基于支持向量回归机的油田异常井预警模型研究

发布时间：2018-03-27 11:41

本文选题：预警　切入点：改进k-均值　出处：《东北石油大学》2015年硕士论文

【摘要】：目前,我国大部分油田已经开采到中后期阶段,开采过程中经常出现机泵断脱、管线穿孔、电箱故障等事故造成油井关井停产,从而导致石油产量趋于不稳定状态。针对上述问题,开发油田异常井预警系统,系统在异常发生初期及时给出预警信息,以便提前采取有效防范措施,避免事故发生造成不必要的损失,确保产量稳定。常用油田预警方法绝大部分是采用单独指标判别方法或利用单一预警模型进行预警,这就容易造成预警结果准确度低、预警结果错误的情况发生。针对这些问题,本文提出了基于支持向量回归机的油田异常井预警模型,模型利用组合预警的方法,主要分为三个部分,首先,对油井异常中的参数进行选择和处理,再对所得参数下的数据进行预处理操作;然后,利用改进k-均值算法对预处理后数据进行聚类分析,形成井况正常、井况预警和井况已异常三种聚类数据;最后,采用支持向量回归机算法,使用基于网格搜索的交叉验证法优化支持向量回归机的参数,并确定处理边缘数据后的井况预警聚类数据,建立基于支持向量回归机的油田异常井预警模型,实现对油田异常井的预警分析。在基于支持向量回归机的油田异常井预警模型的基础上,开发了油田异常井预警系统。对油田异常井预警系统开发过程中所涉及的方法进行了详细的设计,以某采油厂已有遗产数据为样本实例验证了本课题提出的模型相较于已有预警方法预警准确率有所提高,基于模型所开发的系统也具有良好的应用价值。
[Abstract]:At present, most oil fields in our country have been exploited to the middle and late stage. During the process of exploitation, accidents such as machine pump breaking, pipeline perforation, electric box failure, etc., often result in shut down of oil wells. Therefore, the oil production tends to be unstable. In view of the above problems, the early warning system of abnormal wells in oil fields is developed, and the early warning information is given in time in order to take effective preventive measures in advance. To avoid unnecessary loss caused by accidents and to ensure the stability of production, most of the commonly used early warning methods for oil fields are to use single index discriminant method or single early-warning model for early warning, which can easily lead to low accuracy of early warning results. In view of these problems, this paper puts forward an oilfield abnormal well warning model based on support vector regression machine. The model is mainly divided into three parts by using the method of combination early warning. The parameters of oil well anomalies are selected and processed, and then the data under the parameters are preprocessed. Then, the improved k-means algorithm is used to cluster the pretreated data, and the well condition is normal. Finally, support vector regression algorithm is used to optimize the parameters of support vector regression machine, which is based on cross validation method based on grid search. After processing edge data, the well condition early warning clustering data is determined, and an oilfield abnormal well warning model based on support vector regression machine is established. Based on the prediction model of abnormal wells based on support vector regression machine, The early warning system for abnormal wells in oil fields is developed. The methods involved in the development of abnormal wells in oil fields are designed in detail. Taking the existing heritage data of an oil extraction plant as an example, it is verified that the proposed model is more accurate than the existing early warning method, and the system based on the model has good application value.
【学位授予单位】：东北石油大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TE38

【相似文献】