Research On Modeling And Optimization Of Fermentation Process Based On Particle Swarm Optimization And Support Vector Machines

With the quick development of the biological technology, the microorganism ferment has become more and more important in national economy. Because of the strong nonlinear characteristic, time-variation, uncertainty and some essential biologic variables are difficult to be measured directly in fermentative process. It makes the optimization control hard to realize in fermentative process. So modeling and forecasting for the fermentative process by soft sensor measuring technology is one of the methods to overcome the difficulty. Penicillin ferment is the typical one of all microorganism ferment. Therefore, modeling and optimization control of penicillin ferment has important actual value.Artificial Neural Network (ANN) is the most representative one in modeling and forecasting of ferment. But it is based on the Empirical Risk Minimization (ERM). It has some disadvantage such as over fitting, local minimum. It will make the model haven't better forecasting ability and diminish the forecast precision. The method of Support Vector Machine (SVM) which is based on Structure Risk Minimization (SRM) has strong ability of learning minor samples, smaller forecast error and so on.In accordance with modeling of fermentation process, the method of SVM is put forward to the modeling and forecasting for the penicillin ferment. Compare with ANN method, the simulation result indicated that the method of SVM can cause the model to obtain the better forecast effect than ANN in the penicillin fermentation process. But the parameter of model is important to the model, so we must get good value of the parameter by optimization.In accordance with optimization of fermentation process, the method is put forward to find the better parameter value by using particle swarm optimization which has the better search ability. The simulation result indicated that the model has better forecast effect after the parameters have been optimized. And based on the model, for realizing parameter optimization of material makeup fermentation process control, particle swarm optimization is applied to adjust the speed of material makeup, temperature, pH and DO. The simulation result indicated that the method can improve the output of product.