CGA-RBFN模型及其在丙烯产率预测中的应用  

Chaos Genetic Algorithm-Radial Basis Function Networks Model and its Application in Predicting Propylene Yield

作　　者：郑启富[1,2] 陈德钊[1] 俞欢军[1] 

机构地区：[1]浙江大学化学工程与生物工程学系 [2]浙江工业大学浙西分校化工系,浙江衢州324006

出　　处：《高校化学工程学报》2005年第1期71-76,共6页Journal of Chemical Engineering of Chinese Universities

基　　金：国家自然科学基金(20276063)。

摘　　要：利用混沌变量的遍历性和不规则性,将其引入遗传算法,可提高其全局搜优的性能;采用混沌遗传算法(CGA)训练径向基函数网(RBFN),并均衡地兼顾网络的拟合与预报能力,恰当地设计适应度函数,由此建成的CGA-RBFN 模型,其预测能力与稳健性都有提高。将其应用于烃类热裂解丙烯预测,效果良好,与传统方法相比有明显的优越性。Designing and training the radial basis function networks (RBFN), which involves the optimization of the cell-number of hidden layer and the center vector of basis function, and the figuring out the weight between hidden layer and output layer, is always a puzzle that hasn't been well solved. Chaos genetic algorithm (CGA) is an improved algorithm that introduces chaos variable to genetic algorithm by the full use of its ergodic property. It makes the individuals of subgeneration distributing uniformly in the defined space and avoids the prematurity of subgeneration. So CGA has better ability of global searching than simple genetic algorithm (SGA). CGA was applied to train RBFN, and a novel fitness function was designed, which considers the prediction errors of RBFN as well as the fitting errors of RBFN, on this groundwork, the CGA-RBFN model was proposed. To compare the performances of the CGA-RBFN model with those traditional approaches, the model was applied to predict the propylene yield in the course of the thermal cracking of hydrocarbon. The results demonstrate that the CGA-RBFN model possesses better prediction precision and steadiness.

关 键 词：混沌 遗传算法 径向基函数网络 热裂解 丙烯产率 预测 

分 类 号：TP15[自动化与计算机技术—控制理论与控制工程] TQ221.212[自动化与计算机技术—控制科学与工程]