基于差分群时延的物理引导网络同塔双回直流输电线路单端故障定位OA
Physics-guided network with differential group delay for single-ended fault location of tower-sharing double-circuit DC transmission lines
同塔双回直流输电线路导线间电磁耦合关系复杂,准确故障定位存在困难.针对该问题,提出一种基于差分群时延的物理引导网络单端故障定位方案.首先,利用相模变换提取零模与线模分量,并基于行波色散构造频域差分群时延特征,揭示了差分群时延与故障距离间的频率相关映射机制,有效抑制了过渡电阻的干扰.然后,利用边界元件特性,定义边界平滑化指数作为识别判据,实现了区内外故障判别.最后,构建物理引导网络,通过引入物理频率尺度通道、物理调制门控及物理约束学习,将行波传播规律融入网络结构,并结合两阶段测距策略进一步提升了故障测距的精度与稳定性.仿真结果表明,该方案在短数据窗口下定位准确,能够可靠耐受高过渡电阻并抑制强噪声干扰,具有良好的鲁棒性与物理可解释性.
The complex electromagnetic coupling between conductors of tower-sharing double-circuit HVDC transmission lines poses significant challenges to accurate fault location.To address this issue,a single-ended fault-location scheme based on differential group delay and a physics-guided network is proposed.First,phase-mode transformation is used to extract the zero-mode and line-mode components,and frequency-domain differential group delay features are constructed based on traveling-wave dispersion.The frequency-dependent mapping mechanism between differential group delay and fault distance is thereby revealed,effectively suppressing the interference of transition resistance.Then,based on the characteristics of boundary elements,a boundary smoothing index is defined as the identification criterion to discriminate between internal and external faults.Finally,a physics-guided network is developed by introducing physical frequency-scale channels,physics-based modulation gating,and physics-constrained learning,thereby embedding traveling-wave propagation characteristics into the network architecture.Combined with a two-stage fault-distance estimation strategy,the accuracy and stability of fault distance estimation are further improved.Simulation results show that the proposed scheme achieves accurate fault location under short data windows,exhibits reliable tolerance to high transition resistance and strong noise interference,and offers good robustness and physical interpretability.
吴笑天;陈仕龙;杨珠丹;程婧;王健;周世燚
昆明理工大学电力工程学院,云南 昆明 650500昆明理工大学电力工程学院,云南 昆明 650500昆明理工大学电力工程学院,云南 昆明 650500昆明理工大学电力工程学院,云南 昆明 650500昆明理工大学电力工程学院,云南 昆明 650500昆明理工大学电力工程学院,云南 昆明 650500
单端故障定位同塔双回色散特征群时延深度学习
single-ended fault locationtower-sharing double-circuit transmission linesdispersion characteristicsgroup delaydeep learning
《电力系统保护与控制》 2026 (17)
60-71,12
This work is supported by the National Natural Science Foundation of China(No.52067009). 国家自然科学基金项目资助(52067009)
评论