西南石油大学学报(自然科学版) ›› 2012, Vol. 34 ›› Issue (3): 98-104.

• 石油与天然气工程 • 上一篇    下一篇

基于混合差分进化算法的试井分析最优化方法

贺 英1,聂仁仕1,贾永禄1,阮宝涛2,杨思松2   

  1. 1. 西南石油大学石油工程学院,四川 成都 6105002. 中国石油吉林油田分公司勘探开发研究院,吉林 松原 138000
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2012-06-01 发布日期:2012-06-01

An Optimal Method of Well Test Analysis Based on Advanced DifferentialEvolutionary Algorithm

He Ying1, Nie Renshi1, Jia Yonglu1, Ruan Baotao2, Yang Sisong2   

  1. 1. School of Petroleum Engineering,Southwest Petroleum University,Chengdu,Sichuan 610500,China2. Research Insitute of Exploration and Development,Jilin Oilfield Company,PetroChina,Songyuan,Jilin 138000,China
  • Received:1900-01-01 Revised:1900-01-01 Online:2012-06-01 Published:2012-06-01

摘要:
要:试井参数优化是对通过测试得到的油气井井底流压及产量数据得到的油气藏模型参数进行优化处理,得到最
符合油气藏实际情况的参数。试井存在多解性,以往使用的最优化方法要求目标函数连续可微,受初值影响较大,是
属于寻找局部最优解的办法。为此,提出了一种混合差分进化算法的试井分析最优化方法。将标准差分进化算法加
以改进,与模拟退火算法和小生境思想结合在一起,构成一种混合差分进化算法,该算法在保证算法全局搜索能力的
同时,能更快地收敛到非线性问题的最优解。进一步将该算法应用到试井分析中,构建了基于混合差分算法的试井分
析方法,不需要估计井筒和油藏参数的初值,也不要求目标函数连续可微,优于标准差分进化算法。通过实测试井资
料分析,与 L–M 方法相比,效果良好。

关键词: 关键词:试井分析, 自动拟合, 优化, 差分进化

Abstract: Abstract:The purpose of well test parameter optimization is to get parameters which reflect the real condition of oil and gas
reservoir by optimizing the parameter of reservoir model from the well button pressure and production data. Well test analysis is
amultiplicityproblem,buttheoptimizationmethodrequirestheobjectfunctionbedifferentiableandcontinuous,andisinfected
by the initial value. This is the way to find the part optimal value. Standard differential evolutionary algorithm is improved,
combinedwithsimulatedannealingalgorithmandniche,anddevelopedintoamixeddifferentialevolutionaryalgorithm,which
guaranty global network ability and converge to optimal solution. And this algorithm is introduced to well test. An optimal
method of well test analysis based on advanced differential evolutionary algorithm is established.It is not necessary to specify
initial estimates for the parameter values and objective function continuous. The method is superior to standard differential
evolutionary algorithm. Through the comparison between the L–M method and the method mentioned in this paper,well test
data demonstrated the method is efficient.

Key words: Key words:well test analysis, automatch, optimize, differential evolutionary