PV Mid- and Long-Term Power Prediction Method Based on SDCN-TiDE and Power Reconstruction

Mid- and long-term power prediction is critical for power planning, market optimization, and grid stability. However, existing short-term methods fail to capture long-term meteorological patterns, leading to reduced accuracy. This study addresses this limitation by using irradiance data to develop a robust power model and photovoltaic (PV) utilization curve for periodic feature extraction. A mid- and long-term PV power prediction model based on Structured Deep Clustering Network (SDCN) and Time-series Dense Encoder (TiDE) is proposed, incorporating power reconstruction to improve accuracy. Experimental results demonstrate its superior performance compared to traditional approaches traditional methods, validating its effectiveness.