Application of Aesthetic Differential Evolution in Identification of Noisy Sources in a Multi Noise Plant

Abstract

The applications of Differential Evolution (DE) and its variants narrate its success. DE is simple, efficient and powerful stochastic optimization algorithm to solve wide range of optimization problems. But sometimes it gets stuck into local optima that results in slow convergence. To overcome this issue, we have proposed a modified variant of DE called aesthetic DE algorithm (ADEA). In this proposed variant the mutation phase is modified to generate the new positions using the concept of reflection. The position of global best individual is reflected i.e. mirror image to get new positions (solutions). This concept provides perturbation that in later stage helps in getting optimal positions. The proposed variant is tested on a set of 13 benchmark functions consulted from literature. The simulated results are then compared with basic DE and state-of-art algorithms. Non-parametric statistical analysis is performed for the result comparisons. Further ADEA is investigated and also compared with DE and on a real time problem of identification of noisy sources in a multi noise plant. The experimental results show the efficacy of the proposal.