文档介绍：气田集气站安全监测数据异常值识别研究
张子涛1,殷存志2,杨毅1,董彩虹1,向敏1
1、中国石油北京油气调控中心,北京,100007,2、中亚天然气管道有限公司, 北京,100010
摘要:天然气集气站监测数据异常值的准确识别具有重要意义。借助支持向量回归机算法,对传感器建立在线预测模型,利用遗传算法获取最佳SVR参数,避免过拟合问题;基于一步预测误差判断方法,通过遗传算法获取最佳惩罚因子、不敏感损失函数参数和核函数参数以保证SVR回归效果,构建集气站监测数据异常识别方法;通过计算回归值和监测值之间的差值,若其大于阈值,可判断该实测值为异常,用回归值代替监测值,通过中心服务器写入到现场控制执行器中,则可防止安全系统误动作。工程应用表明该方法具有良好的适用性。
关键词:集气站;安全监测数据;异常;识别
[中图分类号]TG174. 1 [文献标识码]A
Study on anomaly identification of safety monitoring data of Gas gathering station
Zhang zitao1, Yin cunzhi2,Yang Yi1,Dong caihong1,Xiang min1
( oil and gas regulation and control center, Beijing, 100007,2. Central Asia natural gas pipeline Co., Ltd., Beijing, 100010)
Abstract: The accurate identification of abnormal value of monitoring data in gas gathering station had an important significance. Based on the support vector regression algorithm, online prediction model of the sensor was established. the optimal SVR parameters were obtained by means of ic algorithm to avoid over fitting problem. based on one step prediction method, anomaly recognition technology of monitoring data was constructed. Applied ic algorithm, the best penalty factor, insensitive loss function parameters and kernel parameters were obtained. engineering application showed that the method had good applicability.
Keywords: gas gathering station; safety monitoring data; anomaly; identification
为保障气田集气站安全高效运行,需对站内工