Research On Time-delay Compensation And Scheduling Methods For Networked Control System

With the development of computer, network, communication technology and control theory, networked control system is more widely used in the field of industrial control. With network used in the closed-loop control, the control system inevitably causes network-induced delay, packet loss and other problems, which makes the research methods of the traditional control system difficult to be applied directly to the networked control system. Thus it is urgent to study how existing control methods can be applied to the networked control system, as well as to develop new ones based on network-based control. Meanwhile, due to the limited resources of network bandwidth, it is also crucial to study how to reasonably allocate resources of network to improve network utilization and performance of the control system through the appropriate scheduling methods. Therefore, the performance of the networked control system not only depends on the merits of the control strategy also relies on the merits of network scheduling policy, both merits of which are associated and indispensable. In this paper, the following aspects are mainly studied:After analyzing the composition and characteristics of time-delay of the networked control system, and after studying the impact of the time-delay on the performance of the networked control system through simulation, a set of delay testing software has been developed, which have been used to analyze and process a large number of samples of time-delay testing data. As a result, time-delay data after pretreatment have better reliability.To deal with the uncertainty of network time-delay, two time-delay prediction methods are proposed:one method based on EMD and LSSVM of small data samples and the other based on Lyapunov-Elman of large data samples. The simulation experiments show that both of the prediction methods have higher precision and predicted results, and that they lay a good foundation for the time-delay compensation.Generalized predictive control algorithm has self-adaptability and strong robust and multiple-step prediction function, but the generalized predictive control needs to work out Diophantine equations and matrix inverse operation, so the computing speed of the algorithm is reduced. Aiming at this problem, a combination of time-delay prediction and fast implicit generalized predictive control delay compensation method is put forward. This method doesn't need to work out Diophantine equations and matrix inverse operation, so the computing of the algorithm has been sped up. Simulation results show that the method has a good compensation effect. Meanwhile, for nonlinear controlled object, predictive control time-delay compensation method is proposed by using an improved free search(IFS) algorithm to optimize nonlinear system least squares support vector machines. This method uses the least squares support vector machines to establish nonlinear system model and predicts the output of the system, and reduces the prediction error by output feedback and offset calibration. Also control variable of nonlinear systems is obtained by IFS rolling optimization. The method can design effective predictive controller for nonlinear systems. The single-input single-output, multi-input multi-output nonlinear system simulation proves the predictive control method is effective and has good adaptability and robustness. A better control effect has also been achieved by applying the method to the networked control system with random time-delay.For a multi-loop resource constrained networked control system, a fuzzy dynamical scheduling method is presented. First different weight coefficients of the control loop is given through a quadratic function mapping dynamic weight adjustment strategy. Next, considering the change of output and output error rate in the system, an improved fuzzy feedback priority scheduling strategy was proposed and a fuzzy feedback scheduler is designed, the priority of each control loop is adjusted dynamically. Finally, True Time toolbox is used to model the networked control system which contains dynamic weight adjustment fuzzy feedback scheduling, and compares it both with EDF and fixed weight fuzzy feedback scheduling method. The simulation results show that the method in the paper causes a smaller time-delay and better control performance under the same network bandwidth usage.An improved variable sampling period algorithm based on the prediction and feedback control of networked control system is proposed. The paper firstly proposes a kind of dynamic weight adjustment strategy of control loop based on transmission error, and then uses the least squares support vector machines algorithm to predict the packet transmission time. Finally, under certain conditions of the network bandwidth, changing system sampling period can constantly optimize the performance of the network control system by network utilization, packet transmission time and dynamic weight coefficient. Simulation results show that this method can make the network utilization fast convergence to the set value. Relatively fixed weight variable sampling scheduling algorithm, the method improves the performance of the output response of control loop, reducing the IAE of control loop. Combined with the fuzzy feedback priority scheduling strategy, a scheduling method based on variable sampling period and fuzzy feedback is proposed, and is verified by simulation.With the analysis based on the overall time-delay of the networked control system for Industrial Ethernet made, switched Ethernet is chosen as the main research field. For IEEE 802.1p priority scheduling switch-based internal queuing time-delay, scheduling improvement ideas is proposed and implementation program is given. A typical tree topology switched Ethernet model is constructed, the maximum delay of periodic real data frame in the respective transmission link between before and after scheduling improvement is calculated. At the same time, the basic condition for the scheduling improvement programs is deduced. Finally, the results of Network calculus and OPNET simulation show that the proper improvement of IEEE802.1p priority scheduling protocol can improve the real-time performance of hard real-time data transmitted in the switched Ethernet.