Toward Real-time Microseismic Event Detection using the YOLOv3 Algorithm

Real-time microseismic monitoring is essential for understanding fractures associated with underground fluid injection in unconventional reservoirs. However, microseismic events recorded on monitoring arrays are usually contaminated with strong noise. With a low signal-to-noise ratio (S/R), the detection of microseismic events is challenging using conventional detection methods such as the short-term average/long-term average (STA/LTA) technique. Common machine learning methods, e.g., feature extraction plus support vector machine (SVM) and convolutional neural networks (CNNs), can achieve higher accuracy with strong noise, but they are usually time-consuming and memory-intensive to run. We propose the use of YOLOv3, a state-of-art real-time object detection system in microseismic event detection. YOLOv3 is a one-stage deep CNN detector that predicts class confidence and bounding boxes for images at high speed and with great precision. With pre-trained weights from the ImageNet 1000-class competition dataset, physics-based training of the YOLOv3 algorithm is performed on a group of forward modeled synthetic microseismic data with varying S/R. We also add randomized forward-modeled surface seismic events and Gaussian white noise to generate ``semi-realistic'' training and testing datasets. YOLOv3 is able to detect weaker microseismic event signals with low signal-to-noise ratios (e.g., S/N=0.1) and achieves a mean average precision of 88.71\% in near real time. Further work is required to test YOLOv3 in field production settings.

Microseismicity during geothermal stimulation at the Ngatamariki geothermal field: New detections via a matched- filter method

<p>The high-temperature, fluid-dominated Ngatamariki geothermal field is located in the central Taupo Volcanic Zone, North Island, New Zealand, and is used to generate electricity via an 82 MW power plant. Injection wells have been in operation since June 2012. During June and July 2012, injection well NM8 was injected with with cold water in order to improve reservoir permeability. Geothermal stimulation and production may trigger microearthquakes by fluid flow through the reservoir. Close clustering of microseismic events’ hypocentres relative to the source-receiver distance results in many events having similar waveforms. We capitalize on this relationship by using a matched-filter detection method in which high-quality seismograms corresponding to a well-recorded earthquake (“templates”) are cross-correlated against continuous data to reveal additional earthquakes with similar characteristics. Clustering of the detections’ hypocenters also implies that small variations in travel times between two events corresponds to small differences in hypocentral locations, which is the foundation of the double-difference relocation method. Using an 11 station seismic network, we detect 863 events via cross-correlation of 110 matched-filter templates during the two months stimulation testing. We locate each of these detections using a double-difference relocation method by which events are relocated based on relative travel times. The locatable seismicity delineates: a northern Ngatamariki cluster, a southern Ngatamariki cluster, and a cluster to the south, at the neighboring Rotokawa field. Seismicity in the northern Ngatamariki cluster (522 events) is of greatest interest for this project due to its proximity to well NM8 and temporal signature relative to injection. The seismicity cluster centers around well NM8 at a depth of 2.1 km below sea level. Events in this cluster extend to up to 2.5 km from the injection well. An increase in seismicity near NM8 lags behind the onset of injection by 4–8 days. In contrast, a seismicity-rate decrease coincides with injection shut-in without any time lag. Local magnitudes in this cluster span the range −0.09 ≤ Ml ≤ 1.66 with a completeness magnitude of 0.25. Seismicity within 200 m of NM8 is induced by thermal stresses caused by the difference in temperature between the injectate and the reservoir. Seismicity further than 200 m, but still within this cluster, from NM8 is induced via pore fluid pressure increases from the injected fluid. The coupled mechanism acts on two different length scales and is known as a thermoporoelastic mechanism. The matched-filter detection of microseismic events allows interpretation of extent of injection well stimulation and the relationship between injection and seismicity.</p>

Microseismicity during geothermal stimulation at the Ngatamariki geothermal field: New detections via a matched- filter method

<p>The high-temperature, fluid-dominated Ngatamariki geothermal field is located in the central Taupo Volcanic Zone, North Island, New Zealand, and is used to generate electricity via an 82 MW power plant. Injection wells have been in operation since June 2012. During June and July 2012, injection well NM8 was injected with with cold water in order to improve reservoir permeability. Geothermal stimulation and production may trigger microearthquakes by fluid flow through the reservoir. Close clustering of microseismic events’ hypocentres relative to the source-receiver distance results in many events having similar waveforms. We capitalize on this relationship by using a matched-filter detection method in which high-quality seismograms corresponding to a well-recorded earthquake (“templates”) are cross-correlated against continuous data to reveal additional earthquakes with similar characteristics. Clustering of the detections’ hypocenters also implies that small variations in travel times between two events corresponds to small differences in hypocentral locations, which is the foundation of the double-difference relocation method. Using an 11 station seismic network, we detect 863 events via cross-correlation of 110 matched-filter templates during the two months stimulation testing. We locate each of these detections using a double-difference relocation method by which events are relocated based on relative travel times. The locatable seismicity delineates: a northern Ngatamariki cluster, a southern Ngatamariki cluster, and a cluster to the south, at the neighboring Rotokawa field. Seismicity in the northern Ngatamariki cluster (522 events) is of greatest interest for this project due to its proximity to well NM8 and temporal signature relative to injection. The seismicity cluster centers around well NM8 at a depth of 2.1 km below sea level. Events in this cluster extend to up to 2.5 km from the injection well. An increase in seismicity near NM8 lags behind the onset of injection by 4–8 days. In contrast, a seismicity-rate decrease coincides with injection shut-in without any time lag. Local magnitudes in this cluster span the range −0.09 ≤ Ml ≤ 1.66 with a completeness magnitude of 0.25. Seismicity within 200 m of NM8 is induced by thermal stresses caused by the difference in temperature between the injectate and the reservoir. Seismicity further than 200 m, but still within this cluster, from NM8 is induced via pore fluid pressure increases from the injected fluid. The coupled mechanism acts on two different length scales and is known as a thermoporoelastic mechanism. The matched-filter detection of microseismic events allows interpretation of extent of injection well stimulation and the relationship between injection and seismicity.</p>

Imaging crustal structures through a passive seismic imaging approach in a mining area in Saxony, Germany

We have used several flooding-induced microseismic events that occurred in an abandoned mining area to image geological structures close to the hypocentres in the vicinity of the mine. The events have been located using a migration-based localization approach. We used the recorded full waveforms of these localized microseismic events and have processed these passive source data as if they resulted from active sources at the known hypocentre location and origin time defined by the applied location approach. The imaging was then performed using a focusing 3D prestack depth migration approach for the secondary P-wave arrivals. The needed 3D migration velocity model was taken from a recent 3D active (controlled-source) seismic survey in that area. We observed several clear and pronounced reflectors in our obtained 3D seismic image cube, some of them related to a major fault zone in that area and some correlating well with information from the nearby mining activities. We compared our results to the 3D seismic image cube obtained directly from the 3D active seismic survey and have found new structures with our approach that were not known yet, probably because of their steep dips which the 3D active seismic survey had not illuminated. The location of the hypocentres at depth with respect to the illumination angles of those structures proved to be favourable in that case, and our 3D passive image complements the 3D active seismic image in an elegant way, thereby revealing new structures that cannot be imaged otherwise with surface seismic configurations alone.

Arrival-time picking of microseismic events based on MSNet

Arrival-time picking of microseismic events is a critical procedure in microseismic data processing. However, as field monitoring data contain many microseismic events with low signal-to-noise ratios (SNRs), traditional arrival-time picking methods based on the instantaneous characteristics of seismic signals cannot meet the picking accuracy and efficiency requirements of microseismic monitoring owing to the large volume of monitoring data. Conversely, methods based on deep neural networks can significantly improve arrival-time picking accuracy and efficiency in low-SNR environments. Therefore, we propose a deep convolutional network that combines the U-net and DenseNet approaches to pick arrival times automatically. This novel network, called MSNet not only retains the spatial information of any input signal or profile based on the U-net, but also extracts and integrates more essential features of events and non-events through dense blocks, thereby further improving the picking accuracy and efficiency. An effective workflow is developed to verify the superiority of the proposed method. First, we describe the structure of MSNet and the workflow of the proposed picking method. Then, datasets are constructed using variable microseismic traces from field microseismic monitoring records and from the finite-difference forward modeling of microseismic data to train the network. Subsequently, hyperparameter tuning is conducted to optimize the MSNet. Finally, we test the MSNet using modeled signals with different SNRs and field microseismic data from different monitoring areas. By comparing the picking results of the proposed method with the results of U-net and short-term average and long-term average (STA/LTA) methods, the effectiveness of the proposed method is verified. The arrival picking results of synthetic data and microseismic field data show that the proposed network has increased adaptability and can achieve high accuracy for picking the arrival-time of microseismic events.