ISI Logo

Smart Proxy Models™

Highly Accurate Proxy Models that Preserve Complex Physics and High Resolution of the Numerical Simulation Using Artificial Intelligence and Machine Learning.

Smart Proxy Model is a realistic, tested, and validated alternative to traditional proxy models known as Reduced Order Models (ROM) or statistical Response Surfaces. Samrt Proxy is a technology that combines reservoir engineering, reservoir modeling, and reservoir management expertise with state of the art Artificial Intelligence and Machine Learning in order to build highly accurate and very fast proxy model for highly complex numerical reservoir simulation models.

Traditional Proxy Modleing

The main problem that contributes to the lack of success of the traditional proxy models is the similarity in the scientific paradigm. In other words, traditional proxy models and numerical simualtion both use computational science for their problem solving purposes. As long as the same paradigm is used, there is only so much that can be accomplished. Smart Proxy Modeling introduces a complete paradigm shift in engineering problem solving. That is why majority of the traditional scientists have called smart proxy modeling as "too good to be true".

Reduced Order Models (ROM): In reduced order models either the complex physics of numerical simulation or its resolution in space and in time is compromised in order to increase the speed of model execution. In other words to gain the speed, the quliaty of the original model is compromised.

Statistical Response Surface: The mani basis of response surfaces are statistics. Unfortunately, many people, fail to realizze the major differences between statistics with AI and machine learning. Statistis have been around for more than a century. If statistics could have been an alternative for engineering problem solving, it would have happened long time ago. Correlation is not the same as Causation.

Well-Level Smart Proxy Modleing

The well-level smart proxy replicates the complete profile of oil, gas, and water production as well as reservoir pressure and water saturation for every well that is modeled by the reservoir simulator, including drilling of new wells throughout the reservoir in the future. The well-level smart proxy accomplishes this task at very high accuracy (more than 95% accuracy) and in fraction of a second. Tens of thousands of simulation runs can be executed for a given well on a laptop (or even on your tablet or smart phone) in minutes. Following is an example of a well-level smart proxy developed using ARAMCO's "Powers" numerical reservoir simulation for an offshore field. Similar well-level smart proxy models have been develoed successfully using all commercial numerical reservoir simulation models.

Cell-Level Smart Proxy Modleing

The cell (grid block)-level smart proxy replicates changes in pressure and saturation for every grid block in the reservoir simulation model. The cell-level smart proxy accomplishes this task at very high accuracy and in seconds. Tens of thousands of simulation runs can be executed for a given well on a laptop (or even on your tablet or smart phone) in minutes. Following is an example of a cell-level smart proxy developed using ECLIPSE numerical reservoir simulation for an onshore field in the Middle East. This example shows the changes in water saturation for in a model with more than one million active grid blocks. Similar cell-level smart proxy models have been develoed successfully using all commercial numerical reservoir simulation models.