基于BP神经网络的油田生产动态分析方法

Analysis method of oilfield production performance based on BP neural network

为克服目前生产动态分析方法所需数据量大、费工费时和应用局限性大等缺点,文中提出了一种基于BP神经网络的油田生产动态分析新方法。该方法使用一些广泛易得的数据(如测井数据、生产历史数据)建立数据集,然后利用神经网络建模技术建立全油藏范围的基于数据驱动的预测模型,进行预测分析。实际油藏应用结果表明,产油速度的最大预测误差低于7%,产水速度的预测误差低于5%。预测效果较好,具有一定的应用和研究价值。

Aiming at the situation that many techniques of production performance analysis require lots of data and are expensive considering the computational and human resources, and their applications are limited, this paper puts forward a new method based on BP neural network, tt builds a dataset with some available measured data such as logging and production history, then, builds a field-wide predication model by neural network technique. The model will be used to predict. Actual application results show that the maximum prediction error of oil production rate is lower than 7% and the maximum prediction error of water production rate of is lower than 5%, having certain application and research value.