基于DA-RNN的电潜泵系统剩余使用寿命预测方法

打开文本图片集

中图分类号：TE952 文献标识码：ADOI：10.12473/CPM.202408075

Abstract：Electric submersible pump（ESP）is the main artificial lift equipment in offshore oilfields.Its operation and maintenance costs are extremely high.Once it gets outof order，it willcause losses tooilfieldoperations.Therefore，a remaining useful life（RUL）prediction method for ESP system based on a Dual-Stage Attention-Based Recurrent Neural Network（DA-RNN）was proposed. First，the DA-RNN was used to perform feature extractiononreal-time data of ESP to build a remaining useful life prediction model of ESP.Then，the model was used to conduct accurate prediction on the RUL of ESP，providing a scientific basis for predictive maintenance of ESP and significantly enhancing equipment reliability and safety.Finall，a case analysis was conducted for Bohai Oilfield，showing thatthe average prediction eroroftheRUL prediction model is within 28 days， verifying the practicality and accuracy of the DA-RNN based model in predicting the RUL of ESPs.The research conclusions provide data supportforfault prevention and maintenance decision-making of ESPs in ofshoreoilfields，and also provide an efficient data-driven strategy for operation management.

Keywords：ESP system；remaining useful life（RUL）；DA-RNN；prediction model;hyperparameter optimization;Pearsoncorrelationcoefficient

0 引言

电潜泵生产系统是一种将电动机和多级离心泵一起下入井内，并利用电缆将地面电能传输至井下电动机，驱动多级离心泵将井筒中的原油举升至地面的人工举升系统[1]。（剩余11584字）