基于模糊最小二乘支持向量机的发酵过程建模研究

青霉素发酵过程具有时变性和高度非线性,对菌体浓度等的在线测量十分困难。最小二乘支持向量机建模,虽然提高了预测速度,但是预测精度有所欠缺。为提高预测精度,本文在最小二乘支持向量机中引入模糊思想,采用一种基于类中心距离的模糊隶属度函数,为青霉素发酵过程菌体浓度建立预测模型。原理分析与仿真结果表明模糊最小二乘支持向量机建模方法相比于单一的最小二乘支持向量机建模,它的预测精度高,性能更加优越。

The research of fermentation process modeling based on fuzzy least square support vector machine

The Penicillin Fermentation process is usually characterized as time varying and nonlinear dynamic. It is very difficult to measure the mycelia concentration online. The least square support vector machine （LS-SVM） modeling method could improve the predict speed, but reduced the predict precision. The idea of fuzzy membership is introduced into the method of LS-SVM. The fuzzy least square support vector machine （FLS-SVM） uses a fuzzy membership which is based on the distance of cluster center. This method is applicated to mycelia predict of penicillin fermentation modeling. The simulation results show that the higher predict accuracy and better performance can be abtained by FLS-SVM than by LS-SVM.