty of ground measurements involves a combination of theoretical instrument uncertainty, the results of quality control procedures, and redundant measurement comparisons. For high-accuracy instruments, such as Class A pyranometers and pyrheliometers, uncertainties for annual solar measurements are typically around ±2.0% for GHI and ±1.0% for DNI when best practices and rigorous quality control are applied.
Pyranometers | RSR | |||
|---|---|---|---|---|
| Class A (Secondary standard) | Class B (First class) | Class C (Second class) | (After data post-processing) |
GHI Hourly | ±3% | ±8% | ±20% | ±3.5% to ±4.5% |
GHI Daily | ±2% | ±5% | ±10% | ±2.5% to ±3.5% |
Theoretically-achievable daily uncertainty of GHI at 95% confidence level
Pyrheliometers | RSR | ||
|---|---|---|---|
| Class A (Secondary standard) | Class B (First class) | (After data post-processing) |
DNI Hourly | ±0.7% | ±1.5% | ±3.5% to ±4.5% |
DNI Daily | ±0.5% | ±1.0% | ±2.5% to ±3.5% |
Theoretically-achievable daily uncertainty of DNI at 95% confidence level
The lowest achievable uncertainty in solar measurements is critical for accurately determining solar resources. High-quality data is essential not only for validating solar resource assessments but also for adapting satellite-based models to local conditions. Poor-quality or unvalidated data can introduce significant errors, undermining confidence in energy production estimates and project feasibility studies.
PV output simulation uncertainty
Simulating the energy output of photovoltaic (PV) systems involves various uncertainties that can affect the accuracy of predictions. These uncertainties stem from the inherent complexity of the system, the variability of its components, and the limitations of simulation tools. Accurately estimating these uncertainties requires comprehensive data, sensitivity analysis, and a thorough understanding of modeling limitations, making the process technically demanding.
The uncertainties in PV output simulations can be broadly categorized into three main groups:
Model assumptions: Simulation tools often rely on simplifying assumptions to model system components such as PV module performance, inverter efficiency, and shading effects. These assumptions can introduce inaccuracies, as they may not perfectly align with real-world conditions.
System component variability: Each component in the PV system has its own uncertainty, such as the efficiency of inverters or module degradation rates. These variations can compound, impacting overall system performance predictions.
User inputs: Values provided or adjusted by users come with inherent uncertainty. Inaccurate or overly generalized inputs can significantly affect simulation outcomes.
Uncertainty models used in PV simulations vary in complexity, ranging from simple deterministic approximations to advanced probabilistic frameworks. While modern simulation software makes energy yield predictions more accessible, uncertainties in these simulations are often underexplored.