基于PSO-ANN逆的发酵过程软测量建模

针对发酵过程中一些关键生化参量难以通过常规仪表实时测量,而制约发酵生产过程优化控制的问题,提出一种基于粒子群神经网络逆(PSO-ANN逆)的发酵软测量建模方法.以青霉素发酵过程为背景,首先建立其虚拟子系统数学模型,并构建发酵过程逆模型;其次,提出PSO-ANN逆的软测量实现方法,以克服解析法逆运算的复杂性甚至难于实现的问题;最终构建PSO-ANN逆软测量模型,并进行试验及仿真.结果表明:该软测量建模方法能够将机理建模与数据驱动建模方法相结合,充分利用对象模型的先验知识和经验数据,有效解决了青霉素发酵过程中不可在线测量的关键参量实时测量难题,其训练和测试误差分别达到0.037 2和0.046 1,模型具有较高的预测精度和较强的预测能力.

Soft sensor modeling of fermentation process based on PSO-ANN inversion

To solve the difficulty of measuring the key biological parameters on line by conventional instruments,a soft sensor modeling method was proposed based on PSO-ANN inversion to realize optimal control in biological fermentation.The mathematical model of assumed subsystem for penicillin fermentation was established to give the inverse model of fermentation.To solve the complex operation of inversion,the soft sensor method based on PSO-ANN inversion was proposed.The PSO-ANN inverse model was finally established to complete simulation and experiments.The results show that the prior knowledge and empirical data can be entirely used by the soft sensor modeling method to measure the key parameters.The proposed model shows high precision and good performance with training error of 0.037 2 and testing error of 0.046 1.