Efficient Algorithms For Large Scale Global Optimization

With the rapid development of society,large-scale global optimization has become more and more widely used in various fields,such as the urban intelligent transportation system,largescale cluster scheduling management,and aircraft wing design.This kind of problem has the characteristics of high dimension of decision variables,space too large to search,and large number of local optimal solutions,which makes the solution of such problems very difficult.Therefore,research on algorithms for solving large-scale global optimization problems is of great practical significance and social value.Evolutionary algorithms are often used to solve optimization problems,however,when the scale of the problem increases,the algorithm's optimization performance drops dramatically because of its slow convergence rate,therefore it is not suitable for solving large-scale global optimization problems,sloving such problems requires improvement of classical evolutionary algorithms.Cooperative co-evolution algorithms and hybrid optimization algorithms are often used in solving large-scale global optimization problems,the former relies on the performance of evolutionary algorithms,it is not suitable for different types of optimization problems.While the latter optimizes all decision variables as a whole,when the problem decision variable increases,it does not have scalability.This paper conducts indepth research on large-scale global optimization problems,comprehensively analyzes existing algorithm defects and deficiencies,and proposes two new algorithms for solving large-scale global optimization problems.The specific content is as follows:(1)In order to solve the problems of low efficiency,and inefficient use of optimization information of the existing differential evolution algorithms in large-scale global optimization,Dynamic Parameters Adjust Differential Evolution algorithm(DPADE)was proposed.A new parameter dynamic adjustment strategy was designed,which statistically weights parameters based on the survival of offspring in the evolutionary process,making it possible to dynamically adjust the search neighborhood in the optimization process.At the same time,in order to make full use of the historical optimization information,a new strategy is proposed in which the search step size can be adaptively scaled,this strategy uses the current state of the population to calculate the entropy value,thus determining the search step length.Finally,in order to prevent from falling into a local optimum,the algorithm saves the suboptimal solution in the selection phase to generate a difference vector and improve the diversity of the search direction.The simulation experiment verifies the efficiency of the proposed new algorithm.(2)Aiming at the problems of low resource utilization efficiency and single solution type existing in solving large-scale problems of existing algorithms,Bi-level Resource Allocation(BRA)large-scale global optimization algorithm was proposed,this algorithm absorbs the “divide and conquer” of cooperative co-evolution algorithms and the advantages of the hybrid optimization algorithm "able to do more." In order to be able to solve hyperscale problems,the new algorithm will first group the problems,so that solving large-scale problems into a number of small-scale sub-problems.In order to improve the resource utilization efficiency of non-equilibrium problems,a Deep Resource Allocation(DRA)strategy is designed so that the sub-problems with the greatest contribution are fully optimized.At the same time,in order to solve different types of large scale optimization problems,a Range Resource Allocation(RRA)strategy is designed to allocate more computing resources and a larger population size for the best performing algorithm in each round of iterative optimization.Finally,simulation experiments are carried out to verify the efficiency and universality of the new algorithm.