Production Assurance with Enhanced Decision Support powered by AI & Digital Twin


Author: Srikanth | Chief Customer Officer, Flutura


Gathering system optimization has become a necessity than a mere option. Systems approach is incorporated to emphasize the need for production assurance by optimizing all aspects of production facilities. Technological advances have enabled the industry to produce natural gas at never-seen-before rates by maximizing the potential from assets and ensuring optimum flow rates across the field. However, these pose plenty of challenges to match the demand at sales point – constantly fluctuating oil and gas price, production inefficiencies, cost pressures squeezing the margins and limited personnel on the field.

The uncertainty in the behavior of the challenges needs a connected system that helps in making dynamic decisions to suit the situation. For consistent and faster problem recognition, AI-driven digital twin is proposed to optimize the production of the gas gathering network.

Cerebra Digital Twin was applied in a Complex Scenario Simulation and Advanced Anomaly Detection to help one of the International Oil and Gas majors, based out of the USA, to reduce production deferment with enhanced decision support.

-        95% reduction in operator’s time to identify network anomalies 

-        97% accurate prediction results across the network

-        5-10% improvement in the simulation results

-        5-7% reduction in unscheduled production deferment

Why is there a need for advanced technologies?

With time, operators face challenges to adhere to the demand, due to inefficient production surveillance, production optimization and capital project evaluation.

Operators are now collecting and managing huge amounts of data streaming from their operations as part of their digital journey. The conventional tools used in practice do not support automated simulation and analysis. The tool lacks to refresh at a faster rate and show any trends of the dynamically changing real-time data. There are times leading to missed insights and details by operators and conventional tools.

We spoke to a Production Optimization Lead at a leading operator, he said "I get 2 AM calls from the operations team to help overcome flow issues in the network , help them with compressor calibration to achieve desired flow rates.. I wish there were tools to detect issues before hand and with possible recommendation on actions that can be taken.”

Why Digital Twin is so effective?

Digital Twin uses Physics, Heuristics and Machine Learning based models for analysis and provides anomaly detection based statistical alarms. It provides situational awareness to the operators that can access data on production or any other objectives to be fulfilled across geography. To achieve an optimized flow across the gathering system during normal and rerouting situations, operators must consider both the scenarios -

  1. Detection of reduced equipment/well performance
  2. The possibility of accumulation across the gathering pipelines

For an effective and faster way to identify the actual causes and achieve right flow in the network, AI-powered Digital Twin is introduced. It is based on automated data flow ingesting data from historian/SCADA system into hybrid models (first principle and machine learning models) to simulate scenarios on the field.

Digital Twin provides a digitally accessible skeleton of the physical network to access operational KPIs and metrics. For an operator, it provides a comprehensive view of the state of the network to improve their situational awareness, detect critical field assets, achieve correctness of simulation models, and decision support by suggesting accurate scenarios to be implemented in the field all in real-time.
This enhances the efficiency, reduces the time of operators’ day-to-day field activities, and promotes overall performance of the gas gathering systems.

Critical AI Systems Need Data Care

Due to the complexity of the gas gathering system operators are not able to get a holistic picture of the network. There is only so much information collected manually based on which decisions are made. As all the data points are not viewed at once, correlating multiple parameters are usually missed out. Ineffective decisions lead to loss of revenue. Most of the issues like pipeline congestions, salespoint flow rate degradation, production deferment, compressor operating sub optimally are generally resolved manually. This is time consuming and cumbersome adding latency in detection, hence reducing the velocity of the decision- making process. A digital model with reliable data helps prevent any losses due to these challenges.

Digital Twin houses AI-driven machine learning models, deep learning-based neural network algorithms for predicting operational parameters and scenario of the field changes, while these models are trained on steady-state simulation results and field sensor data.

The data sets comprise of operations data and stationary information sourced from sensors capturing live data, static data (compressor curves, water-gas ratios), daily and hourly captured data. Pre-processing of these data is performed to handle data quality issues (missing data, the irregular format of data, etc.) improving the model performance upon training.

Value Drivers of the Digital Twin

Overall performance of the network improvement is realised with digital twin as it provides pivotal insights on crucial operational parameters such as operational wellpads, operational compressor stations, sales points flow rate, wellpad anomalies, line pressure anomalies, and more.

AI-powered solutions are generally used to improve reliability, efficiency, and operational excellence in a proactive manner. Thus, predictive, and prescriptive analytics play vital constituents of the Digital Twin. The analysis of crucial operational parameter and field asset behaviour overtime helps in predicting any deviation from the normalcy. Here predictive analytics is used to identify critical field asset or any operational parameter, acting berserk, early and notify the same to the operators. To optimize field trials, parametric settings are suggested by the machine learning models for a better decision support.

In either of the situations, Digital Twin enhances the efficiency of the engineers’ and operators’ decision-making process.

Exception Based Real-time Critical Field Assets Identification.

Operators are intimated to act only when there is a critical alert due to the detection of anomaly. When the flow rate of a wellpad suddenly declines, it indicates subsurface event like liquid loading, sand accumulation, etc. in one of the wellheads. The operator quickly detects this critical wellpad and intervenes to improve the flow rate.

When a pipeline section experiences a sudden drop in differential pressure, it indicates the possibility of accumulation like liquid holdup, wax deposition, etc. After the identification of critical pipeline sections, the operator performs pigging to improve the flow across that section.

There is a high probability that performance of the assets declines for some hours on a daily basis. By considering daily frequency of flow rate and differential pressure in the anomaly detection machine learning module these daily asset declines or unusual variations will not lead to a false alarm.

Cerebra: Advanced Anomaly Detection


Simulate Scenarios with What-If Analysis

Digital twin helps the operator to accurately simulate the future what-if scenarios through machine learning augmented simulation output results. The scenarios considered for what-if field scenarios are:

  1. Compressor lease expiry – Compressor removal/replacement
  2. Offset well shut in
  3. Preventative maintenance - Compressor Maintenance, Pigging Operations, Wellpad/Wellhead Maintenance
  4. Adding new wellhead/wellpads
  5. Sales point contract updates
  6. Pressure/flow adjustment
  7. Contract expired at sales point
  8. Strategic decision to divert to other sales point

The operator can make a single or combination of changes in the network like shutting a wellpad, changing the speed of a compressor at a compressor station, etc. These changes are tracked and stored as a part of scenario log that can be accessed at any point before viewing the simulation results to understand the scenario trying to simulate.

The predicted flow rate of the sales points is visible on the network. The associated changes in the network for any what-if scenario can be viewed by selecting a field asset & viewing its output predicted operational parameter.

Simulation to demonstrate what-if analysis


Business Impact

Production Optimization

The digital twin solution reduced the unscheduled part of production deferment by about 5-15% with the help of real-time anomaly detection leading to early intervention.

Enhanced Decision Support

Accuracy of machine learning model is 10 -15% better than first principle only model. This helps improve the decision-making accuracy.

Model Fidelity Index for Impact Analysis


Improved Productivity

The automation of data workflow across the digital twin led to no manual input required for any simulation. The automation led to improved productivity of the operator in making the changes on the network to investigate the impact of scenario on the network, by reducing the time of simulation from 5 hours to mere 5 minutes.

HSE impact

The digital twin is modeled to capture and predict the behavior of the asset in advance reducing the manual intervention and efforts. Early warning signals help to prevent unwanted critical events from occurring - minimizing operators’ efforts and improving their safety.  For an operator, digital twin aids in improving their situational awareness, detecting critical field assets, achieving correctness of simulation models, and decision support by suggesting accurate scenarios to be implemented in the field all in real-time. This enhances the efficiency, reduces the time of operators’ day-to-day field activities, and operator satisfaction with less frustration.

The gas gathering from the wellheads and processing operations downstream significantly impacts the marginal cost of gas production. AI- and cognition-driven Digital Twin implements new, innovative ideas as well as commonly known concepts into their processing facilities to boost their efficiency and decision making.

In conclusion

Digital Twin enhances the decision making of operators in avoiding any event of contractual obligation in terms of minimal volume commitment (MVP) with the midstream operators.
The automation of data workflow across the digital twin leads to no manual input required for any simulation. The automation aids in improved productivity of the operator in making the changes on the network to investigate the impact of scenario on the network, all at a reduced time.

It reduces the unscheduled part of production deferment with the help of real-time anomaly detection leading to early intervention. The digital twin model is a single platform to view holistically, compare the performance of equipment and perform simulation and prediction across the network.


About Flutura

Flutura Decision Sciences and Analytics, an Industrial AI company tuned for IIoT focuses on core business objectives of "Asset Uptime" and "Operational Efficiency" with Cerebra.

Flutura provides suite of product technologies for Oil & Gas, Process Manufacturing, and Heavy Machinery industries.

  1. Cerebra Digital Assistants has diagnostics & prognostics modules for processes and assets.
  2. Cerebra Vision Intelligence, based on video analytics, has modules for quality inspection, safety violations, and other event detections.
  3. Engineer’s WorkBench equips industrial engineers with Operational and Engineering insights using data science without having to code.

Cerebra is ranked #1 in Gartner Peer Insights Voice of Customer (4.8/5)

To learn more about this amazing technology click here: