The Theoretical Research And Improvement Of The Chemical Reaction Optimization Algorithm

Chemical reaction optimization algorithm is puted forward a new method of natural computation. The algorithm by interaction between the molecules of a chemical reaction to find a solution to the minimum potential energy in the potential energy surface phenomenon, using four elementary reaction, follow the law of conservation of energy. The algorithm not only have simple, universal, robust strongly and the characteristics of self-learning, self-organizing, adaptive, but also suitable for parallel processing. The algorithm in solving the problem of combination optimization, function optimization, especially the high-dimensional multimodal function of single objective optimization problem, the convergence speed, strong robustness, and can effectively avoid falling into local optimum.Broadly speaking, CRO is an algorithm framework, only need to define the general operating agent(molecular) and energy management solutions; The population variability can make the system independent toadapt to the corresponding question; According to user requirements,CRO has strong flexibility. Compared with other natural calculation methods, the CRO algorithm easies to implement, needs to less of the parameters and easies to implement parallel without gradient information.This paper studies the theoretical basis of chemical reaction optimization, expanding its application field, and explore and other properties and the application of the hybrid intelligent algorithm, the main research work is as follows:(1) Explore the working mechanism of chemical reaction optimization. The background of the CRO, its working mechanism and implementation approach, the optimization equation and algorithm flow chart of the algorithm are given out, and compared with other intelligent algorithms, the simulation results verify its superior performance. Then through the search mechanism and experimental results analysis the advantages and disadvantages of chemical reaction optimization, and gives its application range further.(2) Prove the global convergence of the Real-Coded Chemical Reaction Optimization. It has superior performance, but lacks theoretical analysis. Aiming at this problem, we study the convergence of the Real-Coded Chemical Reaction Optimization. First of all, model RCCRO as a finite absorbing Markov chain, through the convergence of the Markov chain, show that RCCRO converges to global optimal solution with probability 1; Second, validate the effectiveness of the elementary reaction by adopting different combinations of the elementary reactions,present some relevant convergence results and then learn that the global convergence is determined by both the elementary reaction and the total energy of the system; Finally the convergence speed and the first time is discussed.(3) Present the chemical reaction ant colony optimization algorithm.On the basis of researching the CRO in-depth, aiming at the problem of feedback information under-utilization lead to solve the low efficiency in the chemical reaction optimization, we present chemical reaction ant colony optimization algorithm(CRACA). Firstly, the hybrid algorithm generates the distribution of the pheromone by chemical reaction optimization(CRO); then converts the solution of chemical reaction optimization to the initial pheromone of ant colony optimization(ACO)via the transformation strategy CSA; finally obtains the optimal solution through the cumulative update the pheromone of ACO. Take the traveling salesman problem(TSP) for example, the simulation results show,compared with the CRO, ACO, simulated annealing algorithm, the algorithm has higher convergence ability, convergence efficiency and computational efficiency.