How To Improve The Predictability Of Your Reservoir Model With The Smallest Effort Possible (Part Two) – Addressing the “Under Modeling” issue
Before I get down to the meat of the post, I’d like to say that THE TIME HAS CHANGED for conventional oil producers.
Yes, you read it right.
With the rise of the unconventional oil, we’ve witnessed a structural change of the industry’s focus. The competition has never been fiercer between conventional and unconventional. The industry will surely boom again, but might not be in a way that many of us expected. Long-term, the $/barrel figure of your company’s mature assets will decide whether it can stay in the game.
In the short term,
The profitability of your mature fields is ruthlessly challenged. The nature decline rate of fields is on average around 8%, whereas further investment is urgently needed for future growth. In addition, you might have also noticed that the cost of developing unconventional oil has been continuously reduced due to the ever-improving efficiency globally.
All these make us want to bang our heads against the wall…
You are not alone today, if you are sharing same feelings. The question is, for us still living on the conventional side of the fence, what should we do to face the challenges from the other?
Unfortunately, there is no simple recipe for this. If there was, it must contain several key elements. On top of which, my vote goes to the superior technical ability to accurately locate the remaining oil.
Because the Pareto Principle, a.k.a. the 80/20 rule, also applies to the mature field development. Interestingly, we see in many cases most of the oil production comes from only a few good wells which are successfully landed on sweet spots. Instead, the rest and majority of wells only contribute little to the production. In other words, the key is just to place the wells at the good locations, while avoiding the bad locations.
It is just that simple, isn’t it?
No! It isn’t!
In reality, it is easy said than done. Here is why.
In mature conventional fields, the initial state of equilibrium has long been disrupted by years of injection and production. The remaining oil is shuffled and redistributed. Moreover, reservoir models, the most widely used tool in the industry for fluid flow simulation, often do not have necessary accuracy and realism to replicate and predict the actual fluid movement, let along the location of remaining oil. As mentioned in the Part I of the article, we refer to this problem as ‘Under Modeling’.
Indeed, the performance of simulation model can be improved by production history matching. Unfortunately, it is time-consuming, and the incremental nature of the data accumulation means history matching is often just a patch-fix. One of the main reasons production history matching fails is the lack of respect to spatial information. It only pays attention to local well performances from reservoir engineering perspectives, e.g. skin, PVT table etc., while the truth is the static model is poorly depicted. It means that a history matched simulation model may suggest very different spatial flow movement than what reality is.
OK, here comes the most important part.
A large amount of field data of various types is available to the mature asset teams. It has been agreed by all of us that the uncertainty is killed by jointly honoring all of them, but how to effectively and efficiently integrate these data into the reservoir model becomes the biggest stone in the way.
Most of the reservoir models for mature fields DO NOT include the abundant field data for uncertainty reduction, resulting in an “under modelled” subsurface.
Now, let us pause here.
If the mature fields are ‘under modeled’, what is an ideal reservoir model should look like?
Before we go further, there are a few important things to bear with.
With the advances in computing power and reservoir simulation technologies, we start seeing more celebrations on successful reservoir simulations over tens of millions or even billion cells.
However, it is very important not to confuse fineness with accuracy.
Figure above Is the fine model on the right necessarily more accurate than the left one?
The answer is No. A highly detailed model can be wrong and uncertain if it is not constrained properly. It only serves as a fine representation of false understanding of subsurface reality. And it is even worse since the results are more misleading due to the apparent attractiveness.
Do not get me wrong, I am not saying that the industry should stop developing powerful simulators. My point is that physics should not be replaced by powerful computers. At the end, it is the combination of accurate modelling + accurate simulation that gives us the ability to accurately locate the remaining oil.
Moreover, it is important not to confuse accuracy with precision. As our geologist advisor Raffik explained in the diagram below, precision is a term describing uncertainty, while accuracy describing the level of realism
To sum up, what we want to support profitable mature fields development is a reservoir model that fits all these descriptions on accuracy, precision, and fineness. Remember the only way to make this happen is by incorporating and honoring all the data available.
Once you overcome these problems, you become able to maximize the field production with least number of wells. The field finance will be improved, and you will regain the competitive edge with a much lower number for $/barrel.
The industry is steadily making progress in this direction. At iRes-Geo, we are contributing the leading technology to serve this idea. In the part 3 of this article, I will talk about some of the trending technologies we are actively developing.
If you think this article is useful, Stay tuned and follow our company page iRes-Geo to read more articles about this topic.
Author: Dr Yi Huang
Formerly Sr. Reservoir Geophysicist Statoil, 10 years + Experience in Integrated Reservoir Geophysics R&D and Technology Implementation
MSc. Reservoir Engineering, PhD in Reservoir Geophysics, Institute of Petroleum Engineering, Heriot-Watt University, Edinburgh
Reference: “The concept of right modelling, toward a more efficient use of the static modelling tool” by Raffik Lazar