Pattern Recognition on 1D Pressure Data: Bottom Hole Pressure Prediction
Pattern Recognition on 1D Pressure Data: Bottom Hole Pressure Prediction project was part of Spot the Trend – A PetroAI hackathon that presented a challenge to recognize the pattern with time series pressure data provided. As a problem statement and its solutions, a bottom-hole pressure was to be predicted with the help of provided time series data to decide whether a well is flowing or not. An AI model was then developed and optimized with 70.1% accuracy in predicting the BHP and well-flowingness.
The data for this project was 1D bottom hole pressure time series from oil and gas wells. By monitoring and analyzing bottom-hole pressures, we can optimize well design, and production rates, and maximize the fluid recovery from each well. Understanding and monitoring these pressures also helps us detect wellbore instability or formation damage.
There were basically two primary tasks to complete:
- Bottom Hole Pressure Prediction with Time Series Data.
- Predict Whether the Well is Flowing or Not.
