This work proposes a new method of diagnosing the Sucker Rod Pump Systems, by analyzing segments between the points of opening and closing valves identified in the Downhole Dynamometer Cards. An analysis of the shape of the pumping condition classes represented in the Downhole Dynamometer Cards, especially regarding the behavior of segments between the valve opening and closing points is presented. The feasibility of the classification of Downhole Dynamometer Cards by the curvature characteristics of segments is demonstrated. Then, methods of classification of 16 pumping condition classes are presented, 4 by simple statistics and 12 of them extracting curvature characteristics of the segments andusing statistics to implement Mamdani fuzzy inference systems. Tests are developed with classifiers created by the method, and the test results over a library of Dynamometer Cards from real wells, previously classified by human experts, demonstrate the feasibility of the approach and the precision of the classification even without a deeper study to choose the better characteristics and statistics. In addition the classifiers were developed using samples of Dynamometer Cards from the literature, showing that the method reduces the need of many samples of the great variety of the Dynamometer Cards Shapes from real wells. Different from other methods, these results indicate the severity degree of many pumping condition classes and demonstrate the concomitant pumping condition detection feature. The classifiers were even able to spot misleading classifications by the human experts thanks to deeper understanding of the behavior of the Dynamometer Cards segments in the case of one or more anomalies in the system. The features to quantify and identify concomitant classes of so many pumping conditionsdemonstrated in this approach are novel in the area. The paper shows the feasibility of the approach and the potential for greater identification precision.

This work proposes a new methodology to identify and validate deep learning models for artificial oil lift systems that use submersible electric pumps. The proposed methodology allows for obtaining the models and evaluating the prediction's uncertainty jointly and systematically. The methodology employs a nonlinear model to generate training and validation data and the Markov Chain Monte Carlo algorithm to assess the neural network's epistemic uncertainty. The nonlinear model was used to overcome the limitations of the need for big datasets for training deep learning models. However, the developed models are validated against experimental data after training and validation with synthetic data. The validation is also performed through the models' uncertainty assessment and experimental data. From the implementation point of view, the method was coded in Python with Tensorflow and Keras libraries used to build the neural Networks and find the hyperparameters. The results show that the proposed methodology obtained models representing both the nonlinear model's dynamic behavior and the experimental data. It provides a most probable value close to the experimental data, and the uncertainty of the generated deep learning models has the same order of magnitude as that of the nonlinear model. This uncertainty assessment shows that the built models were adequately validated. The proposed deep learning models can be applied in several applications requiring a reliable and computationally lighter model. Hence, the obtained AI dynamic models can be employed for digital twin construction, control, and optimization.

To avoid premature operational failure of the ESP systems, their vibration levels must be evaluated for acceptance before installation. The vibration assessment of Electrical Submersible Pump (ESP) systems and other rotating machines considers that the most important amplitude occurs in the radial orientation. Typical vibration amplitude estimation uses radial orthogonal sensors and considers the maximum vibration amplitude between the two sensor signals. However, this approach does not account for the orbit´s shape, leading to an error up to 41,4% between the maximum estimated amplitude in each sensor and the maximum radial vibration amplitude, the orbit semi-major axis. This paper proposes a method to generate a frequency spectrum with invariance to the shape of the vibration orbit, namely the Orbit Semi-major Axis Spectrum. The proposed method is compared to the typical vibration amplitude estimation approach using individual sensor signals and the Full Spectrum method. The methods are applied on vibration signals from two pumps at different wear states, and results are analyzed. The estimation error of the vibration amplitude in ESP systems due to disregarding the shape of the vibration orbits is shown. The results indicate that the proposed method presented the maximum radial vibration amplitudes of each orbit frequency component (orbit semi-major axis), which is not directly possible with the Full Spectrum and is impossible with the individual frequency spectrum. The orbit semi-major axis amplitudes of the pumps were up to 28% higher than the maximum amplitude estimated from the individual orthogonal sensors.

This study presents practical details and results from implementing model-based predictive controllers for artificial lift systems in oil production, which utilise electrical submersible pumps (ESPs). The proposed methodology integrates the existing instrumentation architecture with new control systems and techniques. This work implements two control strategies: a standard MPC and an MPC with input targets. Both were executed in real time, encapsulated in a C++ executable, and deployed within Petrobras's supervisory and control system. The results include an analysis of the instrumentation system, a discussion of operation under unmeasured disturbances and constraints, and a comparison of the controllers. The findings are extensive and indicate that both controllers stabilise the system, ensure constraint satisfaction, and appropriately compensate disturbances.