Study And Application Of Artificial Bee Colony And Differential Evolution Algorithms

As optimization problems exist widely in all domains of scientific research and engineering application,research on optimization method has important theoretical significance and practical value.Following the deep research in all kinds of complex system optimization problems,the traditional top-down deterministic optimization algorithm has met many difficulties.But the stochastic optimization algorithm which comes from the reference and simulation of the natural phenomena and physical phenomena makes many highly complex problems for the human to be perfectly resolved by its own evolution.In stochastic optimization algorithm,swarm intelligence algorithm and evolutionary algorithm are two kinds of important representative.The Paper gives a comprehensive study on artificial bee colony in swarm intelligence algorithm and differential evolution in evolutionary algorithm from the aspects of algorithm mechanism,modifications and their applications.The main contents and contributions given in this dissertation are as follows:(1)Aiming at the shortcomings of the artificial bee colony(ABC)algorithm and its existing improved algorithms,a new variant of ABC algorithm based on the sub-populations(SPABC)is proposed.The initial solutions are generated by piecewise Logistic equation to enhance the convergence speed of the algorithm.This algorithm divides the population into two sub-populations according to the comparison between the individual fitness value and the mean of population fitness values,and adopts the different search method in the different sub-populations to effectively balance exploration and exploitation capability.Comparing with ABC algorithm and other modified ABC algorithms,the numerical simulation result demonstrates the proposed algorithm has better optimization accuracy and convergence speed in solving complex numerical optimization problems.(2)In view of the shortest path problem of the common traffic road,this paper puts forward the new standard rectangular network concept,and analyzes the properties of the shortest path between nodes,and presents a new shortest path algorithm based on the standard rectangular network(SRNSP).Making use of the geometric properties of the standard rectangular network,the algorithm simplifies the search direction and step judgment.Meanwhile this paper points out that some or even all of the common traffic road network can be converted into standard rectangular networks.Compared with the common Dijkstra,Floyd,ACO,A*and PSO algorithms of solving the shortest path problem;,the experimental results demonstrate the proposed algorithm has better optimization accuracy,stability and searching speed for large-scale standard rectangular road network.(3)In order to improve the performance of differential evolution(DE)algorithm effectively,a dynamic multi-subgroups differential evolution(DMSDE)algorithm is presented for solving the economic load dispatch(ELD)problem of power system.From the view of population diversity,DMSDE algorithm proposes a dynamic multi-subgroups strategy to add the probability of jumping out local minima.A random guided mutation operator is designed based on a balance between local search and global search,which is aim at enhance the effectiveness and universality of searching.Meanwhile,global best learning operation is introduced to avoid algorithm premature.It is compared with DE algorithm and its improved variants(DE/rand/1,DE/best/1,SaDE,JADE,ODE)and other intelligent optimization algorithms(CLPSO,IGHS,GABC,CS),the numerical results demonstrated the validity of the proposed DMSDE algorithm.DMSDE algorithm is used to solve the ELD of power systems with 3 units,13 units and 40 units.Experimental results show that DMSDE can obtain better results than other algorithms when solving economic load dispatch problems.(4)A new improved differential evolution algorithm(IDE)is proposed for solving system reliability redundancy allocation problem.The constraints handling method is improved based on punishing function method.The new constraints handling method doesn't need to calculate the every punishing function value in the search process,which greatly speeds up the searching for the optimized solutions.The new method of constraints handling has good generality,which can be introduced completely to other intelligent optimization algorithms.Using the improved algorithm to solve four typical system reliability redundancy allocation problems,the experimental results show that the algorithm has good optimization precision and speed.