基于CatBoost算法的孔隙压力预测方法及其在井壁稳定分析中的应用

Prediction method of pore pressure based on CatBoost algorithm and its application in wellbore stability analysis

为解决传统孔隙压力预测方法(如伊顿法和鲍尔斯法)在预测孔隙压力时,适用范围较小、受人为因素影响较大等问题。基于CatBoost机器学习回归算法建立孔隙压力智能预测模型,并与决策树回归算法和随机森林回归算法进行比较,以某区块2口直井为例验证模型的预测效果。结合CatBoost模型的孔隙压力预测结果,利用数值模拟软件分析孔隙压力对井壁稳定的影响。研究结果表明:CatBoost模型的5个评价指标相对最优,孔隙压力当量密度实测值与预测值的相对误差最小,CatBoost模型具有较强的泛化能力和较高的预测精度;在低孔隙压力条件下,井周等效塑性应变不均匀性明显,井周进入塑性区的围岩区域主要集中在最大主应力方向;在较大孔隙压力作用下,井周等效塑性应变不均匀性有所降低,但井周等效塑性应变的极大值仍存在于最大主应力方向。研究结果可对孔隙压力精确预测和钻井作业安全施工提供一定指导作用。

In order to solve the problems of small applicable scope,large influence by human factors and large limitation when predicting the pore pressure with the traditional pore pressure prediction methods such as Eaton method and Bowles method,an intelligent prediction model of pore pressure was established by using the CatBoost machine learning regression algorithm.It was compared with the decision tree regression algorithm and random forest regression algorithm,and the prediction effect of the model was verified by two vertical wells in a block.Combined with the prediction results of pore pressure by CatBoost model,the influence of pore pressure on the wellbore stability was analyzed by numerical simulation software.The results showed that the five evaluation indexes of the CatBoost model were relatively the optimal,and the prediction error of the measured pore pressure data was relatively the minimal,so the CatBoost model had strong generalization ability and high prediction accuracy.Under the condition of low pore pressure,the non-uniformity of equivalent plastic strain around the wellbore was obvious,and the surrounding rock area entering the plastic zone around the wellbore was mainly concentrated in the direction of the maximum principal stress.Under large pore pressure,the non-uniformity of equivalent plastic strain around the wellbore decreased,but the equivalent plastic strain still existed in the direction of the maximum principal stress.The research results can provide certain guidance for the accurate prediction of pore pressure and safe construction of drilling operation.