基于相对熵改进模糊C均值聚类的溢流预警研究

Research on Kick Warning Based on Relative Entropy Improved Fuzzy C-mean Clustering

钻井过程中溢流的早期发现非常重要,目前国内外基于人工智能的溢流预警模型普遍使用大量先验知识或训练数据,其准确性、实时性、可靠性完全受限于先验知识和训练数据,文章提出了基于相对熵改进模糊C均值聚类的溢流预警模型,采用相对熵理论改进模糊C均值聚类算法,克服传统模糊C均值聚类时聚类数目由用户主动给出的缺点,并结合溢流故障的发生与立压、套压的变化趋势具有相关性的特点,建立了早期溢流智能预警模型,实现对早期溢流的及时发现。通过对现场数据的仿真分析表明,该预警模型能够通过立压和套压的斜率变化及时准确地判断是否发生溢流。

The early kick detection is essential during the drilling process to avoid accidents.At present,the early kick detection models based on artificial intelligence at home and abroad generally adopt a large amount of prior knowledge or training data,and their accuracy,timeliness and reliability are completely limited by prior knowledge and training data.In this paper,a new kick warning model based on relative entropy improved fuzzy C-mean clustering is proposed to overcome these problems.The relative entropy theory is used to improve the FCM clustering algorithm and overcome the disadvantage that the number of clusters is actively given by the user in FCM clustering.Combined with the characteristics of the correlation between the occurrence of overflow fault and the change trend of standpipe pressure(SSP)and casing pressure(CP),an intelligent early kick warning model is established to realize the timely early kick detection.The simulation analysis of the field data shows that the early warning model can determine whether the kick occurs in a timely and accurate manner by the change of the slope of SSP and CP.