Fast 3D elastic micro-seismic source location using new GPU features

https://doi.org/10.1016/j.pepi.2016.08.001Get rights and content

Abstract

In this paper, we describe new GPU features and their applications in passive seismic – micro-seismic location. Locating micro-seismic events is quite important in seismic exploration, especially when searching for unconventional oil and gas resources. Different from the traditional ray-based methods, the wave equation method, such as the method we use in our paper, has a remarkable advantage in adapting to low signal-to-noise ratio conditions and does not need a person to select the data. However, because it has a conspicuous deficiency due to its computation cost, these methods are not widely used in industrial fields. To make the method useful, we implement imaging-like wave equation micro-seismic location in a 3D elastic media and use GPU to accelerate our algorithm. We also introduce some new GPU features into the implementation to solve the data transfer and GPU utilization problems. Numerical and field data experiments show that our method can achieve a more than 30% performance improvement in GPU implementation just by using these new features.

Introduction

Seismic source location is one of the most basic research topics in seismology and geophysics. In recent years, due to hydraulic fracturing becoming a common method applied in unconventional oil and gas fields, there has been increasing interest in research on locating micro-seismic events.

The basic idea in earthquake location studies is to use the seismic wave travel time to inverse the source location. The most traditional methods are based on the works of Geiger (Geiger, 1912), and later geophysicists (Bratt and Bache, 1988). The core of these methods is using the least squares method to solve the linear system between the seismic source location and the wave travel time information. Further development has introduced many nonlinear methods such as the gradient, Newton, global search, and Monte Carlo methods to the location field. In brief, both the theoretical and practical aspects of travel-time inversion methods have been extensively studied. These methods have their limits, however; clear P and S wave travel times must be acquired from seismic records and consequently must have a high signal-to-noise ratio (SNR). However, micro-seismic monitoring data are often huge datasets with low signal-to-noise ratios, which limits the use of traditional Geiger-type methods.

An alternative approach, which requires less user interaction and allows for greater accuracy, is using the concept from reflection seismic exploration to reverse-propagate the wavefield and image the micro-seismic event locations (Artman et al., 2010). In these approaches, we can obtain both the source location and excitation time, What’s more, there is no need to select the data for the first arrival, and the method can be applied to low-SNR data (Douma and Snieder, 2014). All of these advantages make the time reversal method an ideal method for solving the source location problem. However, the inherent problems of the wave equation based method are the cost time. Hence, these types of approaches have not been widely used in real-time field hydraulic monitoring.

However, in recent years, with the development of high-performance graphics processing units (GPUs), we finally have a good way to solve the time-consuming problem of time reversal location methods (Xue et al., 2015). GPU acceleration has been widely applied in various computing-intensive fields. Using a GPU, many algorithms and applications in geophysics, which is traditionally among the most computationally intensive scientific research fields, have been able to run considerably more quickly. However, for the time reversal problem, the data transfer and GPU utilization meet a bottleneck due to the limitations of the GPU memory and the algorithm character, especially for 3D problems. In this paper, we implement two new GPU features, GPU Direct, focusing on data transfer, and Hyper-Q, used for improving the GPU utilization, to attempt to solve these problems.

This article first presents the theory behind the time-reversal imaging microseismic event location method. Then, it outlines the details of the bottleneck of the time-reversal method and the corresponding new GPU features, describes the implementation scheme and discusses the extensibility of the multi-GPU setup. Next, it uses 3D elastic numerical experiments to simulate the real micro-seismic event location progress and uses field data to test the algorithm. Finally, through the implementation of the scheme and a performance analysis, we show the utility of the new GPU features.

Section snippets

Focusing by time-reverse modeling

The time-reverse modeling (TRM) method was used to locate earthquakes and micro-seismic events beginning with the work of (Fink et al., 2000). The TRM method is quite similar to reverse-time migration (RTM) (Levin, 1984), but it has a difference. The main difference between TRM and RTM is that TRM does not require knowledge of the source wave field. Besides, the main process flows are the same: the receivers’ data are used as the source, and the data are reverse-propagated in time at the

New GPU features and their implementation

In this section, we first analyze the process of the time-reversal location and then provide the general GPU implementation flow. Finally, we illustrate how to use the new GPU features Hyper-Q and GPU Direct to solve two bottlenecks in the general GPU implementation.

Application to synthetic data

This section presents the results of applying the time-reversal source location algorithm to the synthetic data set. We use two models in this section, a horizontally layered model and the 3D overthrust model.

Application to field data

This section present the results of applying the time-reversal source location algorithm to the field data. The velocity models for the P-wave and S-wave are given in Table 1. The data acquisition system, containing seven vertical geophones, is shown in Fig. 12, and we use Y = 300 (middle for the totally Y direction) slice to illustrate the entire model. First, we use perforation data to test the flexibility of our algorithm to field data. The first trace of the three-component recorded

Performance analysis

In this section, we will analyze the two features mentioned above to verify that they can solve the main bottlenecks to the general GPU implementation.

Conclusions

In this paper, we introduce a time-reversal method for locating microseismic events with high precision. This method provides relatively high-resolution results from synthetic seismic data and field data. Although the computation times are significantly greater than those of traditional positioning methods, the time-reversal microseismic event location method is efficient and accurate and has excellent development prospects due to the introduction of GPU and its new features to accelerate the

Acknowledgements

The research was supported by the National Basic Research Program of China (Grant No. 2015CB258500) and funded by the National Natural Science Foundation of China (Grant No. 41274112).

References (11)

There are more references available in the full text version of this article.

Cited by (0)

View full text