Data-driven impedance identification of AC catenary

Abstract

The accuracy of catenary impedance identification directly determines the reliability of traction power supply system relay protection and the precision of power supply capability assessment. Traditional impedance identification methods rely on static models and idealized assumptions, making them inadequate for complex scenarios under dynamic vehicle-network coupling conditions. This paper proposes a data-driven catenary impedance identification framework for both direct power supply systems and AT power supply systems. By integrating multi-source measurement data with intelligent algorithms, this approach aims to improve line relay protection setting verification methods and enhance system power supply capability evaluation. The study first analyzes various vehicle-network conditions requiring differentiation: direct-supply mode with only upbound trains, direct-supply mode with only downbound trains, direct-supply mode with both upbound and downbound trains, direct-supply mode with no trains, AT power supply mode with only upbound trains, AT power supply mode with only downbound trains, AT power supply mode with both upbound and downbound trains, and AT power supply mode with no trains. Subsequently, the proposed method applies measured voltage/current data from traction substations and sectional substations, combined with the classification process and total least squares method, to identify catenary impedance. Finally, the analysis of identification errors evaluates the rationality and applicability of the proposed method.