Application of the hybrid genetic particle swarm algorithm to design the linear quadratic regulator controller for the accelerator power supply

Purpose The purpose of this paper is to study a new method to improve the performance of the magnet power supply in the experimental ring of HIRFL-CSR.Methods A hybrid genetic particle swarm optimization algorithm is introduced,and the algorithm is applied to the optimal design of the LQR controller of pulse width modulated power supply.The fitness function of hybrid genetic particle swarm optimization is a multi-objective function,which combined the current and voltage,so that the dynamic performance of the closed-loop system can be better.The hybrid genetic particle swarm algorithm is applied to determine LQR controlling matrices Q and R.Results The simulation results show that adoption of this method leads to good transient responses,and the computational time is shorter than in the traditional trial and error methods.Conclusions The results presented in this paper show that the proposed method is robust,efficient and feasible,and the dynamic and static performance of the accelerator PWM power supply has been considerably improved.

Multi-objective Invasive Weed Optimization of the LQR Controller

The Robogymnast is a triple link underactuated pendulum that mimics a human gymnast hanging from a horizontal bar.In this paper, two multi-objective optimization methods are developed using invasive weed optimization（IWO）. The first method is the weighted criteria method IWO（WCMIWO） and the second method is the fuzzy logic IWO hybrid（FLIWOH）. The two optimization methods were used to investigate the optimum diagonal values for the Q matrix of the linear quadratic regulator（LQR） controller that can balance the Robogymnast in an upright configuration. Two LQR controllers were first developed using the parameters obtained from the two optimization methods. The same process was then repeated, but this time with disturbance applied to the Robogymnast states to develop another set of two LQR controllers. The response of the controllers was then tested in different scenarios using simulation and their performance evaluated. The results show that all four controllers are able to balance the Robogymnast with varying accuracies. It has also been observed that the controllers trained with disturbance achieve faster settling time.