Optimization of PID and sliding mode control parameters with water cycle algorithm

Control systems in the industrial field are one of the important elements of production. With the controls performed at every step of the production process, it is ensured that the product is obtained with the desired properties. One of the necessary conditions to achieve this is that the selected control system parameters are appropriate. In this context, the PID controller is one of the most preferred controllers in the industrial field due to the fact that the number of parameters and the effect of the applied parameters on the system can be observed. Tuning methods using mathematical modeling and the methods based on open and closed loop system responses are available in the literature. However, it is seen that meta-heuristic methods are also used popularly in recent years. In addition, the sliding mode control method, which is an alternative controller for PID, is a type of controller suggested for real systems in the literature. Using both of the controllers, sufficient level of desired output can be obtained if well-tuned in real systems. However, there is no tuning method suitable for the structure of every controlled system. In recent years, the suggested tuning methods to overcome this problem are the meta-heuristic search algorithms. Meta-heuristic algorithms are optimization methods inspired by various events in nature. The most important feature of these algorithms is population-based search algorithms and the mathematical model of the controlled system is not needed. With this aspect, it can give good results in the optimization of complex control systems. In this study, Water Cycle Algorithm, which is one of the new meta-heuristic methods, was chosen to find optimum parameters PID and sliding mode controller. With these algorithms based on a single selected fitness function, the parameters that will make the function minimum are searched. In the literature, by using one or more of the error-based performance functions together, optimum parameters are searched in different search spaces. Since each of the control parameters shows different properties, customized fitness functions are proposed in this study. Thus, results were obtained fast in finding the optimum parameters. The obtained results are presented both graphically and statistically.