摘要
Abstract
Aiming at the problems of relatively low production pump efficiency,high lifting power,and high energy consumption in electric submersible pump wells at present,a method for evaluating and predicting the health of electric submersible pumps based on artificial intelligence was proposed,including data preprocessing,optimization of main control parameters for health,construction of health index,health classification,and health prediction.A health prediction method based on deep learning neural network was established,The health of electric submersible pumps is divided into three stages:health,sub health,and failure;The research on the health prediction model of electric submersible pump wells based on long and short term memory network(LSTM)has been carried out,which is of great significance for guiding equipment selection and working condition diagnosis and early warning.This paper takes the collected data from 10 faulty electric submersible pump wells and 4 normal wells on a platform of CNOOC Tianjin Branch as the research object,determines four main control parameters that can reflect the health level of electric submersible pump wells,calculates and analyzes the health index,and identifies the type of operating conditions and health level of electric submersible pump wells.Research has shown that the health prediction model for electric submersible pump production wells based on long-term and short-term memory neural networks(LSTM)has high prediction accuracy,can achieve real-time and accurate evaluation and prediction of the health of electric submersible pump operation,comprehensively and intuitively display the overall production status of a single well or oil production plant,and make the production management of electric submersible pump wells more refined,It has a good guiding role in the evaluation of production status and condition diagnosis of electric submersible pump wells.关键词
电潜泵/健康指数/健康度/长短时记忆神经网络LSTM/特征分析/预测Key words
electric submersible pump/health index/health level/long and short term memory neural network LSTM/characteristic analysis/forecast分类
能源科技
