 |
| Titel |
Micro seismic event detection based on neural networks in the Groningen area, The Netherlands |
| VerfasserIn |
Bob Paap, Peter-Paul van Maanen, Stefan Carpentier, Sjef Meekes |
| Konferenz |
EGU General Assembly 2017
|
| Medientyp |
Artikel
|
| Sprache |
en
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 19 (2017) |
| Datensatznummer |
250144875
|
| Publikation (Nr.) |
 EGU/EGU2017-8751.pdf |
|
|
|
|
|
| Zusammenfassung |
| Over the past decades, the Groningen gas field has been increasingly faced by induced
earthquakes resulting from gas production. The seismic monitoring network at Groningen has
been densified in order to acquire more accurate information regarding the onset and origin of
seismic events, resulting in increasing amounts of seismic data. Although traditional
automated event detection techniques generally are successful in detecting events from
continuous data, its detection success is challenged in cases of lower signal-to-noise ratios
and often limited availability of seismologists. Besides the recent expansion of the Groningen
seismic network, additional new seismic networks have been deployed at several
geothermal and CO2 storage fields. The data stream coming from these networks
has sparked specific interest in neural networks for automated classification and
interpretation.
Here we explore the feasibility of neural networks in classifying the occurrence of
seismic events. For this purpose a three-layered feedforward neural network was trained
using public data related to a seismic event in the Groningen gas field obtained from the
Royal Netherlands Meteorological Institute (KNMI) data portal. The first arrival times that
were determined by KNMI for a subset of the station data were used to determine
the arrival times for the other station data. Different derivatives, using different
frequency sub-band and STA/LTA settings, were used as input. Based on these data,
the network’s parameters were then optimized to predict arrival times accurately.
Although this study is still ongoing, we anticipate our approach can significantly
increase the performance as compared to detection methods usually applied to the
Groningen gas field. This will clear the way for future real-time micro seismic event
classification. |
|
|
|
|
|
|