不稳定试井分析中参数识别的神经网络方法

在不稳定试井分析中,调整试井解释参数往往花费解释人员大量的时间。特别,当试井解释参数较多时,进行参数调整就更为困难,有时得不到一个合理的参数识别结果。因此,研究新的试井解释参数识别方法势在必行。目前,典型曲线的自动拟合方法是其研究成果之一,但由于数值计算方法的局限,使得该方法难于推广。文中研究了基于神经网络的系统辨识方法在不稳定试井分析参数识别中的应用。通过神经网络对一实际的油气藏系统进行建模和辨识,从而由新的神经网络模型可以获得参数识别结果。着重讨论了均质地层和双重介质地层的压力不稳定测试的参数识别问题,一个实例的分析显示了该识别算法的特性。

On the Neural Network Method of the Parameter Identification in Transient Well Testing

In the well transient testion, it costs a lot of time for people to adjust well testion parameters. Especially, it is much more difficult when many parameters need to be adjusted. Sometimes, people can not get a reasonable results. So,it is imperative to study some new parameter identification methods. Up to mow, the automatic matching method of type curves has been studied by many researchers. But it is difficult to be used widely because of the limit of numerical calculation method. In this paper, system identification method based on neural network in parameter identification of transient testing is presented. After a practice oil or gas reservoir is modeled and identified by a neural network, the results of parameter identification are obtained by the new neural network model. The problem of parameter identification in transient testing of single porous and doube porous reservoirs are studied especially. One example illustrated the performance of the identification algorithm.