CAJ | 학술논문

提出了一种基于人工神经网络模型预测高含CO2天然气的含水量的新方法。网络输入变量CO2摩尔分数、温度、压力，网络的输出为高含CO2天然气的含水量。该人工神经网络模型能够估算温度在20.0～200.0℃，压力在0.1～70.0 MPa，CO2摩尔分数高达70%天然气中水蒸汽的含量。对比文中建立的人工神经网络模型和目前常用的3种预测高含CO2天然气的含水量的经验模型，结果表明，人工神经网络的平均相对误差值最小，为1.275%，3种经验模型在CO2含量较高时，预测精度较低。这就表明，人工神经网络模型在预测高含CO2天然气含水量时，比3种常用的经验模型更具有优势。
제출료일충기우인공신경망락모형예측고함CO2천연기적함수량적신방법。망락수입변량CO2마이분수、온도、압력，망락적수출위고함CO2천연기적함수량。해인공신경망락모형능구고산온도재20.0～200.0℃，압력재0.1～70.0 MPa，CO2마이분수고체70%천연기중수증기적함량。대비문중건립적인공신경망락모형화목전상용적3충예측고함CO2천연기적함수량적경험모형，결과표명，인공신경망락적평균상대오차치최소，위1.275%，3충경험모형재CO2함량교고시，예측정도교저。저취표명，인공신경망락모형재예측고함CO2천연기함수량시，비3충상용적경험모형경구유우세。
In this paper,a new method based on artificial neural network(ANN)for prediction of natural gas mixture water content is presented. CO2 mole fraction,temperature,and pressure have been input variables of the network and water content has been set as network output. The proposed ANN model is able to estimate water content as a function of CO2 composition up to 70%,temperature between 20.0～200.0℃and pressure from 0.1 to 70.0 MPa. Comparisons show average absolute relative error equal to 1.275%between ANN estimations and experimental data,which is smaller than the other three commonly used empirical correlations. Furthermore,there is considerable deviation between experimental data and the other three commonly used empirical correlations for prediction of high CO2 content natural gas water content. But artificial neural network has good prediction results in high CO2 content natural gas. Results show ANN superiority to the common three correlations in literatures.