Hybrid data-driven method for distribution network topology and line parameters joint estimation under small data sets

Abstract

Time-varying characteristics of distribution network makes the measurement data of intelligent instrument in the same state not sufficient, which cannot satisfy the required amount of data in existing topology and parameter estimation methods. This paper proposes a hybrid data-driven method for distribution network topology and line parameters joint estimation under small data sets. Based on the sparse characteristics of distribution network, a partial correlation analysis-based neighbor node selecting mechanism is designed, and the initial topology under small data sets is robustly identified with OR criterion. Combined with approximate linear power flow equation, initial values of line parameters and phase angle are estimated based on the linear regression. Furthermore, initial topology, line parameters, phase angle are optimized iteratively based on Newtonian method, with the decoupled linear power flow model introduced to simplify Jacobian matrix to accelerate iterative speed. Test results show that the proposed method only depends on small data sets to estimate the topology and parameters accurately and stably, adapting to different network scales and types. Compared with existing methods, it has faster calculation speed without loss of accuracy, and is robust to the noise and sampling resolution.