Research On Networked Control Systems Based On Predictive Control Algorithm

With the rapid development of society, controlled objects are becomingincreasingly complicated and the traditional point to point control has been unable tofulfill daily production needs. As the network technology and control theory continuesto evolve and optimize, NCSs(Networked Control Systems) came into being.Compared with other traditional control, networked control systems has beensuccessfully used in industrial fields with less lines, flexible structures, strongreliabilities. But there must have a few new control problems as the same time, thispaper mainly researched on random time-delay in networked control systems deep.An improved GPC(Generalized Predictive Control) algorithm based ontraditional generalized predictive control was proposed in this paper, which inheritedthe basic idea of prediction model, rolling optimization, feedback correction. Theimproved algorithm induced the minimum variance output prediction to avoid solvingDiophantine equations and save a lot of computing time. The networked controlsystems could get higher prediction accuracy, better dynamic and static performanceand superior control effect by this improved algorithm.Considering the actual production process, the parameters of the controlledobject are often unknown, even existed the harsh conditions of parameters change.The paper adopted a control strategy, which combined recursive least squares methodwith improved generalized predictive control to deal with this problem. For thesystems with unknown parameters, use the least squares method with forgetting factorto identify system parameters online firstly, adopt the improved generalized predictivecontrol algorithm to get an effective predictive control performance. For the systemswith changed parameters, use the least squares method with variable forgetting factorto identify changed parameters online, and then combined with the improved GPC tocontrol the system better.Since generalized predictive control algorithm do well in linear systems, andcould not obtain a good control effect in strong nonlinear systems. Therefore, thepaper proposed a control strategy, which combined neural network with generalizedpredictive control to control the nonlinear systems. Firstly, indentify the nonlinearsystems with Elman neural network and establish the predictive model, and thencombined with rolling optimization and feedback correction of generalized predictivecontrol to make system further predictive control.To further validate the researching work, the paper gave the correspondingsimulation analysis for the proposed algorithms. The simulation analysis showed thatthe improved GPC algorithm can not only save a lot of computing time, but also getbetter control effect compared with the traditional generalized predictive control algorithm. The simulation of the control strategy, which corresponding combinedrecursive least squares method with improved generalized predictive control, showedthat the systems could get a good control performance in the condition of unknownparameters or changed parameters. Finally, the nonlinear networked control systemfor simulation analysis, simulation results show that the neural network combinedwith generalized predictive control could build predictive model effectively andnonlinear networked control systems could achieve effective control.