Introduction
SolarAnywhere Probabilistic Forecast provides P1-P99 percentile forecasts that help quantify uncertainty in expected solar production. These forecasts support risk-aware operational and market decisions by providing a range of plausible outcomes rather than a single expected value.
Probabilistic forecasts are available for GHI, POAI, AC Power and AC Energy outputs.
Applications
Typical applications include:
- Energy market bidding strategies
- Reserve planning
- Ramp-risk awareness
- Operational scheduling
- Maintenance planning
- Solar-plus-storage operations
Methodology
SolarAnywhere Probabilistic irradiance forecasts were created using a probabilistic mixture-of-experts model architecture. The model was trained by ingesting features from several numerical weather predictors (NWPs) and minimized the negative log-likelihood loss between predicted irradiance and ground truth irradiance measurements. Feature importance across more than 60 inputs was also evaluated to identify which had the most influence on prediction accuracy.
For site-level probabilistic power forecasts, these irradiance values serve as inputs to a physics-based power model and can be tuned to historical generation. Aggregate or regional probabilistic power forecasts for a portfolio of assets are customized models trained on both historical power generation and weather forecast data.
For more information on the validation of the probabilistic forecast product, see the validation whitepaper. If you would like access to hindcasts of the probabilistic forecast for validation at your locations, these are available back to 2021.
Technical Specifications
The Probabilistic Forecast supports intra-day and day-ahead irradiance and power forecasts for a single PV asset as well as power forecast aggregates for a collection of assets covering a given region. The table below summarizes these forecast specifications and more API details can be found in the Postman Collection.
| Parameter | Value |
|---|---|
| Output Fields |
Global Horizonal Irradiance (GHI) Plane of Array Irradiance (POAI) AC Power AC Energy |
| Forecast Horizon | Up to 40 hours ahead. |
| Spatial Resolution | ~3 km |
| Temporal Resolution |
15 minutes 60 minutes |
| Geographic Coverage | Contiguous United States (CONUS) |
| Percentiles | 1 – 99 |
| Data Delivery | API only. Custom data delivery available upon request. |
How to Access
Access to the SolarAnywhere® probabilistic forecast is available as a trial and through the Advanced Forecast license. Probabilistic forecasts are accessible via the SolarAnywhere API.
The SolarAnywhere API works asynchronously, meaning POST requests are submitted, a unique ID associated with the POST request is returned, and that ID can be used to retrieve the results. The ID returned by the POST request is valid for 14 days; after that period, you must submit a new POST request. It’s important to note that a valid SolarAnywhere API key is required to access data via the API. For instructions on creating API keys, please visit the SolarAnywhere API page of our support center.
Detailed instructions on this data retrieval process are available within the API documentation.
If you need assistance getting started with probabilistic forecasts, please contact us at support@solaranywhere.com.