[1]李春生,邹林浩,张可佳,等.基于 BP 神经网络的录井异常数据检测方法研究[J].计算机技术与发展,2022,32(06):173-178.[doi:10. 3969 / j. issn. 1673-629X. 2022. 06. 029]
　LI Chun-sheng,ZOU Lin-hao,ZHANG Ke-jia,et al.Research on Detection Method of Logging Anomaly Data Based on BP Neural Network[J].,2022,32(06):173-178.[doi:10. 3969 / j. issn. 1673-629X. 2022. 06. 029]
点击复制

基于 BP 神经网络的录井异常数据检测方法研究()

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:
32
期数:
2022年06期
页码:
173-178
栏目:
应用前沿与综合
出版日期:
2022-06-10

文章信息/Info

Title:
Research on Detection Method of Logging Anomaly Data Based on BP Neural Network
文章编号:
1673-629X(2022)06-0173-06
作者:
李春生; 邹林浩; 张可佳; 高雅田; 刘 涛; 豆立宪
东北石油大学 计算机与信息技术学院,黑龙江 大庆 163318
Author(s):
LI Chun-sheng; ZOU Lin-hao; ZHANG Ke-jia; GAO Ya-tian; LIU Tao; DOU Li-xian
School of Computer and Information Technology,Northeast Petroleum University,Daqing 163318,China
关键词:
异常点检测; 录井工程数据; BP 神经网络; 格鲁布斯法; K-means 聚类算法
Keywords:
anomaly detection; logging engineering data; BP neural network; Grubbs method; K-means clustering algorithm
分类号:
TP391
DOI:
10. 3969 / j. issn. 1673-629X. 2022. 06. 029
摘要:
在石油钻井工程中,由于技术和设备的客观因素,导致录井数据经常出现异常值,影响了录井解释评价精度。 针对该问题,提出了一种基于 BP 神经网络的录井异常数据处理方法。 为了在构建数据环节中提供准确且可信的工程数据,研究了录井异常数据的产生原因及异常数据的表征,并且通过对比格鲁布斯法、K-means 聚类算法以及 BP 神经网络等方法的特点,选择 BP 神经网络作为异常值处理的方法。 通过模型预测的录井数据误差平方值与样本数据的均方根误差进行比较,来确定数据的异常情况,保证检测异常点的合理性。 经实验验证和同类算法的比较,表明了 BP 神经网络模型可以实现检测录井异常点数据,且检测异常点的准确率高于同类算法,处理异常点结果可信,能够有效解决因异常点数据所带来的问题。
Abstract:
In the oil drilling engineering,because of the objective factors of technology and equipment,abnormal values often appear in thelogging data,which affects the accuracy of logging interpretation and evaluation. Aiming at this problem,a method of logging anomalydata processing based on BP neural network is proposed. In order to provide accurate and reliable engineering data in the construction ofdata,we study the causes of logging abnormal data and the characterization of abnormal data,and select BP neural network as the methodof outlier processing by comparing the characteristics of Grubbs method, K - means clustering algorithm,BP neural network and othermethods. By comparing the square error of the logging data predicted by the model with the root mean square error of the sample data,the abnormal situation of the data can be determined to ensure the rationality of the abnormal points detected. The experimentalverification and comparison with the similar algorithms show that the BP neural network model can detect logging anomaly data,and theaccuracy of detecting anomaly points is higher than that of the similar algorithms. The results of handling anomaly points are reliable,andit can effectively solve the problems caused by the abnormal point data.

相似文献/References:

[1]程艳云,张守超,杨杨. 基于大数据的时间序列异常点检测研究[J].计算机技术与发展,2016,26(05):139.
　CHENG Yan-yun,ZHANG Shou-chao,YANG Yang. Research on Time Series Outlier Detection Based on Big Data[J].,2016,26(06):139.
[2]李春生,田梦晴,张可佳.基于 Bi-LSTM 网络的管道异常数据检测方法[J].计算机技术与发展,2023,33(06):215.[doi:10. 3969 / j. issn. 1673-629X. 2023. 06. 032]
　LI Chun-sheng,TIAN Meng-qing,ZHANG Ke-jia.Pipeline Anomaly Data Detection Method Based on CNN and Bi-LSTM Network[J].,2023,33(06):215.[doi:10. 3969 / j. issn. 1673-629X. 2023. 06. 032]

更新日期/Last Update: 2022-06-10