基于B-P神经网络的非线性预测控制

Nonlinear Predictive Control Based on B-P Network Model of Sugar Crystallization in Batch Pan

针对用一般的方法控制糖液的过饱和度很难奏效的问题,运用B-P神经网络构造预测模型,将模拟退火算法的局部搜索与遗传算法的全局搜索相结合,进行在线滚动优化,对煮糖结晶过程中的过饱和度进行预测控制。实际运行结果表明,基于B-P神经网络的预测控制算法响应速度快、控制精度高、鲁棒性强,具有很强的实用性。

In the productive process of boiling in cane sugar crystallization, that is a complicated chemical and physical process, and both heat transfer and mass transfer occur, the control of oversaturation is a basic factor in boiling crystallization process, Which can not be solved by ordinary method. A nonlinear predictive control algorithm based on B-P network is employed in the control of oversaturation. In this algorithm, optimization problem is solved by implementing genetic simulated annealing algorithm. The running result shows that the algorithm possesses such advantages as good robustness, quick response, high control accuracy, practicability.