Artificial Intelligence In Reservoir Modeling

Artificial intelligence (AI) is a rapidly developing field that is transforming many industries, and reservoir modeling takes no exception. It is one area that is gaining such much traction, but not enough of it has been said or applied within the oil and gas industry.

Despite my broad knowledge and experience in the geoscience space. I have always been a penchant enthusiast for the tech industry. In recent times, AI has sparked my interest and I have a strong conviction that it is going to revolutionize the tech space and the world at large. So today’s article has been structured on step by step guide of developing AI in reservoir modeling.

I want to make it clear that I am not an expert in artificial intelligence and the opinions and ideas expressed in this article are based on my personal beliefs, perceptions, and understanding.

Shall we proceed?

Over the years, reservoir models have been created using traditional techniques such as statistical analysis and numerical simulations. These methods are; time-consuming, requires a high level of expertise to execute, challenging in accurately predicting the behaviour of reservoirs due to complex interactions between subsurface parameters, challenging in its ability to adapt to changing conditions such as new data and even handling of large datasets, the pitfalls goes on and on.

However, with the advent of AI, there is a strong likelihood to create accurate and reliable reservoir models much more efficiently, reducing human errors to its barest.

There are several ways in which AI is being used in reservoir modeling. One of the most common applications is the use of machine learning algorithms to analyze large datasets. These algorithms can identify patterns and trends in the data that would be difficult for a human to detect, allowing for more accurate modeling of reservoir behavior.

To keep it simple, I like to regard machine learning as a child with super-sonic learning abilities. Just like every conventional learning outfit, there have to be a teacher/instructor/guide etc. I call this phase the human to machine interaction, which is very fundamental. As much as we may not want to admit it, this Child (the machine) will grow to be smarter than us. But its ingeniousness is highly dependent on the early phase codependency between the instructor (humans) and the child (the machine).

1.??????The first expected step in using machine learning for reservoir modeling is to collect and preprocess data, which is similar to what we practice in the industry as data gathering and review. You’d agreed with me that this is a very laborious task. Especially for unsorted and unformatted dataset.

This step includes gathering data on the characteristics of the subsurface, rock properties, data on presence of fluids, data on the geology of the area etc. It also includes and not limited to data on the behavior of the reservoir, production history, pressure data, temperature etc. The data required be cleaned and formatted in a way that is suitable for machine learning algorithms to process. This is done by the teacher (Instructor).

2.??????The second step is to choose a machine learning algorithm. In a more simplified manner, choosing a machine learning algorithm is like training a child and choosing what language you need him/her to interact with. Possibly English, Yoruba, Igbo, Hausa, French etc. There are many different machine learning algorithms to choose from, and the best one will depend on the specific needs of the modeling project. Some common algorithms that are used in reservoir modeling include decision trees, random forests, etc. These are just a few from a spectrum of machine learning algorithms

3.??????Train the algorithm: Once the data has been collected and pre-processed, and the algorithm has been selected, the next step is to train the algorithm. Let me once again simplify this for you. Just like we have letters in English which are meaningless without being structured to form words, we may want to liken this selected algorithm to this analogy. For this third step, we need to feed the algorithm with the preprocessed data and allow it to learn from it. The goal is to teach the algorithm to recognize patterns and trends in the data that will allow it to make accurate predictions about the behavior of the reservoir.

4.??????To validate the acquired intelligence of your child (the AI or Machine), we’d need to put the algorithm to a test. After the algorithm has been trained, it is important to test its performance to ensure that it is accurate and reliable. This can be done by comparing the predictions made by the algorithm with actual data from the reservoir.

Once the algorithm has been trained and satisfactorily tested, it can be used to make predictions about the behavior of the reservoir. These predictions can be used to optimize reservoir management strategies and improve the efficiency of the reservoir.

One of the beauties of an AI is that as more data becomes available, the machine learning model can be updated and improved to make more accurate predictions seamlessly. This may involve retraining the algorithm or incorporating new data into the model.

I can hear someone asking me via ”cyber-mental-telepathy” (I made up that word). How practical is this? Do I need to know how to program?

Yes/No! One may need to learn a programming language. However, while programming experience is generally helpful when building machine learning models, it is may not be a show stopper. With the right resources, motivation and support from tech communities, it is possible to learn the necessary skills and get started in this field. You could also use an existing framework. Though this may limit your creativity. My ultimate advise will be to learn a programming language. Python is a good place to start. I have begun my own journey. You can lookup some good short training videos/courses here on LinkedIn, Udemy, Coursera or even YouTube. Good-luck!

Thanks for taking the time out to read. Feel free to share or make contributions.