Research On Iterative Learning Control And Its Application On Batch Process Based On Generalized Predictive Control

Currently the chemical industries infrastructure has been significantly altered. The features of its development have gradually changed from large scale to fine chemistry. The proportion of batch process in industries increased because the growing demand of market for products that yield a high added value, that of multiple kinds and that of small quantity. As a result of the batch process's instinct characteristic-unstable state, uncertainty, the automation technology of batch process is far behind continuous process. To ensure full economic and social returns, it is necessary to do more researches on advanced control methods of batch process. These methods can make the operation more stable and also can make the quality of end product better.Batch process's repetitive chatacteristic between batches is one of the important reasons that make Iterative Learning Control (ILC) technology wildly used in batch process. ILC gradually became one of the most important branches of intelligent control since 1980s. ILC is suitable for the process which has repetitive chatacteristic and it doesn't depend on detailed process model. ILC exploits every possibility to incorporate past control information: the past tracking error signals and in particular the past control input signals, into the construction of the present control action. More research results were found in recent years. All these new findings mainly focus on the new control strategy which combined ILC with some advanced control methods. One of them is BGPC (GPC for batch process), which is proposed in this paper. It not only can self-adaptive like Adaptive Control, but also can learn iteratively between batches like ILC. BGPC can reduce output error batch by batch and the control strategy can gradually be optimum. Model basis, noise and disturbance can be controlled in the least possible time, this make the reference output can be tracked accurately by output variable.This thesis applies two kinds of Iterative Learning Control algorithms to batch process in a numerical example and an experimental batch reactor system. The main contributions include the following aspects:1. The purpose, the meaning, the characteristics and the state of the batch process and Iterative Learning Control are summarized first. The existing state of ILCapplication in batch process was analyzed at the end of the first paragraph.2. The basic principle of ILC, GPC Control and the connection between them were introduced and some new algorithms in the later paragraph were also introduced.3. BGPC was proposed. It uses the past output tracking error to modify current model predictive value so that new control variable changed value can be calculated which was used together with past control value to generate new control action. At last, PID type error and control variable changed value compensation were added into BGPC, which make BGPC has better control effect.4. PID type ILC and BGPC algorithms geared specifically toward batch process application were demonstrated in a numerical example and an experimental batch reactor system for temperature tracking control. The results of the experiment show that BGPC has the better performance than PID type algorithm. The algorithm combined with BGPC and PID has good robustness and can make the output trajectory agrees very closely with the reference trajectory after a few batch runs.Finally, the work of the thesis and the further research directions are summarized.