Soft Sensor for Glutamate Fermentation Process Using Gray Least Squares Support Vector Machine

Abstract

The on-line control of glutamate fermentation process is difficult, owing to the typical uncertainties of fermentation process and the lack of suitable on-line sensors for primary process variables. A prediction model based on gray relational analysis (GRA) and least squares vector machine (LSSVM) is presented to predict glutamate concentration on-line. First, the correlation analysis of input variables was carried out by grey relational analysis to reduce model dimensionality and improve model performance. Second, in the training process of nonlinear predict model, coupled simulated annealing (CSA) arithmetic combining grid search was adopted to determine model parameter values of LSSVM for better predict accuracy. Simulation results showed that the prediction model proposed for glutamate concentration has obvious prediction performance compared with radial basis function (RBF) neural networks, it can provide effective operation guidance for control and optimization of the glutamate fermentation process.