Artificial intelligence (AI) techniques including the MLP, NN Delay, recurrent Elman NN, NN radial basic function, ANFIS, adaptive resonance theory (ART), and k-NN techniques.
Point and probabilistic forecasts based on multivariate models such as autoregressive (AR), vector AR (VAR), and vector ARX (VARX). Therefore, VARX is the most accurate model with NRMSE = 8.5%.
Random forests (RFs) that are the set of decision RTs. The analysis of the literature in the PV power forecasting showed the best results from RFs in terms of average forecasting for individual trees. In addition, the bagging technique that involved the increasing analysis to understand the complete trees, respectively, with a sample initiated from the entire training set. Nevertheless, the RFs deal with this problematic with a feature encapsulation that involves the choice of an unplanned subgroup of entities at each node. The feature encapsulation, meanwhile, reduced the error of correlation.
