New method estimates train traction power in real-time with 9% accuracy

A team led by Marton Laszlo Ambrus at the University of Birmingham has developed a lightweight estimation method to predict available traction power per train in multi-train AC railway networks, addressing a critical bottleneck caused by rising electrification. As decarbonisation pushes more trains to accelerate simultaneously on feeder lines originally designed for lower loads, contact-line voltage can drop dangerously low, triggering current limitation or protective trips that erode capacity and reliability. Traditional power-flow solvers are too slow for real-time use, so the researchers derive a distance-dependent single-train power envelope—showing that minimum network voltage depends on the product of power and distance, not power alone. They generalise this into a conservative multi-train estimator that uses a calibrated shared-path voltage model.

Calibration requires just two short offline solver runs: one to fix self-impedance and one to capture inter-train coupling via a separation-dependent factor. The model follows EN 50388-1 current-limitation rules and, when tested against full power-flow simulations for two, three, and four trains, averages within 9% accuracy—improving as more trains share the feeder. Crucially, the online computational cost scales linearly with the number of trains rather than the network size, making it practical for real-time deployment. The method gives operators a continuously updated estimate of traction power available to each train, enabling proactive capacity management and preventing voltage collapse without heavy solver infrastructure.