Statistical filtering-based least square method for short-circuit capacity identification of traction power supply system

• An identification method of short-circuit capacity considering randomness is proposed. • Two statistical rules are proposed to reduce identification errors caused by samples. • A modification based on the probability distribution is presented for identification. • Simulations and field test shows great accuracy of the proposed SFLS algorithm. The short-circuit capacity (SCC) of electrified railways extends beyond power supply and voltage stability, encompassing traffic organization, relay protection, and fault diagnosis. However, due to the dual fluctuations of source and loads, it is difficult to establish Thevenin equivalent model (TEM) and calculate the SCC at feeders of traction power supply system (TPSS). This paper proposes a statistical filtering-based least square method to identify SCC. Two statistical filtering rules are described to improve the consistence between voltage and power variations. The data that satisfy two rules are obtained by power threshold and box plots. It can reduce the error of TEM caused by source-side fluctuations and low voltage-power variation consistency. Preliminary results are obtained by the least square regression and evaluated by the fitting quality. The result with poor fitting quality is modified by probability distribution of historical results. Considering the loose thresholds, Kalman filter is introduced to reduce the error caused by obvious outliers. Finally, the proposed algorithm is validated by simulations and field test evaluated by Jensen-Shannon divergence. The SCC is roughly same as the actual value with an improvement of 15.76% compared to conventional methods. Besides, the sensitivity analysis shows the great generalization performance of the proposed methods.