一种基于BP神经网络的气井重复压裂井优选方法

目前我国一些气田经过初次压裂后,增产效果不明显,或生产一段时间后,产量明显下降。为了提高气井的产量,可对气井进行重复压裂。进行重复压裂优选,可使气井得到较好的压裂效果。分析了影响气井重复压裂效果的因素,可分为地质因素和工程因素2部分。地质因素包括孔隙度、渗透率、表皮系数、产层厚度、含气饱和度、地层压力系数及剩余可采储量,工程因素包括前一次压裂是否成功、前一次压裂液用量、前一次压裂加砂量。基于BP神经网络理论,结合气井重复压裂效果影响因素分析,建立了重复压裂井优选模型。使用粒子群算法对其进行了优化,提高收敛速度的同时有效防止了局部最优解情况的发生,预测重复压裂井的日产气量,以此为依据优选重复压裂井。通过对C区重复压裂效果预测表明,基于BP神经网络优选重复压裂井可以提高选井的准确性。

A preferred method for gas well re-fracturing well based on BP neural network

At present,after initial fracturing in some gas fields,the effect of increasing production is not obvious,or after a period of production,the output is significantly reduced. In order to increase the production of gas wells,the gas wells can be repeatedly fractured. It is preferred to carry out repeated fracturing to obtain a better fracturing effect of the gas well. The factors affecting the effect of repeated fracturing of gas wells are analyzed.The influencing factors are divided into two parts:geological factors and engineering factors. Geological factors include porosity,permeability,skin coefficient,production layer thickness,gas saturation,formation pressure coefficient,remaining recoverable reserves and engineering factors include the success of the previous fracturing,the amount of fracturing fluid used in the previous fracturing,and the amount of sand added during the previous fracturing. Based on BP neural network theory,combined with the analysis of influencing factors of gas well repeated fracturing effect,the optimal model of re-fracturing well was established. The particle swarm optimization algorithm was used to optimize it. While increasing the convergence rate,it effectively prevented the occurrence of local optimal solutions,and predicted the daily gas production rate of the repeated fracturing wells. Based on this,the repeated fracturing wells were optimized. By predicting the effect of repeated fracturing in Zone C,it is better to repeat the fracturing well based on BP neural network to improve the accuracy of well selection.