神经网络在彩南油田油藏描述中的应用

利用自组织神经网络,建立了彩南油田彩9井区岩性识别模型,通过检验,分类结果的符合率达93.75%.利用神经网络技术,建立了研究区的孔隙度和渗透率模型,利用该模型对未知样本进行预测,预测结果与实际测量结果具有很好的一致性,孔隙度平均绝对误差为0.54%,平均相对误差3.5%,渗透率平均绝对误差1.68×10-3μm2,平均相对误差32%,其精度较传统的方法有了很大的提高。

Application of Neural Networks to Reservoir Characterization in Cainan Oilfield

Lithology identification model is developed in Well Block Cai 9 of Cainan oilfield by using neural network technology. The classification can reach 93.75% of coincident rate with the test samples. Also, porosity and permeability model in studied area is developed, by which unknown samples are predicted, obtaining average absolute errors of 0.54% for porosity and 1.68×10^-3 μm² for permeability; average relative errors of 3.5% for porosity and 32% for permeability. The accuracy of prediction is greatly improved compared with traditional methods.