深度神经网络在电潜泵井稳定生产中的应用

Application of Deep Neural Networks in Steady Production of ESP Wells

为了保证电潜泵油井稳定生产和提高电潜泵的运转时长,对电潜泵运行过程中的实时参数变化进行了研究。首先对渤海某油田电潜泵井生产过程中出现的气锁、油嘴堵塞、乳化和出砂时实时电机、生产和泵工况的数据特征进行分析,总结电潜泵异常时的参数特征;收集电潜泵异常生产时的参数作为训练样本和测试样本,并利用深度神经网络方法对这些样本数据进行模拟学习;最后得到用于预测电潜泵异常生产情况的深度神经网络模型。该模型通过分析电潜泵井实时生产数据,对其运行状态进行监控预警,对历史数据进行分析,辅助建立生产状态案例库,进一步保证电潜泵稳定运转。

In order to ensure the stable production of ESP oil wells and improve the operating time of ESP,the real-time parameter changes during the operation of ESP were studied.The data characteristics of real-time motor,production and pump conditions during gas lock,oil nozzle blockage,emulsification and sand production in the production process of ESP well in an oil field in Bohai Sea were analyzed,and the parameter characteristics of electric submersible pump were summarized.The parameters of abnormal production of ESP were collected as training samples and test samples,and the deep neural network method was used to simulate and learn these sample data.A deep neural network model for predicting abnormal production of ESP is obtained.This model can monitor and forecast the state of ESP running by analyzing the real-time data of ESP wells,and can also build a case library of ESP wells state by analyzing history data to maintain the steady running of ESP.