Observer-Based Model-Free Adaptive Predictive Control for Discrete-Time Nonlinear System

In the setting of discrete-time nonlinear systems, a data-modeling-based control technique that combines observer and predictive control is presented in this study, where a mix of model predictive control (MPC) and dynamic linearization approaches is used to construct a data-driven strategy focused on output optimization. First, the equivalent conversion from nonlinear to linear systems is realized by compact format dynamic linearization(CFDL) method, and a pseudo-partial derivative (PPD) method of estimation with only data information is created using an adaptive observer. Second, a predictive control function containing the observer output, control input and observation error is defined and combined with continuous optimization of actions to generate the optimal control action. The convergence performance of the parameters arising from the application of the suggested method to a nonlinear system, such as PPD estimate error, observer error, and systematic error, is obtained mathematically. Ultimately, numerical simulation studies were carried out to verify the suggested algorithm's value in raising control precision.