Neural Networks Predictive Control For Superheated Steam Temperature Based On Modified Particle Swarm Optimization

The superheated steam temperature of boiler superheated system in fired power station is a typical object, which has nonlinear, uncertain, large delay and inertial characteristics. At present, general PID cascade control system is mainly adopted, but it can't work well when more interference come or cycles are changed. Neural network predictive control (NNPC) will take full advantage of the nonlinear, self-organizing, self-learning performance of neural networks and the rolling optimizing, feedback adjusting effectiveness of predictive control. It will be more suitable for this uncertain boiler superheated system control.In order to improve the performance of NNPC, by combining modified particle swarm optimization (MPSO) with NNPC, a model predictive controller based-on MPSO and radial basis function (RBF) hybrid optimization strategy (MPSO-RBF) is presented, also a non-linear optimization controller based-on MPSO is presented. In allusion to the superheated steam temperature control, a cascade control system based-on neural network predictive control is constructed, also its predictive model & rolling optimizing algorithm & feedback adjusting & simulation parameters setting problems are analyzed, at the same time particle encoded format of MPSO & operating design method and hybrid optimization algorithm step are presented. Finally the simulation experiments are done for superheated steam temperature system control of super critical 600MW direct current boiler, also this method will be a contrast to conventional NNPC, the simulation results have shown this method's validity and superior control performance index and applied foreground.