应用人工神经网络方法确定岩石压缩系数  

Predicting rock compressibility by artificial neural network

作　　者：王安辉[1] 张英魁[2] 高景龙[2] 邵文勇[2] 阮宝涛[2] 朱明文[3] 王海忠[2] 刘丙辉[2] 

机构地区：[1]中国地质大学 [2]吉林油田分公司勘探开发研究院 [3]吉林油田分公司试采公司

出　　处：《石油勘探与开发》2003年第4期105-107,共3页Petroleum Exploration and Development

摘　　要：应用吉林油区现有的实际岩石压缩系数资料对前人的经验公式进行验证的结果 ,总相对误差为 5 0 4.2 5 %～681.15 % ,由此发现这些经验公式对吉林油区并不适用。采用人工神经网络BP算法 ,以压力、孔隙度为输入层参数 ,以岩石压缩系数为输出层参数 ,分别预测了吉林油区两个地区油藏的岩石压缩系数 ,应用实际资料验证 ,相对误差仅为 12 .8% ,表明用此方法预测岩石压缩系数的可靠性。图 1表 5参Due to the requirements of the mass balance calculation, elasticity energy calculation, well testing interpretation, and overcoming the difficulty in measuring the rock compressibility in the oil gas reservoir engineering in Jilin Oilfield,east China, earlier experience formula is tested and verified by using the present data of practice rock compressibility in Jilin Oilfield, east China, resulting in 504.20 % 681.15 % of the total relative error. The study has found that these experience formulas are hardly to be used and applied in the Jilin Oilfield. So, taking pressure and porosity as input layer parameters and rock compressibility as output layer parameter,the rock compressibility is calculated by using the error back propagation of artificial neural network,which is named as a BP method. The study has predicted the rock compressibility in the Block Ⅰ and Ⅱ of Jilin Oilfield,respectively, and tested and verified the reliability of this method by using practice data, resulting in 12.8 % of the relative error only.

关 键 词：岩石压缩系数 相关公式 人工神经网络 BP算法 

分 类 号：TE135[石油与天然气工程—油气勘探]