What Is a Solar Power Forecast?

A solar power forecast, or PV forecast, predicts how much electricity a photovoltaic plant will generate in the coming minutes, hours or days. Like wind power forecasts, it is usually provided in 15-minute steps, for single plants or whole portfolios.

How a solar power forecast is created

The key input is solar irradiance, above all the global horizontal irradiance (GHI) from weather models and, for the first few hours, from satellite and nowcasting data. From there, a PV forecast typically goes through three physical steps:

1. Decomposition of the global irradiance into direct and diffuse components.
2. Transposition onto the tilted module plane, using the azimuth and tilt of each module field (plane-of-array irradiance).
3. Conversion into AC power, taking module temperature, inverter limits and losses into account.

The biggest source of error is clouds: their position and thickness are hard to predict, especially on days with broken cloud cover. That's why, as with wind, physics alone isn't enough. Combining several weather models and training a model on each plant's own measurements captures what no datasheet contains, such as shading, soiling or the real behaviour of trackers. Curtailments must be kept separate, and intraday updates bring in the latest observations.

What solar power forecasts are used for

With rising PV capacity, solar generation dominates the midday hours in many markets. Direct marketers need PV forecasts for schedule nominations and quarter-hour trading, utilities for procurement and balancing group management, and grid operators for redispatch. Every percentage point of forecast error turns into balancing energy or short-term trading costs.

In a case study for a 15 MWp PV plant, our day-ahead forecast reached an nRMSE of 5.7 % over 13 days. More on how it works, including pricing and a free trial, is on our solar power forecast page.