基于差分进化优化随机森林模型的油层结垢预测方法

Prediction method of oil layer scaling based on differential evolution optimization random forest model

注水开发极大地提高了油田的采收率,但是注入水与地层水的配伍性会导致储层和管道结垢,建立高效准确的油层结垢预测方法对保障安全生产较为重要。结合油层结垢的化学机理和人工智能技术,提出了基于差分进化算法(Differential Evolution)优化的随机森林模型(Random Forest)的油层结垢预测方法(RF-DE)。以离子浓度和储层性质作为输入变量,结垢等级作为输出变量,建立了油层结垢的RF-DE预测模型。在鄂尔多斯盆地的延长组和延安组油层进行应用,结果表明建立的RF-DE方法预测油层结垢准确度高,是一种能够较好预测油层结垢的方法。

Water injection development greatly improves the oil recovery of the oilfield,but the compatibility of injection water and formation water will lead to scaling of reservoir and pipeline.It is important to establish an efficient and accurate prediction method for oil reservoir scaling to ensure safe production.Combining the chemical mechanism of oil layer scaling and artificial intelligence technology,a method for predicting oil layer scaling(RF-DE)based on the random forest model optimized by differential evolution algorithm is proposed.With ion concentration and reservoir property as input variables and scaling grade as output variables,the RF-DE prediction model for reservoir scaling is established.The results of application in Yanchang Formation and Yan’an Formation in Ordos Basin show that the established RF-DE method has high accuracy in predicting oil layer scaling,and is a good method for predicting oil layer scaling.