电力系统自动化2026，Vol.50Issue(1)：74-85,12.DOI:10.7500/AEPS20250324002
面向可信度提升的数模混合驱动快速机组组合求解方法
Hybrid Data-model-driven Fast Unit Commitment Solution Method for Credibility Enhancement
摘要
Abstract
With the increasing complexity of the power source structure and grid topology in new power systems,the number of system nodes and units continues to rise.The solution for the security constrained unit commitment(SCUC)model using traditional optimization methods faces problems such as the curse of dimensionality and slow calculation speed.Although the data-driven decision methods can quickly solve the SCUC model,the lack of interpretability makes the decision results unusable.To address these issues,a hybrid data-model-driven fast unit commitment solution method for the credibility enhancement is proposed.First,an SCUC solution model based on deep reinforcement learning(DRL)is constructed to achieve fast pre-solution for the unit start-up/shut-down decision results.Then,by comprehensively considering DRL behavior-level interpretability indicators and strategy-level interpretability indicators,a credibility evaluation system for the start-up/shut-down decisions is constructed to identify high-credibility unit start-up/shut-down results,and enhance the interpretability of decision results.Finally,the hybrid data-model-driven SCUC is constructed to achieve fast model solution and optimize and adjust the low-credibility decision results.The simulation verification based on a 748-bus system of a provincial power grid shows that the proposed method achieves fast SCUC solution on the premise of enhancing the interpretability of unit start-up/shut-down decision results.关键词
机组组合/深度强化学习/数据驱动/可解释性/可信度Key words
unit commitment/deep reinforcement learning(DRL)/data-driven/interpretability/credibility引用本文复制引用
WANG Wenye,FENG Chuan,GUAN Yuxiang,MA Wenhao,CHE Liang..面向可信度提升的数模混合驱动快速机组组合求解方法[J].电力系统自动化,2026,50(1):74-85,12.基金项目
国家重点研发计划资助项目(2023YFB2407601) （2023YFB2407601）
江西省重点研发计划资助项目(20223BBE51013). This work is supported by National Key R&D Program of China(No.2023YFB2407601)and Key R&D Program of Jiangxi Province(No.20223BBE51013). （20223BBE51013）
