基于多基因遗传编程的凝析气藏露点压力估算方法

Calculation method of dew point pressure of condensate gas reservoir based on multigene genetic programming

黄鑫,张芮菡,张烈辉,鲁友常,袁山,敬晓锋

油气藏地质及开发工程国家重点实验室，四川 成都 610500 四川页岩气勘探开发有限责任公司，四川 成都 610041 中国石油西南油气田公司勘探开发研究院，四川 成都 610041 中国石油玉门油田分公司勘探开发研究院，甘肃 酒泉735000

为准确预测凝析气藏露点压力，降低反凝析现象对凝析气藏开发带来的伤害，文中基于多基因遗传编程理论，选取储层温度、流体组分和露点压力为变量，建立模型数据库，模拟自然界物种进化机制，使用矿场实际数据演化生成非线性显式露点压力预测模型。研究结果表明：使用27组实测数据组建模型数据库演化生成的预测模型种群中，有13个非支配模型解，最优非支配模型解在模型复杂度和预测精度上均具有良好的适应性表现，模型结构紧凑，物理意义明确。经15组矿场实际数据验证，最大预测误差为7.99%，最小预测误差为1.15%，平均预测误差为4.36%，表现出良好的泛化能力和鲁棒性，为快速高效估算凝析气藏露点压力提供了一种有效方法。

In order to accurately predict the dew point pressure of condensate gas reservoir and reduce the damage caused by the reverse condensate to the condensate gas reservoir development, based on multigene genetic programming theory in this paper, the model database is established by selecting reservoir temperature, fluid composition and dew point pressure as variables. By simulating the biological evolution mechanism of nature, the nonlinear explicit dew point pressure prediction model is generated by using the actual field data evolution to carry out the research on dew point pressure prediction. The research results show that there are 13 non dominated model solutions in prediction models generated and evoluted by 27 groups of actual field data. The optimal non dominated model solution has good adaptability in model complexity and prediction accuracy, and the model structure is compact and the physical meaning is clear. By the verification of 15 groups of field data, the maximum prediction error is 7.99%, the minimum prediction error is 1.15%, the average prediction error is 4.36%, which show good generalization and robustness. The research provides an effective method for quickly and efficiently calculating dew point pressure of condensate gas reservoir.

10.6056/dkyqt202106021