超稠油油藏水平井产量的预测方法  

Prediction method of output in horizontal wells of extra-heavy oil reservoir

作　　者：杜殿发[1] 石达友[2] 师耀利[1] 孙永杰[1] 邱增法[3] 

机构地区：[1]中国石油大学(华东)石油工程学院,山东东营257061 [2]中国石化股份胜利油田分公司地质科学研究院,山东东营257015 [3]中国石化股份胜利油田分公司现河采油厂,山东东营257000

出　　处：《油气地质与采收率》2009年第5期76-78,共3页Petroleum Geology and Recovery Efficiency

摘　　要：水平井产量的预测工作越来越受到人们的重视,应用基于误差反向传播算法的人工神经网络和数学统计2种方法对超稠油油藏水平井产量进行了研究。第1种方法综合考虑地质、生产、人为等因素对水平井产量的影响,在对原有水平井产量数据拟合的基础上,利用基于误差反向传播算法的人工神经网络,对超稠油油藏水平井未来周期的产量进行预测;第2种方法是回归出水平井截至某周期的累积产量与第1周期产量的关系,在已知第1个周期产量的前提下,用此关系来预测后几个周期的产量。结果表明,2种方法得到的预测产量与实际产量的相对误差均小于5%。The prediction on horizontal well output is getting more and more attention. BP neural network and mathematical statistics methods were applied for studying the horizontal well output in extra-heavy oil reservoir. The first method is:based on the data fitting of the horizontal well output,BP neural networks was applied to forecast of the horizontal well output in the future cycle in extra-heavy oil reservoir with the consideration of geology,production and man-made factors. The second method is:if the first cycle's output is known,a relationship between the cumulative oil production and the first cycle's output will be regressed for the forecast of the last cycle's output. The forecast result is that the above error is less than 5% ,which has a stronger reference value for decision-making worker.

关 键 词：人工神经网络 超稠油 水平井 产量预测 统计方法 

分 类 号：TE313.8[石油与天然气工程—油气田开发工程]