How does MEERA CENTRUM work ?
TRULY UNDERSTAND
You can’t achieve the highest return unless you know the actual capacity
MEERA CENTRUM's Evergreen model combines the conventional numerical simulation with advanced deep learning techniques. The former provides a physics-based foundation that is lacking in pure data-driven approaches and the latter delivers insights with learning the production history rather than tedious history matching. In this way, a better understanding of the asset can be achieved in a significantly shorter time period with more realistic forecasts.
MEERACENTRUM
Oil & Gas Asset Management
1. ANALYSIS & SCOPING OF CLIENT NEEDS
Understanding the asset is crucial in studying the reservoir and forecasting the production. In this initial step, Asset owners’/Operators’ problems and challenges besides the practical solutions will be clarified. Also, in this step, the client shares the reservoir model.
2. DATA ENTRY
ECLIPSE and CMG format can be imported into MEERA Centrum. For MEERA Centrum to have a reliable production forecast, it is vital to collect high-quality observed hydrocarbons and water production data from the field. The more available data, the better the AI model can be tuned.
3. BUILDING THE PHYSICS-BASED FOUNDATION
MEERA Centrum utilizes numerical simulation to provide physics-based support to avoid the pitfalls of relying on data-driven approaches. However, employing excellent AI capabilities, numerical simulation can be released from the task of capturing physics accurately. That is why it is recommended to upscale the fine-grid model to a coarse one to reduce the problems and accelerate the simulation run.
4. DATA MINING
The next step is to create a data set in which different parts of data, including static, geometric, history observations and, simulation results, will be set in a well-wise manner. For data mining, MEERA CENTRUM offers several Data Analysis tools to analyze and perform some operations on the data for feeding it into the hybrid model, including MEERA Centrum’s unique Correlation analysis.
5. TRAINING THE AI-PHYSIC HYBRID MODEL
It is time to build a machine learning pillar after preparing the dataset as fuel for MEERA Centrum’s hybrid engine. In this step, there should be some features for training. In addition, machine learning hyper-parameters tuning should be performed, though wisely selected default settings provided in MEERA CENTRUM work appropriately in most cases. EverGreen model repeatedly trains until it gets precisely close to reality.
6. PREDICTIVE VALIDITY
Standard performance metrics such as learning curves, accuracy, precision, and recall will be used to evaluate the trained model. Furthermore, the predictability of the trained model can be more assessed with a blind forecast. In other words, 60-70% of observations will be used in training, and the rest of the data will be checked with trained model responses. Finally, a pretty Healthy Asset will be obtained.
Now Significant
Reduction in Production Forecast
Error will be achieved
0.62%
Actual Vs ExpectedIn oil production error
0.01%
Actual Vs ExpectedIn water production error
0.87%
Actual Vs ExpectedIn gas production error
“We tested Meera Centrum on a highly complex field with many reservoirs and were impressed by its fast turn-around in delivering good history match in several weeks which has been achieved in several months with traditional conventional simulation workflows.”
Subsurface Lead, Petoro AS.
