基于数字孪生的生产分离器分离效果混合模型

A Hybrid Model for Separating Effect of Production Separator Based on Digital Twin

生产分离器作为海上油气生产平台的关键处理设备,其处理效果直接影响油田生产能力。由于分离器内部流动特性及化学药剂作用机理复杂,现有机理模型难以对生产分离器实际处理效果进行准确评估。为此,基于机理与数据融合的数字孪生理念,提出了适用于不同原油含水工况的生产分离器实际分离效果混合模型构建方法。对于低含水原油,提出了基于试验数据的卷积神经网络模型构建方法,并通过多种数据驱模型计算结果比选,验证该模型在表征多因素与分离效果之间关系的有效性;针对实验室乳化效果不佳的高含水原油,集成重力分离机理与代理模型建立了分离效果评估模型,该模型将原油乳状液的液滴尺寸分布、破乳剂质量分数及含水体积分数相关的黏温特性对分离效果的影响考虑在内,通过蒙特卡罗模拟表征模型参数及工艺参数的不确定性对分离效果的影响。将建立的混合模型应用于实际海上生产分离器,通过与现场化验数据对比及多工况分析对模型的有效性进行验证。结果表明,构建的混合模型能够实现对海上油气生产平台各级分离器实际处理效果的在线孪生。该模型有效解决了数据驱模型对于高含水原油分离效果预测缺乏泛化性的问题;克服了传统计算流体力学模拟效率低、模型偏差无法修正的缺陷。该混合模型能够用于分析生产分离器在多种工况下的实际分离效果,可为海上生产平台实际处理能力估计及生产参数调优提供指导。

Production separator is a key processing equipment on the offshore oil and gas production platforms,and its processing effect directly affects the oilfield production capacity.Due to the complex internal flow characteristics and chemicals active mechanisms of the separator,the existing mechanism models are difficult to accurately evaluate the actual processing effect of the production separator.Based on the digital twin concept of mechanism and data fusion,a construction method of hybrid model for the actual separating effect of production separator suitable for different water cut conditions of crude oil was proposed.For low water cut crude oil,a convolutional neural network model construction method based on test data was proposed,and the effectiveness of the model in characterizing the relationship between multiple factors and separating effect was verified by comparing the calculation results of multiple data driven models.For high water cut crude oil with poor emulsification performance in the laboratory,a separating effect evaluation model was built by integrating gravity separation mechanism and surrogate model;the evaluation model took into account the influence of viscosity temperature characteristics related to droplet size distribution,demulsifier mass fraction and volume fraction of water on separating effect,and characterized the influence of uncertainty of model and process parameters on separating effect through Monte Carlo simulation.The established hybrid model was applied to actual offshore production separator,and its effectiveness was verified through comparison with field assaying data and multiple working conditions analysis.The results show that the constructed hybrid model can achieve online twinning on the actual processing effect of all stages of separators on offshore oil and gas production platforms.It effectively solves the problem of lack of generalization in predicting the separating effect of high water cut crude oil by data driven models,and overcomes the shortcomings of traditional computational fluid dynamics simulation such as low efficiency and inability to correct model deviations.It can also be used to analyze the actual separating effect of production separator under various working conditions,providing guidance for estimating the actual processing capacity and optimizing the production parameters of offshore production platforms.