In this paper, self-learning sliding mode control (SLSMC) method is proposed for a class of linear time-varying systems (LTV) based on extended state observer (ESO). This method firstly designed two kinds of nonlinear smooth function (NSF), and then two smooth function were applied to the design of the extended state observer and the sliding mode reaching law. In order to further improve adaptive control ability of linear time-varying systems, the steepest descent method is used to update gain parameters of sliding mode controller through self-learning algorithm. Simulation results show that the proposed control method has not only fast response and high control precision, but also effectively solves the problem that the existing theoretical methods are difficult to solve. Hence it is an effective control method independent of the model of the controlled LTV systems.