Receiver Orientation
The horizontal components of a downhole geophone array rotate freely when being sent into a monitoring well (such that they point in unknown directions upon arrival at target depth). The industry-standard practice is to use a single perforation shot to calibrate the receivers’ orientation. This single-shot method suffers from noise in the seismograms, and rotation matrices obtained from different perforation shots are likely to be inconsistent with each other. This means that different perforation shots can lead to dramatically different orientation results – even when the receiver is not being moved at all. Large errors emerge when simple averaging is applied to this standard method in order to arrive at a final orientation solution. The industry-standard single shot method will also yield receiver orientation errors when the lateral distance between a perforation shot and an observation well is small. This is due to weak horizontal-component signal, and traditional calibration will not be stable. Because receiver orientation is one of the first steps in any downhole microseismic data processing workflow, any receiver orientation error is compounded exponentially with each subsequent step. |
|
Borehole Seismic™ Receiver Orientation incorporates all available perforation shots into our calibration. Each perforation shot can be weighted by signal-to-noise ratio¹ and hodogram analysis quality² to compensate for errors imparted by noise and geometry. The optimized rotation matrix is calculated inversely by minimizing the misfit between calculated and weighted polarity records from all perforation shots. For deviated wells, receivers are reoriented in the plane perpendicular to well direction.
|
¹ Signal-to-Noise Ratio (“SNR”) is the ratio between the energy of wave arrival and the static noise within the seismogram. Low SNR means that the data is less reliable, and such traces carry lower weight in the calculation to reduce error.
² Hodogram Analysis Quality (“HAQ”) is determined by comparing hodogram analysis polarities with theoretical polarities. A hodogram analysis is a plot of the recorded time history of a particle’s motion, and is used to determine the polarity of a wave arrival. Such polarities are then input into a receiver reorientation script and compared with theoretical polarities calculated by independent methods (e.g., survey geometry for 2D rotation). If the polarity given by hodogram analysis is of bad quality (“low HAQ”), the corresponding perforation shots are weighted less to reduce error. |
|
Triggering
Borehole Seismic™ Triggering locates event data in the time domain using the popular short-term average/long-term average (“STA/LTA”) method – with picks refined manually by degreed geophysicists applying proprietary techniques. Our geophysicists are trained to find and refine wave picks to maximize pick accuracy, thereby increasing the accuracy of calculated event locations.
Using the STA/LTA method, the ratio of energy within two windows is first calculated and stored (the “STA/LTA ratio”). The derivative of the STA/LTA ratio between each successive window is then calculated and used as the indicator function. For theoretical noise-free data, the maximum value of the derivative of the STA/LTA ratio is generally close to the event arrival time. For field data, the derivative of a smoothed STA/LTA ratio is used to suppress the contribution from real-world noise.
Borehole Seismic™ Triggering locates event data in the time domain using the popular short-term average/long-term average (“STA/LTA”) method – with picks refined manually by degreed geophysicists applying proprietary techniques. Our geophysicists are trained to find and refine wave picks to maximize pick accuracy, thereby increasing the accuracy of calculated event locations.
Using the STA/LTA method, the ratio of energy within two windows is first calculated and stored (the “STA/LTA ratio”). The derivative of the STA/LTA ratio between each successive window is then calculated and used as the indicator function. For theoretical noise-free data, the maximum value of the derivative of the STA/LTA ratio is generally close to the event arrival time. For field data, the derivative of a smoothed STA/LTA ratio is used to suppress the contribution from real-world noise.
The STA/LTA method is a single-trace operation, but a single event is likely to appear on multiple receivers such that the recorded events will display certain trends (“moveout trends”) which help to constrain triggering of event data in the time domain. When receiver coverage is sufficient, Radon transform (or equivalently, tau-p transform or slant-stack) of the traces will be calculated, helping to evaluate such moveout behavior. For observation wells, the trace around each pick can be plotted and used as a quality control check.
Real-world data usually contains coherent noise which will lower triggering quality. Spectrograms around triggered events can be calculated to help further constrain the solution. This is possible because microseismic events usually emit much broader frequency bands than field noise. Events can accordingly be triggered when there is a burst of frequency content on the spectrogram.
Real-world data usually contains coherent noise which will lower triggering quality. Spectrograms around triggered events can be calculated to help further constrain the solution. This is possible because microseismic events usually emit much broader frequency bands than field noise. Events can accordingly be triggered when there is a burst of frequency content on the spectrogram.
Surface - Relative Trace Stacking
The relative trace stacking method is designed to overcome the difficulty in estimating the move-out and polarity correction among large receiver arrays. At the beginning, the continuous traces need to be cut into small segments. Then, we create a group of relative traces for each segment by:
where
is the ith relative trace while
and
are the ith and (i+d)th measured traces.
The symbol “*” stands for cross-correlation and the relative trace is formed by cross-correlating two of the selected traces. The index i ranges within the number of selected traces and d is the separation between two traces. An event can be identified once the stacking of the relative traces exceeds the posed threshold.
The first advantage of operating with relative traces is that the cross-correlation corrects the move-out. For example, if the trace separation d is set to one, cross-correlating the nearby traces will form the relative trace as seen below.
The first advantage of operating with relative traces is that the cross-correlation corrects the move-out. For example, if the trace separation d is set to one, cross-correlating the nearby traces will form the relative trace as seen below.
As a result, the relative trace containing an event will have a finite-length wavelet. The peak of the relative trace will consistently be located at the shift between nearby receivers, which is supposed to be close to the zero lag as seen below.
In this case, the cross-correlation throughout the entire trace is supposed to be stackable. The second advantage of operating with relative traces is that the peak polarity will be corrected. The polarity is flipped to a positive within the receivers located in the same lobe of a radiation pattern. Although we might lose a few traces when the polarity is flipped, most of the relative traces will still show a positive peak and the energy will be increased once they are stacked as seen below.
The advantage of this method is that travel time estimation is no longer needed and polarity flipping is taken care of automatically. The relative trace stacking is able to suppress the random and narrow-band noises during the cross-correlation since they are supposed to be independent from each receiver. The narrow-band noise will display low-frequency correlation on the relative traces and their influence will be decreased once the relative traces are stacked. The move-out estimated from the velocity model helps suppress the coherent noises as well. If the traces are corrected from the estimated move-out, then the increasing separation d will generally decrease the contribution from coherent noises. However, the quality of such an operation is dependent on the acquisition geometry. Extension of this method can include grid search, but this is not as efficient as stacking-energy when coherent noise is present.
Surface - Stacking Energy Method
Passive seismic data acquired at surface is often accompanied by relatively strong, coherent noises. Borehole Seismic, LLC utilizes the stacked energy method to trigger microseismic events within such survey geometries. This method is based on the distinct move-out pattern of an event, which is generated at a deeper treatment area than the coherent noises. Coherent noises are often generated at the wellhead on the surface. During triggering, grids are pre-defined to a reasonable area and the travel time from each grid can be computed to serve as the estimated moveout. The continuous seismic records will be shifted according to these computed moveouts and the event will be aligned among receivers. Closer receivers will be stacked to enhance SNR and a moving average filter will be applied to enhance further stacking. Since the polarity the entire survey area will not be consistent, direct stacking across all the receivers cannot guarantee SNR enhancement. As an alternative method, Borehole Seismic, LLC computes the energy of each stacked patch, removing a constant level of energy to suppress random noise. The de-trended energy among all patches will be stacked and the prominence of the stacked trace will be the indicator of an event.
The top left figure shows raw data. The bottom left shows data after moveout correction. The bottom right is the data stacked within each “patch.” This dataset is from a “star” array, but we still stack the nearby receivers as if they were a “patch” in order to enhance SNR. The upper right figure is the indicator function, which is computed via stacking energy of the data from all patches. We compute the area of the contour and that area is called prominence. If the prominence exceeds a threshold, an event will be triggered.
Would you like to know more? Consider having us take a second look at your data.
³ Grechka, V. and Yaskevich, S. (2014), Azimuthal anisotropy in microseismic monitoring: A Bakken case study. 79 Geophysics 1, 11-15. doi: 10.1190/geo2013-0211.1
⁴ See Grechka, V. and Yaskevich, S. (2013), Inversion of microseismic data for triclinic velocity models. Geophysical Prospecting 61: 1159–1170. doi: 10.1111/1365-2478.12042
⁵ Grechka, V. and Yaskevich, S. (2014), Anisotropic Velocity Model Building in Microseismic Monitoring. GSH Journal 4.5, 11-15.
⁴ See Grechka, V. and Yaskevich, S. (2013), Inversion of microseismic data for triclinic velocity models. Geophysical Prospecting 61: 1159–1170. doi: 10.1111/1365-2478.12042
⁵ Grechka, V. and Yaskevich, S. (2014), Anisotropic Velocity Model Building in Microseismic Monitoring. GSH Journal 4.5, 11-15.
