Research On Particle Swarm Optimization Algorithm With Complex Network Topology And Fuzzy Logic

Optimization is an important problem in many practical applications,which exists in all fields of science and engineering.For structured optimization problems,the size of the solution space can be controlled,and the optimal solution can be obtained by using precise algorithms.However,when the size of the solution space of optimization problems increases gradually,it will become very difficult or even impossible to solve these problems by using precise algorithms.Therefore,intelligent optimization algorithm came into being.Intelligent optimization algorithms include many,such as genetic algorithm,differential evolution algorithm,particle swarm optimization algorithm,ant colony optimization algorithm,and so on.As an important part of intelligent optimization algorithm,particle swarm optimization algorithm is well known for its advantages of less adjustment parameters,simple implementation and high search efficiency.However,when solving a complex problem with multiple local optimal solutions,the diversity of the population of the algorithm will gradually decrease with the progress of the search process,resulting in the reduction of the global exploration ability of the algorithm and premature convergence stagnation.At the same time,the optimization performance of the algorithm often depends on the guidance of the population leader,and the selection of appropriate individuals and leaders has an important impact on achieving the balance between the convergence ability of the algorithm and the diversity of the population.Therefore,this paper has conducted in-depth research on particle swarm optimization algorithm,including the improvement of the algorithm in the field of single-objective optimization and the improvement in the field of constraint optimization.The main research work is as follows:(1)A single objective particle swarm optimization algorithm using scale-free network topology is proposed.In view of the problem that the algorithm usually has slow convergence speed when solving some complex problems,based on the characteristics of scale-free network topology power-law distribution,this paper constructs a corresponding neighborhood for each particle,selects elite particles from the neighborhood to participate in the particle evolution process,and introduces the speed difference update operation to give full play to the guiding role of elite particles in the population evolution process.In addition,a novel inertia weight adaptive strategy is proposed to achieve the balance of global exploration and local exploitation capabilities in the algorithm search process.In this paper,18 benchmark test functions are used to test the accuracy and stability of the solution.The experimental results show that the proposed adaptive particle swarm optimization algorithm based on scale-free network topology has good robustness and can obtain competitive solutions.(2)A single objective particle swarm optimization algorithm based on complex network topology adaptation is proposed.In the process of optimization,the particle swarm optimization algorithm only learns from the global optimal particles,resulting in premature convergence and poor accuracy.In this paper,the fitness distance correlation is introduced as the basis to judge the difficulty of the problem,and the degree of dispersion of different network topologies in the process of particle swarm optimization is obtained through analysis,which is the adaptive network neighborhood topology of particle swarm optimization,effectively balancing the global exploration and local exploitation capabilities.The complex network neighborhood topology construction strategy is used to build neighborhood topology for each particle,so that the local optimal particles in the neighborhood can participate in the search process of particle swarm,get rid of the situation of only learning the global optimal particles,and achieve the goal of improving the accuracy of particle swarm optimization.In addition,this paper introduces a random drift strategy to make the particles drift slightly randomly and reduce the risk of particle swarm falling into premature convergence.The experimental results on 24 benchmark test functions show that CNAPSO has a great improvement in solving accuracy and convergence speed compared with six representative PSO algorithms.(3)A constrained particle swarm optimization algorithm based on correlation analysis and fuzzy logic is proposed.When particle swarm optimization algorithm is used to solve constrained optimization problems,there is always an imbalance between the objective function and the constraint processing technology.Aiming at the above problems,we propose an adaptive algorithm with correlation analysis ε.The constraint processing method is used to adjust the utilization ratio between the constraint and the objective function information,and fully search the boundary space of the feasible region and the infeasible region.Secondly,a fuzzy logic rule integrating individual feasibility is proposed,which uses fuzzy logic to control inertia weight and individual cognitive and social cognitive coefficients to enhance the optimization ability of search space.Finally,to better balance the relationship between the constraint and the objective function,an individual learning mechanism with stagnation detection is proposed.The search state of the algorithm is judged by the threshold,the stagnant particles are restarted,and the individual is pardoned by the amnesty criterion,so as to avoid the particle swarm falling into local optimization.Use CEC2017 constraint optimization kit to compare with well-known constraint evolutionary algorithm and PSO variants.The experimental results show that the algorithm has certain advantages in processing constraints and optimizing.