Ultra-short-term PV power prediction using optimal ELM and improved variational mode decomposition

The development of photovoltaic (PV) power forecast technology that is accurate utmost importance for ensuring the reliability and cost-effective functioning system. However, meteorological factors make solar energy have strong intermittent random fluctuation characteristics, which brings challenges to prediction. This work proposes, a new ultra-short-term PV prediction using an improved sparrow search algorithm (ISSA) optimize key parameters variational mode decomposition (VMD) extreme learning machine (ELM). ISSA’s global capability enhanced by levy flight logical chaotic mapping optimal number penalty factor VMD, VMD adaptively decomposes into sub-sequences with different center frequencies. Then ISSA used initial weight threshold ELM improve performance ELM, optimized predicts each subsequence reconstructs results component obtain final result. Furthermore, isolated forest (IF) Spearman correlation coefficient (SCC) are respectively in data preprocessing stage eliminate outliers original determine appropriate input features. actual plants show proposed model can effectively mine information historical more predictions, has good robustness various weather conditions.