Research on Correlation Aggregation Method of Multiple Wind Farms Based on FA-LSTM

The aggregation method of wind power time series data is mainly separate aggregation. For this reason, this paper proposes a multi-wind farm time series data aggregation method considering correlation. First, based on factor analysis, the standardized wind power raw data is decomposed into horizontal components and fluctuating components; second, the horizontal components are clustered by Affinity Propagation (AP) to obtain 2n typical daily scenes are obtained through stratified sampling. Then the 2n corresponding daily scenes are selected from the standardized wind power original sequence, the training set and prediction set are taken. The fitted wind power sequence is obtained through the LSTM algorithm. Finally, the fitted sequence is de-normalized to obtain the aggregate sequence of n days. Probability and statistical indicators and simulation results show that the aggregate sequence generated by this method can accurately reflect the correlation between the wind power, which has certain engineering application value.