Research And Application Of Heuristic Intelligent Optimization Based On Batalgorithm

Swarm intelligence optimization algorithm is a kind of stochastic optimal method, which simulates the behavior of animal groups. It solves the problems by simulating the swarm intelligence of animals. So far scholars have put forward various swarm intelligence optimization algorithms, which simulate the visual,hearing, smell of animals, etc. However their local search performance is generally poor. Meanwhile, they have the problem of premature convergence. The bats have a completely different way of foraging—echo location. Therefore, the swarm intelligent algorithm which simulates the behavior microbats echolocation provides an effective method.In this paper, we research more efficient algorithm based on the bat algorithm(BA) to further improve its performance and have a better prospects of application.Compared with the existing research results, we make a systematic research from various perspectives including performance analysis, theoretical analysis, operators design, algorithms fusion, knowledge learning, multi-objective optimization. And we applied the bat algorithm to bioinformatics, wireless sensor network, and other engineering problems. The main research results and contributions of this dissertation are stated as follows:(1) Focus on the speed oscillation of bat algorithm, in order to make the bats can adaptively adjust their velocity according to the fitness of position, we design a series of weights of velocity. This way can make bats perform better in the search space. In view of the shortcomings, for example, the divergence of bat algorithm search method and incomplete search area, we adjust the frequency range so that bats can search whole search space. This strategy improve the exploration capability of bat algorithm.(2) Focus on the poor local search of bat algorithm. The optimization mechanism of different strategies and intelligent algorithms are analyzed. The bat algorithm is optimized by coupling with the advantages from pure mathematical theory and other intelligent optimization algorithms. In mathematical theory, the Powell method is used to improve the local search capability of the standard bat algorithm. In the different intelligent optimization algorithms fusion, we analysis genetic algorithm, simulated annealing algorithm and distributed estimation algorithm and draw their key operators to fuse into the standard bat algorithm. These methodsimprove the local search of bat algorithm.(3) Knowledge learning has been applied to the bat algorithm to overcome low utilization of information. Firstly, bat individuals constantly adjust optimization in their search by individual history knowledge and group knowledge. It is good for the group members to move to a better position, so as to accelerate the convergence rate of the algorithm. Then for the higher-dimension and multi-modal nonlinear optimization questions, we divide the bat population into different bat clusters by dynamic clustering algorithm based on similarity function to make the bats have a better regional knowledge learning. Finally the preference knowledge dimension is introduced, and it improved the individual learning ability.(4) The fast non-dominated sorting approach is introduced to multi-objective bat algorithm. This strategy can not only select the individuals who are close to the real frontier, but also can make the individuals to be evenly distributed on the edge of the real frontier. The introduction of preference polyhedron strategy further reduces the difficulty of selecting non-inferior solutions. The strategy uses the non-decreasing quasi concave function to replace the decision makers to select pareto solution. The results can effectively respond to the preferences of the decision makers.(5) The bat algorithm and its variants are applied to different fields, among them,RNA secondary structure prediction problem, protein folding structure prediction problem, the node location algorithm and coverage problem in wireless sensor networks and unbiased optimization of large Lennard-Jones clusters are included. In view of the different problems, the bat algorithm is discretized, and the different objective optimization models are designed.