Skip to main content

Predicting Multiphase Flow from Well Logs and Reservoir Data

The AI Hackathon is a collaborative platform for universities, organizations and individuals to demonstrate their expertise in machine learning by predicting crucial oil and gas production logging outcomes. Specifically, participants are tasked with predicting Water and Oil Holdups, Flow Rates (Q), Pressure, etc., using the provided OH, CH logs, and other well information data. The AI algorithm learns from one of the wells' data to build a model, tests the model with another well, and then uses the model to predict other wells' (for validation) production profiles and recommend the best location to target drilling new well/s. One of the (validation wells) is used for blind test validation by the AI Hackathon Committee.

Key Elements

Data, Usage, and Reporting

Participants receive a comprehensive dataset with historical PLT logs associated with water and oil holdup measurements, Q values, and other Open Hole (OH) and Cased Hole (CH) logs, Pressure, Perforation, inclination, formations, and the well x,y,z data for training their machine learning models in well # 1. Participants also get a testing dataset without well # 6 for participants' ability to self-evaluate their models. At the same time, you would predict the production (PLT output) for wells # 2, 3, 4, and 5. Keeping in mind that the blind test is going to be among one of these four wells. The AI Hackathon committee keeps some data hidden for fair machine-learning model validation. The Participants would also recommend the best location to target drilling new wells and briefly explain the selected target. The Participants must write a report about the algorithms used, the parameters and hyperparameters, and the feature engineering (input data, augmented data, data created out of the given data), the model results (Mean Absolute Error (MAE)) and the run time of the models), also provide a discussion and observation for each predicted well, and the location selection of the new target well to drill.


The objective of the Hackathon is to understand these main things below:

a. The ability of AI to predict the production profile within a particular area from one well to another well.

b. The ability of AI to identify inter-well communication.

c. The ability to identify the optimal new well target location to drill.

Note-1: All these objectives must be honored by participants, and in the report, explain how these objectives have been achieved, MAE, and the run time, and provide recommendations for further actions.

Evaluation Criteria

  • Mean Absolute Error (MAE): MAE is calculated as the average of the absolute errors between predicted and actual values. The error is the difference between the actual value and the predicted value. Therefore, MAE would be considered the primary criterion for ranking the models.
  • Model Run Time: This is the time needed to run a model on the validation data set. Model Run Time is the second criterion when two algorithms deliver the same MAE. Therefore, it might be used in the MAE tie between different participant models.

Rules and Guidelines

Participants must exclusively use the data provided for model development, and unethical practices will result in disqualification. The Hackathon is open to universities, research institutions, and organizations, fostering inclusivity and collaboration. Documentation of the model development process is required.

  • Participants must use the provided training data to develop their machine-learning models. External data sources are not allowed. However, data augmentation or creating new data from the provided data is welcomed. If additional data is required, the participants should request it via email to the listed addresses in the Contacts section.
  • The Hackathon will be open to universities, research institutions, organizations, and individuals. Teams may consist of individuals or groups from these entities or in collaborations with others.
  • Participants must submit their predictions for the testing data within the specified timeframe.
  • The use of any unethical or unfair practices, including plagiarism or data leakage, will result in disqualification.
  • Participants are encouraged to document their model development process and submit an explanation of their approach and predictions as described in the Data, Usage, Reporting, and Objective section Note-1.

Prediction Deliverables

  • Qt: Total Flow Rate
  • Qo: Oil Flow Rate
  • Qw: Water Flow Rate
  • Qg: Gas Flow Rate
  • Hw: Water Holdup
  • Ho: Oil Holdup
  • Hg: Gas Holdup
  • Pressure
  • Temperature
  • Identify the area to drill a new well between these two wells


The Hackathon offers prizes, recognition certificates, and potential collaboration opportunities to three winners and high-performing participants, incentivizing cutting-edge contributions to the oil and gas industry. This would include the team members who participated in the winning submission. The registration fees for GOTECH 2024 will be waived-off for winners.

Winners will be announced by Thursday, 25 April 2024.

Note: The winners to bear all the travel and accommodation-related expenses.


Based on the GOTECH 2024 program time availability, the top three winners might be invited to present at GOTECH.

Important Links

Deadline - Monday, 22 April 2024

application form
Application Form (please use this form to submit your application, and once you click submit, a new page will appear that contains the link for downloading the data)


upload response

Upload Response (please use this link to upload your final response in one zip file as mentioned in the Additional Notes for Responding section)

and email the zip file to Dr. Omar Alfarisi -


Additional Notes for Responding:

Once a participant submits the application form and downloads the data, another email will be sent to the participant after a few weeks with a link where the participant should:

  • Upload the AI Models. The models must be in a format that can run in a Python IDE or Python compiler; therefore, all the dependent libraries to run the model must be provided.
  • Upload the Prediction Work Report. The report should contain the following sections [Title, Participants & Affiliation, Abstract, Introduction, Method, Results, Discussion, Conclusion].
  • Upload the Input/Output Data used in the Prediction.
  • All the data, models, reports should be included in one zip file following this file naming format: [“Participant last name”-“Participant first name”-“MMM-DD”]

Hackathon Chair

Dr Omar
Omar Alfarisi
Senior AI Technology Advisor
Dragon Oil

Hackathon Champions

Mohammed Al Kobaisi
Associate Professor
Khalifa University

Jasem AlMansoori
Reservoir Engineer
Dragon Oil

Hadi Belhaj
Associate Professor
Khalifa University

Magdi ElDali
Senior Geologist
Dragon Oil

Muhammad Gibrata
Senior Petrophysicist
Dragon Oil

Alaa Hassan
Data Analytics Specialist
Dragon Oil

Qingfeng Huang
Senior Reservoir Simulation Engineer
Dragon Oil

Lamia Rouis
Manager Reservoir
Dragon Oil

Safiya Sherman
Innovation Process Analyst
Dragon Oil

For any additional information or query, please contact: