Loading...

Convex Methods for the Design of Low Order Dynamic Pre-Compensators

Shamgah, Laya | 2011

469 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 42715 (05)
  4. University: Sharif University of Technology
  5. Department: Electrical Engieering
  6. Advisor(s): Nobakhti, Amin
  7. Abstract:
  8. Decentralized control is a typical method for the design of industrial MIMO control systems. However, in most multivariable systems, the existence of large interactions limit the achievable performance of such controllers. In the Nyquist Array design method proposed by Rosenbrock, the system interactions can be reduced using a precompensator, such that a decentralized controller may be used with acceptable performance. A well designed pre-compensator will effectively decouple the system and the multivariable systems may be considered as a set of SISO systems. It is therefore obvious that the key element in the Nyquist Array design method is the design of the pre-compensator. There are two important factors, for the pre-compensators: performance and dynamical order. If the pre-compensator has sufficient decoupling, the design of decentralized controller would be less conservative. However, higher performance is usually achieved at the expense of increased dynamical order which will lead to difficulties in the implementation. There are various methods proposed for the design of pre-compensators, but these methods generally do not have a good balance between performance and complexity. Traditionally heuristic or cut-and-try methods were used for the design of dynamic pre-compensators. Although they usually yield a low-order pre-compensator, the performance is often lacking. More recently, utilizing optimization algorithms, methods have been proposed which have better decoupling performance, but with high dynamical order (at least same as the order of system). This thesis presents an algorithm for decoupling multivariable systems based on Quadratic Programming (QP). A single framework is presented which can be used to design centralized, decentralized, and sparse structures of arbitrary dynamical order. Sparsity is important in implementation of controller in large-scale systems, where the connection between some inputs and outputs may not be possible. Moreover the implementation of these controllers is more economical
  9. Keywords:
  10. Quadratic Programming ; Dynamic Pre-compensation ; Covex Method ; Simply Structured Control

 Digital Object List

 Bookmark

No TOC