特高压三端混合直流输电系统作为直流输电的一种重要形式，存在传输距离较长而导致的线路故障率较高的问题，对其进行准确的故障定位是系统稳定运行的基础。针对现有故障定位方法应用于输电线路单极接地故障时存在的高阻接地故障下定位模糊、精度较低的问题，提出了一种基于小波包熵特征融合提取故障特征，再由改进麻雀搜索算法（improved sparrow search algorithm, ISSA）优化的双向时域卷积网络（bidirectional time-domain convolution network, BiTCN）模型的故障定位方法。首先，利用小波包变换提取线模电压行波信号，利用信息熵刻画电压波形中的深层故障特征，形成熵特征融合特征向量构成的特征矩阵作为BiTCN模型的输入；其次，搭建并训练BiTCN模型，并利用ISSA的迭代寻优对其进行优化，最终实现三端混合直流输电线路故障的精确定位；最后，在PSCAD/EMTDC仿真平台中搭建系统模型，验证所提方法的可实施性。结果表明该方法定位精度较高，具有较好的泛化能力和鲁棒性，对高阻故障耐受能力较好。

The issue of a high line failure rate due to long transmission distances is faced by the UHV three-terminal hybrid DC transmission system, which is an important form of DC transmission. The basis of stable system operation is formed by accurate fault location. A fault location method was proposed that addressed the challenge where existing fault location methods were applied to the unipolar grounding fault of the transmission line, leading to issues of fuzzy location and low accuracy under high resistance grounding fault conditions. The method proposed using wavelet packet entropy feature fusion to extract fault features and employed a BiTCN model optimized by ISSA. Firstly, the line-mode voltage traveling wave signal was extracted using wavelet packet transform, and the deep fault characteristics in the voltage waveform were described using information entropy. The feature matrix, composed of the entropy feature fusion feature vector, was formed as the input of the BiTCN model. Secondly, the BiTCN model was constructed and trained, and it was optimized using iterative ISSA optimization. Finally, accurate fault location of the three-terminal hybrid DC transmission line was achieved. The feasibility of the proposed method is verified by constructing the system model in the PSCAD/EMTDC simulation platform. The results indicate that the method is characterized by high positioning accuracy, good generalization ability, and robustness, and high resistance faults are effectively tolerated.