基于卷积神经网络的原油组分光谱分析研究

The Crude Oil Components Spectral Analysis Based on Convolution Neural Network

由于近红外光谱能实现快速实时检测,使其成为研究井下原油组分和性质的有效方法。针对传统方法在处理光谱时存在的不足,提出了基于卷积神经网络的原油组分光谱分析方法。该方法用卷积神经网络替代传统的预处理方法,并将预处理步骤融入模型当中,达到在训练过程中综合优化的目的。实验结果表明,基于卷积神经网络的原油组分光谱分析方法的预测效果普遍优于传统方法,具有较好的应用价值。

Due to fast real-time detection character,near-infrared spectroscopy has become an effective method for studying the composition and properties of oil products.In view of the shortcomings of traditional methods in dealing with spectra,this paper presents a method for quantitative analysis of crude oil light hydrocarbons based on convolutional neural networks.This method replaces the traditional preprocessing method with a convolutional neural network,and integrates the preprocessing steps into the model to achieve the purpose of comprehensive optimization in the training process.The experimental results show that the method proposed in this paper is generally better than traditional method in the prediction of light hydrocarbon,and has good practical value.