Announcing our latest forecast accuracy validation study: Delivering trustworthy solar data at scale

As part of our ongoing commitment to provide accurate, high-resolution solar forecasts to the energy industry, we are pleased to share the results of our latest forecast accuracy validation study. Our team is constantly working to improve and refining our modelling approach and forecasting accuracy, as demonstrated by our leading results in the independent forecasting trial run by EPRI.

Competitive and comparative forecast studies like that are hugely valuable to the industry - but expensive, lengthy and challenging to run at large or global scale. Whilst the DNV Bankability study into Solcast historical data accuracy was large scale, and global - the scope did not cover forecast data. So through the course of 2024 our team have been using Solcast forecast archives and hindcast data (retrospective forecast modelling) to validate and measure our forecast accuracy using all the available high quality measurement data we can get our hands on. 

This study measures Solcast forecast performance across 135 sites globally, each with 1-4 years of site data. This work allows us to validate the performance of our models with precision, delivering the metrics the industry needs to make informed, data-backed decisions. We’re proud of these results, they demonstrate the challenges of constant global forecasting and the hard work of our team to improve our forecast quality. 

Forecasting solar power is a unique challenge, one that requires constant improvement and iteration. While every model update brings new levels of accuracy, it also sets a new standard for future performance. This ongoing process of validation, refinement, and comparison helps ensure that our models are not only reliable but are continuously evolving to meet the growing needs of the energy sector.

Large-Scale Forecast Accuracy Validation to Support Operational Decisions

We consistently hear from customers that they are trying to understand or quantify the relative accuracy of forecasts for their use case, their assets, or their internal requirements. Partly, this is because forecast trials are lengthy, arduous, resource intensive processes. But also, it’s because of a lack of large scale validations published by forecast vendors. So we’re publishing our own analysis of our forecast data, compared to GFS weather data, and a smart persistence model. We’re publishing that data across multiple parameters, for a global data set, and making it dynamic and interactive with the goal of supporting those looking for accurate, high resolution forecasts made for solar applications.

This large-scale validation effort provides detailed accuracy metrics for solar forecasts, enabling operators to reduce uncertainty in their day-to-day operations. By evaluating performance across 135 sites with 1-4 years of data each, we have built a robust foundation of insights that solar operators can rely on to improve generation management, load forecasting, and grid integration.

Accurate forecasting plays a critical role in developing grids that deliver reliable power and harness the weather as fuel. Operators must balance the need for renewable energy generation with the variability inherent to solar power. We are sharing this extensive validation to provide those grid operators, and asset operators, the data they need to make more informed decisions, allowing for smoother integration of renewable energy into the grid.

Comparative Data to Quantify Forecast Accuracy Gains

To truly understand the value of our high-resolution forecasts, we compared our performance against widely-available models such as the Global Forecast System (GFS) and a smart persistence model, which assumes no change in weather conditions. These benchmarks provide clear, quantifiable evidence of how much more accurate a forecast model that is built for solar can be, highlighting the significant potential for efficiency and optimisation that can be made with forecast models designed for solar energy applications.

By offering this comparative data, we aim to give operators a clear understanding of the value that high-resolution, solar-specific forecasting can bring. It’s not just about improving accuracy—it's about demonstrating the measurable gains in forecast skill that directly translate into more efficient and reliable renewable energy management.

Global and Climate-Zone Forecast Accuracy Validation for Precise Regional Forecasting

In addition to large-scale accuracy validation, our study also evaluates forecast performance at global, regional, and climate-zone levels. This allows us to tailor our insights to the specific needs of operators in different regions and climates. Whether you're working in tropical, desert, or temperate climates, our localized validation helps ensure that you have the data you need to optimize solar generation and plan for grid integration with greater confidence.

Our data sources include both public datasets and the DNV Bankability Study, such as BSRN, SURFRAD, BoM, DWD, KNMI, and more. Each of these sources undergoes strict quality control to ensure recency and relevance, which is vital when dealing with solar forecasting where both historical data and real-time forecasts are essential.

Continuous Forecast Enhancements

As we continue to refine and expand our validation processes, this study represents an important milestone in our journey towards even greater forecast accuracy. The detailed metrics provided in this study not only validate our current models but also set the stage for further improvements and innovations in the future.

The energy industry needs forecasts that are not only reliable but also adaptable to the evolving challenges of grid integration. By continuing to validate our models at scale and across diverse regions, we are proud to contribute to that goal.

We encourage our users to stay tuned for ongoing updates, as this validation process is continuous, and we will be publishing new data as we incorporate additional sources and improvements.