Improved Particle Swarm Optimization Algorithms For Multi-dimensional Optimization Problems

As a new kind of swarm intelligence optimization algorithm, particle swarm optimization algorithm originated in the imitation of the behavior of birds’ foraging. It uses the information exchange and cooperation of the individuals in the group to achieve its purpose of seeking the best objective. Compared with other intelligent optimization algorithms, PSO, with the characteristics of simple realization, adjustable parameters, fast convergence speed, has been widely used in biomedicine, image processing, target detection and location, and engineering optimization. Although some results has been achieved under the current research on particle swarm optimization algorithm, with the continuous improvement of the level of the model, the optimization model is more and more complicated, that is, its dimensions are higher and higher. This makes the extension of original PSO algorithm from low dimension to higher dimension infeasible and unsatisfactory.This work was partially supported by The National Natural Science Foundation of China, The Natural Science Foundation of Zhejiang Province and Graduate Innovation Project of Zhejiang Sci-Tech University. The main research contents and achievements are as follows:(1)Aiming at the difficulty of premature local convergence of standard particle swarm optimization(SPSO) algorithm exposed in tackling multi-dimensional and multimodal optimization issues. In this article, a new MDDCIW_PSO algorithm(Multi-Dimensional Descending Chaotic Inertia Weight based PSO) is proposed. The main idea of the algorithm is as follows: That is to say, vertically,the value of the inertia weight linearly decreases as the number of iterations increases; horizontally, every dimension of each particle is given an independent chaotic inertia weight within current attenuation radius. Thus, from both vertical and horizontal directions, the proposed MDDCIW_PSO algorithm tries its best to enhance the group activity and local search ability in the late period of evolution to avoid premature convergence risk. The tests on several typical benchmark functions show the MDDCIW_PSO algorithm outperforms the other classic inertia weight adaptation strategies in terms of searching precision.(2) Based on particle swarm optimization algorithm and mechanism of parallelism, a new MPPSO algorithm(Multipopulation Parallel PSO based on the island model,MPPSO)is proposed. Firstly, the multipopulation size sampling function is constructed to provide a reference to the number of the sub populations. Secondly, the K-means++ clustering method is introduced to enhance the search efficiency. Then, a strategy of information interaction between populations is proposed based on the network topology structure. The simulation results show that the algorithm can improve the performance of the algorithm and avoid premature convergence.(3) The proposed algorithm is applied to energy consumption optimization model of the printing and dyeing setting machine, which can provide a reference for the optimal working points of temperature of the oven at all levels and cloth speed. The satisfactory results show the proposed algorithm’s feasibility.