A data-driven model-free adaptive fault-tolerant control algorithm based on partial form dynamic linearization(PFDL-MFAFTC) is proposed to solve the problems of traction/braking force constraint and actuator faults for high-speed train operation control. Firstly, using the concept of pseudo gradient in the model-free adaptive control framework, the dynamic model of a high-speed train, which is difficult to accurately obtain parameters such as train mass, resistance and actuator faults, is transformed into a partial format dynamic linearization data model. Secondly, the radial basis function neural network (RBFNN) is used to deal with the nonlinear function caused by actuator faults. Then, the convergence of the PFDL-MFAFTC algorithm is guaranteed by utilizing the contraction mapping method. Finally, the effectiveness of the PFDL-MFAFTC algorithm is verified by a high-speed train numerical simulation.