Online learning neural network inverse controller of the multivariable fermentation process

The fermentation process is nonlinear and multivariable coupling. To improve the performance of neural network inverse (NNI) controller, an online learning neural network inverse (OLNNI) control method is proposed in this paper. First, considering the strict inverse system theory, the inverse system of the fermentation process is obtained. Then, neural network is offline trained and used to approximate the inverse system. Online learning algorithm of the parameters is designed based on the basis function theory. And at last, the proof of the convergence of online learning neural network is given based on Lyapunov stability theory. The designed controller is successfully applied to the multivariable fermentation process control. Simulations show that OLNNI controller has higher performance comparing with NNI controller offline trained in previous works.