Solar forecasting can help solve a number of problems associated with solar energy harvesting:
High variability of the solar resource, due to movement of
the sun through the sky
clouds over the sun
Efficient and cost effective operation requires reliability and predictability.
Sharp changes can cause local voltage flicker.
Underestimating output can lead to balancing and frequency issues.
A number of solar forecasting methods currently exist. Their relevance varies
with the size of the photovoltaic application and the desired forecast lead time.
Sky-imager forecasting, or cloud forecasting,
is one of these methods. Images of the sky are captured from the ground at
regular intervals and forecasts are made in real-time from the series of images.
Forecasting method
Pros
Cons
Physically-based
- User-defined inputs and components - Sharing made easy (invariant to geolocation and training)
- Suffers from insufficient data (inputs) - Coarse resolution O(10km)
Satellite imaging
- More direct measure of irradiance vs NWP - Assimilates NWP
- Coarse resolution O(km) - Requires visible light (IR would help with morning forecast accuracy)
Sky imaging
- High spatial and temporal resolution O(min)
- Does not account for cloud development and dissipation - Limited to the SI FOV
Stochastic learning
- Can assimilate all discussed methods
- Extraneous or redundant inputs will give inaccurate results (ANNs and GAs can select inputs).