This paper presents the development of an Artificial Neural Network system for Dynamometer Card pattern recognition in oil well rod pump systems. It covers the establishment of pattern classes and a set of standards for training and validation, the study of descriptors which allow the design and the implementation of features extractor, training, analysis and finally the validation and performance test with a real data base.

No abstract

Currently, hundreds of petroleum wells, mainly in the northeast region of Brazil, are equipped with Intermittent Gas-Lift systems (IGL), due to the high number of mature fields with low reservoir static pressure. Over the years, mathematical models based on integral analysis were developed to predict the behavior of the entire production system of IGL wells. Although these models have evolved, they do not include, in their current forms, models for some important parts and components of the well installation. This paper extends the IGL mathematical model, which was first laid out by T. Liao (1991) and later developed by Carvalho Filho (2004), including specific formulations for topics such as the throttling flow regime for the gas-lift valve, the two-phase flow on the production line, the behavior of the pressure upstream of the motor valve during the injection stage, the behavior of the flowing bottom-hole pressure while the standing valve is closed and evaluation of the gas velocity during the decompression stage. A mathematical model for the conventional IGL was built, based on Liao’s and Carvalho’s models, including the aforementioned new formulations. This new model was used as the core of a computer program, which was extended to simulate various other variants of the IGL commonly employed by the petroleum producers in Brazil: the Gas-Lift with Plunger, the Inverted Gas-Lift and the Gas-Lift with Chamber. An algorithm was written for each variant and simulations were carried out using hypothetical and real data to test the code and the models. A graphical user interface was also designed and built, making it easier for the users to input the well completion parameters and to read and display the results of the IGL simulations.