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