A novel adaptive neuro linear quadratic regulator (ANLQR) controller design on DC-DC buck converter

Abstract

This paper proposes an Adaptive Neuro Linear Quadratic Regulator (ANLQR) controller for Buck converter operating under harmful disturbances. Considering the real-time condition of a converter with regular variations, Neural Network is adopted to improve and tune the gain of the LQR strategy with an adaptive mechanism. This strategy assumes the system as a Gray-box process without the need for the exact mathematical model of the system which can result in lower computational burden, faster dynamics, and ease of implementation. ANLQR control strategy is a suitable alternative considering its significant robustness against different disturbances, particularly noise. It should be mentioned that the number of neurons is limited to 2 and 4 in each layer to decrease the computational burden with lower complexity. The advantages of this ANLQR method is justified for various operating situations through experimental and simulation outcomes. Moreover, LQR and PSO-PID controllers are designed and compared with the presented strategy to carry on a comparison with the proposed method. Furthermore, this approach provides better outcomes with faster dynamics and better frequency adaption in the real-time environments.