Research On An Improved Cultural Algorithm And Its Application

Cultural Algorithm is an improved algorithm based on Genetic Algorithms. Ithas faster convergence speed and is more targeted to converge to global optimalsolution than conventional evolutionary algorithm since it introduces belief space onthe basis of population space. If we think of Genetic Algorithms as a simulation ofevolutionary process of living creatures except human beings, we can think ofCultural Algorithm as a simulation of evolutionary process of human beings. Thereason that human beings have faster convergence speed is cultural heritage in humanevolution. Similarly, Cultural Algorithm is faster than the algorithms that don’t havebelief space after it introduce belief space and let it affect the evolution. Since thestudy of CA is still at an early age and CA is not widely used, it is necessary to have adeeper study of CA which includes parameter selection and application on practicalproblems. This paper mainly study CA’s application on benchmark functions test,parameter selection, CA’s improvement and it’s application on parameteridentification and controller parameter optimization., which is specifically elaboratedas follows:Benchmark function test is done to test CA’s optimize performance and the resultshows that CA demonstrates good performance in dealing with constrained andunconstrained optimization problems. The paper also dose test how it affect CA’sperformance to change CA’s parameters, which provides basis for parameterselection.An improved CA with intelligent stretching factor is designed to deal withunconstrained optimization problems and the result shows that it makes theoptimization much faster. The paper also puts forward an improved CA which dividesbelief cells adaptively to deal with constrained optimization problems and the resultshow that the number of iterations is apparently reduced.Apply CA and improved CA in control system to verify its performance. AfterCA is used in parameter identification and controller parameter optimization,wecontrast the result with PSO and GA and conclude that CA has a speed advantage indealing with this kind of problem.