Perpustakaan Digital ITB

Oil production rates are frequently monitored for better reservoir management. Oil-flow measurements can be performed using several metering systems and techniques such as orifice meters and a multi-phase flow system. Production tests are often brief, in many cases, they are not representatives. If the test system is serving a large number of wells, the production well tests are infrequent. Inferred oil production rates are generated from real-time data from rod pump controller and dynamometer card. Previously measured flow rates are collected with the parameter from the rod pump controller during the test. The correlations of these attributes are captured using an artificial neural network. The application of machine learning and artificial intelligence is developed to enhance a quantitative understanding of complex data using Tensorflow, a high-level neural network programming interface. The data consist of eight attributes and one response, barrel oil per day, from with artificial lift sucker rod pump wells. The Artificial neural network involves the process of training, testing, and developing a model at end-stage. The perfect architecture is very challenging because it can influence the error and higher error reflects worst stability. The proposed model is evaluated by mean absolute error and the coefficient of determination of cross-validation to estimate the model skill on unseen data then will be used to infer the production rates.