多模型预测控制及其在火电厂过热汽温控制中的应用

针对火电厂存在的过热汽温问题,设计了多模型预测控制系统.根据若干建模工况点,离线训练局部人工神经网络模型,利用贝叶斯估计的方法在线计算每个局部神经网络模型概率,加权计算出模型预测输出值.根据预测控制的原理,利用Newton-Raphson迭代法得到控制信号,从而得到了仅含一个控制器的多模型预测控制系统.仿真结果表明,在负荷大范围变化的工况下仍能保持良好的控制性能,具有较强的鲁棒性.

Multiple Model Predictive Control and Its Application to Control of Super-heated Steam

Based on the characteristics of super-heated steam of a thermal power plant, a multiple model predictive control system is designed. The local artificial neural network models are trained offline based on different operating points. A Bayesian estimation method is introduced to compute the probabilities of local artificial neural network models online. Then, the model predictive output is synthetized by weighting. According to the predictive theory, the Newton- raphson iterated method is used to obtain the final control law. Hence, the multiple model predictive control system is established where only one controller is needed. The simulation shows that the control system of super-heated steam temperature can maintain a better co-performance under a larger range of load change while a strong robustness is kept.