首页|期刊导航|电力系统保护与控制|基于控制响应解耦与多源信息自适应融合的直流输电故障原因辨识方法

基于控制响应解耦与多源信息自适应融合的直流输电故障原因辨识方法OA

A fault cause identification method for HVDC transmission systems based on control response decoupling and adaptive multi-source information fusion

中文摘要英文摘要

针对高压直流输电线路故障电气量受换流站控制系统非线性重塑影响,导致暂态特征呈现同质化且难以精确辨识的难题,提出了一种基于控制响应解耦与多源信息自适应融合的故障原因辨识方法.首先,分析了雷击、山火、风偏、异物诱因下直流线路故障暂态特征的演化差异,揭示了故障特征同质化的成因.其次,通过构建考虑触发角动态演变的解析模型,提出了计及控制响应的直流电压和电流故障固有特征量提取方法.进一步,构建了融合卷积神经网络(convolutional neural network,CNN)、双向长短期记忆(bidirectional long short-term memory,BiLSTM)网络与空间注意力机制(spatial attention mechanism,SAM)的辨识模型,用以挖掘故障固有电气量的多尺度空间细节与双向全局时序演化特征.在此基础上,引入基于气象环境先验的贝叶斯概率模型,并提出基于JS散度的多源信息动态融合策略,实现了数据驱动特征与外部物理约束的自适应权重协同.算例分析表明,所提方法能够针对各种故障原因实现高精度辨识,成功解决了特征同质化严重的辨识难题.

To address the difficulty of accurately identifying the causes of HVDC transmission line faults due to the homogenization of transient features caused by the nonlinear reshaping of electrical quantities by converter-station control systems,a fault cause identification method is proposed based on control response decoupling and adaptive multi-source information fusion.First,the differences in the evolution transient characteristics of DC line faults caused by lightning strikes,wildfires,wind-induced conductor deviations,and foreign objects are analyzed,and the underlying causes of fault characteristic homogenization are revealed.Second,an analytical model is constructed considering the dynamic evolution of the firing angle.An extraction method for inherent fault characteristic quantities of DC voltage and current while considering the control response is proposed.Furthermore,an identification model is constructed by fusing a convolutional neural network(CNN),a bidirectional long short-term memory(BiLSTM)network,and a spatial attention mechanism(SAM).Multi-scale spatial details and bidirectional global temporal evolution characteristics of the inherent fault electrical quantities are mined by the constructed model.On this basis,a Bayesian probability model is introduced based on meteorological environment priors,and a dynamic fusion strategy of multi-source information is proposed based on Jensen-Shannon divergence to achieve adaptive weight coordination between data-driven features and external physical constraints.Case studies demonstrate that the proposed method achieves high-precision identification for different fault causes,effectively addressing the challenge of fault cause identification under severe feature homogenization.

孙可桢;肖泰宇;于冰;欧阳金鑫;孙嘉琪;陈启迪;杨亚明

国网江苏省电力有限公司超高压分公司,江苏 南京 210000输变电装备技术全国重点实验室(重庆大学),重庆 400044国网新疆电力有限公司,新疆 乌鲁木齐 830018输变电装备技术全国重点实验室(重庆大学),重庆 400044国网新疆电力有限公司,新疆 乌鲁木齐 830018国网新疆电力有限公司,新疆 乌鲁木齐 830018乌鲁木齐职业大学,新疆 乌鲁木齐 830002

直流输电系统故障辨识换流站控制直流线路故障

HVDC transmission systemfault identificationconverter station controlDC line fault

《电力系统保护与控制》 2026 (19)

14-27,14

This work is supported by the National Natural Science Foundation of China(No.52577085). 国家自然科学基金项目资助(52577085)国网新疆电力有限公司科技项目资助(SGXJ0000TKJS2500296)

10.19783/j.cnki.pspc.260248

评论