To address the issues of accurate short-term power prediction for distributed photovoltaic (PV) systems, this paper proposes a hybrid method integrating derivative dynamic time warping (DDTW) clustering and a composite attention prediction network (SK TCN-GC BiGRU). First, DDTW is applied to historical data for similarity day analysis, constructing targeted training sets. Second, the temporal convolutional network (TCN) and bidirectional gated recurrent unit (BiGRU) are optimized by combining a selective kernel network (SKNet) and a global context block (GC Block), enhancing their capabilities in extracting multi-scale features and global contextual information. Simulation results validate the superiority of the proposed model, demonstrating strong robustness under significant meteorological data fluctuations.
YANG D Z,WANG W T,GUEYMARD C A,et al.A review of solar forecasting,its dependence on atmospheric sciences and implications for grid integration: towards carbon neutrality [J].Renewable and Sustainable Energy Reviews,2022,161:112348
SCHINKE-NENDZA A,VON LOEPER F,OSINSKI P,et al.Probabilistic forecasting of photovoltaic power supply—A hybrid approach using D-vine copulas to model spatial dependencies[J].Applied Energy,2021,304: 117599.
SON H,KIM C,KANG Y,et al.Prediction of government-owned building energy consumption based on an RReliefF and support vector machine model[J].Journal of Civil Engineering and Management,2015,21(6):748-760
JEONG Y S,JEONG M K.OMITAOMU O A.Weighted dynamic time warping for time series classification[J].Pattern Recognition,2011,44(9):2231-224.
LI X,WANG W,HU X,et al.Selective kernel networks[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition,Long Beach,USA:IEEE,2019:510-519.