论文部分内容阅读
现代油藏参数反演中遇到的复杂方程和定解条件使得参数的反演问题高度非线性,存在多局部极值。文章提出了用于油藏参数反演全局优化的改进差分进化算法(MDE),利用DE算法在一定进化代数后出现的种群聚类特性,将种群识别为不同的聚类区域,然后以每个聚类的中心为起始点,利用基于梯度的局部搜索算法可以快速找到该聚类区域的最小极值。算法可以避免在搜索过的聚类区域中出现重复搜索,有效地缩小了搜索空间、提高了种群的多样性,防止出现早熟现象。该改进算法应用于油田地层参数反演,对求压力数据的正问题采用数值解析方法得到,并对选定压力分析段曲线进行最小二乘拟合的实验结果表明,该方法迭代次数较少,收敛精度高。
The complex equations and solution conditions encountered in the inversion of modern reservoir parameters make the inversion of parameters highly nonlinear and have multiple local extremums. In this paper, we propose an improved differential evolution algorithm (MDE) for global optimization of reservoir parameters inversion. By using the DE algorithm, population clustering features appear after a certain evolutionary algebra. The population is identified as different clustering regions. Then, The center of clustering is the starting point. Using the gradient-based local search algorithm, the minimum extremum of the clustering area can be quickly found. The algorithm can avoid repeated searching in the searched clustering area, effectively narrow the search space, improve the diversity of the population and prevent premature phenomenon. The improved algorithm is applied to inversion of formation parameters in oilfield. The positive analysis of pressure data is obtained by numerical analysis. The experimental results of the least squares fitting of selected pressure analysis section curves show that this method has fewer iterations, High convergence accuracy.