Overview / Usage

Creation of deep smart proxy in the Oil sector, more precisely in producing wells. For this purpose, the simulation of intelligent completion data was used to predict the production of oil, water and gas. The database consists in historical data series of production and the position (open or closed) of 3 valves in the injector well. For the prediction, deep learning recurrent neural networks algorithms such as LSTM and GRU was used and all the tests performed in the Intel Dev Cloud.

Methodology / Approach

In order to create a deep smart proxy to reduce computacional time in oil production simulations, tests were performed using machine learning techniques, more specifically deep learning recurrent neural networks such as LSTM and GRU , in order to have a prediction of oil, water and gas production. The objective is to have a network with the lowest root mean square error (RMSE) in the production forecast.

Technologies Used

Matplotlib, Pandas, Sklearn, numpy, Math, Keras