पेट्रोलियम और पर्यावरण जैव प्रौद्योगिकी जर्नल

अमूर्त

Oil and Gas Well Production Forecasting Based on Machine Learning Models: The Volve Field Case

Moreno Millan

The current techniques for predicting the oil and gas production flow rates at well and reservoir scales include from the classical decline curves analysis thru numerical simulation models. The present work proposes the use of the following Machine Learning Models (MLM): Linear Regression (LR), Support Vector Machines (SVM), Random Forest (RF), and an Artificial Neural Network (ANN), as an alternative to the conventional methods for forecasting oil and gas production flow rates. The application of this proposal is demonstrated based on production data recorded along 8 years in wells from Volve field, located in the Norwegian continental shelf. Thus, the benefits for each MLM above mentioned are discussed, concluding based on a practical experience that not always the more complex algorithm is the best choice. It is demonstrated that the alternative of SVM yield best results, and it is also a simpler and easier model to be implemented in comparison to RF or ANN alternatives.