Application of an adaptive Bayesian-based model for probabilistic and deterministic PV forecasting

Abstract

Accurate prediction of solar photovoltaic plant energy generation is essential for optimal planning and operation of modern power systems, and incorporating such plants into the energy sector. In this study, an adaptive Gaussian mixture method (AGM) and a developed variational Bayesian model (VBM) inference through multikernel regression (MkR) are utilized to assist desirable precise prediction. In this model, the MkR processes the multiresolution solar energy signal, and then the AGM models the complex signals forecasting error. Finally, the proposed model can be optimized, and the concurrent output of the solar energy signal in both probabilistic and deterministic status can be attained through the introduction of the VBM. The solar energy output of an actual plant, including four measurement sites provided the data for the study. The results confirmed that the proposed model delivers higher prediction accuracy for both probabilistic and deterministic forecasts when compared with other well-known models.