In recent years, with the development of digital computers/semiconductor devices and the establishment of various control theories, control technology has grown remarkably. However, for controllers on the industrial sites, economic feasibility, performance, reliability, and easiness of maintenance need to be essentially considered, and thus, PID controllers, which are based on a classical control technique, are still widely used, accounting for more than about 90%.
A PID controller basically has a simple structure that consists of three parameters, and has relatively superior control characteristics even in a nonlinear system and a system where an accurate mathematical model cannot be obtained. Also, an engineer on the sites can easily tune controller parameters, and the establishment and maintenance of a system are convenient. However, the performance of a PID controller, which is widely used on general industrial sites, varies significantly depending on the degree of optimization of parameter tuning. On the sites, the parameters of a controller are mostly tuned based on the experience of an engineer, rather than an analytical method for a system, and it is inevitably vulnerable to the changes and uncertainties of a system.
In this situation, the development of a PID controller parameter tuning technique that has superior performance and can be applied to various types of systems is required, rather than a controller design based on the experience of individuals.
A number of tuning rules have been suggested until recently, and the most representative PID tuning method is the Ziegler-Nichols tuning method, which is widely used on the sites. Also, there are the Cohen-Coon method, the IMC method, and the Lopez ITAE method, which tune parameters by simplifying a hight-order system into a First-Order Plus Time Delay(FOPTD) system with a time delay. Recently, controller design methods that introduce behavioral patterns found in the natural world into an optimization technique have been studied. The representative method includes a Genetic Algorithm (GA), which has been implemented using natural selection and evolutionary mechanism. Also, a Particle Swarm Optimization (PSO) algorithm, which has recently been suggested, simulates social behavior patterns found in the communities of insects, birds, and fish. This introduces a concept, where a number of individuals find an optimal solution in a search area based on the information on each individual and the entire community, into an optimization search algorithm. Despite the relative simplicity of the algorithm, many studies have been performed based on its superior control performance.
In this study, a PSO-based PID controller that optimally tunes the parameters of a controller using a PSO algorithm that is based on the social behavior patterns of organisms was proposed. To appropriately tune the three kinds of gains of the proposed PID controller (proportional gain, integral gain, and derivative gain), IAE was used as the objective function so that the sum of the absolute values of errors, which are the difference between the input and the output, could be minimized. Also, to strengthen the global search of particles in the early stage of search and to strengthen the local search in the convergence stage, the inertial load was linearly decreased as the generation number increased.
To examine the validity of the proposed tuning method, simulations were performed by applying the proposed method to three kinds of systems (first-order, second-order, and fifth-order systems with a time delay) and the superiority of the proposed PSO-based PID controller was demonstrated by comparing its response characteristics with those of the Z-N tuning method, the Cohen-Coon method, the IMC method, and the Lopez ITAE method, which have frequently been used.