应用多属性神经网络方法预测油气

地震属性包含的地球物理信息十分丰富，但地震属性种类繁多，并且与储层特征对应关系复杂，单属性分析难以确保预测的准确性。人工神经网络方法具备较强的非线性映射能力，使用该方法可以综合利用多属性进行油气预测，提高预测精度。该文采用梯度下降神经网络算法，避免陷入局部极小值，有效加快网络收敛速度，使网络达到全局最优，提高了预测效果。

Application of multi-attribute and neural network method to hydrocarbon reservoir prediction

Seismic attributes contain abundant geophysical information.There are so many seismic attributes and the relationship between attributes and reservoir characteristics is complicated.Single attribute analysis cannot assure the prediction accuracy.Artificial neural network technology has strong nonlinear mapping ability,so it can be applied to improve hydrocarbon prediction accuracy.The gradient descent learning algorithm is used in neural network.It can avoid local minimum and effectively speed up the network convergence to achieve the global optimal network and improve its forecasting performance.