College of Information and Electrical Engineering, China Agricultural University, Beijing 100083, China

Aiming at the problem that the separate aggregation of wind power and photovoltaic (PV) time series data changes the correlation between wind power and PV series, the factor analysis-extreme learning machine (ELM) aggregation method is proposed. Firstly, the z-score standardized wind power-PV original daily scene set is decomposed into horizontal components and fluctuating components of daily scene sets. Secondly, affinity propagation (AP) clustering is carried out for the horizontal components of daily scene sets to obtain K types of scene cluster. 2n-day horizontal components of daily scene sets are obtained through stratified sampling. The corresponding 2n-day original daily scene sets is selected from the standardized wind power-PV original daily scene sets, and n-day training set and n-day testing set are obtained respectively. Then, the mapping relationship between the horizontal components of daily scene sets and the original daily scene sets is obtained through the ELM, and the fitted power daily scene set is output. Finally, through the reverse standardization, the n-day wind power and PV power aggregation series are obtained, respectively. The accuracy and feasibility of the proposed method are verified by probability statistical indices, correlation coefficients and simulation results.