摘要
本文讨论利用BP网络进行储层油气预测的缺陷和遗传算法优化神经网络连接权的局限性,通过融合变尺度混沌搜索策略的思想,构成一种新的混合遗传算法———混沌遗传算法。此法不仅能对遗传算法的初始种群进行优化,筛选出优化种群,而且可对遗传学习过程进行优化,从而明显减少了遗传算法的搜索空间,提高了遗传算法优化神经网络初始权值的计算效率,改善了遗传算法的性能。此法充分利用了变尺度混沌优化算法和遗传算法的各自优点。将此新方法用于储层油气预测,取得了较好的效果。
The paper discussed the failures of using BP neural network to carry out prediction of oil/gas reservoir and limitations of optimizing the connection weights of neural network in genetic algorithm. Combining with the idea of scale-variable chaos searching strategy, a new hybrid genetic algorithm—chaos genetic algorithm is formed. The algorithm can optimize not only the initial population of genetic algorithm, choosing optimum population, but also the genetic study process, significantly reducing the searched space of genetic algorithm, increasing the computational efficiency of initial weights of optimized neural network in genetic algorithm and improving the property of genetic algorithm. The approach uses respective advantages of scale-variable chaos optimum algorithm and genetic algorithm. Using the new hybrid genetic algorithm for prediction of oil/gas reservoir achieved better results.
出处
《石油地球物理勘探》
EI
CSCD
北大核心
2006年第6期651-655,共5页
Oil Geophysical Prospecting
关键词
油气预测
BP神经网络
遗传算法
变尺度混沌算法
oil/gas prediction, BP neural network, genetic algorithm, scale-variable chaos algorithm