用广义回归神经网络和遗传算法分析产量递减

Application of Generalized Regression Neural Network and Genetic Algorithm to Production Decline Analaysis

我国大多数油田开发已经进入到产量递减阶段,对这一阶段的相关理论有着迫切的需求。产量递减阶段的传统理论存在着不少缺点,使得它的应用受到了很大限制,应用效果也不是很好,主要的问题集中在对递减指数的求解上。广义回归神经网络(GRNN)和遗传算法(GA)都是在模拟人的生理活动进而提出的人工智能技术。GRNN对数据样本没有太多的要求,可以逼近任意类型的函数;GA可以进行全局搜优,也可以进行局部搜优。它们的联合应用可以克服传统理论的缺陷。首先建立GRNN神经网络,然后利用一种改进的GA搜索全局最优的平滑因子,最终建立模型,并把它们应用于孤岛油田产量递减期,取得了良好的效果。

Most oilfiehls developed in China up to now have already entered into the stage of production decline, so the theories about this stage are in urgent demand. But the shortages of conventional theory about it cause its application to be greatly restricted for the key problem of solution for decline index. The generalized regression neural network （GRNN） and the genetic algorithm （GA） are regarded as the artificial intelligence techniques. GRNN has little demand on data sampling, easily approaching to any type of functions; GA can be used to look for the best results in full and partial ranges. The combined application of these two techniques will overeome the shortages of conventional theory. This paper established the GRNN and uses an improved GA to search for the optimum smoothing factor in full range, hence proposes a model. By application of this model to Gudao oilfield in the stage of production decline, good effects are gained.