基于深度生成模型的区域能源系统频率安全能量与备用调度

Abstract With the increasing penetration of variable renewable energy sources such as wind and solar power, the frequency security management of regional integrated energy systems has been fundamentally altered due to reduced physical inertia and increased operational uncertainty. These changes introduce significant intra-day variability in system frequency dynamics, thereby challenging conventional frequency regulation approaches that rely on pre-scheduled reserve capacities. To address this challenge, this paper hypothesizes a decentralized frequency regulation scheme that integrates deep generative learning technology to maintain frequency security and achieve cost-effective operation under uncertainty in regional power systems. The proposed method coordinates multiple regulation resources and employs a modified generative adversarial network with a parallel classifier to assess the mismatch between dynamically evolving frequency regulation requirements and the pre-defined reserve capabilities. Simulation results on a modified IEEE 118-bus system demonstrate that the proposed method enhances frequency security by maintaining the system frequency within ±0.5 Hz across all regions and reducing the related frequency regulation cost by up to 22.4% in Zone C. These results indicate that the proposed frequency regulation framework achieves frequency security enhancement with lower operational costs in regional power systems with high renewable energy penetration.