Introducing SunSolve P90: A Free Solar Yield Forecast Uncertainty API

We are excited to share SunSolve P90, our contribution to the growing effort around better uncertainty quantification in Energy Yield Assessments (EYAs). This free Python API for calculating yield uncertainty builds on years of research and community discussions about improving confidence in our forecasts.

Building on a Decade of Physics-Based Simulation

SunSolve P90 draws on more than 10 years of work with the solar industry developing physics-based simulation tools. Our foundation in SunSolve Power for solar cell and module design, and SunSolve Yield for utility-scale energy modeling, has given us the experience to bring a pragmatic and diligent approach to uncertainty analysis.

This new tool wouldn’t exist without extensive consultation with the industry. We ran a roundtable discussion at PVPMC 2024, surveyed practitioners at ACP Resource and Tech, and have been learning from colleagues working on similar challenges. Their insights about what uncertainty quantification needs to accomplish in practice have been fundamental to shaping this work.

The Challenge For Yield Forecasters

An Energy Yield Assessment (EYA) is one of the most important documents in a solar project, yet uncertainty quantification remains challenging for our industry.

Banks and other investors don’t just want to hear “this project will generate 100 GWh per year.” They need to understand the uncertainty around that estimate. What confidence intervals can you provide? What’s the P90 to P50 ratio? Without proper quantification, yield forecasters can’t demonstrate the value of improvements or trade-offs in project design.

Currently, most EYAs apply a relatively simple approach to calculating P90. This approach omits the asymmetries and dependencies of the sources of uncertainty, and even omits some sources of uncertainty altogether.

A New Foundation for Advancing Confidence

With support from ARENA, we have developed SunSolve P90, which we believe is the next critical step toward improving investor confidence in EYAs: a tool that makes it practical to account for the complexity of uncertainty in solar forecasting.

Our approach tackles three fundamental challenges that current methods struggle with:

1. Accounting for asymmetrical sources of uncertainty - Simple Gaussian distributions fail to represent the true shape of input distributions such as availability, curtailment and diffuse irradiance fraction.
2. Combining uncertainties correctly – accounting for their codependencies rather than treating them as independent.
3. Capturing a wide range of sources of uncertainty – from weather variability and degradation to availability and operational factors.

Monte Carlo: Bringing a Proven Approach to Yield Uncertainty

Monte Carlo simulation has been used to quantify uncertainty since the 1960s in finance, and more recently by some practitioners in solar energy. The technique is both powerful and elegantly simple, which is why we’ve chosen it as the foundation for SunSolve P90.

Instead of generating a single forecast and then trying to analytically calculate its uncertainty, the Monte Carlo approach generates thousands of forecasts — each representing a plausible combination of weather patterns, degradation rates, availability scenarios, and other variables. The result is a distribution that naturally captures the complex interactions between different uncertainty sources.

Until now, the PV industry has shied away from applying the Monte Carlo approach to EYAs because the time to conduct and combine thousands of EYAs was prohibitive. Companies were also unsure what inputs to apply for many uncertainties. SunSolve P90 can handle this complexity within minutes, making it practical for routine use in yield assessments.

Get Started Today

We believe rigorous uncertainty analysis should be accessible to everyone in the solar industry — which is why we’ve made this tool free to use.

Ready to try it? Follow our simple tutorial in the documentation and you’ll be up and running in less than 30 minutes. The tutorial walks you through running our online sample, loading your first project data, and generating uncertainty distributions.

Share Your Experience

We’re eager to hear how SunSolve P90 works for you in practice. Whether you’re using it on real projects or exploring it for the first time, your feedback helps us improve:

• What worked well and what could be better in your workflow;
• Contributing to a shared library of default uncertainty distributions for factors like availability, degradation, soiling, and more;
• How the approach performs across different project types and regions.

For researchers, tool developers, and independent engineers working on uncertainty quantification, we’d love to explore how SunSolve P90 might support or integrate with your work. We’re active participants in the PVPMC Uncertainty Working Group and committed to advancing industry-wide approaches to this challenge.

Meet us at ACP Peak

We’re presenting SunSolve P90 at ACP Peak 2025 this week and would love to hear about your experiences with uncertainty quantification. Whether you’re already working on similar tools, have tried the tutorial, or are interested in collaborative approaches, we’re eager to connect and learn from each other.

Come find us at the conference to discuss your projects and how Monte Carlo uncertainty analysis might fit into your workflow.

The path to better forecasts starts with honest uncertainty quantification. We are excited to share this tool with the community and to work together on building the trust that our industry needs to thrive.

Acknowledgments

Thanks to the PVPMC for supporting the original panel discussion that helped guide this project, and to everyone who participated in our industry survey at ACP Resource and Tech 2024. We are grateful to be part of the PVPMC Uncertainty Working Group, where ongoing collaboration continues to advance best practices for uncertainty quantification across the industry.

We are also grateful for the encouragement and feedback from colleagues across industry and academia — too many to name individually — whose insights have been invaluable in shaping this work.

This work was supported by funding from the Australian Renewable Energy Agency (ARENA). The views expressed herein are not necessarily the views of the Australian Government, and the Australian Government does not accept responsibility for any information or advice contained herein.