含抑制剂体系下天然气水合物形成预测模型的建立与应用

油气水混相在开采与运输过程中易生成天然气水合物,预测其形成条件可保障油气田的安全运行。建立了包含超过1500个数据点的广泛数据库和基于PSO-LSSVM的含抑制剂天然气水合物形成解离模型,与BP神经网络模型模拟结果进行了对比分析,并在荔湾气田现场管道进行了预测应用。结果表明,LSSVM模型通过PSO算法优化后确定的正则化参数为321.1523642521、核参数为1.0210368871948。BP神经网络模型和PSO-LSSVM模型的平均相对误差分别为1.92%和0.49%,说明PSO-LSSVM模型的预测能力更强。以纯水为基础介质时,BP神经网络模型和PSO-LSSVM模型均可较准确地预测水合物形成,但以含热力学抑制剂水溶液为基础介质时,PSO-LSSVM模型具有更优秀的表现。PSO-LSSVM模型对于荔湾气田现场生产具有较强的适用性,可为甲醇注入量的确定和现场安全运行策略的制定提供理论依据。

Establishment and application of prediction model of natural gas hydrate formation in inhibitor-containing system

Natural gas hydrate is easily formed in the process of exploitation and transportation of oil-gas-water miscible phase,and the prediction of its formation conditions can ensure the safe operation of oil and gas fields.An extensive database of over 1500 data points and a PSO-LSSVM-based gas hydrate containing inhibitors formation and dissociation model were established.The simulation results were compared with BP Neral Network model and predictably applied in Liwan gas field pipeline.The results show that the regularization parameter of LSSVM model optimized by PSO algorithm is 321.1523642521 and the kernel parameter is 1.0210368871948.The mean absolute relative errors of the BP Neral Network model and the PSO-LSSVM model are 1.92%and 0.49%,respectively,indicating that the PSO-LSSVM model has stronger predictive ability.When pure water is used as the basic medium,both the BP Neral Network model and the PSO-LSSVM model can predict hydrate formation accurately.However,when the aqueous solution containing thermodynamic inhibitors is used as the basic medium,the PSO-LSSVM model has better performance.The PSO-LSSVM model has strong applicability for on-site production in Liwan gas field,and can provide a theoretical basis for the determination of methanol injection volume and the formulation of on-site safe operation strategies.