Beruflich Dokumente
Kultur Dokumente
Revolutionising the scenarios in minutes, all based on the experience of the asset’s 400 (a)
MEERA Al
petroleum engineers. A reservoir model that is constantly 350
Simulator
being updated has the major advantage of allowing reliable
Prediction of Reservoir
Observed
Performance
SINA MOHAJERI and shareholders. 200
Reservoir models are vitally important in the oil and gas subsurface data with machine learning acquired through the reservoir may impact the production rates. 100
(b) MEERA Al
industry, for example for estimating remaining oil and gas stream of ‘big data’ coming from production. 90 Simulator
reserves and providing production forecasts. Managing the Case Study 1: History Matching Observed
80
operational and financial performance of an E&P asset is A Revolutionary Breakthrough The benefits of the utilisation of a system like this can be seen by
reliable prediction of the dynamic reservoir behaviour. All 3D, 3-phase numerical reservoir simulator coupled with a field that had started producing in 1985 and now has more than 60
this information is crucial for creating development and complex deep learning engine. The first step is to simply 80 production wells, with strong aquifer support. In this case 50
operational plans and providing an auditable cash flow to integrate an asset’s optimum dynamic reservoir model into study, production data was matched up to 2010 and the period 40
shareholders. It is therefore important to have an up-to-date Target’s platform MEERATM, and its AI-Simulation framework 2010–2018 was used for blind test prediction. Each full field 30
and accurate reservoir model. therein. This is done by re-assembling the components of the simulation run took 7 to 8 hours using commercial simulators,
20
However, E&P asset operators have an issue, in that their original conventional model, so it meshes into the machine but when using MEERA Simulation, each history simulation
optimum history matched dynamic reservoir models are learning framework. Various parameters are extracted took 12 minutes, while prediction simulations took 50–70 10
becoming quickly out-of-date and ‘dormant’, due to the length from the original reservoir model, some of which, based on milliseconds, all on the same IT infrastructure. 0
1996 1998 2001 2004 2006 2009 2012 2015 2017
of time it takes to update models. This means that frequent a comprehensive correlation analysis, are selected as key These faster results meant the engineers could run
40
model updates are simply unaffordable. Engineers therefore reservoir elements to be used for machine learning training multiple simulations in minutes to generate lots of different (c) MEERA Al
Simulator
often resort to using simpler approximation models to and prediction. New production data is then incorporated, potential scenarios, thereby allowing for better optimisation 35
Observed
forecast production and remaining reserves, as these methods creating a rapidly self-updating live dynamic reservoir model of development strategies. In this case study, the field history
30
are faster and cheaper to use, although at the price of losing that can be used as a reliable production forecasting and proved to be matched better using this system than with the
This industry dilemma led engineers and scientists of It takes just minutes to train the AI using historic production test prediction was more accurate than the traditional simulation
20
the international technology and services company Target data, while successive prediction simulations take merely historical-matching of the same wells. The better the history
Energy Solutions to develop a new hybrid modelling method milliseconds. matching quality, the better the understanding of the reservoir, 15
face. Confronted with this problem, a company that had been Case study 1: Comparison of MEERA vs. conventional simulator in
using a model built on a reservoir modelling tool suite widely predicting oil production rates per well : (a) example in history match
used in the industry decided to test using MEERA Simulation mode only, i.e. prior to 2010; (b) in both history match and prediction
mode; and (c) well example in prediction mode only.
to build their models. As a result of the machine learning built
into the software, the lack of data was not such an issue and the
client found that not only was it quicker to build models through Evergreen Model Predictions
this system, but that the history matching was more reliable than Field development decision making needs reliable
before. However, it must be emphasised that although machine predictive calculations, so a hybrid approach to reservoir
learning can fi ll in data gaps, the more data that is available, the modelling, using a fully coupled simulator with an AI/
more accurate the outcome will be. machine learning engine, is an obvious way forward. The
A key feature seen across both these case studies is that a result benefits from the history matching capabilities of
tool like this, that includes machine learning, has the ability to artificial intelligence combined with the proven forecasting
integrate with other industry reservoir modelling applications, capabilities of a conventional reservoir simulator, while
allowing it to read and write from these applications using data- preserving local and global physical consistency.
exchange industry standards such as RESCUE. In addition, the With their evergreen model predictions and machine
tool also possesses a wide range of professional and easy-to-use learning abilities, tools like MEERA Simulation have the
data-handling utilities and visualisers, including 3D and 2D potential to revolutionise the way that reservoir models
Target Energy Solutions models, a chart visualiser and a histogram visualiser. are built and producing fields are developed.