Combined fitness function based particle swarm optimization algorithm for system identification

• We focus on the limitation of fitness function with the squared error in PSO.
• A combined fitness function based PSO algorithm is proposed.
• It employs entropy method to combine two measures, MSE and GARG.
• Results show that it’s feasible for system identification.

An improved particle swarm optimization (PSO) algorithm, called combined fitness function based particle swarm optimization algorithm is presented in this investigation. PSO algorithm originated from bird flocking models and is effective in solving system identification problems. However in the identification process, single measure like the squared error between the measured values and the modeled ones may be not a sufficient criterion. The improved PSO algorithm adopts a combined fitness function to solve this problem. Mean Square Error (MSE) and Grey Absolute Relational Grade (GARG) are employed as evaluation measures, and entropy method is used to determine the relative weights of the two measures. Numerical simulations and experiments are carried out to evaluate the performance of the improved PSO. Consistent results demonstrate that combined fitness function based PSO algorithm is feasible and efficient for system identification, and can achieve better performance over conventional PSO algorithm.