Research And Application Of Krill Herd Optimization Algorithm In Constraint Optimization

Nowadays,with the development of science and technology,the research of heuristic optimization algorithm is widely concerned by scholars,and its stronger robustness,versatility,ease of use are expected to solve more complex optimization problems.Compared with traditional optimization algorithms,heuristic optimization algorithm is a kind of optimization algorithm which simulates biological habits or laws of nature,and has better optimization processing ability.At present,there are a variety of heuristic optimization algorithms,and the research on the performance improvement and practical application of heuristic optimization algorithms has achieved relatively good results.However,the following three difficulties are still common in heuristic optimization algorithm process:(1)Falling into a local optimal solution leads to a reduction in search accuracy.(2)Unable to deal effectively with constraints may result the final solution is not feasible.(3)Its performance is seriously affected in high dimensional scenarios and cause algorithm's search ability decline.Therefore,aiming at the above three problems,this thesis studies the krill herd algorithm in heuristic optimization algorithm.The main research contents and contributions are as follows:First,an Improved Fuzzy Krill Herd Algorithm is proposed for the problem of algorithm falling into local optimal solution in this thesis.In the Standard Krill Herd Algorithm,the Foraging Inertia Weight,Induced Inertia Weight and Step Scaling Factor are part of the algorithm Search Step Size.The innovation of this algorithm is that introducing more detailed and accurate fuzzy control logic to the original algorithm.Based on the degree of variation of particle fitness and the number of iterations during optimization,krill algorithm dynamically adjusts the Inertial Weight of algorithm and Step Scale Factor so as to achieve the purpose of effectively balancing the Global Search and Local Search to avoid the algorithm falling into the local optimal value.Secondly,a new constraint processing method for the difficulty of the algorithm to deal with constraints effectively is proposed in this thesis.The innovation of this method is that proposed a new adaptive penalty function method which is divide the state of the previous G round of iterative optimal particles into five different categories to avoid degenerating the adaptive penalty function into static penalty function method;At the same time,in order to avoid the capability degradation of the adaptive penalty function method,this method deals with the constraint condition by using dynamic penalty function method which is according to the feature that most of the particles are feasible the later stage.This constraint processing mechanism can give full play to the optimization performance of the krill herd algorithm so that it can deal with the constraint optimization problem effectively.Finally,a Gene Krill Herd Algorithm is proposed for the degradation of search capability in high-dimensional scenarios in this thesis.In high-dimensional scenarios,the invalid iterative effect of Krill Herd Algorithm will seriously affect the performance of the algorithm.The innovation of this algorithm is that introducing Genetic Algorithm to algorithm optimization,then the particles in populations are devided into Particles with variable fitness values and Particles with better fitness two categories.For worse particles,the krill algorithm uses genetic algorithms to cross-variate particles until the particles get a better solution...