作者/Adam Wilson 译者/周诗雨 编辑/魏亚蒙
Introduction and Background
Seismic imaging of hydrocarbon reservoirs with small impedance contrasts relative to bounding lithologies is challenging. Ideally, an infinite bandwidth wavelet would detect and resolve subsurface geology as observed in the resolution of a typical wireline log. Improvements in seismic acquisition and processing continue to strive toward this ideal condition, and recent advancements in seismic acquisition—namely, the broadband acquisition technique—have attempted an approximation of the infinite bandwidth wavelet through the removal of the receiver or source ghosts. This advancement is of particular interest in oil and gas development areas where known thin or poorquality reservoirs that are difficult to detect on conventional seismic records have been penetrated by previous wells.
The study area is located in the shallow (less than 200 ft) waters off peninsular Malaysia. Seven exploration and appraisal wells have targeted shallow and deep reservoirs, although only three wells have been drilled to deep targets. Generally speaking, seismic imaging of shallow reservoirs is good, while deep targets are more challenging to image.
The primary motivation for acquiring a new broadband seismic survey stemmed from imaging difficulties of thin (less than 15 ft) deep gas reservoirs on conventional seismic data. Many of these reservoirs are vertically juxtaposed by coals, which tend to mask the acoustic response of the reservoir. Even after later reprocessing of legacy data through anisotropic PSDM, imaging difficulties remained.
Seismic Imaging Challenges
The study field consists of multiple, stacked reservoirs. Coal and organic shale layers are prevalent throughout the entire section. Although very thin (less than 10 ft), they have large impedance contrasts with bounding rocks and are usually detected on seismic. Often, the seismic response from these layers overwhelms the response from the gas-charged reservoir facies.
Adding further to the imaging complexity are three geophysical data-quality issues common to the basin.
– Natural amplitude and frequency decay—The absorption of seismic energy with depth has been welldocumented. In the study area, strong frequency and amplitude decay with depth is present in the seismic record, having direct implications for imaging deeper targets.
– Shallow gas—Pockets of shallow gas, ranging in thickness from 0 to approximately 150 ft, are prolific in the basin. They are generally elongated and broad in map view and can be up to 20 km in length. Acoustically, they are very “soft” because their velocities and densities, which have been measured in well logs, are very slow and low, respectively. The bodies absorb seismic signals, causing areas beneath the gas to have both amplitude and frequency loss, although this is usually repaired deeper down as the gas is undershot by farther offsets. If the slow velocities of the shallow gas bodies are not incorporated in the velocity model, then structural sags are observed beneath them. Traditionally, it has been very difficult to incorporate the shallow gas bodies into the velocity model because of low fold in the common depth point gathers at very shallow times.
– Fault shadow—A third challenge to imaging occurs in what are known as fault shadows. A fault shadow is generally expressed as a triangular zone of distortion on the footwall side of a normal fault. Usually, the larger the fault, the more severe the fault-shadow effect is. In the study area, faults are most prevalent at the crest of the structure where imaging is most important. It is believed that fault shadows are the result of not fully capturing velocity differences across the faults.
Seismic Acquisition, Processing, and Stack Comparison
There are approximately 500 km2 of full-fold seismic overlap and three wells common to the 2003 legacy survey and the 2012 broadband survey. The general processing flows between the two surveys are similar; however, important differences in short-period multiple- attenuation techniques, the handling of Qattenuation, and velocity-model building exist between the two surveys.
In the cross-section view, the broadband section appears richer and of lower frequency than the legacy data. Quantitatively, the actual frequency range at target depths is larger in the broadband data because of the added low-frequency content. It appears that the legacy spectrum contains stronger midrange frequencies. However, signal/noise analysis suggests that legacy noise levels are the highest in the midrange frequencies; thus, the legacy midrange frequencies are interpreted to be noise-dominated. Qualitative comparison of the two data sets under gas clouds and fault shadows shows significant imaging improvement on the broadband data, although imaging challenges still remain.
Interpretation Work Flow and Results
Full geophysical evaluations were conducted on the legacy and broadband data sets across the study area. Two static geological models (geomodels) were constructed with input from both of these evaluations. In terms of the contribution to the geomodel, the evaluation work flow is twofold: structural interpretation and stratigraphic assessment. The former methodology consists of well ties, mapping, and depth conversion, resulting in the depth surfaces that make up the model framework. The latter method involves investigating the rock properties from available well data, which provides a link between the rocks and seismic-derived attributes. When there is high confidence that a specific attribute corresponds to the presence of sand, for example, then that attribute is used to condition the distribution of sand in the geomodel qualitatively.
Seven common seismic surfaces were mapped in two-way time and subsequently transformed to depth by use of the respective anisotropic PSDM velocity calibrated to sonic and checkshot velocities measured in the wells. These surfaces cover a two-way time window of approximately 1.5 seconds and bound the structural framework for the geomodel. In general, the broadband interpretation is more continuous and there is less manual interpretation required for the broadband surfaces, although manual interpretation is still required in highernoise areas beneath shallow gas bodies or under fault shadows.
Comparison of Legacy and Broadband Data
Considering the target reservoir previously used for 1D modeling(Fig. 1),the far stack amplitude map for this particular interval is indicative of gas reservoir and is used by the geomodeler to condition facies distribution. Similar stratigraphic features are observed on the legacy- and broadband-data-set attribute maps. There is an east/west-trending channel geometry that contains amplitude anomalies against a set of faults to the east. High-frequency noise contamination is evident on the legacy map, while the broadband map appears much smoother. Outside of the main channel axis, amplitude anomalies associated with coal are observed. However, on the broadband map, there also appear to be additional channels to the north, representing upside resources, which are not seen on the legacy map. A likely explanation for this is that the broadband signal is able to analyze reservoirs and distinguish them from coal and organic shale when the two facies are in close proximity.
Seismic data contain signals and some amount of coherent and random noise. Signal/noise analysis at our study area indicates that the broadband data are cleaner overall and have a higher signal/noise ratio relative to conventional streamer seismic. A high signal/ noise ratio enables the potential for high- frequency spectral shaping to try to enhance thin-bed resolution, especially if the noise floor is low on the high end.
Broadband seismic acquisition is an improvement over conventional streamer acquisition in terms of imaging complex, fluvial/tidal geologic layers. Before acquiring broadband data, it is crucial to understand what the specific problems are for imaging. In this case, the main causes for seismic-imaging problems stemmed from shallow gas bodies, fault shadows, coal seams, and the natural absorption of energy with depth. Synthetic forward modeling is an excellent tool for feasibility studies. In terms of frequency content, the actual broadband data contain more lower and higher frequencies than conventional data and have a higher signal/noise ratio. These factors have resulted in better resolution and detection of thin reservoirs in complex impedance environments, which, in turn, have bolstered confidence in the reservoir model used for field development.