K-means聚类优化井底压力修正模型研究

Optimization of a Bottom-hole Pressure Correction Model Using K-means Clustering

在钻井过程中,由于井底压力计算模型误差大,而井下实测井底压力时数据容易失真、甚至无数据,因而不能准确测定井底压力,对钻井作业带来极大的安全风险。针对此类问题,提出了一种用K-means聚类方法优化朴素贝叶斯模型,结合井底压力监测原理,形成一套实现井底压力智能动态分析K-means聚类优化的朴素贝叶斯模型,利用该模型修正传统水力学模型井底压力计算值,根据修正结果与实测井底压力值进行对比,使得计算结果误差最小。利用现场数据进行对比分析,结果表明,应用基于K-means聚类的朴素贝叶斯模型修正井底压力计算值其误差较小,处于安全钻井压力监测误差范围内,能够满足现场正常钻井要求。

In the process of drilling,owing to problems such as data distortion during the measurement ofbottom-hole pressure,absence of return data,and incapacity of bottom-hole pressure calculation models to accurately reflect the measurements,the bottom-hole pressure cannot be accurately monitored,and this results in a substantial safety risk for drilling operations.To provide an effective overall monitoring of the bottom-hole pressure,a K-means clustering optimization method was established to improve the Naive Bayesian model.Combined with the principle of bottom-hole pressure monitoring,a Naive Bayesian model optimized by K-means clustering was designed,which could perform intelligent dynamic analysis of the bottom-hole pressure.This model was adopted to correct the bottom-hole pressure calculated by the traditional hydraulic model,and the results ofboth these models were compared to minimize the calculation error.Analysis of field data suggested that the calculated bottom-hole pressure corrected using the optimized model demonstrated a smaller error.This error was within the safe range of the pressure monitoring of drilling operation,indicating that the model was able to satisfy the requirements of regular drilling practices.