基于神经网络模型的致密气藏分段压裂井产能预测

致密气藏储层致密，开采难度较大，应用分段压裂水平井技术可以更高效的开采致密气藏，如何准确预测水平井分段压裂井的产能也成为气藏实际生产开发的关注重点。本文通过对前人致密气藏分段压裂水平井解析模型的探究确定影响产能因素；运用灰色关联分析法可视化分析影响压裂水平井无阻流量的主要因素；选取不同输入层个数影响因素利用python编程程序建立神经网络模型，对比选取高精度的训练模型结构；利用训练好的神经网络模型对实际气田待压裂井进行产能预测。结果表明，利用神经网络模型建立的压裂水平井产能预测方法具有误差小、简便高效的优点，对实际气田生产开发制度有一定指导意义。

Productivity prediction of staged fractured Wells in tight gas reservoir based on neural network model

The tight gas reservoir is dense and difficult to be exploited. The application of staged fracturing horizontal well technology can produce tight gas reservoir more efficiently. How to accurately predict the productivity of staged fracturing horizontal well has become the focus of practical production and development of gas reservoir. In this paper, the factors affecting the productivity of horizontal Wells in tight gas reservoirs are determined by exploring the analytical model of previous staged fractured horizontal Wells. The main factors affecting the open flow rate of fractured horizontal Wells are analyzed visually by using grey correlation analysis method. The influence factors of different input layers were selected and the neural network model was established by Python programming program, and the high-precision training model structure was selected by comparison. The trained neural network model is used to predict the productivity of actual gas field Wells to be fractured. The results show that the prediction method of productivity of fractured horizontal Wells based on neural network model has the advantages of small error, simple and efficient, and has certain guiding significance for practical gas field production and development system.