Traditionally, the closed-loop controller of Buck converters is designed on the premise of ignoring the nonlinearity of the switching devices and linearizing the converter with the small signal model. As a result, when the parameters are perturbed, or the load is disturbed, the dynamic and the static performances and the disturbance rejection performance of the controller will be degraded. Therefore, the switching control based on the hybrid system theory is introduced into Buck converter control to overcome the disadvantages mentioned above. However, the algorithm of the switching control law involves many matrix operations and still inevitably requires the object parameters. Considering the powerful ability of neural networks to approximate complex nonlinear functions, and the fact that the parameters of the object model are not needed when training, a backpropagation neural network (BPNN) is designed to replace the switching control. To achieve simpler neural network structure and better learning performance, the BPNN only learns the continuous input-output parts in the switching control scheme, keeping the last part which generating discrete driving signal. The simulation results indicate that the BPNN control scheme inherits the excellent dynamic and static performances of the switching control and weakens the model parameter dependence. In addition, the generalization ability of neural networks enables the proposed BPNN control scheme to exhibit stronger parameter robustness.