To improve service quality and energy efficiency in urban rail transit networks under frequent operational disturbances, this paper proposes a real-time train rescheduling and speed profile selection approach considering transfer coordination and energy saving. First, a mixed-integer nonlinear programming model is developed by introducing decision variables related to train rescheduling and speed profile selection. The model aims to minimize timetable deviations, total passenger waiting time, and train energy consumption. Subsequently, a rolling horizon optimization approach is employed to dynamically adjust train timetables and speed profiles. An efficient decomposition algorithm based on passenger flow estimation is designed to solve the optimization problem for each decision stage. This algorithm decomposes the complex network-level problem into a series of smaller line-level subproblems that can be solved in parallel. It also effectively addresses the computational challenges of nonconvexity and nonlinearity, significantly improving computational efficiency. Finally, simulation experiments are conducted on the Beijing Metro network under various disturbance and passenger demand scenarios. The results validate the feasibility and effectiveness of the proposed model and algorithm in large-scale real-world applications. Compared with heuristic rule-based train rescheduling methods, the proposed method reduces timetable deviations, passenger waiting time, and energy consumption by 14.18%, 6.85%, and 2.35%, respectively, across different disturbance and demand scenarios, effectively improving operational efficiency and service levels. Moreover, the decomposition algorithm achieves solutions with an optimality gap of less than 5% within 3 seconds, meeting real-time requirements. The proposed approach provides decision-making support for train rescheduling of urban rail transit networks under disturbances.