Solar Generation Forecasting
The primary driver of solar generation is cloud cover. When large bands of clouds pass over major population centres, the output of all of the rooftop solar PV systems can suddenly fall or rise at the same time — leading to fluctuations equivalent to turning a large power plant on and off.
This can make it very difficult to manage energy flows within the grid, and it can have a significant impact on energy prices.
New data, new techniques
Fortunately, we now have access to new types of satellite imagery that provide highly detailed information on the locations of clouds. There are also new techniques arising out of the artificial intelligence motion prediction community that are incredibly effective at forecasting movement in imagery.
Our proprietary forecasting solution applies these latest AI techniques to satellite imagery — enabling us to forecast the formation and movement of clouds at a high resolution. Such cloud forecasts can improve the accuracy of solar generation forecasts by 20% over conventional solar forecasting approaches.
Highly localized forecasts
Cloud formation and movement is a highly localised process that can be heavily affected by conditions specific to a given site – for example how close it is to the coast, or to mountains, or to whatever other micro-climatic effects it may be subject to. In other words, cloud motion forecasts become more accurate if they are targeted at and trained on data for a specific site.
We train models specific to your use case, whether it is a solar farm, a region of your network, or a larger market region — taking these local effects into account.