Design of A Novel Neural Network-based Adaptive Backstepping Controller On Buck Converter

A Neural Network-based Adaptive Backstepping Control (NN-ABSC) is designed for a DC\DC Buck converter. A Lyapunov definition-based adaptive mechanism is presented for the Backstepping technique, which enhances robustness and stability of this scheme under various distortions. Also, without requiring a perfect mathematical modeling of the system, its performance is not suitable for practical applications, and a more rapid convergence rate for the Lyapunov function and robustness to the controller’s coefficient variations is needed. To achieve that, the parameters must be tuned again for more reliable functions. Therefore, to satisfy this need, a Neural Network adaptive mechanism has been adopted that can optimize the gains of the ABSM in challenging conditions resulting in higher efficiency and lower sensitivity to disturbances. The main issue considered by the NNs is their high computational burden and complexity, which have been addressed here with a single layer NN and a limited number of neurons in the layer. To examine the strengths of this method, conventional BSM and PID control techniques are also designed to be compared with this work. Finally, the results related to both simulations and experimental outputs are examined to show significant robustness and faster dynamics of NN-ABSC in different situations.