The Improved Estimation Of Distribution Algorithm And Its Application In Chemical Process

Evolutionary algorithm is a kind of global optimization algorithm with high robustness and wide applicability. With its excellent optimization property, evolutionary algorithm will not be restricted by the specific problems condition. Thus it can effectively solve some complex problems which are difficult for some traditional intelligent optimization algorithms. It has been widely applied in many fields.As a research hotspot in evolutionary algorithm recently, estimation of distribution algorithm (EDA) shows good optimization effect in dealing with multidimensional, nonlinear, non convex problems, it has been studied widely in recent years. This paper will do research on improvement of algorithms based on estimation of distribution algorithm. Then the improved algorithm will be applied to solve the problems of specific chemical process optimization. Mainly completed the following work:1. By introducing the theory of cloud model this paper proposed a new method to construct the probability model of the problems’solution space. Using normal cloud model’s forward cloud generator to construct the EDA probability model and it’s backward cloud model to sample new population. In order to make the model more accurate, a hybrid cloud model which combined the theory of affinity propagation clustering is proposed for estimation of distribution algorithms. The improved algorithms are compared with the Gauss probability models of EDA algorithm which are based on the similar idea of modeling method, can achieve fast convergence in the early stage of the optimization process and increase the population diversity in the late process, not easy to fall into local optimal. The optimization results could be got precisely and rapidly.2. The Kriging agent model which has high prediction precision is applied to improve the EDA algorithm. According to the specific Kriging model, calculate model extreme point to compete and combination with the best individuals in the population, affecting the population evolution process, improve the convergence speed and search ability.3. Combining the specific process optimization problems in the fields of chemical production, this paper puts forward some new ideas for solving optimization problems. The improved EDA algorithm based on cloud model has been respectively applied to the optimization of industrial processes, using the improved algorithm to calculate thermodynamics physical properties parameters for olefin polymerization system. This method will be used to calculate binary interaction parameters of PC-SAFT state equation in supercritical coordination polymerization system. Put the optimization results to parameter the state equation, can accurately describe the vapor-liquid phase equilibrium between components; The improved EDA algorithm which combined with Kriging agent model is used to process optimization of styrene, effectively resolve the contradictions between precision and efficiency. The optimization results show that, the optimization effect is good.