Task is to optimise ROP (Rate of Penetration) for Oil and Drilling. Since the cost of daily penetration is about $40,000 to $300,000.
Refer Dataset for rate of penetration benchmarking under Journal of Petroleum Science and Engineering by Elsevier to know the Problem faced while calculating ROP and why it still hasn't been perfected.
For Reference, one can also refer to the code and visualization by Andrzej Tunkiel.
Task was to achieve the highest MAE on an open and closed dataset. Data for 7 oil wells were provided. I achieved an accuracy of 99.76% using MSE criteria and 0.0736 MAE using MAE criteria. MAE recieved on open dataset on a random well was 7.85 & on closed dataset recieved MAE 8.22. Received rank 2 on both open and closed dataset.
Final Notebook will be uploaded after ending of competition.
- PetroCoder_Round3_ByPradhuman.ipynb - A standalone file for reproduction of Graphs and Models.
- exploratoryDataAnalysis_1.ipynb - Explains out the Data Preprocessing step with some visualizations.
- graphMinMaxTrans.xlsx - Contains models, their properties and plots for MinMaxScaler. Produced using RandomSearchCV.
- graphPowerTrans.xlsx - Contains models, their properties and plots for PowerTransform Scaler. Produced using RandomSearchCV.
- graphStandardScalar.xlsx - Contains models, their properties and plots for StandardScaler. Produced using RandomSearchCV.
- tech_challenge2021_train.zip - Contains training dataset provided by Hackathon Organisers.
- tech_challenge2021_test.zip - Contains testing dataset provided by Hackathon Organisers.