基于神经网络信息融合技术预测气藏水锁

Prediction of water lock in gas reservoirs based on neural network information fusion

快速、准确地诊断与预测气藏水锁问题一直是气藏开发中的技术难点。常规研究方法需使用大量岩心进行测试,且时间较长,无法快速应用于生产中。在充分分析水锁产生机理和各种影响因素的基础上,利用神经网络信息融合技术的三层前馈神经网络系统,以孔隙度等11个对水锁有影响的因素为网络的输入层节点,建立气藏水锁识别、诊断、评价和预测的信息融合模型,并编制软件,进行气藏水锁模拟预测。研究结果表明,该方法可快速给出气藏水锁程度,对比预测结果与实测结果有较高符合率。

Quick and accurate diagnosis and prediction of water lock have always been a technical issue in the process of gas reservoir development.The conventional method uses a large amount of cores,takes longer time,and is not applicable to production.Based on full analysis of the mechanism and influencing factors of water lock,a model of information fusion is built up by using three-layer feedforward neural network system,with 11 influencing factors of water lock including porosity as input nodes to identify,diagnose,evaluate and predict water lock in gas reservoirs.Meanwhile,related software has been developed.This method can fast predict the degree of water lock in a gas reservoir,and the prediction result highly coincides with the measured result.