Reservoir imaging using induced microseismicity

Abstract

Production activities within a hydrocarbon reservoir, such as extracting oil or injecting fluid, result in changes in stress which consequently cause micro-earthquakes. The induced micro-seismic events are small earthquakes producing high frequency waves which can be used to give a hi-resolution image of the hydrocarbon reservoir. However, induced micro-seismic events are usually too small in magnitude to be detected on the surface due to seismic wave attenuation through the overburden rocks. In addition, we lack information about their hypocentres and origin times. Besides, because the ray-path depends on the slowness model, the relationship between the arrival time and the slowness is nonlinear. Therefore, it is important to deploy many sensors well positioned within the hydrocarbon reservoir in order to make use of such induced micro-seismic events for monitoring, characterizing and/or imaging of the hydrocarbon reservoir.The current practice uses a fixed slowness model to obtain the origin times and hypocentres of induced micro-seismic events within a hydrocarbon reservoir. This, on the one hand, assumes that the velocity model is not changing, which may introduce errors into the hypocentres and origin times. It also ignores the information carried by the waves through the inactive zone. This, on the other hand, cannot replace the conventional 4D seismic time-lapse monitoring method to monitor the dynamic changes within a carbonate hydrocarbon reservoir.In this thesis, I present an iterative two-stage integrated framework to incorporate arrival times in order to address the problem. First, to estimate the hypocentre and origin time for each micro-seismic event, I have developed and implemented a systematic grid search algorithm to obtain the global minimiser of a nonlinear and multimodal objective function. The algorithm can also be applied to SWD (seismic while drilling) data to locate the drilling bit. Second, to reconstruct an improved velocity model, I have developed and implemented a two-phase algorithm to initially construct an objective function with its gradient for all the micro-seismic events and then apply the variable metric method to optimise the objective function. The algorithm can also be applied to VSP (vertical seismic profiling) data to construct the velocity model. The procedure is iterated until an acceptable match between observed data and computed synthetics is achieved. There are two main reasons for such a choice. First is the fact that both the position coordinates and origin time are unique for each particular micro-seismic event, whereas the slowness is common to all sources. Second is that we start with a good velocity model resulting from the prior information within a hydrocarbon reservoir and this can lead to a very accurate source positions coordinates and origin times. The framework can be used for either P-wave or S-wave.The methodology could lead to enhanced understanding and hence efficient management of the hydrocarbon reservoir. This in turn would enhance the understanding of fluid movements resulting in improved petroleum recovery from the reservoir.