What is solar power forecasting?
Solar power forecasting predicts how much electricity a specific asset will generate in a given period, enabling asset operators to submit optimal day-ahead and real-time bids and manage solar curtailment risk.
Unlike grid-level net demand forecasts, solar power forecasting software models site-specific conditions such as panel configuration and hyper-local weather. AI-powered solar asset forecasting continuously adapts to changing weather conditions and rapid output swings.
Solar asset short-term forecast
(Solar STF)
(Solar STF)
Short-term solar generation forecasts for solar assets and portfolios across day-ahead and multi-day horizons. Models are trained on each site’s historical generation, physical configuration, historical solar curtailments, and hyper-local weather data to support bidding, scheduling, and portfolio reporting.
4-6%
Target cnMAE
Solar asset sub-hourly short-term forecast
(Solar Sub-hourly STF)
(Solar Sub-hourly STF)
Intra-day solar generation forecasts with sub-hourly resolution for day-ahead bidding, real-time dispatch, and solar curtailment management. Continuously updated using real-time irradiance data to capture rapid output swings caused by changing cloud cover.
5
Minute resolution available
Who uses solar forecasting software?
Solar asset operators at IPPs, utilities, and certain retailers and commercial and industrial operations.
Independent Power Producers
IPPs operating solar farms, including hybrid battery-tied assets.
Utilities and Gentailers
Utilities and energy retailers who also own solar assets.
Commercial & Industrial Customers
Commercial & industrial facilities and data centers with market-exposed solar assets.
4-6%
Target solar forecast cnMAE
200+
Solar assets forecasted
3 Weeks
Average time from contract to solar forecast deployment
1
Data inputs
We identify and collect relevant data streams like coordinates, historical generation, and physical asset attributes.
Hybrid modeling
A blend of physics-based and machine learning models produces our forecasts and updates them hourly as new data arrives.
2
3
Flexible outputs for any workflow
The forecasts integrate seamlessly with Amperon customers’ preferred analytical tools.
Actionable insights
Asset operators make better-informed decisions that improve margins and achieve strategic goals.
4
Frequently asked questions
Here are some answers about our platform, implementation process and pricing.
How does Amperon forecast solar generation?
Amperon builds asset-specific solar generation forecast models that combine physics-based representations of each asset's performance characteristics with machine learning trained on historical generation and weather data. Models retrain hourly on the latest observational inputs, producing sub-hourly forecasts that reflect current atmospheric conditions rather than static assumptions. Probabilistic outputs quantify the range of likely generation outcomes across weather scenarios, giving operators visibility into forecast uncertainty at the time horizons that matter most for bidding and scheduling decisions.
How accurate are Amperon’s solar asset forecasts?
We target 4-6% cnMAE for solar asset forecasts. Actual results vary by site.
For more on the cnMAE accuracy metric, see cnMAE: The Right Metric for Evaluating Solar and Wind Forecasts.
For more on the cnMAE accuracy metric, see cnMAE: The Right Metric for Evaluating Solar and Wind Forecasts.
What is cnMAE and why is it the right metric for evaluating solar power forecasts?
Capacity-normalized Mean Absolute Error, or cnMAE, measures forecast error as a percentage of an asset's installed capacity rather than its actual generation output. This distinction matters because solar generation frequently operates at low output levels, particularly during morning and evening hours, and expressing error as a percentage of actual output during those periods can make forecast performance appear far worse than it functionally is.
cnMAE provides a stable, asset-scale denominator that allows meaningful accuracy comparisons across assets, seasons, and time horizons, making it the most reliable basis for evaluating and benchmarking renewable generation forecast quality. For more on the cnMAE accuracy metric, see cnMAE: The Right Metric for Evaluating Solar and Wind Forecasts.
cnMAE provides a stable, asset-scale denominator that allows meaningful accuracy comparisons across assets, seasons, and time horizons, making it the most reliable basis for evaluating and benchmarking renewable generation forecast quality. For more on the cnMAE accuracy metric, see cnMAE: The Right Metric for Evaluating Solar and Wind Forecasts.
Which weather models does Amperon use for solar generation forecasting?
Amperon’s wind forecasts use a dynamic blend of multiple leading weather models from vendors such as ECMWF, DTN, AG2, GFS, NWS (US only), and DWD (Europe only). Learn more about Amperon’s approach to weather ensembling here.
How often do Amperon's solar generation models update?
Amperon’s solar asset generation forecasts update hourly, which continuously takes into account changing weather and generation dynamics and helps to capture crucial ramping periods that 6-hour updates miss. Hourly updates refer to model retraining, meaning new sub-hourly output predictions are available every hour.
Can solar energy forecasts predict output during peak times?
Yes, Amperon’s solar forecasts provide predictions at hourly or sub-hourly granularity, up to every 5 minutes, which is more than enough to forecast output during morning and evening load ramps, peak pricing periods, and coincident peak events.
What data do you need to produce a solar generation forecast?
Amperon ingests each site’s coordinates, historical generation data, and asset metadata such as nameplate capacity, panel tilt, azimuth, tracking, and historical curtailments. Our team works collaboratively with each client to identify and collect relevant data streams for each solar farm.
How long does it take to deploy a solar forecast?
Amperon works collaboratively with asset operators to identify and collect relevant data streams. If all needed data is readily available, Amperon can stand up solar asset forecasts in a matter of days. Typically, it takes a few weeks to collect and integrate data streams and deploy and test a solar forecast.
How does probabilistic solar forecasting work?
Probabilistic solar forecasting takes the guesswork out of renewable energy bidding by quantifying not only the range of possible outcomes, but the associated likelihood of each scenario. This helps asset operators calibrate expectations against real-world probabilities and compare scenarios against each other. For example, a P50 forecast is the median–the actual value is equally likely to come in above or below it–whereas a P90 is a conservative estimate, with the actual expected to land at or below it 90% of the time, making it useful for downside planning. For more, see Quantifying Solar and Wind Uncertainty with Probabilistic Asset Forecasting.
