数据挖掘技术在钻井液优化配方设计中应用

The application of data-mining technology in the optimization of drilling fluid program

利用数据挖掘技术建立了一套快速预测钻井液优化配方的新方法。验证结果表明,应用该方法预测出的钻井液优化配方与实验室优选出的钻井液配方是一致的。该套数据挖掘工具能快速给出符合要求的钻井液配方,指导用户有针对性地进行配方实验,缩短实验时间,降低实验成本,同时极大地提高实验成功率。其基本步骤为:①数据准备,从历史数据中选取数据项和行记录作为进行挖掘的数据集,其中包含钻井液性能指标和配方;②根据所选数据特征进行模型训练前的处理;③数据集训练,先依据钻井液常规性能进行聚类分析挖掘,再对挖掘结果进行预测分析挖掘;④利用数据集中没有参与训练的部分数据集对挖掘进行验证;⑤根据给定的钻井液性能要求,挖掘工具给出相应的钻井液配方。

A new method for fast predicting of drilling fluid optimization program is established based on the data-mining technology. Tests show that the drilling fluid formulation predicted by this method coincides with that of by laboratory selection. The data-mining tool can provide the required formulation rapidly, and direct the formulation experiment, thus shorten the experiment time and cut down the cost. This method is consisted of the following processes, 1. data preparation, including drilling fluid performance indexes and formulations; 2. pre-processing before model training based on the selected data; 3. data training; 4. verifying of mining using the untrained data; 5. present the suitable drilling fluid formulation, according to the required drilling fluid properties.