Model-Free Control Research Based On Cross-Entropy Algorithm

In practical systems, many factors lead to system disturbance. These disturbances may be caused by known time delay phenomenon, parameter time-varying characteristics, internal system coupling, etc., or stem from unknown factors such as signal noise and climate change. The existences of disturbances will seriously affect the precision of the system modeling, and bring a challenge to model-based control methods. Model-free control (MFC) algorithm, which works without complex mathematical model, could estimate system disturbance by using the input and output information, and make up for defects of model based control method in terms of disturbance rejecting. The contributions of this thesis are described as follows:1. Multi-variable time-delay systems widely exist in industrial plants. It is difficult to establish accurate mathematical model for the complex industrial plants. In addition, the existing time-delay needs to be pre-estimated. All these factors make model based control method for multi-variable time-delay systems hard to use. In this thesis, model-free control (MFC) strategy is applied to design multi-variable closed-loop system, and the improved cross-entropy algorithm is employed to tune controller parameters. The simulation result shows that the proposed control strategy could cope with multi-variable time-delay problem with good robustness and disturbance rejection.2. Model-free control can pre-estimate system disturbance by using online estimation, and improve the disturbance rejecting level of the system. Via tuning error parameters a and PID parameters, the model-free control system can achieve higher control accuracy. According to the properties of the fractional order calculus, this thesis investigates the model-free control problem of fractional order system and propose the fractional order model-free control (FOMFC) strategy. Compared with the traditional MFC scheme, FOMFC possesses more tuning parameters and increases the degree of freedom of controller design. The simulation result demonstrates that the proposed method improves the dynamic performance of control system and ensures the control quality while the disturbance resistance is remained.3. The control system design of boiler drum system is a typical multi-objective optimization problem. On the one hand, the stability of the boiler liquid level need to be guaranteed. On the other hand, the supply should be controlled strictly. Meanwhile, the boiler system possesses the characteristics of false liquid level and time-varying parameters, which increase the difficulty of control. This thesis investigate the problem of multi-objectives optimization by taking the boiler drum system as an example. A multi-objective cross-entropy algorithm (MOCE) is deduced from the single objective cross-entropy algorithm by using frequency histogram and Pareto sorting method. Furthermore, the proposed algorithm is verified via a standard multi-objective problem (MOO). The simulation result shows that MOCE can be applied to optimize the multi-objective problem and improve the performance of the control system.