Oil Production Prediction Using Time Series Forecasting and Machine Learning Techniques

Abstract

Abstract Prediction of oil production is critical for the oil and gas industry, as it helps production engineers plan and execute strategic decisions. In the past, various empirical correlations and mathematical models have been utilized for this purpose. However, with the advent of data-driven techniques, machine learning algorithms such as Random Forest (RF), Artificial Neural Network (ANN), Long Short-Term Memory neural network (LSTM), Recurrent Neural Network (RNN), DeepAR, and others have been adopted for predicting oil production. This paper presents a comparative analysis of time series forecasting and machine learning techniques for predicting oil production, using the ARIMA, Prophet, Random Forest, CatBoost, and XGBoost Algorithms. Time series forecasting involves building models based on historical data and using them to make predictions for the future, while machine learning algorithms use data to train models that can accurately predict future outcomes. The study aims to develop a prediction model for oil production using daily production data obtained from the Volve production field in Norway. Results from this study demonstrate that while time series forecasting had a larger error margin and a negative coefficient of determination (R2 Score), while machine learning techniques improved the accuracy of the prediction, with the stacked regressor algorithm having an R2 score of 97.5%. Feature selection was done for the prediction, and features such as bottom-hole pressure, bottom-hole temperature, annulus pressure, choke size, and tubing downhole pressure contributed to the more accurate prediction of oil production. In conclusion, the satisfactory results of the comparative analysis demonstrate the effectiveness of machine learning algorithms in predicting oil production. This study can serve as a reference for production engineers looking to develop more accurate oil production prediction models using machine learning techniques.