# Data-Driven Permeability Determination with Logs

Updated: Mar 28, 2019

#permeability #logs #petrophysics #prediction #reservoir

Determining the permeability within reservoirs can help us understand more about how the fluid moves, which direction it’s going and how fast or slow it flows. There have been various methods in estimating permeability across the years, starting from regression methods to database approach, fuzzy clustering techniques and artificial neural networks. Within this article, we discuss an approach to predict the permeability within reservoirs using log data and a pattern recognition algorithm, AiSara.

## Lets us show you how it's done

Here we have a data from a well that was drilled in a field. The graph displays the permeability of the formation at the given depth of the reservoir. The depth shown is between 9632ft to 11900ft, and we do not know the permeability of the formation between 10300ft to 10800ft. Thus, we would like to show how we can determine the permeability with the data that we have, and AiSara.

Let us look into our data. It consists of

Depth, ft

Water Saturation at the specified depth, fraction

Porosity of formation at the specified depth, fraction

Gamma Ray reading at the specified depth, API

Finally as the output, permeability in mD

The permeability output is the actual permeability of the reservoir respective of depth whereas the input data are taken from the logs of the well.

There is one variable that has no relation with the permeability of the formation, and that is the depth.

Thus, porosity, water saturation and gamma ray has relations/patterns between them that AiSara can learn from. Thus, with AiSara, you must almost always start with Learn.

We learn the inputs which are water saturation, porosity and gamma ray to find the relation with permeability, which is the output. Here is an extra tip to get the better prediction with AiSara. The better prediction for AiSara can be achieved when AiSara predicts any value between its maximum and minimum data that it learns from. For example, if the input data of porosity minimum is 0.2 and maximum is 0.3, then AiSara can provide good prediction of the permeability, which is the outcome variable, when the porosity input is between 0.2 to 0.3. Otherwise, if the porosity input is below the minimum, or above the maximum, then AiSara will extrapolate the values, reducing the accuracy permeability prediction of the formation.

Therefore, to achieve a good model for AiSara to learn, remember to provide data that is within the maximum and minimum values.

Once AiSara learns, we can proceed to predicting the permeability of the formation between 11500ft to 12150ft. Let us see how the results look like.

## Objective Achieved

And here we have it, AiSara has managed to predict the permeability of the formation between 10300ft and 10800ft.

Bear in mind the model of AiSara can be changed based on the inputs that was given to her. For example, there are more logs that could help determine the permeability of the formation such as logs that determine the type of formation within the reservoir. Since the type of formation has its relation with the permeability of the rock, it can further improve the permeability prediction model of AiSara once she learns from the data.

These are just one of the many examples of what AiSara can do. Thank you for spending the time to ready this article, and we hope there is benefit in this article for you.

If you have any question don't hesitate to contact us. You can email us at __enquiry@aisara.ai__