This paper presents an integrated digital approach for diagnosing the technical condition of traction power supply systems, including 27.5 kV traction substations and overhead contact networks. The proposed methodology combines state estimation models, residual-based monitoring using the deviation vector Δy(t), predictive analytics, and control algorithms into a unified closed-loop diagnostic system. Unlike conventional approaches that treat monitoring and control separately, the proposed framework integrates real-time deviation analysis, threshold-based decision logic, and trend-based prediction of parameter behavior. A functional digital architecture is developed, including measurement transformers, PLC-based data acquisition, analog-to-digital conversion, IEC 61850 communication, and SCADA-based visualization and control. The diagnostic approach is supported by mathematical models describing system dynamics, residual generation, and minimization of a quality functional. Predictive analysis is performed using trend extrapolation of monitored parameters, enabling early detection of abnormal conditions. The proposed system is validated using retrospective operational data from traction power supply equipment. The results demonstrate the capability to detect pre-failure states prior to protective shutdowns, reduce false alarms, and support predictive maintenance. The developed approach improves the reliability and operational efficiency of traction power supply systems and provides a foundation for the implementation of intelligent diagnostic solutions within Smart Grid environments.