应用广义回归神经网络预测油井含水率

Using General Regression Neural Networks Method to Predict Water Cut in Oil Well

提出一种广义回归神经网络模型(GRNN)预测油井含水率模型,该模型利用不同阶段油井累计注水量和累计产油量,模型简单且具有实用性。相对于拟合模型,应用GRNN模型预测结果精度更高,且可避免新增样本数据后需要重新预测的问题。而相比于BP神经网络,GRNN模型预测结果更加快捷和准确。

This paper presents general regression neural networks method to predict water cut in oil well.The new method is developed taking full advantage of the cumulative water injection and cumulative oil production in different time,and it can also simplify the problem.The calculated results show that the developed mode can provide high accuracy in predicting water cut in oil well than fitting mode,and it can also avoid a new calculating for the adding sample dates.What's more,compared with Back Propagation neural networks,the developed mode can provide a higher accuracy and higher speed result.