基于SVM和D-S证据理论的早期溢流智能识别方法

Research on Intelligent Early Kick Identification Method Based on SVM and D-S Evidence Theory

早期溢流监测是预防井喷事故发生的重要手段。目前的溢流识别方法大多基于单一的监测手段,可靠性不高,现场需要结合多种监测手段对溢流发生进行综合研判。直接应用多种手段监测溢流时,存在由于各种手段监测结果不一致甚至出现矛盾冲突的问题。为此,提出了一种基于支持向量机后验概率模型和D-S证据理论的早期溢监测方法,结合钻井液微流量参数、综合录井参数、井底随钻测量参数对溢流发生进行综合判别;有效解决应用各类监测参数识别溢流时出现溢流识别结果矛盾冲突的问题。采用仿真及现场实测数据进行了溢流识别实验,结果表明,该方法能有效处理多源信息间矛盾冲突,提高溢流监测的可靠性具有较高的现场应用价值。

Early kick detection plays an important role in preventing blowout accident. Most existing kick detection methods rely only on single detection tool,which is limited by drilling conditions and can only produce low reliable results. Therefore,a synthetic identification method by combining multiple detection means to comprehensively determine the occurrence of kick is required. However,the direct application of multiple monitoring tools may produce inconsistent monitoring results. In order to address this problem,an early kick detection method based on support vector machine posterior probability model and D-S evidence theory is proposed,which integrates the drilling fluid,the comprehensive logging parameters and the PWD measurement parameters to identify kick. Meanwhile,it can deal with contradictory or even conflict problems during applying various types of monitoring parameters. Experimental results with both simulated and real data show that the proposed method can effectively deal with the conflicts among multiple monitoring results and thus improve the reliability of kick detection,which behaves potential practical application value.