- Home
- Conferences
- Conference Proceedings
- Conferences
EAGE GeoTech 2021 First EAGE Workshop on Induced Seismicity
- Conference date: March 1-5, 2021
- Location: Online
- Published: 01 March 2021
-
-
Towards Synthetically Trained Convolutional Neural Networks for Induced Earthquake Detection and Localization
Authors N. Vinard, G.G Drijkoningen and D.J VerschuurSummaryThe seismic hazard associated to induced seismicity becomes increasingly important as the sites for the exploitation of natural resources, geothermal energy and CO2 injections move closer to urban areas. Successful, reliable and fast detection and earthquake localization methods are a necessity for optimal safety and productivity. The many advances in artificial intelligence in various fields and their capability to apply their learned features to new data in real time is intriguing. We propose to use synthetic earthquake response data modelled using the velocity model and same receiver configuration as in the field to train a convolutional neural network to classify the source to its corresponding 3-D cluster. Once training is completed the network is applied to the corresponding field data. The results show promise that this is indeed feasible as 4 out of 5 events were correctly classified and the missed event was only missed by a little. This method returns a possible cluster inside which the exact hypocenter is located. Therefore, this approach can be used in combination with grid-search type hypocenter location methods to speed up the search by limiting the possible search space.
-