Font Size: a A A

Research On Data-Driven Control Scheme Based On Subspace Identification

Posted on:2012-12-28Degree:MasterType:Thesis
Country:ChinaCandidate:L F HouFull Text:PDF
GTID:2298330467478611Subject:Navigation, guidance and control
Abstract/Summary:PDF Full Text Request
The subspace-based method plays a vitally important role in the related field of approximation, optimization, signal processing, system science, etc. Polynomial approximation to the transfer function or the state space equations of complicated systems can be achieved via subspace technique. Even wh en system modeling is difficult or modeling mismatch occurs, satisfying data-driven controllers can be designed by combining the fore-mentioned method with other control schemes. Based on subspace identification technique, this thesis, in the framework of predictive control and H∞control approach, studies the related data-driven control design scheme as follows:(1) A data model for linear time-invariant systems is introduced by subspace identification technique, and also given is a data-driven predictive controller design scheme. The subspace identification problem of innovation processes with noise in the integrated form, which is commonly seen in the process industries, is subsequently considered, and the corresponding form of controller is investigated as well. It is also discussed how to design a feed-forward controller in the presence of measured noise. A simulation result is provided in order to illustrate the feasibility of the proposed control method with a guaranteed tracking performance.(2) The subspace identification problem as well as the data-driven controller design problem for a class of time-varying systems is discussed. In order to overcome the time-varying characteristic due to the system change, the Givens matrix rotaition method is used to update the subspace-based data equation. A predictive controller is then derived thereby. Simulation results indicate the feasibility of the proposed method.(3) Both the subspace identification problem and the corresponding data-driven predictive controller design problem for Hammerstein systems are considered. The input-output subspace-based data model for Hammerstein systems is constructed by identifying static input nonlinearity and estimating Markov parameter. The parameter is obtained via kernel methods, and the corresponding data-driven predictive controller is given thereafter.(4) The form of data-driven controller with a pre-specified H∞performance index is studied. A subspace predictor is identified from input-output data beforehand. Additionally, a data-driven controller and the corresponding adaptive implementation algorithm are given as well. Simulation results are provided to verify the robustness and the feasibility of the proposed method.
Keywords/Search Tags:data-driven control, subspace identification, predictive control, H_∞control, parameter estimation, iterative update algorithm
PDF Full Text Request
Related items