Research and Application of Big Data Production Measurement Method for SRP Wells Based on Electrical Parameters

Abstract

Well metering is an important part of daily oilfield management. For wells in a block, production metering can help reservoir managers fully understand the changes in the reservoir and provide a basis for reservoir dynamics analysis and scientific field development planning. For single-well metering, accurate producing rate can help oil well operators optimize the well production system, improve the efficiency of oil wells, and even discover abnormal conditions in oil wells based on changes in production. In order to obtain accurate well production, over 300 SRP wells in an experimental area of an oil field in northeastern China are tracked and measured in this paper. Easily available continuous electrical parameter data (including electrical power, current and voltage) and real-time output of the wells were selected as training parameters. We separated the SRP well electrical curves and corresponding real-time production data into a set of samples by one-stroke time, and obtained a total of 200,000 valid samples. The production status of the pumping wells was classified by deep learning, and the electric curves were Fourier transformed to extract statistical features. Then, we performed deep learning on these samples, using production parameters as input vectors and well fluid production as output results. Finally, good results were obtained by training and a model for calculating SRP well production based on big data was developed. The model was used to calculate the production of SRP wells in an experimental area of an oil field in northeastern China and compared with the actual production data. For low-producing wells with daily production less than 6 m3, the error of the model was less than 0.5 m3 /d, and for wells with daily production greater than 6 m3, the relative error of the wells was less than 10%, which met the expectation of managers. Compared with the methods mentioned in this paper, the currently used measurement methods, such as flowmeter measurement and volumetric measurement, have limitations in terms of instrumental measurement range and real-time measurement, respectively. In addition, both of these methods increase the construction cost of flow measurement systems. The big data production measurement model provides operators with a method for optimizing the production system of oil wells and also provides signals for early warning of oil well failures. This method can help managers achieve cost reduction and efficiency increase. The processing and application methods of electrical parameters in this paper can also provide ideas for production prediction of PCP o ESP wells.