基于信息调控和MATCN的超短期风电功率多步预测OA北大核心CSTPCD

The effective forecasting of fluctuating wind power is an important guarantee for the balance of power supply and demand and the stable operation of the system.Therefore,a method of ultra-short-term wind power multi-step forecasting based on information regulation and multi-head attention temporal convolution network(MATCN)is proposed.High-order items and interactive items are derived from the existing data to increase the proportion of the number of feature sequences and effective features.The variational mode decomposition(VMD)is used to split the complex wind power data structure,the important sequence components are retained according to the calculation results of sub-sequence correlation and variance contribution rate,and other components are aggregated to reduce the calculation burden,and shorter training time.attention mechanism is introduced to construct the MATCN,and the transmitted information between convolution units within the network is adjusted by the attention score,so as to realize the prediction of each sequence component of the model.The sequence component prediction values are reconstructed to obtain the final output result.The proposed model is compared and verified on example data,and the results show that the model has excellent prediction effects under different strides.