AGA和L-M算法联合预测生产油井油水流动剖面

Joint Prediction of Oil-water Two-phase Profiles with Genetic Algorithm(AGA) and Levenberg Marquardt(L-M) Algorithm

生产测井是目前油井动态监测的主要手段,生产测井产液剖面解释方法是油井动态评价的关键,具有十分重要的意义。提出了AGA和L-M联合优化技术处理油水两相流产出剖面资料的方法。该方法以生产层段单元空间中流体的能量守恒为基础,根据非线性加权最小二乘法原理和误差理论,以理论流体温度和温度计测量值的差值建立求最小值的目标函数,由地面计量并根据流体体积换算得到的井下各相流量及控制体底部的流量测井值为约束条件。为了提高解释精度,采用遗传算法的全局搜索法(AGA)与L-M算法的直接式局域搜索法相结合的方法求解最优解,即先利用自适应遗传算法进行启发式全局搜索,然后直接把该结果作为初始参数,再进行L-M法处理来接近最优解。联合反演出来的解释结果稳定、可靠。该方法不仅适用于直井也适用于斜井。

At present,production logging was a major method for well performance monitoring,and the production logging flow profile interpretation is the key of well performance analysis it was of great significance.A new method of oil and water flow production logging interpretation was put forward based on energy conservation and optimization mathematics.Based on the energy conservation of flow fluid in the control volume of production interval,it built the object function which combines theoretical value of fluid temperature and its measured value according to the principle of nonlinear least square method and error theory.In order to improve the inversion accuracy,an integrated method based on adaptive genetic algorithm(AGA) and Levenberg-Marquardt algorithm(L-M) was presented.By using variable global results obtained by the inversion depended on AGA as initial parameters;additional local search is carried out based on the L-M algorithm to approach the optimal solution,the result is reasonable and reliable.This method is not only appropriate for vertical wells but also for deviated wells.