Iterative Dynamic Programming Methodologies With Batch Process Optimal Control Applications

Dynamic programming is a classical approach for multilevel decision making applied in optimization control, economy and other engineering fields. As an algorithmic extension initialized by Luus from Toronto University in 1989, iterative dynamic programming (IDP) is able to solve non-linear and multivariable problems by discretizing the system in time and space.Characterized by a discrete algorithm, the optimizing capability of IDP largely depends on the discretizing merits. However, there have not been any intelligent methods concerning the instruction of how to discretize ever since. Based on an in-depth research on IDP, an improved IDP algorithm is proposed in this thesis, along with an application in batch process optimal control.The main work of the paper is as follows.A deep research on the performance and implementing steps of IDP are performed, extensively explaining and illustrating the principle and characteristic of IDP. With appropriate values of iterative times, constriction factors, grid points and other parameters, the IDP algorithm has been employed to the stimulations of two typical batch reactions.In response to the intractable discretizing problem, a self-adaptive variable-step IDP approach is originated, which optimizes the time intervals, interval numbers and switching points with two criterion functions. Subsequently, this approach is applied in two batch processes, demonstrating an improved optimizing performance index.With an in-depth investigation on the reaction mechanism and mathematical models, the penicillin fermentation process is used as the test-bed for carrying out the study of the engineering application. Simulation results show that the new IDP algorithm is of an excellent optimization ability which is expected to enhance the productivity in real batch process industry.