Big Data Advanced Anlytics to Forecast Operational Upsets in Upstream Production System

Abstract

This paper highlights the development and results of an innovative tool for prediction of process upsets and hazard events associated with production operations of an oil and gas field. Summarily, this software can give recommendations on actions to mitigate or avoid operational issues, maximizing the asset value, while maintaining the highest safety and environmental quality. This in-house developed tool is based on big data analytics techniques such as machine and deep learning algorithms. The workflow developed allows predicting future events and the related influencing variables. This is done thanks to a powerful machine-learning algorithm specifically selected for the physical problem analyzed. The inputs come from a heterogeneous data-lake, composed by historical data, real-time series, maintenance reports, chemical analysis and operator experience. The workflow developed starts processing and enhancing this huge amount of data in order to train and validate the selected algorithm. Finally, the tool is fed with real-time data from the field, predicting potential events and prescribing possible actions to avoid problems that jeopardize the production and the integrity of the asset. The tool has demonstrated the capability to predict in advance operational upsets occurring within the entire production system avoiding issues, maximizing the field availability. The case illustrated in this paper focuses the attention on the process section of an upstream oil field. In particular, process upsets of the sweetening unit, such as H2S out of specification, are analyzed since they affect not only the field production, but also the asset integrity and the environmental emissions. Several Big Data Analytics have been tested and presented in this paper, along with different methodologies of input-data pre-conditioning. Results related to the application of the tool on normal operations show a significant impact in terms of down-time reduction and production optimization. The possibility to have alerts and information a few hours in advance gives to the operator the ability to reach the asset operational target, which is not only related to the management of critical events but also to the achievement of the maximum level of production thanks to the definition of an optimal configuration of operating parameters. The tool highlights also the main parameters affecting the prediction suggesting corrective actions to prevent and mitigate risks and occurring critical events. The innovative characteristics of the tool are the ability to take advantage of a huge amount of field data and to simulate complex phenomenon through mathematical-statistical methodologies, based on machine learning algorithms. Thanks to this innovative approach, it is possible to quickly predict possible hazardous events and consequently find the optimum asset configuration. This produces positive effects in the field short-term production optimization and the long-term maintenance strategies, maximizing its value and minimizing associated risks.

Supervised Anomaly Detection in Crude Oil Stabilization

“…As we have pointed out in our work, to maximize the effectivenss of the method, a business-oriented evaluation is required. In [5] the authors make similar considerations and propose an evaluation procedure for Machine Learning models trained to predict peaks in the SO2 emissions of an oil and gas treatment plant.…”

A Comprehensive Review on Big Data for Industries: Challenges and Opportunities

“…The bootstrapping algo-rithm has been used for replication of sample data and in this work the authors utilized approximately 1,000 replicate wells' data. Cadei et al developed a model that can forecast the H2S trespassing events and provide a rapid and broad solution by analyzing the main cause to early troubleshoot the fault [17]. They also discussed the effectiveness of various ML models as binary classifiers such as logistic regression, decision tree, and neural networks for forecasting the H2S trespassing events and they found that neural networks achieve high accuracy where logistic regression and decision tree improve the transparency of the forecasting.…”

“…Mineral extraction [130] Reservoirs modeling [42], management [104], evaluation [35], Upset and hazard prediction [17], [97], Casing damage prediction [78] Metal loss detection [82] Plants' health and safety [62] Business intelligence [12] techniques [64]. The architecture and process scheduling techniques of the manufacturing industry in the cloud service platform have been investigated in the context of big data analytics in the paper [51].…”

“…Ak et al [14] Adrian et al [15] Alam et al [16] Cadei [17] Desai et al [8] Wang et al [18] Alfaleh et al [19] Alguliyev et al [20] Rollins [21] Chan et al [22] Tsanousa et al […”

Human Centric Digital Transformation and Operator 4.0 for the Oil and Gas Industry

“…They analysed the collected data using eXtreme Gradient Boosting (XGBoost) and Multi-Layer Perceptron (MLP) neural networks to identify the potential operational gains. A range of studies analysed realtime data, historical data, maintenance reports, and operator data to improve O&G occupational safety and identify the underlying hidden trends [128]- [131].…”