A Novel Empiric Model-Based Classification Algorithm for Fault Detection in DC Railway Systems

During the early stages of Railway Systems, the implementation of DC technology facilitated the deployment of lightweight and efficient traction locomotives. At that time, DC systems offered considerable advantages over AC systems, allowing an easy control of the motor torque and avoiding the need for heavy step-down transformers and rectifier bridges. Additionally, substations connected in parallel along the line ensured a consistent contact line voltage, contributing to high supply reliability and balanced load for the AC power grid. However, safeguarding this system against short circuits presented challenges due to potential configuration changes and the short-circuit currents, whose magnitude is often comparable to the substation rated current. While the current protection system effectively clears faults, it might be subjected to wrong fault detections in the case of major transient events not linked to faults, causing unnecessary protection trips. The current protection system, being a very mature technology, has reached its performance limit nowadays. To enhance system performance, new protection strategies must be explored. This paper proposes a new advanced Fault Detection Algorithm (FDA) capable of overcoming the limitations of the protection system nowadays deployed in DC railways systems.