Comparison of different optimization algorithms with two stage initialization for PID controller tuning in automatic voltage regulator system

PID controller tuning plays a very important role in the proper functioning of the plant in which the controller is incorporated. A modification to the already existing algorithms like the differential evolution (DE), particle swarm optimization (PSO) and teaching-learning-based optimization (TLBO) algorithms is done in this paper. The modification is done in the initialization method and is called the two stage initialization (TSI). In TSI, the population vector generation is done randomly in two stages. Then the newly generated population vector from TSI would go through the rest of the phases involved in the algorithms. Further, these two stage initialized optimization algorithms and their original versions are used to tune the PID controller for automatic voltage regulator (AVR) system. Comparison of the results so obtained due to the application of six optimization algorithms out of which three are already existing ones and the other three are their TSI versions to PID-AVR system is also done. It can be realized that TSI helps the existing algorithms to have greater convergence rate and also to converge at a lesser minimum value of the objective function.