Dynamic opposite learning enhanced artificial ecosystem optimizer for IIR system identification

The contrive principle of the adaptive infinite impulse response (IIR) filter is to find the filter parameters based on the error function, thus obtaining the best model of the unbeknown plant. Since the error function has a multimodal error surface, it is challenging to get the ideal identification result by traditional methods. In this work, a modified artificial ecosystem optimizer based on the novel dynamic opposite learning (DOL) strategy and a well-designed nonlinear adaptive weight coefficient, called the DAEO, is proposed to minimize the error function. The DOL adopted a random model to dynamically generate the asymmetric opposite solutions of the current population for generation jumping and population formation. To obtain more chances to find the optimal parametric solution, the DAEO is formed from two phases: The first phase produces the initial population by adopting DOL strategy, and the second phase is that DOL is employed as an extra phase to renew the AEO population in each iteration. The asymmetric search area of DOL holistically enhances the exploitation ability of DAEO, and the dynamically changing feature increases the diversity of the swarm, improving the exploration capability of the algorithm. Meanwhile, introducing the well-designed nonlinear adaptive weight coefficient makes search agents explore search space adaptively and poises exploration and exploitation phases. The classical set of benchmark problems is employed to test the performance of DAEO. The experimental results indicate that DAEO ranked first in terms of mean and variance values compared with other algorithms, except f(13). Furthermore, the DAEO algorithm is also applied to the IIR system identification problem. Simulation results on five benchmarked IIR systems show DAEO outperforms the comparison approach in improving the accuracy of recognition results and can obtain the minimum values of 0 and 1.69E-05 for mean square error (MSE) in the same-order and reduced-order system, respectively, which proves that DAEO is effective and valuable.