Performance of Intelligent Optimization Methods in IIR System Identification Problems

Intelligent optimization methods effectively explore and review the response space using past experiences a population of search agents. These population-based techniques can solve complex optimization problems with a defined number of iterations. This paper evaluates the performance of different types of common and powerful optimization algorithms in the system identification problem in order to optimal design and modeling of Infinite Impulse Response digital filters. Assumed methods include the Genetic Algorithm and Differential Evolution both based on evolutionary strategy along with six swarm intelligence algorithms, Particle Swarm Optimization, Gravitational Search Algorithm, Inclined Planes system Optimization, Ant Lion Optimizer, Teaching-Learning-Based Optimization, and, for the first time, Biogeography-Based Optimization. In the present study, the IIR system identification problem is assumed as a single-objective optimization function. It is evaluated for two experimental and challenging IIR models for the equivalent and reduced order modeling. To evaluate the efficiency and performance of the algorithms, the simulation results are evaluated in terms of Indicator of Success and Degree of Reliability with Mean Square Error. Also, the effect of reducing search agents on the performance of algorithms is analyzed. The overall estimation of the results confirms the acknowledgment of the effectiveness of the proposed evaluation indexes and the desirable performance of the proposed methods, especially for the PSO, IPO and BBO algorithms in terms of convergence characteristics, average runtime, average values of MSE, and IoS and DoR indices; GA and GSA algorithms in terms of convergence, runtime and DoR; DE method for running time; ALO algorithm for the mean of MSE, and TLBO algorithm in terms of the convergence characteristics, mean IoS, and DoR percent