摘要
石油、天然气开采、加工和运输过程中在一定温度和压力下形成的天然气水合物,会堵塞井筒、地层孔隙,从而影响油气井产能。由于在实验室或现场要对形成水合物的影响因素进行综合判定具有相当难度,故从影响水合物形成的相关因素出发,针对前人研究天然气水合物生成预测方法的优缺点,引入了具有解决复杂系统问题能力的人工神经网络(ANN)理论,运用Matlab(r)语言编程建立了一个包含6个神经元(CH4、C2H6、C3H8、C4+、其他气含量及压力)输入的三层向前BP网络模型,对天然气水合物生成条件进行了预测。采用引自不同文献的有关天然气水合物生成的实验室测试值建立适应本模型的样本,将经过学习后的神经网络模型运用于实际气田水合物预测,实验值与模型预测值具有很好的一致性,取得了令人满意的效果。实践证明该模型准确、可靠,具有良好的推广性。
Gas hydrate may appear under certain temperature and pressure in the course of oil/gas production, processing and transport. The formed hydrate may plug the borehole and formation and so influence the productivity of oil/gas well. Owing to much difficulty for estimating comprehensively the hydrate formation factors in lab or on site, the authors, after considering the hydrate formation factors and the merits and disadvantages of previous hydration prediction methods, use Matlab? computer language to establish a three-layer forward BP model incorporating six nerve cell inputs(CH_4, C_2H_6, C_3H_8, C_4+, other composition content and pressure)based on artificial neural network theory (ANN) to predict the generating conditions for gas hydrate. The model uses the lab-tested data of generating conditions of hydrate from different literature as the training sample. After learning by the network, actual gas hydrate generating conditions can be predicted based on six nerve cell inputs. As a result, the forecast value is in good consistency with lab value, showing satisfactory effects. The results indicate that the model is a reliable and accurate and suitable for popular application.
出处
《天然气工业》
EI
CAS
CSCD
北大核心
2006年第7期85-87,共3页
Natural Gas Industry
基金
中国石油天然气集团公司创新基金项目"天然气水合物相态行为及动力学研究"资助(编号:04E7047)。
关键词
天然气
水合物
神经网络
计算机程序
预测
natural gas, hydrate, artificial neural network, computer program, prediction
