Optimal Controller Parameter Tuning from Multi/Many-objective Optimization Algorithms

Abstract

Controller performance is evaluated with the properties of steady-state and transient response of the system at the time domain. The compensators and conventional controllers like PID are designed, so that the desired performance is reached only by adjusting the controller parameters; this adjustment mechanism is called tuning. Even many approaches are proposed for tuning; still, it remains one of the problems of control theory due to the imperfect modeling, disturbance, and problem complexity. At this stage, optimization algorithms have helped the researchers to find these parameters via designing objective function concerning the time response characteristics of the system. As the expectations (for example, steady-state error, overshoot, rise time, settling time) related to the system performance are increased in number, the number of objectives is also increased. As a result, multi-objective optimization algorithms have applied to solve these problems. In this chapter, a set of benchmark tuning problems is defined as the test—benchmark—problems. Then, by using the multi- and many-objective optimization algorithms, the performance of the controlled system with respect to the Pareto approximate set is compared with each other. The multi/many-objective optimization algorithms evaluated in this chapter are as follows: Multi-objective Evolutionary Algorithm based on Decomposition (MOEAD), Non-dominated Sorting Genetic Algorithm II (NSGA-II), Multi-objective Particle Swarm Optimization (MOPSO), Strength Pareto Evolutionary Algorithm 2 (SPEA2), Approximation-guided Evolutionary Algorithm II (AGE-II), and Reference Vector Guided Evolutionary Algorithm (RVEA).