Research On The Improvement Of Harmony Search Algorithm

Harmony Search (HS), as a heuristic optimization method, mimics the musicians’improvisation behavior. It has been widely used in many areas. However, the algorithm is not anuniversally method, so it needs to improve when in different conditions.This paper conducts a series of studies on harmony search algorithm. First of all, the original,principle and development of the HS algorithm are given respectively. Secondly, the HS algorithmis used to solve the single and multi objective optimization problems. Finally, the HS algorithm isused to solve the constrained multi objective optimization problems. Main contents of this articlecan be summarized in the following.(1) The original, principle, development and application of the HS algorithm are given indetails. This part is for further researches.(2) For the purpose of avoiding the disadvantage of the improved harmony search (IHS)algorithm, chaos adaptive harmony search algorithm is presented. The algorithm utilizes chaostechnique initialling a population, and then adaptive harmony memory consideration rate, pitchadjusting rate and bandwidth are used to produce new solutions. Several solutions are generatedevery iteration. It makes full use of information of harmony memory. If the algorithm arriving atthe stagnating state, chaotic mutation was adopted in order to increase the diversity of thesolutions. The test results show the favorable abilities of accuracy and escaping local minimums.(3) A self-adaptive harmony search algorithm incorporates Pareto dominance to solvemulti-objective optimization problems is presented. The algorithm adopts an external archive tokeep non-dominated solutions. In order to maintain the diversity of the non-dominated solutions, acrowding measure is proposed in this article. The crowding strategy can measure the crowdingdegree accurately. The experimental results show that, the proposed MOSAHS algorithm is aneffective multi-objective harmony search algorithm with fine performance in both convergenceand diversity.(4) An improved harmony search algorithm for constrained multi-objective optimizationproblem is proposed in this paper (CMOHS). Inspired from Particle Swarm Optimization (PSO),the global extremism is introduced to speed up the convergence rate. In the proposed algorithm,two populations are adopted to increase the chance of finding the optimal solutions. Beside this, anew distance metric is constructed to measure the group density. Numerical experiments aredivided into two parts, the first one compares our new algorithm with NSGA-II by using fourbenchmark test functions, experimental results show that, the proposed algorithm is more effectivethan NSGA-II. The second part compares our new algorithm with the algorithm which not uses theglobal extremism to supervise the search process by using six benchmark test functions.Experimental results show that the global extremism improved the convergence and diversity ofthe algorithm.In general, the HS algorithm is analyzed comprehensively. Finally, whole research contentsare summarized, and further research directions are indicated.