Detection and location of cable aging in underground distribution network with complex topologies

Abstract Precise detection and localization of aged cable segment in distribution networks are essential for ensuring the stable operation of power systems. However, the nonlinearity introduced by the complex topology of distribution networks significantly complicates the monitoring and localization of aged cable segments. To address these issues, this paper presents a novel approach integrating broadband transfer function (TF) measurement with a topology-feature-fused neural network (TFFNN). Specifically, an adjacency matrix transformation is employed to derive the transfer function spectrum of a complex topological cable network. The obtained transfer function spectrum, along with topological information, is then used to train the TFF-NN. The trained model is subsequently leveraged to estimate both the location and severity of aged cable segments. Experimental results under various topological configurations and load conditions demonstrate that incorporating topological information enhances the accuracy of the proposed method in complex network scenarios. Compared to existing approaches, the proposed method achieves superior performance in localization accuracy and severity assessment. In complex topologies, the relative errors in evaluating the defect start position, end position, and severity level are 5.51%, 4.2%, and 3.17%, respectively.