Wavelet adaptive and predictive control with applications to chemical looping system

Wavelet Network;Adaptive Control;Predictive Control;Multiresolution Analysis;Chemical Looping

Zhang, Shu

关键词: Wavelet Network;    Adaptive Control;    Predictive Control;    Multiresolution Analysis;    Chemical Looping;   

Others  :  https://www.ideals.illinois.edu/bitstream/handle/2142/49510/Shu_Zhang.pdf?sequence=1&isAllowed=y

美国|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

Chemical looping process is a novel technology to separate oxygen from nitrogen using solid oxygen carrier to facilitate carbon dioxide capture in the design of next generation clean coal power plants. Application of the available control techniques to this process to guarantee its tight and robust operation and control has not resulted in satisfactory performance due to its highly uncertain (200% actuator uncertainty) nonlinear multiphase multiscale behavior, bringing in a need for developing novel model-based nonlinear multiscale control solutions – the subject of this thesis. An evolving structure wavelet network adaptive robust state and output feedback control, based on multiresolution analysis, is proposed for a class of nonlinear uncertain dynamical system. The robust control technique has been introduced to attenuate the effects of external disturbances. To meet transient performance requirements, wavelet network technique has been incorporated into L1 adaptive control architecture, extending the latter to a class of nonlinear infinite dimensional systems subject to bounded input operator and unknown Lipschitz nonlinearities. Projection based adaptation law recently introduced for linear infinite dimensional systems with constant and time-varying matched uncertainties has been extended to this class. The proposed extension inherits L1 adaptive control guaranteed transient performance for both input and output signals attained through the use of a low-pass filter in the feedback loop. Next, the multistep adaptive generalized predictive control (GPC) scheme based on online identification of multi-resolution wavelet model structure is designed for the single-input-single-output nonlinear autoregressive exogenous (NARX) models without state or input constraints. The control inputs and wavelet model parameters are calculated by optimizing the cost function using gradient descent method. The convergence and stability of the proposed GPC scheme are proved using Lyapunov stability theorem. Then, the wavelet MRA modeling formalism is combined with design of robust nonlinear non-ad-hoc constrained MPC with guaranteed stability under mild assumptions. The benefits of all the proposed control techniques, such as fast approximation rate and stable tracking of reference trajectories are demonstrated by simulation. Finally, the real-time implementation results of the wavelet MRA based predictive control law with rate constraint on the single loop configuration of Alstom’s cold gas/solid flow chemical loop test facility in Windsor, Connecticut, are presented.