The Application Of GA-RBF Network Generalized Predictive Control Strategy In Circulating Fluidized Bed Unit

Currently, in the thermal power industry, pulverized coal boiler is always dominant.However, the use of clean energy is an inevitable requirement and trend for socialdevelopment. Thus the circulating fluidized bed boiler (CFB) combustion technologyemerges as the times require. CFB has the following advantages due to its distinctivecombustion, such as the wide fuel adaptability, high combustion efficiency and wide loadadjustment range. Unlike the coal boiler, the coordinated control system still needs to addthe bed temperature control, and only in this way can it ensure the economical and safeoperation of the unit. In this article, it takes the300MW CFB’s bed temperature controlsystem and coordinated control system as a whole to study.The bed temperature and coordinated control of the large-scale circulating fluidizedbed boiler unit have the features of complex, nonlinear, large time delay and difficult toestablish accurate model, etc. The conventional control strategy is no longer meeting therequest of the electric network for design and the quality of control to the bedtemperature and the Coordinated Control System of the CFB unit. Therefore, on the basisof referring to a lot of literature, the article adopts the generalized predictive controlstrategy of RBF neural network. The essence is to identify model with RBF neuralnetwork optimized through genetic algorithm, and using multivariable generalizedpredictive control strategy to achieve predictive control. RBF network determines hiddenlayer centers randomly, and make use of the global search capability of geneticalgorithms to achieve the optimal neural network. This algorithm not only increases theprecision and speed of model identification, but also avoids online solving ofDiophantine equation which greatly reduces the online calculating amount of generalizedpredictive control algorithm, and enhances the real-time requirements of GPC when itapplied to this system.Finally, we take the simulation study of the bed temperature and coordinated controlsystem of300MW CFB unit and the result shows that this algorithm can overcome theadverse influence of the time delay on control system. This algorithm not only ensuresthat the output power can track its setting quickly and smoothly, but also maintains themain steam pressure and bed temperature in the range of reasonable. When the CFB unitworking condition is changed greatly, this algorithm still maintains good control performance. Compared with the smith-PID control strategy, the algorithm has goodadaptability and robustness. It provides an effective way to solve bed temperature andcoordinated control problem of large CFB unit.