No abstract

Typically, rod pump system failures are determined using the dynamometer card which may miss some early warnings. This paper presents a novel approach for early failure detection in rod pump wells using more than 14 parameters that indicate the daily functions of rod pump wells and employs advanced machine learning techniques. Our system recognizes failing, failed as well as normal situations by learning their patterns/signature from historical pump data, that include card area, peak-surface load, minimum-surface load, daily run-time, and production data. These data are automatically pre-processed using expert domain knowledge to reduce noise and to fill-in missing data. Our approach is novel in two ways. First, our machine learning algorithm AdaBNet uses boosting to learn several Bayesian Network models and then combines these models with different weights to form a stronger boosted model. Second, our approach generates this single boosted model that is applicable across all the wells in a field, as opposed to well-specific approaches that generate one model per well. This model detects anomalies, prefailure and failure signals and generates corresponding alerts. Early fault detection in rod pump wells is useful for automatic monitoring of large number of assets remotely, and could be extended to other artificial lift systems. We used a training data set of 12 wells to construct the learning model for the AdaBNet algorithm and tested the algorithm on 426 wells from the same field. The results show that our algorithm detects failures with accuracy higher than 90%. This framework can help field operators not only to remotely recognize and predict failures in advance, but also to help prioritize the available manpower, save significant time, reduce operating expense (OPEX), downtime and lost production. Early fault detection in rod pump systems can allow for proactive maintenance that can delay and even prevent future well failures. The proposed algorithm can enable production engineers remotely detect failures and anomalies before they occur, and assess the situation at control centers before taking any remedial or corrective actions. This approach to using a single model for an entire field is superior to other approaches with individual model for each well.

In the petroleum industry, multivariate time series data is commonly used to monitor the performance of their assets, in which wells artificial lift systems are among the key assets that bring oil up to the surface. Failures frequently occur among these artificial lift systems, and they can greatly increase the operational expense due to loss of production and cost of repairs (also known as workovers). Predicting these failures before they occur can dramatically improve operational performance, such as by adjusting operating parameters to forestall failures or by scheduling maintenance to reduce unplanned repairs and to minimize downtime. Artificial lift failure prediction problem poses interesting challenges to data mining algorithms, because of the many real-world data issues, such as noise, missing data, delay of failure event logs, and large variability among normally functioning well artificial lift units. This paper presents the Smart Engineering Apprentice (SEA) framework that incorporates robust feature extraction algorithm, clustering and semi-supervised learning techniques, to enable learning of failure/normal patterns from noisy and poorly labeled multivariate time series, while achieving a high recall and precision for failures for real-world dataset.

This paper presents a new generalized global model approach for failure prediction for rod pumps. By embedding domain knowledge into an Expectation Maximization clustering algorithm, the proposed global model is able to statistically recognize pre-failure and failure patterns from normal patterns during the training stage. Compared with previous field-specific models, the enriched training set for the global model learns from much larger scale of normal and failure examples from all fields with which a generalized Support Vector Machine (SVM) is trained and is able to predict failures for all fields. The data set for this paper is taken from five real-world assets using rod pump artificial lift systems which contain nearly 2,000 rod pumps. The results show that the global model of failure prediction is capable of capturing future rod pump and tubing failures that produce acceptable precision and recall. The resulting global model is scalable and can be used for predicting failures for proactive maintenance to reduce lost oil production. Results from our case studies with multiple fields data show that precision and recall are better than 65% with this global model. Our prior work [1] used machine learning techniques to generate high quality failure prediction models with good accuracy. However, these efforts suffer from two major drawbacks. First, this used machine learning techniques that require labeled datasets for training the model. Generating these labeled datasets is human-intensive and time-consuming. Second, this model is field-specific which is only applicable to the specific field from which the labeled dataset is derived. These field-specific models generally perform poorly on other fields because of the differences in the data characteristic caused by field geology, operational procedure, etc. Moreover, these models have to be maintained independently, which accordingly raises nontrivial maintenance costs.

Electric Submersible Pumps (ESPs) are complex, high power electromechanical systems that operate in extreme environments under very tight geometric constraints. Despite the efforts of suppliers and operators ESPs will eventually fail, requiring replacement. In addition, it is sometimes necessary to intervene in a well to optimize or remediate production – a process not possible, or limited, with an ESP restricting access to the reservoir or lower completion. The process of installing or removing an ESP has changed little since their introduction almost 90 years ago. Removing (or installing) a conventional ESP requires pulling the tubing string, replacing the ESP and reinstalling the tubing string with the use of a rig or pulling unit (a "heavy workover" or HWO). This expensive, time consuming process creates significant logistical challenges and QHSE risks along with significant lost production. Increasing use of ESPs, along with today's difficult economic environment has brought a renewed interest in the potential of alternatively deployed ("rigless") ESP systems to improve project economics by reducing the cost of interventions while simultaneously increasing the oil produced per well. Evaluation of the economics associated with alternative deployed systems is not straightforward, with many nuances which are not obvious. This paper seeks to develop a framework which can be used to compare the economics of wireline retrievable ESP systems (WRESP) to conventional, tubing deployed ESP systems, taking into account the uncertainties surrounding the reliability of the various system components (both conventional ESP and WRESP) through the use of Monte Carlo simulation. This framework will be used to compare the likely economic outcomes in six ‘typical’ operating environments: West Africa Offshore, Middle East Land (high rate and moderate rate), Asia Offshore, Alaska, and US Land (lower 48) using (to the greatest extent possible) actual reliability data and operating parameters in these environments. The production and economic outcomes of both conventional ESP and WRESP deployments will be compared and presented along with the sensitivity of these outcomes to typical uncertainties in each of these environments.