基于PLSR的压裂增油量变权重组合预测方法研究

A Research on the Method of Variable Weight Parameter Combination Method Based of PLSR in Incremental Oil Production Prediction of Fractured Wells

针对传统低渗透断块油藏压裂增油量预测模型对多参数与增油量之间变化关系描述能力不足的问题,提出基于偏最小二乘回归的变权重组合预测方法。考虑参数间存在的多重共线性对模型的影响,利用该算法主成分分析特性提取与自变量、因变量关联度高的解释成分,结合典型相关分析特性开展油井压裂后不同阶段的主要影响因子规律研究。偏最小二乘算法变权重组合法能够克服参数之间的冗余和多重共线性问题,通过压缩模型维度提取各阶段特征信息建立分段回归模型。结果表明,经油井压裂后期增油量实测数据验证,该方法能够实现增油量有效预测以及油井压裂后综合状态的跟踪评估,揭示模型中主要因素中参数权重变化规律,为压裂后期优化调整时机提供理论指导。

Aiming at the problem of insufficient capacity to describe the relationship between muhi-parameter and oil-increasing amount in the fracturing oil-yield prediction model of traditional low-permeability fault block oil reservoirs, a combined method based on partial least-squares regression is proposed. Considering the influence of muhiple co-linearity between the parameters on the model, the main influencing factors in different stages after oil well fracturing are analyzed by using the components of the principal component analysis and the explanatory components with high correlations with independent variables and dependent variables Law research. The method of variable weight parameter combination method based on PLSR can overcome the redundancy and muhiple co- linearity between parameters. It extracts the information of each stage through the compression model dimension to establish the subsection regression model. The results show that the method can effectively predict oil yield and reveal the variation of parameter weight among the main factors in the model to provide theoretical guidance for optimizing the timing of post-fracturing.