Research On Swarm Intelligence Algorithm Based On Direction And Individual Difference Evolution Strategy And Its Applications

In the system of production and lift, optimization problems is very common, whose high efficiency has very important theoretical and practical significance to be researched. Traditional optimization methods have many advantages, such as complete theory, stable calculation results and high efficiency. However, it is difficult for them to solve the optimization problems independently, which have been becoming more and more complex with the development of society and technology.Swarm intelligence is a global optimization method based on the population research, which is very suitable for solving complex single object optimization, multi-object optimization and constrained optimization. Therefore, the swarm intelligence algorithms, based on swarm intelligence, have been getting more and more attention of researchers at home and abroad, which have become one of the research focus in the field of intelligence optimization. Especially, differential evolutionary algorithm and particle swarm optimization algorithm have become the hottest research direction. Compared with the traditional optimization methods, swarm intelligence algorithms are simple and easy for realize, parallel, expansion and intelligent search. Nevertheless, it has its own disadvantages such as premature convergence and local optimum as a kind of evolutionary algorithms.To solve the problem of premature convergence and falling into local optimum easily in swarm intelligence algorithms, this paper introduce evolutional direction and evolution strategy based on individual difference into swarm intelligence algorithms, and apply them to improve the performs of algorithms. They can effectively avoid the premature convergence and trapping into local optimum by balance the global exploitation ability and local exploration ability, which enhance the diversity of the entire population to improve the search efficiency. Generally speaking,the main research work of this dissertation is summarized as follows:(1) The traditional differential evolution algorithm just gets the differential vectors of individuals in the current generation, but ignore the differential vectors between the adjacent generations, which is defined the evolutionary direction between the parent and its offspring. To avoid the evolutionary process is too arbitrary to lose the excellent information of parents, this paper introduce the directed evolution strategy into differential evolution algorithm. In this paper, the adaptive strategy is designed and implemented to protect the diversity of the population in the evolution process effectively. The whole swarm is divided into three subgroups, whose evolutionary model is selected from a constructed mutation and crossover operator pool according to their respective performance. And the specific operation of each individual depends on its own performance, which decides the control parameter of evolutionary model. That's to say, this paper realizes the dynamic adjustment of evolution model and parameters setting. In addition, as an important supplement of population diversity, this paper introduces the cloud model in the selection of the next generation population to enhance the rationality and diversity of the next generation by learning the characteristics of the current swarm. Based on the analysis and design in the above content, this paper proposes an adaptive differential evolution algorithm with directional strategy and cloud model. Experiment results have verified the convergence, efficiency and robustness of the proposed algorithm.(2) In view that the standard particle swarm optimization pay more attention to the information share in the foraging behaviors of birds without considering the self-learning, self-organizing and self-thinking abilities of individual fully, this paper designed and implement an evolution mechanism based on individual difference. It quantifies the individual difference to reflect competitive ability of each individual, which influence and even decide the self- reflection and self- learning. The essence of this mechanism is adaptive, but it more emphasis on the self-decision ability of each particle to select suitable evolution model by self-judgment. It reflects the individual intelligence in the process of policy decision through making full use of the competition mechanism between each individual. To further distinguish the individual differences and to maintain the diversity of the population, this paper utilizes the catastrophe strategy to give the poor performance of individual disaster and re-generate the equivalent scale of new particles at the same time. In addition, this paper implements the adaptive adjustment of control parameters based on individual phenotype, including inertia weight and learning factor. Based on the analysis and design of the model, an improved particle swarm optimization algorithm based on individual differences and catastrophe strategy is presented, and experiments results affirmed the efficiency of the proposed algorithm.(3) To solve the singular and ill conditioned problems of matrix inverse in the traditional load identification method by least square generalized inverse, this paper converted the uncorrelated multi-source load identification in the frequency domain into a single objective optimization problem, then design and implement a multi strategy improved particle swarm optimization algorithm to solve this application problem. The presented improved algorithm includes three aspects: based on the specific application, initialize the swarm with the particle seed of domain knowledge, which has a certain direction; without the excessive human intervention, propose a nonlinear asymmetric time-varying parameter adaptive strategy, which reflects the individual differences of evolutionary way partly; before the entire swarm iterate to the next generation, all candidate particles have pretreated with crossover and mutation of genetic algorithm to enhance the diversity of the population. Comparison numerical experiment shows the high efficiency of the proposed algorithm, and simulation results confirm its applicability and accuracy in identification the uncorrelated multi-source load identification in the frequency domain. Meanwhile, it can be seen that the research of swarm intelligence algorithm has wide application prospect and excellent performance.