How Forecasting Will Help Integrate Solar Energy

2687

A webinar hosted by the Solar Electric Power Association (SEPA) takes a close look at solar forecasting. Models that predict sunshine can help the grid manage fluctuating loads from distributed renewables, and can also guide long-term grid planning. 

Variability is frequently called a shortcoming of solar energy, but new forecasting techniques could change that. On Thursday, the Solar Electric Power Association (SEPA) hosted an online discussion of the topic with presentations by California ISO (CAISO) Forecasting Manager James Blatchford and Clean Power Research president Tom Hoff.

Accurately predicting PV solar output in an area helps with load scheduling as well as longer-term planning. Many data sets are needed, with perhaps the most important data being behind-the-meter PV system specs like orientation, tilt, and inverter type. Blatchford and Hoff both cited Power Clerk as a good data resource for California. 

Not surprisingly, the most accurate forecasts are those that are near real time, meaning less than 15 minutes out. This state-of-the-art analysis uses actual satellite images of local clouds combined with real time wind data to predict solar irradiance. Clean Power Research tested their forecasting model against actual data from ground-based sensors on a variably cloudy day and made an accurate prediction of irradiance (pictured top). Longer-term forecasts are made using National Weather Service predictions. 

One big benefit of all this is that it makes it easier for utilities to manage variable load from renewables, but there are long-term planning benefits as well. For example, the same models can be used to estimate load profiles for the year 2020 based on current market predictions (pictured below).

2020 Demand Forecast

 

The model predicts changes in peak system demand by 2020 depending on PV capacity. 

Currently the models perform with about a 5% margin of error compared to actual ground-based sensor data, which is quite good mathematically speaking. Clean Power Research and CAISO hope to improve the models by incorporating dynamic accuracy validation and probabilistic ramp rate forecasting, and through scaled-up testing. 

SEPA released a report on the topic of solar forecasting on Wednesday. Currently these forecasting efforts cover California, with different models for each of five climate regions that comprise the Golden State.