基于双隐层动态递归神经网络的航煤比重软测量

Soft-sensor of Aviation Kerosene Specific Gravity based on DRNN with Double Hidden Layers

针对原油蒸馏装置常压塔航煤比重模型具有动态特性的特点 ,提出采用双隐层动态递归神经网络 (DRNN)实现比重的软测量 ,推导了双隐层 DRNN的权值学习算法 ,并利用在线比重分析仪构成了航煤比重软测量模型的在线校正。在某炼油厂常压塔装置实际投用表明 ,基于双隐层 DRNN比重软测量模型具有较高的测量精度。

In accordance with that the model of aviation kerosene specific gravity(AKSP) in atmospheric distillation tower exists dynamic characteristic,the soft sensor model of AKSP based on diagonal recurrent neural network(DRNN)with double hidden layers is proposed in this paper The weighting learning algorithm of DRNN with double hidden layers has been deduced,and on line correction method of soft sensor based on analysis value of AKSP has also been designed The soft sensor model has been applied on an atmospheric distillation tower successfully,and the measurement precision is better than soft sensor model based on BP network