A New Event-Based PI Controller Using Evolutionary Algorithms

Abstract

This paper firstly tries to address the issue of high computational load of a PI controller and secondly tries to find the best method among three existing methods for tuning the PI controller. The three existing methods are Ziegler–Nichols (Z–N), genetic algorithm (GA), and particle swarm optimization (PSO). To address the first issue, that is reducing the computational load of a PI controller, the event-based approach is proposed. To find the best tuning method of the three, three PI controllers, each using one of the methods, are simulated; the data is collected and comparison of results is made. The paper first introduces the event-based approach, PI controller tuning, GA, and PSO. Next, three different controllers, each tuned using one of the methods, are investigated. To assess each controller, two benchmarks are used. Furthermore, data for each controller is collected at first without implementing the event-based approach and then with the event-based approach implemented in the controllers. The controlled system in this experiment is the simulated model of a DC motor; therefore, all the data and figures generated are based on this model. Finally, the results are analyzed and discussed in all the event-based controllers; a significant reduction in computational effort is observed. Moreover, compared to the conventional Z–N method, a noticeable improvement in key performance areas, especially with the use of PSO, is observed and the data indicates the superiority of the PSO method over Z–N and GA.