基于人工神经网络的凝析气藏动态预测

Performance prediction of condensate gas reservoir using artificial neural networks

常规的动态预测方法因适应的开发阶段和范围不同,在应用过程中有其局限性。BP人工网络则能克服这些缺点,不仅能描述油田开发的整个过程,而且还可以考虑单一变量和多变量影响因素,把能影响动态预测指标的各种因素自行组织起来,加以训练和学习,建立起广义的、精确的动态预侧模型。针对凝析气藏的开发,提出一种基于BP人工神经网络的凝析气藏产油量和产气量的动态预报方法,该方法对气藏开发过程的时变性和各种随机干扰因素具有自适应性。基于BP人工神经网络的模型,设计了BP网络算法的计算机实现的流程。在W indows XP环境下,采用面向对象的编程方法,以V isual C++.NET为编程语言成功开发了凝析气藏的单井动态产量预测软件。进行了牙哈凝析气藏2口井的产油量和产气量的预测对比,结果表明,模型预测值与实际生产值具有很好的一致性,该预测软件具有较高的预测精度和可靠性,适合于牙哈凝析气藏各个阶段的产油量、产气量的动态预报,具有良好的推广价值。

Deficiencies of conventional performance prediction methods emerged during the process of application owning to development stages and ranges variation. However, BP neural networks, which may overcome that restriction can not only represent the whole development process, but also include signal variable and multiple variables that may influence the prediction. It could self - organize all sorts of factors that have an effect on the behavior prediction, and establish comprehensive accurate model for dynamic prediction through training and researching. Based on the BP neural networks, a new method of performance prediction of oil and gas production in a condensate gas reservoir is presented. The method is adaptive to various random factors in the process of oilfield exploitation. Computerized flow chart was designed for BP neural calculation. Under the operation environment of Windows XP, single well dynamic performance prediction software was constructed by using the Object - Oriented program method on Visual C ＋ ＋. NET. The method has been applied to 2 wells in Yaha condensate gas reservoir for oil and gas production prediction, results showed that the forecast coincided with real production. The prediction software is accurate and reliable and can be used for oil and gas production prediction during different stages in Yaha condensate gas reservoir.