基于神经网络和灰色模型的非线性预估

以某己内酰胺厂磷酸羟胺(HPO)的制备的现场数据为基础,利用贝叶斯正则化神经网络和灰色模型建立了磷酸羟胺中的H+浓度的预测模型;比较了神经网络和灰色模型的差异,并把两者结合起来,建立模型进行预测。最后验证了用神经网络和灰色模型相结合建立起来的磷酸羟胺模型可以迅速有效的预测信息,从而为实现质量指标的实时预估和获取专家系统知识奠定了基础。

State Prediction on Nonlinear System Based on Neural network and Grey model

Based on the local datas from the preparation of the HPO in the hexanolactam factory. The predictive models using Bayesian regularization neural network and grey model are established for the H+ concentration in the HPO, two kinds of model are compared and associated to establish model and to predict. At last it proves that the proposed predictive models can predict effective information on the HPO model efficiently and rapidly. So it proves a method for the quality predict and getting the knowledge of the expert systerm.