Research On Cultural Algorithm And Its Application In Constrained Optimization Problems

Cultural algorithm (CA) is a dual inheritance system that models evolution in human society at both the macro-evolutionary level and the micro-evolutionary level, which correspond to belief space and population space. Any intelligent optimization algorithm can be integrated into the population space and induced by the belief space which can abstract evolutionary information or experience from population space. In this way, the performance of these algorithms can be improved. In this dissertation, harmony search algorithm (HS) and particle swarm optimization (PSO) algorithm are embedded into the framework of CA. Two new algorithms (a CA based on harmony search algorithm, HSCA; an improved particle swarm optimization algorithm based on cultural algorithm, IPSOCA) are proposed which can solve constrained optimization problems effectively. In a real-world refinery, IPSOCA is applied in a type of component-uncertain naphtha blending optimization problem. The contents are summarized as follows.(1) The origin, development and application of the existing HS and PSO are first introduced. Then, CA is presented in detail which including its origin, framework, characteristics, applications, and the designing method. Finally, the development of oil blending optimization methods and general researches on naphtha are reviewed.(2) In order to improve the global search capability of HS, a HSCA algorithm is proposed in this paper. The algorithm embeds HS into the lower layer (population space) of CA framework, the upper layer (belief space) extracts evolution knowledge from perfect harmony in population space and in return makes full use of the knowledge to guide the evolution of population space, such as adjustment in some harmony of bad state in population space. In this way, it ensures the diversity of population and speeds up the evolutionary process. Simulation experiments of several classical constrained functions show that HSCA performs better than HS. (3) An IPSOCA algorithm is developed to solve constrained optimization problems. The algorithm takes advantage of the dual inheritance framework of CA, the particles in population space evolves with an improved PSO algorithm, fixed proportion elite particles are selected from population space to construct the basic elite-swarm and update evolution knowledge again in belief space. After that, the belief space passes down the new knowledge which has been updated twice to give better and further guidance to all the particles in the population space. In this way, it not only improves the diversity of population but also increases the speed of updating knowledge. Simulation results verify the advantages of IPSOCA in converging speed, precision and global searching ability.(4) A type of uncertain-component naphtha blending optimization problem is considered. A blending recipe optimization model based on combinatorial mathematics and constrained optimization is firstly constructed. IPSOCA is then used to solve the model and obtain the best recipe. The simulation results suggest the feasibility of the optimization and the validity of IPSOCA again.