BP神经网络在气井储量计算及动态分析中的应用

APPLICATION OF BP NEURAL NETWORK TO THE RESERVE ESTIMATION AND PERFORMANCE ANALYSIS OF GAS WELL

将油（气）藏看作一个系统，而将产量认为是对系统的激励信号（输入信号），那么压力则为系统的输出信号。输入信号和输出信号之间的关系，隐含着油（气）藏系统内在的信息，如储层和油（气）藏结构、单井控制面积以及井的状况等综合信息。将气井生产记录资料进行适当的统计作为模式特征，输入ＢＰ神经网络，通过向模式学习，ＢＰ网络便可将生产资料所隐含的信息以权矩阵的形式记录下来。应用ＢＰ网络做动态分析和储量计算，以气井日常生产记录资料为基础，计算单井动态储量，并预测气井未来的生产动态。实例证明该方法可行。该方法的应用为利用大量的井口生产记录做了有益的探索。

By taking the oil (gas) reservoir as a system,the production and pressure of gas well can be thought as the input signal and output signal of the system respectively.The relation between the input and output signals conceals the information inherent in the oil (gas) reservoir system,such as reservoir bed structure,oil (gas) reservoir structure,single well controlled area and hole condition,etc.By taking the statistical daily production data of gas well as the pattern attributes into the BP neural network,the concealed information in the production data will be recorded in the form of weight matrix by the BP neural network through learning from the pattern attributes.Based on the daily production data of gas well,the single well controlled reserves may be estimated and the future gas well performance can be predicted by use of the BP neural network.It is proved by two examples that the method is feasible.Therefore a beneficial probing of applying a great deal of daily production data to the reserve estimation and performance analysis has been done.