Research And Application Of Estimation Of Distribution Algorithms

Estimation of distribution algorithms(EDA) is a new stochastic optimization algorithm in the field of evolutionary computation which is a research hotspot in the field of current international evolutionary computation. This algorithm has caused extensive concern of international academia in recent years due to its explicit representation to the relationship between the variables, it can solve a class of optimization problems fast, accurate and reliable which traditional genetic algorithm is not. Also become a popular tools to solve the practical engineering optimization problems gradually. The algorithm is widely used in all kinds of optimization problems, so it is of great research value.The development of EDA in this paper in the historical background and theoretical basis,introduces the EDA in multiple research status in the field of application briefly. EDA is improved by introducing chaos mutation into distribution estimation algorithm. Analyses the EDA in solving complex nonlinear optimization problem and the MPPT problem emphatically.Based on the study of the theory of EDA, its application is used in the control system theory which focus on the key research contents as following:(1) The background, significance and research status at home and abroad of EDA research has made in a detailed description. The brief illumination of the calculation principle of EDA is explained based on the probability model illustration.(2) Four commonly used typical benchmark test functions(Rastrigin function, Rosenbrock function, Ackley function, Schwefel function) are used to do the test in order to verify the validity of EDA. The experimental results confirm the validity of the EDA. Using the EDA with diversity preservation(EDA-DP) to the contrast benchmark functions test results show that the simulation results demonstrate the EDA- DP for nonlinear system identification has important significance.(3) Aim at the weight optimization of the basis function of C hebyshev neural network for nonlinear systems, using EDA- DP and BP networks to identify it, simulation results demonstrate that EDA- DP has important significance for nonlinear system identification.(4) EDA is adopted to realize maximum power point tracking of photovoltaic system in a dynamic environment. An MPPT method based on estimation of distribution algorithms is proposed aiming at traditional method contradiction of slow tracking speed and low tracking p recision in a dynamic environment. Then compared with PSO algorithm. EDA algorithm’s identification precision and speed are better than PSO algorithm.(5) Summary and outlook. Make a systematic summary to the full text of work and the outlook for the next phase of work on the basis of the present study.