Gao, Yuan ， Zhao, Yucan ， Hu, Sile ， Tahir, Mustafa ， Yuan, Wang ， Yang, Jiaqiang
2025-03-15 ENERGY 2025 319(卷), null(期), (null页)
In pursuit of carbon neutrality, renewable energy exploitation in desert regions offers a compelling alternative to coal-fired power generation. However, managing such energy bases presents challenges, including limited wind and solar data, variable grid tariffs, and renewable energy output uncertainties. This study introduces a ThreeStage Adjustable Robust Optimization (TRARO) framework, integrating a Temporal Convolutional Network with Attention and Gated Recurrent Unit (TCNA-GRU) model for enhanced wind and solar prediction using transfer learning. The TRARO framework addresses uncertainties in energy management through three stages: optimizing capacity configuration, managing power exchanges, and scheduling operations. Simulation results demonstrate significant reductions in photovoltaic and wind turbine capacities by 50 % and 32.26 %, respectively, compared to Two-Stage Robust Optimization, alongside a 41.45 % decrease in grid transaction costs. These findings underscore the economic efficiency and reliability of the TRARO model in addressing uncertainties for large-scale energy bases, offering practical implications for sustainable energy planning.