论文部分内容阅读
通过对近年来储层敏感性预测方法的分析研究,认为神经网络方法是一种较理想的预测储层敏感性的新方法.但是常规的BP算法存在收敛速度慢、局部根小值等缺点。为此，采用了动量自适应学习率调整方法和L-M优化算法，效果明显改善，其中L-M优化算法效果最好，收敛速度快，误差最小。利用L-M优化方法建立的储层速敏网络模型，预测渗透率损害程度的准确率达93%，基本上满足了油气层敏感性预测的需要。
Based on the analysis of reservoir sensitivity prediction methods in recent years, it is considered that neural network method is a new ideal method to predict reservoir sensitivity, but the conventional BP algorithm has some disadvantages such as slow convergence speed and small local root value. For this reason, the method of adjusting the momentum adaptive learning rate and the L-M optimization algorithm are adopted, and the effect is obviously improved. The L-M optimization algorithm has the best effect, the fast convergence rate and the minimum error. The velocity-sensitive network model established by L-M optimization method can predict the damage degree of permeability to 93%, basically meeting the need of reservoir sensitivity prediction.