Predictive Control Of Networked Control Systems

With the development of computer, network, communication and control theory, networked control systems (NCS) are getting more applications in industry control field. There are a lot of questions to be studied because of the introduction of network in the closed-loop system. It makes the system analysis and synthesis difficult due to the existence of stochastic delay in the NCS, the system performance goes to bad and unstable if the delay long enough. Based on mechanism of multiple-step prediction, rolling optimization and feedback correction, predictive control can overcome the model uncertainty and have good control performance, strong robustness, so it is widely applied and studied. Based on short delay and long delay of the NCS using predictive control in systems analysis and design, sereral new control strategies are proposed, the system stability is also investigated in this dissertation. The main research contents of this paper are as follows:Delay prediction internal model control strategy is proposed in NCS which sensor is time-drived, controller and actuator are event-driven and the delay is less than one smpling period. This method is comprised of time-stamped BP neural network delay predictor, the historical measured delay data was used to predict the next period delay value, according to which the NCS with internal model control is designed. The delay loop is described by one-order Pade expended method approximately. The delay prediction value effect on the system is also analyzed. Simulation results illustrate that the precision of delay prediction is guaranteed, the system performance are affected by the filter coefficientαand delay prediction precision, the performance of NCS is greatly improved.In the practical engineering system, the choice of sampling frequency is decided by the system signal's top frequency and the network communication delays rely on the network structure. Consequently, there is no corresponding relation between the network transmitted delay and sampling period, nevertheless most papers about NCS usually assume the network delay is less than a sampling period. To avoid the illogical premise, we firstly eatabilished a model of a class of long delay NCS, then two novel predictive control strategies of long delay NCS are put forword. The first one is long delay NCS predivtive functional control algorithm by this algorithm derives the predictive control law of the long delay NCS based on the model of long delay NCS. The simulation results and experiment of an electrical-resistance stove realtime control illustrate that the proposed scheme is effective for a class of long delay NCS. The other one is long delay NCS dynamic matrix control algorithm, this method derives the long delay NCS's control law by using dynamic matrix control theory which contains properties of simple algorithm, less calculation, strong robustness. The simulation results illustrate that the proposed method is effective for a class of long delay NCS.Because of the stochastic delay in the NCS, the system performance goes to bad and even unstable if the delay long enough, the effect of delay on internal model networked control systems (IMNCS) performance is analyzed. The predictive control algorithms include generalized predictive control, model predictive control, dynamic matrix control etc can be equivalently turn over to generalized internal model control, stabilities of continuous IMNCS and discrete IMNCS are studied respectively. The IMNCS stability was discussed by using Routh criterion, state space Lyapunov stability theory and Z plane synthesis (Jury stability criterion) , therefor the maximum allowed time-delay that keep IMNCS stable was presented. In order to make better performance of IMNCS, the delay prediction loop was added and the system stability influenced by prediction was pointed. Simulation results are illustrated that the proposed scheme is very effective.