ABSTRACT
Irradiance forecasts are to be converted to photovoltaic power forecasts in accord with the system specifications, such as location, panel orientation, choice of panels and inverters, as well as the meteorological conditions that can affect the power output, such as incidence angle, spectral mismatch, or soiling. Irradiance-to-power conversion procedure, which is also known as the solar power curve, can be data-driven, physical, or a combination of both. Whereas regression forms the basis of data-driven irradiance-to-power conversion, the physical approach is known as the model chain, which leverages a series of energy meteorology models in cascade, as to arrive at the power output in a step-by-step fashion, considering as much physics as possible. Finally, hybrid irradiance-to-power conversion uses model chain for the first half of the conversion procedure, and leaves the remain half to statistics and machine learning; it is particularly useful when the system specification is incomplete or unknow. This chapter also elucidates the concept of a probabilistic model chain, which represents the latest and possibly the most preferred approach for irradiance-to-power conversion moving forward.
