Sub-optimal Control Of Two-time-scale System Driven By Data And Model

Two-time-scale systems widely exist in complex industrial fields such as aerospace,electricity,chemistry,machinery and other practical industries.The existing research results on sub-optimal control of two-time-scale systems are generally based on mathematical models,which can not be used when it is difficult to establish accurate model.Adaptive dynamic programming(ADP)can be used to design optimal controllers for systems with unknown dynamics.However,it can not be directly applied to two-time-scale systems because of coupling of fast and slow states.Aiming at sloving the sub-optimal control problem of two-time-scale systems with incompletely known models,the quadratic regulator design problem(LQR)and the quadratic tracker design problem(LQT)of linear two-time-scale systems are studied by using model and data.Also,the plant-wide optimal control of complex industrial processes is studied represented by grinding processes.The main work and achievements are summarized as follows:1?A suboptimal control method for two-time-scale systems with partially known mathematical model is proposed.Firstly,the linear two-time scale system is decomposed into fast subsystem and slow subsystem.For the fast subsystem with known model,the optimal regulator is designed by solving the underlying Riccati equation.For the slow subsystem with unknown model,the state variable and performance index are transformed into the standard LQR problem by substitution of variables.According to singular perturbation theory,a suboptimal control of the original system is obtained.Secondly,the stability and sub-optimality of the system are theoretically analyzed.Finally,the effectiveness of the proposed method is verified by simulation.2?A tracking control method for two-time-scale systems with partially known mathematical model is proposed.Firstly,the linear two-time-scale system is decomposed fast subsystem and slow subsystem.For the fast subsystem with known model,the optimal regulator design based on model is adopted to ensure the stability of the system.For the slow subsystem with unknown model,a data-driven tracking controller is designed to track the reference trajectory.Using singular perturbation theory,a suboptimal control of the original system is designed.The suboptimal control enables the system output to track the reference output.Secondly,the stability and other performance of the system are analyzed.Finally,the effectiveness of the proposed method is verified by simulation.3?By the obtain results mentioned above,a plant-wide optimal control method of complex industrial processes is proposed.Firstly,the complex industrial process is transformed into sub-optimal control problem of two-time-scale system.Secondly,the data-driven ADP algorithm is implemented to solve the plant-wide optimal control problem of complex industrial process.Finally,the obtained method is applied to the grinding process to verify the feasibility and validity.