The Artificial Neural Network has been used in a variety of complex conditions where the non-linear relationships between existing variables can be difficult to understand. Forecasting the reservoir production along with resembling the reservoir behavior are the main goals of using simulators. To reduce the number of runs by a simulator, the Artificial Neural Network is used to predict the relationship between variables using a history matching technique. The Artificial Neural Network (ANN) mainly consists of two solution parts. The first part is called forward solution and is used to obtain production properties from parameters generated using the simulator. Using an inverse solution enables users to obtain reservoir or well design data based on production data such as production rate and cumulative production.
The artificial neurons are mathematical functions that handle multiplying, summing and activating the neurons to be transformed to a desired goal, an output. Different scenarios have been implemented following the Artificial Neural Network application to the petroleum industry. As the number of layers in a multilayer reservoir increases, neural network requires more time and computations. By using optimization methods and applying them to the ANN, the tool decides on the best design in which the least error is achieved. The developed tool has been assessed based on the comparison of simulator parameters and ANN learned variables.
To enhance the utilization of the tool, a graphic user interface (GUI) has been created to help users access the data and results efficiently. The GUI is itself considered to be an auxiliary tool and based on the data provided as inputs, results would be available.
