Research Of Rice Wine Fermentation Control And Optimization Based On Iterative Learning Control Algorithm

Iterative learning control has got fast development in last two decades as a new intelligent control method. Based on the fact that many complex industrial process has the property of repeatability, iterative learning technique imports learning mechanism and accumulates the information of controlled objects to design and improve the controller online. That is to say, it combines the online learning and the improvement of the control system into one algorithm, and then achieves the controlling by industrial process's repeating. On the basis of the problem of object nonlinearity or not easy to build accurate model, iterative learning control has good improvement, it can make up lacking prior knowledge during learning process and perfect the system performance.However, ordinary iterative learning control algorithm can't meet the control demand of practical industrial very well. Many studies both at home and abroad combined the control algorithm with other advanced control technologies, such as the optimal control, neural network, adaptive control and so on.Based on iterative learning control, this dissertation is aim to control and optimize the rice wine fermentation to improve the product quality, as the rice wine fermentation belongs to batch process which is repetitive. The contributions of the dissertation are listed below.(1)An iterative learning control algorithm based on parameter identification is proposed for linear time varying model. System model is updated by using parameter identification method, parameters of the learning gain are updated correspondingly by using identification algorithm when model parameters variation occurs. As a result, the proposed algorithm is adaptive. The algorithm performs well in a simulated typical batch reactor.(2)An iterative learning control algorithm optimized by using neural network is proposed. The algorithm adopts neural network's optimization to achieve restrain and solve of the controller, its basic idea is as follows: it learns the controlled object's property online through each iterative learning, and makes the output to follow the expected value very well under certain control input, then it uses BP neural network to improve the gain of PID iterative learning schemes. In addition, after each time of iterative learning, it adopts neural network to optimize the current output and gets the optimal learning gain to replace the original one, then gets the faster learning speed and matches the control property requirements under less iterative times. The algorithm is used to control the temperature of rice wine fermentation process, simulation has obtained a good control effect, and show the feasibility of the algorithm.