Introduction
A hindcast (historical forecast) is a time series of simulated forecasts for historical periods. Hindcasts simulate the site-specific solar forecast that would have been available at the time for operational decision-making.
PV plant developers can use hindcasts to evaluate how solar forecasts can improve dispatch algorithms and reduce hardware requirements to deliver firm power. Hindcasts also make it easy to estimate forecast accuracy for any location.
Applications
Solar forecasts may be used in plant operation to inform energy trading, scheduling and the control of energy storage. Considering the operational strategy of a solar PV plant during the pre-construction phase can reduce capital costs and project risk. The use of hindcasts makes this possible.
Specifically, hindcasts can be used to:
- Understand the value of forecasts in operational decision-making
- Assess site- and condition-specific forecast performance
- Optimize PV and storage plant design to deliver firm power
Methodology
With hindcasting, the forecast for each time step is generated using only the information that would have been available at that time, including operational delays. As with SolarAnywhere forecasts for time-horizons beyond 5 hours, SolarAnywhere hindcasts are created using a blend of Numerical Weather Prediction (NWP) models.
SolarAnywhere operational forecasts may be more accurate than the hindcast simulation. Operational forecasts may also leverage satellite images and cloud motion vector (CMV) processing to increase solar forecast accuracy for short horizons. Due to technical reasons, the CMV has not been incorporated into the hindcast at this time.
In addition, users should note that because solar forecasting technology is constantly improving, hindcast time series are not versioned.
Using SolarAnywhere hindcasts
Hindcasts are included with a SolarAnywhere Sites license and available upon request during forecast trials. Hindcast data is accessible through the SolarAnywhere API or upon request.
For deterministic forecasts, hindcast data is available for the last two calendar years prior to the present year and up to ~2 months prior to the present. The forecast horizon can be chosen from 1 hour up to 168 hours. Hindcast data includes the following datasets: GHI, DNI, DHI, Temperature, Wind Speed, DC Power, AC Power, AC Energy, and POAI. Temporal resolutions of 5, 10, 15, 30, and 60 minutes are supported with Hindcast data requests.
Hindcasts are also available for Probabilistic Forecasts and can support historical forecast validation efforts and historical uncertainty analysis for a single PV project or a collection of assets.
Getting hindcast data through the SolarAnywhere API
Hindcast data can be requested through the SolarAnywhere API. Visit the following pages of our API documentation for example hindcast irradiance, weather and PV production requests: