Research On Model-Free Adaptive Control Method Based On Disturbance Observer

Model-free adaptive control(MFAC)is a technique that directly designs controllers using input and output data from a system,without the need to establish a mathematical model of the system.This method has certain advantages in practical control system applications,as it may be difficult to obtain or there may exist uncertainties in the mathematical model of the system.However,in real-world applications,systems are often subject to external disturbances such as noise,interference,and uncertainties,which can affect the effectiveness of MFAC.Therefore,studying the effects of disturbances on the system can help us better under stand the applicability and limitations of MFAC,as well as how to design more robust controllers to improve the control performance and stability of the system.In the presence of disturbances,disturbance observer based control(DOBC)is an effective method for improving system performance.This article proposes a DOBC-based MFAC(MFAC-DOBC)method for controlling a class of nonlinear systems with disturbances,in order to improve the system’s disturbance resistance and robustness,thereby enhancing the control performance of the system.The main contents of the paper are as follows:(1)A model-free adaptive control method based on a centralized Kalman filter disturbance observer is proposed for a class of discrete-time nonlinear systems with measurement disturbances.Firstly,the linearized data model of the controlled system is constructed by dynamic linearization method;Then,according to the linearized data model and the measurement data of the sensor,an optimal centralized Kalman filter disturbance observer is designed;Finally,using the output of the observer to adjust the pseudo partial derivatives online,the control update scheme of the system is proposed.The design and analysis of the proposed scheme do not depend on any model information except input and output data,which can avoid the conventional model-free adaptive control methods being susceptible to measurement disturbances.Simulation results show that,compared with the model-free adaptive control method based on the single-sensor Kalman filter disturbance observer,the proposed modelfree adaptive control method based on the multi-sensor optimal centralized Kalman filter disturbance observer has better tracking performance and larger data signal-tonoise ratio.(2)A model-free adaptive control method based on sliding mode disturbance observer is proposed for nonlinear discrete systems subject to external disturbances.Firstly,dynamic linearization technique is used to transform the nonlinear system into a linear system.Then,a sliding mode disturbance observer is designed to estimate the disturbance affecting the system,and the estimated value is used as a compensating term in the controller.Based on a performance index function that depends on the control input,a control law based on the disturbance observer is proposed.Mathematical derivation is carried out to prove the bounds of the tracking error.Finally,numerical simulations and a steam-water heat exchanger model are used to validate the effectiveness and superiority of the proposed approach.(3)A model-free adaptive iterative learning control scheme based on a radial basis function neural network disturbance observer(MFAILC-RBFNNDO)is proposed for a class of unknown nonlinear nonaffine discrete systems with non-repetitive disturbances.Firstly,a dynamic linearization method along the iteration axis direction is introduced to convert the repetitive nonlinear system into a linear system,where the uncertain terms and disturbance terms are treated as a total disturbance.Then,an RBFNNDO is designed to estimate the total disturbance and applied to the controller design.Finally,a performance index function is proposed to design the control law with the ability to learn along the iteration direction.Through rigorous theoretical analysis,it is proved that the proposed improved iterative learning control method has consistent and ultimately bounded tracking error.