论文部分内容阅读
根据南京炼油厂所提供的柴油调和凝点和冷滤点基础数据,用人工神经网络(ANN)的反向传播BP算法对凝点和冷滤点进行预测。提出了适宜的人工神经网络拓扑结构,讨论了BP算法中学习速率、动量系数及过拟合现象对网络的影响,通过实验数据的检验,证明了用ANN方法建立的柴油调和模型能有效地给出预测信息。研究表明,ANN方法比常用的调和系数模型、凝点指数模型、凝点换算因子模型等更能准确地关联和预报调和柴油的凝点和冷滤点。
According to the data of diesel reconcile freezing point and cold filter point provided by Nanjing Refinery, the artificial neural network (BP) backpropagation BP algorithm was used to predict the freezing point and the cold filter point. The suitable artificial neural network topology is proposed. The influence of learning rate, momentum coefficient and overfitting on the network is discussed. The experimental data show that the diesel blending model established by ANN method can effectively Forecast information. The research shows that the ANN method can correlate and predict the condensing point and the cold filter point of reconciling diesel oil more accurately than the common harmonic coefficient model, the freezing point index model and the freezing point conversion factor model.