Improvements And Applications Of Harmony Search Algorithm

Harmony search algorithm,mimicking the improvisation process of music player,is a new swarm intelligent optimization algorithm,which generates a new better vector after considering all of the existing vectors.The harmony search algorithm with incomplete theoretical basis mainly applied in continuous optimization problems,while there are some disadvantages,such as low precision,premature convergence,and poor effect in solving discrete problem.Therefore,the algorithm will be improved in order to solve discrete problems with better effects,which has important theoretical significance and practical application values.Based on the previous research,the algorithm is improved according to these ideas,parameter adjustment,improving search strategy and adopting new optimization idea.The main research work is summarized as follows:(1)The opposite improved harmony search(OIHS)is proposed in order to improve the optimization ability.The harmony memory consider rate is improved in harmony search algorithm with introducing the opposite learning idea,and the search space is widened with adding opposite random solution under certain probability;in order to meet the requirements of searching process and improve the effectiveness of algorithm,some parameters are defined to measure the changes of information in harmony memory,and used to adaptively adjust the important parameters in harmony search algorithm,such as harmony memory considering rate(HMCR),pitch adjusting rate(PAR)and fine-tuning distance bandwidth(bw).The numerical test results show that OIHS algorithm has better optimization quality.(2)The hybrid harmony search algorithm(MHHS)with multiple subgroups is proposed to improve the searching effectiveness of harmony search algorithm.In the proposed algorithm,the distances of each harmony from best harmony are sorted and layered with sorting results,and then each layer is regard as a single subgroup.In order to broaden searching range,different subgroups compromise different differential adjustment strategy;and a communication mechanism is established to exchange information and promote coevolution within each subgroup.The numerical test results show that proposed algorithm has better effectiveness in precision,convergence and robustness compared with reported algorithms,such as HS?EHS?NGHS?MPSO?CLPSO?DE?ODE and IABC.(3)In order to solve the complex,discontinuous,non-differentiable and nonlinear power system Economic Load Dispatch(ELD)problem,we proposed a improved harmony search(MHS)algorithm.And in this algorithm we improve two important parameters,pitch adjusting rate(PAR)and fine-tuning distance bandwidth(bw),to effectively balance global and local searching ability.Considering the value point effect and ignoring the network loss,simulations are made to verify the effectiveness of the algorithm.(4)A global competitive harmony search algorithm(GCHS)is proposed in this paper.In this algorithm,the conceptions of stochastic local mean and global mean are given;Competition search mechanism is built to realize two harmony vectors competition selection,and the two harmony vectors both generated in the each iteration.Adaptive global pitch adjustment and local learning strategy are designed to balance the global search and local search.The effects that varying the parameter HMS,HMCR and PAR have on the performance of the GCHS algorithm is also analyzed in detail.The numerical results demonstrated the superiority of the proposed GCHS algorithm in terms of accuracy,convergence speed,and robustness when compared with harmony search algorithm and other seven state-of-the-art harmony search variants.(5)For the high dimension knapsack problem,we proposed a discrete global harmony search algorithm in order to avoid local optimum.Combining the probability model of distribution estimation method,we design multiple-selection adaptively adjustment strategy based on the features of knapsack problem to balance global and local searching.Furthermore,the elite cultivation strategy is introduced to the harmony search algorithm with key training the current global optimal harmony to improve the ability of jumping local optimum.Random repair strategy is adopted to improve the algorithm feasibility,and a random permutation strategy is proposed to add the diversity of the solution.For the high dimension knapsack problem,numerical simulations are carried out to test the feasibility compared the proposed algorithm with the four classical discrete harmony search algorithm.(6)For the Quadratic Knapsack Problem(QKP),the hybrid multiple-strategy harmony search algorithm(HMHS)is proposed.And four modified strategies are proposed bases on the characteristics of the QKP:? The binary code is adopted in the HMHS algorithm,which can solve appropriately the Quadratic Knapsack Problem.? A Harmony memory considering strategy is proposed,which compare the excellent local average with the global variable probability distribution.It applies the excellent individual to guide the searching direction and adds the survival rate of the excellent variables by means of the probability distribution of the global variables,which effectively increases the searching feasibility.Meanwhile,a dynamical adaptive adjustment method is proposed to confirm the value of the parameter HMCR for satisfying the needs of searching process.? A new strategy is proposed with changing the pitch adjustment of the original harmony and hybrid optimal teaching thoughts.By means of global optimization and random harmony operation,the speed of algorithm can be increased.? A simple and effective repair method is proposed to assure the feasibility and strengthen the development ability and convergence rate based on the specificity of QKP.For multiple-unit QKP,a number of simulation tests are made,and the numerical result indicates the feasibility of HMHS algorithm.Finally,the work of this paper is summarized and the future development and prospect are given.