MEERA SIMULATION

harnesses the power of conventional

simulation techniques with modern
machine learning approaches
Evergreen Production Forecasting made effortless.
 Simulation

An international technology and services company, focused on enabling

digital transformation solutions for data-driven industries.

We aim to deliver top-tier solutions tailored to the needs of organization and companies of all sizes. Our mission is to
help you drive high-impact digital transformation by taking full advantage of exponential technology change and getting
to the core of multiple opportunities quickly and efficiently

2
 Simulation

MEERA Simulation, First Evergreen AI-Physics

Based Reservoir Simulator
Challenge AI-Physics Based Hybrid Solution

E&P asset operators invest significant resources on Builds Evergreen reservoir model by
producing an “optimum” history matched dynamic combining conventional simulation techniques
model for Development Planning. with modern machine learning approaches.

The “initially optimum” model becomes “dormant”

in Operation Planning which effects operational and Continuously incorporates streaming
financial performance of the asset. production data into the evergreen reservoir
model.

The “initially optimum” forecast models are

replaced by simplified Analytical Tools leading to Evergreen Forecasting environment provides
major discrepancies between expected performance super-fast self-updating “Optimum and Live”
and the field’s ongoing observed behavior. model for Operation Planning and Budgeting.

3
 Simulation

MEERA Simulation Methodology

1. Build your dynamic model 2. De-noise and prepare your data 3. Choose most effecting parameters

PETREL/ECLIPSE models can easily be imported All required data operation services are available Parameter correlation matrix helps rank parameters
upon their influence on production
4. Train history 5. Do evergreen production forecasting

Training is performed using neural networks supported Model update can be done on a daily basis using new 6
by coarse grid simulation observed data
 Simulation

MEERA Simulation Computational Core

Data Mining
Hybrid Engine

Numerical Simulator Deep Learning Training

(Analytics)

Time Successive
Prediction
GL
ESP
Reservoir Decline Input (tn) Trained Output (tn+1)
Network

De-noising &
Smoothing Well 1
Successive Substitution

Brings Intelligence Mitigates

Uncertainty Impact
9
 Simulation

MEERA Simulation Hybrid Engine

Cycle in Every Time Step

Simulator Outputs
Pressure &
Saturation Maps
Input Data For Simulator Input Data For ML
• Up Scaled Static Data • Wells’ Static Data From Reservoir
(To be prepared for the required (To be extracted by MEERA from
size of grid)
Dynamic Neural fine grid model)
Simulator network
• Wells’ Dynamic Data (using Coarser • Wells’ Dynamic Data
✓ Historical Production Profiles
Engine (To be extracted by MEERA from
Grid)
✓ Completion History the constructed or imported
✓ Future Control Modes model)

ML Outputs
Production Rates

Main Results
10
 Simulation

Tracing the error sources in conventional

simulation approach
Cumulative Absolute Error in Conventional Stages

Physics Continuum Model & Formulation Discretization Uncertain Inputs: Lack of Numerical Solution History Maching with
Navier Stokes Equations Assumptions like Data & Stochastic (without reaching Limited Observations
Convection Based Property Distribution convergence point)
Formulation, 1D Well
Model, etc.
11
 Simulation

MEERA Simulation Hybrid Engine

Cycle in Every Time Step
Less accurate than Less dependent on
conventional Simulator Simulator Outputs uncertain property
(Due to Coarser Grids) Pressure & maps
Saturation Maps

Dynamic Neural
Simulator network
(using Coarser
Grid)
Engine

ML Outputs
AI adjusts production
Production Rates profiles (and pressures &
saturations accordingly)

Main Results
12
 Simulation

MEERA Simulation Completed Projects

Client Case Specification Existing Model Key Advantage of MEERA Simulation

Brown Field, Strong Aquifer Support, Speed, Better History Matching and Forecasts,
Schlumberger
Client 1 High Water-Cut Wells, History More Accurate Remaining Oil Maps, Evergreen
ECLIPSE / PETREL
Matching Issues Model Updates

Client 2 Brown Field, Heavy Oil CMG-IMEX Speed, Evergreen Model Updates

Schlumberger Speed, More Reliable History Matching,

Client 3 Limitation of Data
ECLIPSE / PETREL Evergreen Model Updates
Schlumberger
Client 4 Brown Field, Long Simulation Times Speed, Evergreen Model Updates
ECLIPSE / PETREL
Brown Field, Fractured Structure, Schlumberger Speed, Better History Matching, Acceptable
Other
History Matching Issues ECLIPSE / PETREL Forecasts, Evergreen Model Updates

13
 Simulation

MEERA Simulation Real Case Study

Case Study Description MEERA Simulation Results

The field started production in 1985 History simulation takes 12 minutes

More than 80 production wells drilled Each prediction simulation takes 50-70
milliseconds
In this study, production data matched until
2010 and period 2010-2018 used as “blind
test” prediction The history is matched better than with the
traditional commercial simulator and in
particular the “blind test” prediction for some
The field has strong aquifer support
wells is better than the traditional simulation
HM of the same wells.
Each full field simulation takes 7-8 hours using
commercial simulators

14
 Simulation

Regression plot of cumulative oil productions VS

observed data
Production in both history & prediction Productions just in prediction

▪ History simulation is more accurate in MEERA due to better alignment of ML engine with historical data
▪ Prediction simulation is more accurate in MEERA due to mitigation of uncertainty impact

15
 Simulation

Regression plot of cumulative water productions VS

observed data
Production in both history & prediction Productions just in prediction

▪ History simulation is more accurate in MEERA due to better alignment of ML engine with historical data
▪ Prediction simulation is more accurate in MEERA due to mitigation of uncertainty impact

16
 Simulation

Comparison of STOIP maps at year 2023, MEERA VS.

commercial simulator
MEERA STOIP map Commercial simulator STOIP map

▪ More realistic results are expected from MEERA due to the more accurate oil production rates used to generate
these maps.

17
 Simulation

Oil production comparison in different wells

Producing in history Producing in both history and prediction Producing in prediction

We11 31 We11 38 We11 61

18
 Simulation

Internal Pressure Distribution

@2010 @2016

@2019

19
 Simulation

Comparison of internal pressure with MDT pressures

and Eclipse results

Meera captures pressure trends which, even if it is not accurate, is provided as input in the neural network and in
turn the neural network applies a weighing factor to each input as correction

20
 Simulation

SERVICES
Fast Model Parameter Screening Evergreen Production Forecast Locating The Remaining Oil

Use MEERA data analysis environment for identifying Use evergreen forecasting environment for simple and Use MEERA LTRO service for more reliable remaining
most influencing parameters on reservoir behavior regular model updates to better trace field development oil estimations

21
 Simulation

KEY ADVANTAGES

Evergreen Production Forecasting & Reserves Tracking: Leading to better alignment

between development and operation plans

Super-Fast Field History Matching & Production Forecasting: By combining AI with

numerical reservoir simulation

Live Infill Drilling Optimization: By generating up-to-date remaining hydrocarbon maps

Day-To-Day Field Performance Monitoring: Providing stable & auditable cash flow prediction
to the shareholders

23
For additional information email:
[email protected]

24