凝析气藏的高效开发需要准确的流体相态性质数据,其中准确预测露点压力是凝析气藏开发过程中的重要问题。针对凝析气藏露点压力传统预测方法精度较低的问题,基于最优化理论和应用统计分析,通过实测数据资料拟合,提出了一种利用交替条件期望变换方法（ACE）确定的凝析气藏露点压力非参数回归模型,获取了具有统计意义的凝析气藏露点压力显式关联式。在皮尔逊（Pearson）关联性分析基础上,该模型自变量选取气藏温度、（C₁、C₂-C₆、C₇₊）摩尔分数、C₇₊相对分子质量、C₇₊相对密度。采用公开发表的27组露点压力数据探索自变量和因变量之间的潜在函数关系,并对9组TLM油田实测露点压力数据进行预测。结果表明：该模型精度较高,具有良好的泛化能力,模型回归的平均绝对相对误差（AARD）为2.16 %,模型预测AARD仅为4.8 %,其中最大绝对相对误差（ARD）为9.21 %,最小ARD为0.34 %,本研究为凝析气藏露点压力预测提供了一种参考方法。

The efficient development of condensate gas reservoirs requires accurate fluid phase properties data, among which accurate prediction of dew point pressure is an important issue in the development of condensate gas reservoirs. In order to solve the problem of low accuracy of traditional prediction methods for dew point pressure of condensate gas reservoirs, based on optimization theory and applied statistical analysis, and by fitting measured data, a non-parametric regression model determined by alternating conditional expectation transformation（ACE） is proposed, and an explicit correlation of dew point pressure with statistical significance is obtained. Based on Pearson correlation analysis, the independent variables of the model are gas reservoir temperature, mole fraction of （C₁, C₂-C₆, C₇₊）, and molecular weight and relative density of C₇₊. The potential function relation between independent and dependent variables is analyzed by 27 sets of experimental data for published dew point pressure, and 9 groups of measured dew point pressure data of TLM oilfields are predicted. The results show that the model has high precision and good generalization ability. The average absolute relative deviation（AARD） of model regression is 2.16 %, and the predicted AARD is only 4.8 %. The maximum absolute relative deviation（ARD） is 9.21 % and the minimum is 0.34 %. This study provides a reference method for dew point pressure prediction of condensate gas reservoirs.