用基于实数编码的自适应遗传算法求解产量预测模型  

SOLUTION FOR PRODUCTION PREDICTION MODEL BY ADAPTIVE REAL-CODED GENETIC ALGORITHM

作　　者：侯健[1] 曲昌学[2] 陈月明[1] 王玉斗[1] 周波 

机构地区：[1]石油大学石油工程学院,山东东营257061 [2]胜利油田有限公司采油工艺研究院,山东东营257000 [3]辽河油田有限公司进出口公司,辽宁盘锦124010

出　　处：《石油大学学报（自然科学版）》2002年第3期55-58,共4页Journal of the University of Petroleum,China(Edition of Natural Science)

摘　　要：产量预测模型本质上都是具有若干待定参数的非线性模型 ,对它的求解可表示为一个连续参数优化问题。用基于实数编码的自适应遗传算法对产量预测模型进行了求解 ,并针对模型求解的特点 ,提出了自适应变异、最优个体保留、分段定义变加权适应度函数等若干改进措施 ,从而加快了运算速度 ,保证了问题的收敛 ,并可按需要区别对待不同时间阶段动态数据对拟合效果的影响。该方法具有通用性 ,几乎所有的模型都可用其求解。用实测数据对Weng旋回模型和产量双曲递减曲线进行了求解 ,实例分析表明该算法具有较强的实用性。The production prediction model is essentially a nonlinear model containing some unknown parameters. Its solving process may be regarded as an optimization of floating functions. An adaptive real coded genetic algorithm was applied to solve the model. According to its solution characteristics, some improved strategies, such as adaptive mutation operator, best individual preservation, section change weighting adaptative function definition were proposed. These strategies make arithmetic speed accelerated and convergence guaranteed. The users can discript the influences of dynamic data on the fitting degrees at different time phase referring to their need. This measure has universality and can suit for almost all models. The measured values were used to solve Weng cycle model and the hyperbolic decline curve. The application results indicate that this solution method is practical.

关 键 词：实数编码 自适应遗传算法 产量预测 生产动态 非线性模型 求解 油气田开发 

分 类 号：TE328[石油与天然气工程—油气田开发工程]