BP网络预测H_2在煤液化油混合组分中的溶解度

Solubility prediction of H_2 in mixed solvent of coal liquefaction oil by BP neural network

为满足煤制油工业化过程中设计和操作需要,以H_2在神华煤液化油模型组分混合溶剂中实测溶解度为基础,考察利用人工神经网络法预测H_2在该系统中溶解度的能力。结果表明,神经网络的计算精度随着循环次数的增加而提高;对于不同种类的混合溶剂,随着隐藏层个数的增加,计算值与试验值之间的相对误差呈现逐渐减小的趋势,从减小计算量的角度考虑,选定为4个隐藏层;3-4-1网络结构对于H_2在不同混合溶剂中溶解度的计算值与试验值最大相对误差为4.48%,这表明该模型能够满足H_2在该系统中溶解度的预测需要。

In order to meet the requirement of design and operation during coal oil industrialization,the practical H_2 solubility in the mixed solvent of Shenhua coal liquefaction oil was tested first,then the capacity of predicting H_2 solubility in system was investigated by artificial neural network.The results showed that the calculation precision of neural network increased with the increase of cling times. For different mixed solvent,the relative error between calculated value and experimental value gradually decreased with the increase of hidden layer quantity.To reduce calculation amount,the hidden layer quantity were set as four.The maximum relative error of 3-4-1 network for solubility of hydrogen in different mixed solvent was 4.48%.The model could meet the need of predicting solubility of H_2.