计及风电不确定性与尾部风险的孤岛微电网可靠性快速评估OA
Fast reliability assessment of islanded microgrids considering wind power uncertainty and tail risk
针对传统可靠性评估难以兼顾不确定性刻画、尾部风险度量与计算效率的问题,提出尾部风险可靠性快速评估框架.首先,基于实测数据构造小时级容量因子序列,结合"月-时"划分与共形校准,建立有限样本风电条件椭球不确定集,刻画风电波动范围.其次,以直流最优潮流最小切负荷模型为核心,引入条件风险价值(conditional value-at-risk,CVaR)构造尾部风险指标,并提出FAST-CVaR近似策略减少最优潮流(optimal power flow,OPF)调用.进而设计卷积神经网络(convolutional neural network,CNN)分类-回归代理模型,识别样本级尾部风险状态并预测强度,聚合估计可靠性指标.最后,在改进的RTS-GMLC区域1孤岛微电网算例中进行验证.结果表明,FAST-CVaR将 OPF 调用次数由 60 000 次减少至 400 次;CNN 代理模型对两类可靠性尾部风险指标的估计相对误差分别为0.21%和0.51%,样本评估时间由743.504 s降至19.523 s.所提方法能够在保持尾部风险指标精度的同时显著降低计算开销,可为孤岛微电网尾部风险快速评估提供参考.
To address the difficulty of balancing uncertainty representation,tail-risk evaluation,and computational efficiency in conventional reliability assessment,this paper develops a fast tail-risk reliability assessment framework.First,hourly capacity-factor series are constructed from measured data.By combining month-hour partitioning and conformal calibration,a conditional ellipsoidal uncertainty set for wind power under limited samples is established to characterize wind power fluctuations.Second,based on a DC optimal power flow(DC-OPF)minimum load shedding model,conditional value-at-risk(CVaR)is introduced to construct tail-risk indicators,and a FAST-CVaR approximation strategy is proposed to reduce the number of OPF evaluations.Furthermore,a convolutional neural network(CNN)-based classification-regression surrogate model is developed to identify sample-level tail-risk states and predict risk severity,enabling efficient estimation of reliability indicators through sample aggregation.Finally,the proposed method is validated on a modified RTS-GMLC Area 1 islanded microgrid.The results show that FAST-CVaR reduces the number of OPF evaluations from 60 000 to 400.The relative errors of the CNN surrogate model in estimating the two types of reliability tail-risk indicators are 0.21%and 0.51%,respectively,and the sample evaluation time is reduced from 743.504 s to 19.523 s.The proposed method can significantly reduce computational cost while maintaining the accuracy of tail-risk indicators,providing an effective approach for fast tail-risk reliability assessment of islanded microgrids.
岳大为;李冰新;柴园园;刘浩然;李益波
智能配用电装备与系统全国重点实验室(河北工业大学),天津 300401智能配用电装备与系统全国重点实验室(河北工业大学),天津 300401智能配用电装备与系统全国重点实验室(河北工业大学),天津 300401智能配用电装备与系统全国重点实验室(河北工业大学),天津 300401智能配用电装备与系统全国重点实验室(河北工业大学),天津 300401
风电不确定性孤岛微电网尾部风险可靠性评估共形校准条件风险价值分类-回归代理模型
wind power uncertaintyislanded microgridtail-risk reliability assessmentconformal calibrationconditional value-at-riskclassification-regression surrogate model
《电力系统保护与控制》 2026 (20)
74-84,11
This work is supported by the National Natural Science Foundation of China(No.52207089). 国家自然科学基金项目资助(52207089)
评论