论文部分内容阅读
提出反映炼油厂分馏装置动态特性的前馈神经网络模型 ,根据工厂的生产实际及数据特点建立了一种基于时间序列的、适合油品质量指标监测的动态系统前馈神经网络 (DBPNN)结构 .通过用实验室模拟的动态过程数据和炼油厂分馏装置的生产数据分别建模并与传统静态前馈神经网络模型比较 ,结果表明 ,DBPNN模型能够反映动态过程的特性 ,并具有更高的可靠性和适应性 .
A feedforward neural network model is proposed to reflect the dynamic characteristics of the fractionation plant. A dynamic feedforward neural network (DBPNN) structure based on time series and suitable for oil quality monitoring is established according to the actual production and data characteristics of the plant. By modeling the dynamic process data in laboratory and the production data in refinery fractionation plant and comparing with the traditional static feedforward neural network model, the results show that DBPNN model can reflect the characteristics of dynamic process and has higher reliability And adaptability.