Towards an Improved DAS Workflow for Geothermal Resource Development

Abstract

I have created a novel seismic data processing method and an algorithm that can drastically reduce computation times of finite difference methods (FDM) applied to acoustic wave equations. The former is a data decomposition method designed specifically so that its output can efficiently and accurately describe seismic signals. The method, referred to as shifted-matrix decomposition (SMD), was used to reduce the memory requirements of seismic data, improve signal-to-noise ratio (SNR), and detect seismic events. For compression and denoising, SMD was tested on marine seismic gathers, which contained a large number of reflected waves and noise with high coherence that resembled seismic signals. Shifted-matrix decomposition reduced the memory requirement by 80% and improved the visibility of weak reflections that were obscured by noise. For event detection, SMD was applied to detect microseismic events from distributed acoustic sensing (DAS) recordings during fracture stimulation at a geothermal experimental site. Regarding acoustic wave equations, I developed an algorithm that can be applied to standard finite difference methods to decrease the computational cost of forward modeling. An important feature of the algorithm is the calculation, at each time step, of the pressure in only a moving subdomain which contains the grid-points across which waves are propagating. The computation is skipped on grid-points at which the waves are negligibly small or non-existent. The novelty in this work comes from flexibility of the subdomain, namely its ability to closely follow the developing wavefield. When applied to a standard 2D finite difference scheme it reduced the computation time for wave propagation simulations by over 50% while maintaining low errors.