Algorithm Research On Large Scale Global Optimization Problem

With the development of science and technology,large-scale global optimization has become a hot research field,which is widely used in various fields,such as vehicle routing,gene identification,work scheduling and network topology.The problem has the characteristics of high dimension of decision variables and large number of local optimal solutions,which makes it very difficult to solve.Therefore,it is of great significance to study how to solve large-scale global optimization problem efficiently.Although the evolutionary algorithm has the advantages of simple operation and strong search power,it is only suitable for solving some small and medium-scale problems.When the scale of the problem increases,the evolutionary algorithm is easy to fall into a local optimal solution during the evolution process.At present,there are two types of methods for solving large-scale global optimization problems.The first one is based on decomposition method,which decomposes large-scale problems into small-scale ones.The second one is based on non-decomposition method,which solves the original problem as a whole.Although these methods are effective,there are still some problems,such as how to allocate computing resources reasonably,how to combine various operators efficiently and how to jump out of local optimum.In view of these difficulties,this paper proposes some new solutions: 1.In order to avoid falling into the local optimal solution,a clustering-based population classification algorithm is proposed in this paper.The algorithm dynamically and adaptively divides the population according to the Euclidean distance between individuals in the population,and then optimizes each sub-population separately to achieve the purpose of having multiple evolutionary directions in multiple groups and maintain the diversity of the population.In addition,in order to improve search efficiency and save computing resources,the paper also designs a computational resource allocation algorithm based on contribution capability.It fully considers the optimization information of the current generation and the previous generation of each sub-component.More computing resources can be allocated to sub-components with greater contribution capability.In this way,the waste of computing resources is avoided and a better optimal solution can be obtained under the limited computing resources.By combining population classification algorithm with resource allocation algorithm,a clustering-contribution-based algorithm for large scale global optimization is proposed.The experimental results show that the algorithm is effective.2.There are two problems in the existing two methods when solving large-scale problem: poor universality and low efficiency of local search algorithms.In order to solve these problems,this paper proposes a memory chain based two-stage hybrid optimization algorithm.The algorithm saves the parameters in the memory chain,which can effectively prevent wasting computing resources in the process of parameter adjustment adaptively.It is divided into two stages: optimize and restart.In the first stage,a new hybrid algorithm framework is proposed,which combines multiple evolutionary algorithms and local search algorithms,and adaptively selects better operators to make global search and local optimization cooperate and co-evolution.In the second stage,a restart mechanism based on memory pool is designed.The memory pool is constructed by using the historical optimal solution information.When the algorithm is stagnant,the individual is re-selected from the memory pool to start a new optimization process.In addition,this paper proposes a two-population-based differential evolution algorithm,including new mutation,crossover and selection operators.And it is applied to the hybrid algorithm proposed in this paper.Finally,the efficiency of the algorithm is verified by simulation experiments.