多层合采产能配比的算法研究及应用  

A STUDY AND APPLICATION TO ARITHMETIC OF PRODUCTION ALLOCATION OF MULTILAYER COMMINGLED PRODUCTION

作　　者：王跃文[1] 卢双舫[1] 方伟[2] 张居和[2] 黄修平[3] 

机构地区：[1]大庆石油学院地科院,黑龙江大庆163318 [2]大庆油田公司勘探开发研究院,黑龙江大庆163712 [3]大庆油田有限责任公司第六采油厂,黑龙江大庆163110

出　　处：《石油实验地质》2005年第6期630-634,共5页Petroleum Geology & Experiment

基　　金：黑龙江省杰出青年基金项目(200003)

摘　　要：针对气相色谱指纹技术目前还难以被应用到3层及3层以上合采油层的局限,作者提出了一套人工神经网络学习算法,该算法具有非线性、精度高、适用于多层合采的优点。应用此方法,在实验室通过原油的全烃气相色谱和色谱-质谱分析,确定反映单层和配比的原油特征气相色谱指纹,建立特征指纹数据库,用人工神经网络算法对数据进行分析处理,找出内在规律,就可以对实际合采的原油进行分析应用。通过对喇嘛甸油田及萨尔图油田试验区合采油井的单层产量进行实验室配比计算和实际合采油层原油分析验证,计算结果具有很高的精度并与单井实际产量MFE测试结果吻合很好。该项计算技术为油田利用地球化学方法进行多层合采的单层产能配比计算提供了一个经济适用的途径。Aim at the localization of the technique of gas chromatographic fingerprint, which is difficult of being applied to commingled production of three and more layers oil. The authors study and application an arithmetic of artificial nerve network. It has higher precision, and it is nonlinear and applicable to commingled production of more than three layers. Applying this arithmetic, through the analysis of whole hydrocarbon gas chromatographic fingerprint of crude oil in the laboratory, and pick out some characteristic fingerprint which reflect the character of crude oil in single and commingled layer, Setting up characteristic fingerprint data base, and then analyzing and processing them with ANN arithmetic, finding the rules, then the authors apply the arithmetic to the Lamadian and the Saertu oil field. As a result, even to the problem of more than three layers, the results also have high accuracy, and are close to the MFE test results. The application of the arithmetic offers a economical and applied path to production allocation calculations of multilayer commingled production with the geochemistry means.

关 键 词：人工神经网络 色谱指纹 非线性 产能配比 合采 

分 类 号：TE3[石油与天然气工程—油气田开发工程]