油气藏评价与开发 ›› 2018, Vol. 8 ›› Issue (6): 28-32.

• 油气藏评价 • 上一篇    下一篇

基于RBF神经网络的井筒流动工况预测

王浩儒,李祖友,鲁光亮,唐雷   

  1. 中国石化西南油气分公司采气一厂,四川 德阳 618000
  • 收稿日期:2018-01-07 出版日期:2018-12-26 发布日期:2018-12-13
  • 作者简介:王浩儒(1990-),工程师,天然气开采技术研究。
  • 基金资助:
    国家科技重大专项"低压低产气藏复杂结构井排水采气关键技术"(2016ZX05048-004-004)

Prediction of wellbore flow condition based on RBF neural network

Wang Haoru,Li Zuyou,Lu Guangliang,Tang Lei   

  1. No.1 Gas Production Plant, Southwest oil and gas Company, SINOPEC, Deyang, Sichuan 618000, China
  • Received:2018-01-07 Online:2018-12-26 Published:2018-12-13

摘要:

川西气田气井普遍产水,对气井稳产影响较大。及时掌握气井井筒流动工况可以判断气井积液情况,指导排水采气措施的制定。常规地采用井下压力计的气井工况判断方法存在时效性差、成本较高等缺点,而常规两相流理论对于不能连续携液气井压力分布预测亦存在较大偏差。通过搜集大量气井生产数据和流压测试资料训练形成了基于RBF神经网络的井筒流动工况预测模型,RBF网络具有结构自适应性,输出不依赖于初始权值的特性,15口井预测结果与实际结果符合率为86.67 %,表明应用神经网络模型预测气井井筒流动工况较为可靠,可以用来指导生产。

关键词: 神经网络, 流动工况, 气井积液, 排水采气

Abstract:

Water generally exist in the gas well in west Sichuan gas field, which has a great influence on the steady production of gas wells. The timely grasp of the flow conditions of the gas wellbore can judge the fluid accumulation in gas well and guide the formulation of the drainage gas recovery measures. By using the pressure meter to judge the downhole conditions of the gas well had the disadvantages of poor timeliness and high cost, and the prediction of two phase flow theory often had deviations from the actual existence. So we used a large number of gas well production data and flow pressure test data to establish the prediction model of wellbore flow condition based on RBF neural network. The RBF network had the self adaptability and its output do not dependent on the initial weight. The coincidence rate between prediction results and actual results among 15 wells was 86.67 %, which indicated that the neural network model was reliable for predicting the wellbore flow conditions of gas well, and could be used to guide production.

Key words: neural network, flow condition, gas well liquid loading, drainage gas recovery

中图分类号: 

  • TE319