Abstract: For the measurement of dynamic liquid level of oil wells, the traditional manual work has many shortcomings, such as low precision, poor real-time performance etc. According to analysis of actual production data, a soft sensor method based on empirical mode decomposition (EMD) and black hole-least squares support vector machine (BH-LSSVM) was proposed to realize soft sensor modeling of dynamic liquid level. In oilfield production, static model cannot fully reflect the production conditions which may lead to model failure. Therefore a dynamic model was proposed by building a performance evaluation module, which could improve adaptive ability and prediction precision. The proposed method had higher measurement accuracy of dynamic liquid level and stronger adaptive ability for production fluctuation, which met requirement of oil production. Automation in oil production was improved.

Key words: principal component analysis, dynamic modeling, optimization, EMD, dynamic fluid level, LSSVM

摘要： 针对传统人工检测方法在测量动液面时存在精度低、实时性差等问题,采用软测量技术来完成对动液面的测量工作.根据对现场数据特性的分析,提出采用经验模态分解和基于黑洞的最小二乘支持向量机预测相结合的算法来实现动液面软测量建模;通过构建模型性能评价模块,动态更新模型,解决在油田生产过程中,静态模型不能完全反映生产工况导致模型失效的问题,提高算法的自适应能力及预测量精度.最后通过对油田生产现场监测数据进行实验验证,结果表明,该方法对油田动液面测量精度高,对生产波动的自适应能力强,满足油田现场测试使用要求,提高油田生产自动化程度.

关键词: 主元分析, 动态建模, 优化, 经验模态分解, 动液面, 最小二乘支持向量机