Data-driven topology identification and line parameter estimation for distribution network: A two-stage approach

Accurate power network estimation is critical for the efficient operation and management of distribution networks, especially in the context of integrating renewable energy sources and maintaining grid stability. This paper presents a novel two-stage approach for topology identification and line parameter estimation in distribution networks. The proposed method integrates variational mode decomposition into feature-enhancing techniques to extract meaningful features from noisy, volatile nodal data, effectively addressing the challenges posed by measurement disturbance. Additionally, a graph convolutional network based model is introduced to accurately capture both local and global dependencies within the network topology, enhancing the scalability and robustness of the estimation process. Experimental results demonstrate a significant improvement in accuracy, achieving a 62.3% reduction in identification errors compared to mainstream methods. The proposed framework effectively handles networks under different scales, offering a robust and scalable solution for distribution network analysis and real-time operational applications.