Seismic reservoir monitoring with encoded permanent seismic arrays

Introduction

Time-lapse seismic reservoir monitoring is an established technology. By repeating the seismic experiment over an evolving reservoir, changes in reservoir properties can be estimated from seismic amplitude and travel-time changes. Many successful case studies demonstrate the technical considerations and business impact of time-lapse seismic (Rickett and Lumley, 2001; Whitcombe et al., 2004; Ebaid et al., 2009).

By enabling seismic recordings at small time intervals, permanent seismic arrays can make near-real-time reservoir monitoring possible. Lumley (2001,2004) discusses important business and technical drivers for permanent seismic arrays. Several field experiments have been published (Meunier et al., 2001; Forgues et al., 2006; Smit et al., 2005). Because permanent arrays do not suffer from positioning errors, seismic experiments can be repeated with high accuracy. However, in addition to the high operation and storage costs of the recorded data volumes, conventional processing cost of the recorded data can be high. Although, under certain conditions, simple (e.g. NMO) processing can give satisfactory results (Forgues et al., 2006), such methods are inadequate in many geological environments. In this paper, I show that direct wave-equation migration of encoded data sets from such permanent recording systems can provide high-quality time-lapse images at relatively low cost.

Encoded seismic data recording with permanent seismic arrays straddles conventional and passive data recording. Although our understanding of passive data imaging has improved over the past decade, several limitations still exist. Direct imaging of passive data suffers several pitfalls (Artman, 2006), and interferometric Green's function retrieval is computationally expensive (de Ridder, 2009). In many scenarios, seismic reservoir monitoring with interferometric Green's function from surface passive seismic arrays is difficult (Lu et al., 2009). However, reservoir monitoring with active (virtual) source and ambient-noise interferometric Green's functions have been shown for borehole sensors (Bakulin and Calvert, 2004; Lu et al., 2009). Furthermore, it has been demonstrated that interferometric Green's functions from borehole systems may give satisfactory time-lapse responses in the well vicinity (Bakulin and Calvert, 2004; Lu et al., 2009). The proposed recording approach may overcome some of the current limitations in reservoir monitoring with pure passive data or well-bore virtual source methods.

Although encoded seismic recording is not new (Womack et al., 1990), recent advances in acquisition and processing technology have increased interest in the subject (Hampson et al., 2008; Berkhout et al., 2008; Howe et al., 2009; Beasley, 2008). Direct imaging of such encoded data is possible but suffers from cross-talk between data sets from different shots (Romero et al., 2000; Artman, 2006). To directly image field-encoded time-lapse data sets from non-permanent seismic arrays, a linearized inversion method can be used to attenuate artifacts caused by non-repeatable geometry and relative shot delays (Ayeni et al., 2009). Because permanent seismic arrays enable excellent repeatability of the geometry and encoding function, cross-term artifacts are similar between consecutive surveys, and linearized inversion is unnecessary.

To ensure good repeatability over the monitoring period, to limit operational cost, and to limit environmental impact, low-energy, low-footprint seismic sources are desirable. Each source waveform may be a long-duration sweep (Forgues et al., 2006), or intermittent sweeps from an idealized source. By stacking data from several low-energy sources, the signal-to-noise (S/N) ratio is increased and sufficiently high-quality data and images can be obtained. Encoding is important, because it reduces the total recording time for several shots, each requiring a long recording duration. In this paper, it is assumed that these conceptual low-energy sources are randomly and intermittently ignited over a long time period.

Using a phase-encoding migration operator and the relative time-delays between sources, the encoded data are migrated without any separation or interferometric Green's function retrieval. Because all the data are migrated with a baseline velocity model, images from different vintages are not aligned and must be cross-equalized. In this paper, the data are cross-equalized using a cyclic 1D correlation algorithm and an optimized local-matching method (Ayeni, 2010).

First, we give a conceptual description of the proposed data recording and imaging methods. Next, we summarize the cross-equalization methodology that is applied. Finally, using five data sets from a 2D numerical model, we show that the proposed method gives good-quality time-lapse images.

Seismic reservoir monitoring with encoded permanent seismic arrays