A Predictive Control Approach For The Design Of Networked Control Systems

With the wide application of the computers in industrial control systems, the information networking has brought the original automatic control technology unprecedented challenges, network-induced delays is one of the most important aspects. Time delays may severely degrade the performance of real-time control systems, and it may even cause the instability of the systems. Therefore, how to reduce the effects of the network-induced delay on the system performance is one of the most urgent problems that need to be solved.Based on the predictive control theory, network-induced delays are considered in this thesis. The main works and outcomes of this thesis are as follows:Firstly, predictive control theory based compensation for the delays of non-integral multiple of sampling period is studied. Suppose that the sensor is time driven, the controller is event driven and the actuator is also time driven, but with a smaller period of time than the sensor's. The control sequences of non-integral multiples of sampling period can be derived by using linear interpolation method. Numerical simulation demonstrates the effectiveness of the proposed method.Secondly, the compensation for the systems with model uncertainties is studied. The effects of the network-induced delay are modeled as system uncertainties, in order to solve this problem, two methods are used to improve the predictive control strategy, one of them is to add a low pass filter before the controller, and the other one is using a BP (Back-Propagation) neural network. Meanwhile, the sensor and the actuator still work in the same mode as described in the previous chapter. Contrast of the control performances of those three methods is derived via simulation.Finally, the veracity of the control signals of the predictive-based time delay compensation is studied. All kinds of possibilities of the forward time delay is preset at the controller site, with that delay, the actual output of the system is calculated by using the built-in model. The control sequences are then calculated by those outputs. Numerical simulations demonstrate the effectiveness of this method.