改进BP神经网络在裂缝预测中的应用

常规基于线性的测井计算方法不能很好地表征复杂的裂缝系统,而现有的成像测井技术虽然分辨率高,能较好地记录裂缝特征,但其成本高,实用性不强。因此,本文首先研究了改进BP(Back Propagation)神经网络算法,极大地提升了BP神经网络算法的收敛速度和全局收缩能力,然后将该算法引入到裂缝预测中,使利用常规测井曲线就能较好地达到预测裂缝的目的。结合实际的生产资料进行了效果分析,结果表明,对常规测井资料应用神经网络技术能较好地识别裂缝。

Application of Improved BP Neural Network to Fracture Prediction

The conventional calculation method based on linear logging can be a good characteristic of complex fracture system,but the existing imaging logging technology has high resolution and can record the characteristics of fracture,but its cost is high,so it is not practical.Therefore this paper firstly studies the improvement of BP(Back Propagation) neural network algorithm,which greatly improved the BP neural network algorithm in the convergence speed and global contractile capacity,then the algorithm was introduced to fracture prediction,enabling the use of conventional logging curve to achieve better prediction of fracture ends.Combined with the actual production data of the effect analysis,the results show that the conventional logging data and the application of neural network technology can better identify fracture.