The invention discloses an adaptive control system of a double closed-loop RBF neural network sliding mode variable structure. A controller outputs a driving signal to control a Buck converter to stabilize output voltage to preset reference output voltage according to a set desired output voltage and obtained real-time feedback output voltage and real-time feedback output current. Compared with the prior art, the boundary layer sliding mode control technology is used to reduce chattering generated by sliding mode variable structure control during a control process. At the same time, for the uncertainty of the Buck converter during the working process due to a system parameter change and external unknown interference, and the state equation of the Buck converter is described more accurately. For the uncertainty of a system, the adaptive control is added based on a designed sliding mode variable structure controller, while an external environment is adapted automatically, various influences in the external environment on the Buck converter can be reduced to the largest extent, and robustness is not lost.