Title:

Prediction Modeling for Geothermal Reservoirs Using Deep Learning

Authors:

Halldora GUDMUNDSDOTTIR, Roland N. HORNE

Key Words:

deep learning, machine learning, neural networks, injection, tracer

Conference:

Stanford Geothermal Workshop

Year:

2020

Session:

Reservoir Engineering

Language:

English

Paper Number:

Gudmundsdottir

File Size:

2054 KB

View File:

Abstract:

Reservoir characterization and prediction modeling are among the more challenging tasks in geothermal reservoir engineering. Thermal breakthrough in producers can occur due to injection of cold wastewater and therefore an understanding of how production is influenced by injection is essential for sustainable management of geothermal fields. The fractured nature of geothermal reservoirs causes the relationship between production and injection wells to be highly complex and nonlinear. In this work, we investigate models to serve as a substitute for full reservoir simulations of geothermal reservoirs. Deep learning algorithms are used for this purpose and two different architectures, standard feedforward neural networks and recurrent neural networks, explored. The deep learning model maps the input to an output through series of layers, nodes and activation functions. Here, the mapping is accomplished with injection rates at the injectors as input and tracer concentration data at the producers as output. The models were trained on a synthetic geothermal reservoir and our analysis concluded that for this case the simple feedforward neural network outperformed the more complex recurrent neural network.


ec2-54-175-120-161.compute-1.amazonaws.com, you have accessed 0 records today.

Press the Back button in your browser, or search again.

Copyright 2020, Stanford Geothermal Program: Readers who download papers from this site should honor the copyright of the original authors and may not copy or distribute the work further without the permission of the original publisher.


Attend the nwxt Stanford Geothermal Workshop, click here for details.

Accessed by: ec2-54-175-120-161.compute-1.amazonaws.com (54.175.120.161)
Accessed: Friday 29th of March 2024 12:14:02 AM