A Working Condition Diagnosis Model of Sucker Rod Pumping Wells Based on Big Data Deep Learning

Abstract

Abstract Monitoring the working conditions of the sucker rod pumping wells in a timely and accurate manner is important for oil production. With the development of smart oil fields, more and more sensors are installed on the well. The variety and volume of the monitoring data are big. The oilfield big data can be utilized to improve the diagnostic performance. In this work, we aim to utilize the big data collected during oil well production and a deep learning technique to build a new generation of intelligent diagnosis model to monitor working condition of sucker rod pumping wells. Over 5 million of well monitoring records, which covers information about one year of an oil field block, are collected and preprocessed. To show the dynamic changes of the working conditions for the wells, the overlay dynamometer card is proposed and plotted for each data record. The overlay dynamometer card stacks two dynamometer curve at different times. Based on the overlay dynamometer cards, the working conditions are divided into 30 types, and the corresponding dataset are created. An intelligent diagnosis model using the convolutional neural network (CNN), one of the deep learning framework, is proposed. By the convolution and pooling operation, CNN can extract features of an image implicitly without human effort and prior knowledge. That makes CNN very suitable for the recognition of the overlay dynamometer cards. The architecture for working condition diagnosis CNN model is designed. The CNN model consists of 14 layers with six convolutional layers, three pooling layers, and three fully connected layers. The total number of neurons is over 1.7 million. The overlay dynamometer card dataset is used to train and validate the CNN model. The accuracy and efficiency of the model are evaluated. Both the training and validation accuracies of the CNN model are over 99% after ten training epochs. The average training elapsed time for an epoch is 8909.5 seconds and the average time to diagnosis a sample is 1.3 milliseconds. Based on the trained CNN model, a working condition monitoring software for suck rod pumping well is developed. The software runs 7×24 hours to diagnosis the working conditions of wells and post warning to users. It also has a feedback learning workflow to update the CNN model regularly to improve its performance. Through 3 months of on-site test run show that the actual accuracy of the CNN model is over 90%.