What this category does
Renewables forecasting turns weather predictions into expected power output for solar and wind plants. Operators use it to schedule maintenance and to commit energy to the market. Utilities and grid operators use it to balance supply and demand, and asset managers use it to judge whether a plant is performing as expected.
The product is more than a weather forecast. It includes a plant model that links irradiance or wind speed to output, and it often learns from the plant's own meter data. Two providers can use the same weather input and still deliver different production forecasts because of how they model the plant.
How buyers evaluate
Buyers typically run a bake-off on one or a few plants. The provider forecasts output for a few weeks or months, and the buyer compares that against metered generation. Because the plant model needs history to learn from, ask how much historical data the provider needs and how quickly the forecast improves as it receives yours.
Compare errors at the horizons that match your decisions. A day-ahead forecast used for market bids has different needs than a forecast for the next few hours used for balancing. Pay attention to the hours when errors cost the most, such as fast ramps at sunrise and sunset, cloud events and wind shifts, rather than only to the daily average.
RFP criteria commonly include forecast horizon, update frequency, the metric used to score accuracy, how curtailment and outages are handled, integration with your scheduling systems, and contract length. Trials of one to three months are common, and a longer one covers more seasons.
Ask how the provider treats a new plant with no history. Some start with a generic plant model and improve it as data arrives, and some require months of metered output before they deliver a forecast. The answer affects whether a trial on a plant that has just been commissioned is fair to either side.
Questions to ask a vendor
- Which weather models feed the forecast, and do you blend more than one?
- How does the plant model treat curtailment, outages and maintenance periods?
- What historical data do you need from my plant, and what if I have little?
- What error metric do you use, and will you score the trial on the one I choose?
- How often do you update the forecast, and how do I receive it?
- Do you forecast probabilities or ranges as well as a single value?
- What happens to my data if I leave?
How to read the capability facts
On a profile, check the parameters for solar irradiance or wind, the forecast types for the horizons you need, and the regions for coverage of your sites. The industries field shows whether a vendor works with plant operators, utilities or asset managers. The delivery field matters because a forecast has to reach your trading or scheduling system in a format that system accepts.
The facts describe the offer and not the forecast accuracy at your site. Ask any vendor you shortlist to run a trial on your plant and to agree on the error metric before it starts.
Several vendors in this field also offer services beyond the forecast, such as monitoring or bidding support. Those can be useful, but evaluate the forecast on its own terms first, since it feeds every decision that follows.
Where to go next
Agree on the plant, the horizon and the error metric before you start a trial, then use the list below to find providers.
Browse the Renewables Forecasting providers, filter them by region, parameter and pricing, or compare several side by side. The full list of categories is the place to start if your need spans more than one.