Nonlinear predictive control using genetic algorithms.

摘要

In recent years, the requirements for the quality of automatic control in the process industries increased significantly due to increased complexity of the plants and sharper specifications of product. At the same time, the available computing pourer increased to a very high level. As a result, computer models that were computationally expensive became applicable even to rather complex problems. Model Predictive Control techniques were developed to obtain tighter control and it was introduced successfully in several industrial plants. MPC can provide robust control for processes with variable gain and dynamics, multivariable interaction, measured loads, and unmeasured disturbances. The implementation of MPC requires the solution of a constrained optimization problem at each sampling time. Different optimization techniques including linear programming (LP), quadratic programming (QP) and dynamic programming (DP) have been used in MPC.; In this thesis a new approach for the implementation of MPC will be proposed using Genetic Algorithm GA. The proposed method formulates the MPC as an optimization problem and genetic algorithms are used in the optimization process. The advantages of using genetic algorithms include: applicability to any process model, possibility of defining any control objective and capability of handling any process constraints. The proposed method is applied to both SISO and MIMO systems with different types of process models, disturbance models, cost functions and process constraints. Application of the proposed algorithm to chemical processes is emphasized.