J. Cent. South Univ. (2016) 23: 729-739

DOI: 10.1007/s11771-016-3118-9

Applications of fluid substitution effect analysis on seismic interpretation

CHEN Wei(陈伟)^{1, 2}, WANG Shang-xu(王尚旭)^{1, 2}, CHUAI Xiao-yu(啜晓宇)³, LIU Yong(刘勇)^{1, 2}

1. State Key Laboratory of Petroleum Resources and Prospecting (China University of Petroleum),

Beijing 102249, China;

2. College of Geophysics and Information Engineering, China University of Petroleum, Beijing 102249, China;

3. Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing 100029, China

Central South University Press and Springer-Verlag Berlin Heidelberg 2016

Abstract: The root mean square (RMS) difference of time-lapse seismic amplitudes is routinely used to identify the substituted fluid type in a reservoir during oil field production and recovery. By a time-lapse seismic method, we study the effects of fluid substitution in a physical model, which is an analogy of the three-dimensional inhomogeneous reservoir. For a weak inhomogeneous medium, gas/oil substitution results in positive anomalies in the reservoir layers, and negative anomalies below the bottom of the reservoir layers; while water/oil substitution causes only weak variations in the reservoir layers, but positive anomalies below the bottom of the reservoir layers. For the strong inhomogeneous medium, no matter what kind of fluid substitution (gas/oil or water/oil), there are significant anomalies in seismic amplitude difference attributes both in and below the reservoir layers. Therefore, for weak inhomogeneous media, such as tight sandstone or thin interbedded layers, the RMS amplitude difference attributes can be used to monitor fluid changes and predict the drilling direction; for inhomogeneous medium with karst carves or fractures, it is difficult to accurately determine the distribution of fluids with the RMS amplitude difference attributes.

Key words: inhomogeneous medium; physical modeling; time-lapse seismic technique; fluid substitution effect; seismic attribute

1 Introduction

Along with the increase of exploration and development, it is not only important to carry out precise imaging on the reservoir filled with potential oil or gas, but also to locate and identify the fluid in the reservoir, which could make reservoir development work more efficient and economic [1-6]. In recent years, researchers have done a lot of useful work in the reservoir fluid recognition based on seismic data. Korneev et al [7] studied the influence of reservoir filled with water or oil on the amplitude and traveling of seismic reflection in the low frequency condition, and found that the response of the reservoir filled with water is strong. Li and Chen [8] put forward a new reservoir fluid identification factor based on seismic data according to rock physics model, which reduced the uncertainty of reservoir fluid identification and improved the reservoir drilling success rate, but could only be used in the calculation of the petrophysical parameters of the rock containing a single fluid. Others used the Gassmann and Biot theory and the knowledge of rock physics to do some work in fluid identification such as AVO technology [9-12], and their work is only used to calculate petrophysical parameters of the reservoirs filled with a single fluid. Time- frequency analysis methods are widely applied in reservoir fluid recognition. Especially, after the instantaneous spectrum decomposition technique was proposed by Castagna [13], some people applied the wavelet transform [14] and Hilbert-Huang transform [15-16] into the instantaneous spectrum decomposition. Instantaneous spectrum decomposition method can accurately identify the gas-bearing reservoir area through the low-frequency shadows phenomenon. The other researchers carried out the reservoir fluid identification work from the perspective of experiment. Wandler et al [17] proved that AVO response characteristics of the holes can be used as a distinguishing factor of reservoir fluid identification by physical modeling technique, which coincide with the numerical calculation results of predecessors, thus showed that the physical modeling technique is effective in the reservoir fluid identification.

In fact, the seismic attributes play an important role in the fluid identification [18]. On account of the complexity of actual geology, seismic attribute technology brings us direct cognition as well as some pitfalls in the practical seismic data interpretation. For example the location of the reservoir may be misjudged. Amplitude attribute is the most common attribute in the seismic exploration field, which shows the structure of formation more clearly and more accurately. But whether reservoir characteristics that seismic attributes reflect is consistent with actual situations as far as inhomogeneous medium still needs to be validated by experiments.

The reservoir fluid identification technologies mentioned above take no account of the effects of inhomogeneous medium which may be affected strongly because of the existence of caves or cracks, thus the indirect method to identify reservoir fluid may be a good strategy. Time-lapse seismic technique is used widely to identify and monitor reservoir fluid indirectly, and this method can judge fluid changes through fluid substitution effect which is analyzed from the data of different times in the same region [19]. As we all know, reservoir fluid substitution effect analysis is widely used in the geophysics [20]. This work is a follow-up study built on the research of Wang et al [21]. Wang et al [21] applied P-wave seismic surveys on a realistic physical model formed of interbedded sand and shale layers and filled with gas, water and oil to show anomalous changes which can be used to distinguish between fluids, with significant implications for the interpretation of time-lapse experiments. They thought that the reason for the unique anomalies was the presence of multiples inside the interbedded layers. This work designs a 3-D physical model，in which there are two kinds of inhomogeneous media (weak and strong), to discuss the anomalies of different fluid substitutions. In our experiment, there are few multiples inside the layers, but multiply-scattered waves are obvious. This work will clearly demonstrate how the unique anomalies come out in the RMS amplitude difference attributes.

2 Data acquisition and processing

2.1 Preparation of model

The study of scattering in the geophysical exploration has a long history, which is taken seriously for being widely applied in modern industry. Wu and Aki [22] summarized the scattered seismic wave theory, and classified the seismic wave field according to the inhomogeneous body scale under the condition of the wave number being a constant, as shown in Fig. 1. The classification approach is based on the seismic wave number (k) and the inhomogeneous body scale (a). From Fig. 1, when the size of ka is around 1 (wave number is comparable to inhomogeneous body scale), the scattered wave field intensity is very strong; when ka is between 0.001 and 0.1, the scattered wave field intensity is weak; when ka is very small (less than 0.001), scattered characteristics of wave field will disappear; when ka is large (more than 10), the diffracted wave field intensity is very high; when ka is large enough (more than 100), the diffracted wave field intensity becomes weak and reflection intensity is high, and we can use ray theory to explain the related physical phenomena.

The physical model in this work aims to replicate a reservoir comprised of three lateral layers which are made of sandstone, epoxy resin and steel balls (see Table 1 and Fig. 2). The sizes of particles in the model are designed according to the classification of wave field in Fig. 1. The overlying ( layer 1 in Fig. 2(a)) and bottom (layer 3 in Fig. 2(a)) layers of this model are made by the same pure epoxy resin, and have the same compressional velocity and represent homogeneous media; the middle layer (layer 2 in Fig. 2(a)) represents a inhomogeneous medium, whose right half (0-250 mm) is made of sandstone (particle size is between 0.1 mm and 0.125 mm) and epoxy resin (a weak inhomogeneous medium), and the left half (250-500 mm) is made of steel balls with the diameter of 6mm and the same sandstone and epoxy resin (a strong inhomogeneous medium). The model is acquired at a scale of 1:10000, with the frequencies scaled up by 10000:1. The dimensions of layers 1 and 3 are both 500 mm×500 mm×30 mm, and that of layer 2 is 500 mm×500 mm×49 mm. However, geometry errors will appear in the actual model production. The steel balls can simulate strong inhomogeneous medium, while sandstone medium simulates weak inhomogeneous medium. In addition, there are no complicated structures in the physical model, and only few flat layers, which is mainly to rule out other factors and only considers the fluid’s influence on the heterogeneity.

Fig. 1 Scattering types based on geometry size of medium

Table1 designed and actual speeds of physical model

2.2 Model parameters

The model is disassembled to measure the layer parameters. The model parameters are as follows.

1) The horizontal cross section of the model is square with 500 mm length and its maximum error is 0.2 mm (similarly hereinafter).

2) The thickness, density and compressional velocity of the overlying and bottom layers are 30 mm, 1.15 g/cm³ and 2595 m/s, respectively.

3) The thickness of the middle layer is 49 mm, and the P-wave velocity of the right half is 2304 m/s and the left half is 2549 m/s.

The seismic physical model simulates a realistic geological model, with the corresponding parameters listed in Table 2.

2.3 Seismic physical modeling experiment

The whole physical model is immersed in a water tank, as illustrated in Fig. 2(b). The water surface to the top of the physical model is 78 mm. The wave sources and receivers are located 0.5 mm under the water surface. The 3-D observation system is an 8-line-8-shot spread, with the source on the right and the hydrophone on the left (see Fig. 3). The parameters of this system are given in Table 3. The source consists of an adjustable impulse generator, which is a 23-bit analog-digital converter. The dominant frequency of the pulse is approximately 260 kHz. The numbers that specify the dimensions are in millimeter, and the model is scaled down by 1:10000 and frequency scaled up by 10000:1.

Fig. 2 3-D physical model (a), a photograph of modeling system (b) (The maximum movements in x, y and z directions are 500 mm×500 mm×109 mm, respectively, and the position error is less than 0.2 mm. Square waves pulser/receiver Model 5077PR (a trademark of Square waves) is used, and the dominant frequency is 260 kHz with a data bandwidth of 120-360 kHz) and principle diagram of seismic data acquisition (c)

Table 2 Relation between physical model and real geological model

Fig. 3 Source–receive layout with an aerial spread (a) (the shot points are located at the centre of the spread and fired into 256 receivers) and full-fold domain, weak scattering domain and strong scattering domain map (b)

Seismic experiments were conducted with the reservoir in the physical model filled with gas, water or oil, keeping all other parameters identical. Efforts were made to ensure the positions of survey lines to be the same in all experiments.

The seismic experiments were conducted as follows.

Table 3 Parameters of seismic observation system

1) With air filling the sand layers, the model is put into a vacuum thermotank and baked for 24 h at 50 °C. The sides of the model are sealed to prevent fluid from exchange. The model is then immersed in the water tank and located at a depth of 78 mm. Seismic data are acquired by the observation system listed in Table 3.

2) The seal is removed after the first experiment. The middle layer is fully saturated with water through vacuum pumping, and this process is repeated until the mass of the model does not increase anymore. The total mass of water is 623 g. All open sides are sealed again before the model is put into the water tank. Seismic measurements are carried out in the same way as for the first experiment.

3) The seal is removed after the second experiment. The middle layer is put into a vacuum thermotank and the model is baked at 50 °C and vacuum pumped every 2 h. It is weighed for 24 h later to determine the amount of discharged water. This process is repeated for 4 d when the total amount of water discharged from the model is 592 g. In the last 2 d very little water is discharged, so there is still about 30 g in the middle layer. Then the model is saturated with oil by the same method as the second experiment. The total mass of oil is 615 g. The model is sealed again and put into the water tank, and the data acquisition is repeated.

Figure 4 shows three raw common shot gathers data (1-95 gathers of the 8th) in the third experiment (with the middle layer saturated with oil). From the gathers, we note the strong reflection of the top of the underlying stratum, and the direct wave, random noise, and reflection of the model bottom clearly.

2.4 Data processing

In order to analyze seismic reflection data of physical models filled with gas, water and oil, we upscale the acquisition system according to Table 1.Filtering is not used to suppress noise because of the presence of weak scattering signals in the seismic data. Only common seismic processings, such as velocity analysis and prestack migration, are applied in the data processing (see Fig. 5). In order to avoid the boundary refraction and reflection, only the traces among CDP 88-168 in crossline direction at inline 106 and the traces among CDP 66-146 in inline direction at crossline 88 are used in reservoir characteristic analysis. Figure 6 shows the migration sections from the gas, water, and oil-filled models, respectively. The reservoirs of the left side of Fig. 6(a), 6(c) and 6(e) are weak inhomogeneous medium and the right side is strong inhomogeneous medium. The reservoirs in Figs. 6(b), 6(d) and 6(f) are weak inhomogeneous medium. Acquisition and processing parameters are the same for each experiment. The migrated sections in Fig. 6 show the following characteristics:

1) Horizontal events from the top of the overlying layer are almost identical because this layer does not change between experiments.

2) The seismic responses from the reservoir filled with water, oil and gas are clear, and the seismic amplitudes of the strong inhomogeneous reservoir are stronger than those of the weak inhomogeneous reservoir.

3) Horizontal events from the top of the gas-, water- and oil-filled reservoirs are clear, while the bottom of the reservoirs cannot be identified.

4) As the reflections from the top of the overburden layer are rather strong, the difference of the migrated sections corresponding to the water-, oil- and gas-filled reservoirs (for both the weak and strong heterogeneity) cannot be discriminated; therefore, seismic attributes are introduced for the amplitude differences analysis in the later sections.

Fig. 4 Three raw common shot gathers (shots 17, 25, 34) from oil-filled model (numbered wave arrivals are (1) direct wave, (2) reflection of water bottom/model interface, (3) random noise, (4) reflection at reservoir, (5) multiple wave, and (6) scattering wave of reservoir)

Fig. 5 Amplitude difference attributes calculating process

3 Data analysis

To study the different reflection characteristics of different reservoirs, this work applied the time-lapse seismic technique to the seismic data [21]. Before that, cross-equalization processing has been done to ensure the consistency of non-reservoir seismic data. Then, in order to study the reservoir reflection differences we take the data of the models saturated with gas and water as the monitoring data, using the seismic data of the model saturated with oil as the base data,.

In cross-equalization processing, a reference horizon whose reflection characteristics are free from the effect of reservoir fluid will be selected from the non- reservoir. In our study, we chose the interface between water and the top of model as the reference plane, which ranges from 1000 ms to 1185 ms (Fig. 6). Developing a matched filter by the reference plane, we applied it to the monitoring data, not only to keep reflection consistency in non-reservoir, but to accurately depict reflection differences in the reservoir where fluid substitution occurs (at time depth 1280-1650 ms). The same six methods are applied in this work as Wang et al [21] ever did to show the seismic amplitude differences between the cross-equalized monitoring sections and the cross- equalized base section.

Firstly, we applied the time-lapse seismic attributes analysis technique to the models saturated with gas and water. Figures 7(a) and (b) show the difference sections of the cross-equalized base data and monitoring data, in which few differences can be observed in non-reservoir, thus illustrating repeatability of non-reservoir reflection. However, reflection differences in non-reservoir from 1280 ms to 1650 ms are much more significant (Fig. 7).

Fig. 6 Migration sections for models filled with gas ((a) and (b)), water ((c) and (d)), and oil ((e) and (f)) (Crossline direction goes along x axis of model, showing cross-sections of both weak and strong inhomogeneous reservoirs. Inline direction goes along y axis of model, showing weak inhomogeneous reservoir)

Fig. 7 direct difference sections between base and monitoring surveys after cross-equalization:

Horizontal time slices of the RMS difference in the reservoir and below are shown in Fig. 8, with Figs. 8(a) and (b) representing amplitude difference slice for the substitution of gas/oil at 1300 ms and 1680 ms, respectively, and Figs. 8(c) and (d) representing amplitude difference slice for the substitution of water/oil at 1300 ms and 1680 ms, respectively. In the case of substitution of gas/oil, positive maximum amplitude anomaly appears in the reservoir (at 1300 ms) of weak inhomogeneity, which can also be found in other amplitude difference attributes slices that are not shown here (Fig. 8(a)). In the meantime, negative maximum amplitude anomaly occurs below the reservoir (at 1680 ms) as shown in Fig. 8(b), which can also be observed at depths ranging from 1650 ms to 1700 ms. When it comes to the reservoir of strong inhomogeneity, the amplitude differences are not significant within or below the reservoir. This result suggests that in the process of gas injection production, time-lapse response of oil and gas reservoir filled with inhomogeneous media, such as tight sand and thin interbeds, can be observed. Fluid substitution (gas/oil) occurs where positive maximum amplitude anomaly appears in the RMS amplitude difference attribute sections, whereas fluid substitution will not occur when the negative maximum amplitude anomaly appears. This observation is of great use in oil extraction process driven by gas. However, it is difficult to apply time-lapse seismic technique to monitor time- lapse responses of the inhomogeneous layer where caves and cracks are well developed.

Then, we applied the same technique to the models saturated with water and oil. Data are shown in Figs. 6(c), 6(d) and 6(e), 6(f) separately. Similar reflection characteristics with clear top reflection of the reservoir are found in profiles of both models. A matched filtering operator was specifically developed for both models, and applied to the water-filled model. Figures 7(c) and 7(d) show amplitude difference sections that both models reveal the similar reflection characteristics. However, obvious anomalous response does not show up in the reservoir.

Figures 8(b) and 8(d) show horizontal time slices of the RMS difference in and below the reservoir for the substitution of water/oil. In these difference sections, amplitude responses of inhomogeneous reservoir at a depth of about 1300 ms are weak, as well as those of other time slices are not displayed. Meanwhile, the positive maximum amplitude appears below the inhomogeneous reservoir at a depth of 1680 ms (generally from 1650 ms to 1700 ms). Thus, we come to a conclusion that for a reservoir composed by strong inhomogeneity, amplitude variation due to the change from oil to water is insignificant no matter within or below the reservoir. Similar phenomenon goes for other kinds of amplitude difference attribute sections. Therefore, in the process of water injection in the oil field development, our physical modeling experiment has proved that time-lapse responses can be monitored in the inhomogeneous reservoir filled with oil and gas, such as tight sand and thin interbeds. Specifically, the fluid substitution of water/oil, is likely to take place in and above the depth where the positive maximum amplitude anomaly appears in the RMS amplitude difference attribute sections. Moreover, the amplitude of the reservoir where fluid substitution occurs is approaching to 0. This is an important feature, which may provide guidance for well drilling. Fluid substitution of gas/oil occurs where the positive maximum amplitude anomaly appears in the RMS amplitude difference attribute profiles. Nonetheless, no fluid substitution occurs below the negative maximum amplitude anomaly. These characteristics are of great use in oil extracting driven by gas. However, after gas/oil substitution, it is difficult to apply time-lapse seismic technique to monitoring dynamic responses of the inhomogeneous layers where caves and cracks are well developed. However, after water/oil substitution, it is difficult to apply time-lapse seismic technique to monitoring the seismic response of inhomogeneous layer where caves and cracks are well developed.

Fig. 8 RMS P-wave amplitude difference slices:

Subtracting the oil-filled mode data from that of gas-filled model and water-filled model data respectively, we get amplitude difference sections in and below the reservoirs of weak and strong inhomogeneity. Constant variations of normalized RMS amplitude difference of seismic responses are shown in Fig. 9. The selected inline number of weak inhomogeneity is from 50 to 100 and crossline number is from 50 to 100; while the inline number of strong inhomogeneity is from 50 to 100, and crossline number is from 150 to 200. The time range of the selected area is from 1280 ms to 1730 ms. The reservoir lies from 1280 ms to 1650 ms, and the time after 1650 ms represents layers below the reservoir. Figure 9 clearly shows that, in the area of weak inhomogeneity, the RMS difference amplitude between the gas-filled data and the oil-filled data are generally positive before 1650 ms and obviously negative after that. And the RMS difference of the amplitude value between the water-filled model and the oil-filled model fluctuates around 0 before 1650 ms, but it is positive after 1650 ms.When it comes to the area of strong inhomogeneity, the wave field amplitudes in and below the reservoir are relatively strong.

Fig. 9 RMS amplitude differences for amplitudes recorded in gas- and oil- filled reservoirs (blue lines), and water- and oil- filled reservoirs (red lines) from 1280 ms to 1730 ms

For direct subtraction, elastic characteristics of the reservoir will change with variation of fluid in the strata, which will affect the travel time, amplitude, frequency spectrum and phase of reflected wave in and beneath the reservoir. In the case of the slightly inhomogeneous reservoir, the balance process cannot estimate the differences in travel time accurately. The direct subtraction of two sections will produce errors because of the time difference. As a result, for amplitude difference attribute of direct subtraction, the abnormal responses generating from the reservoir are very apparent (Fig. 7). However, the calculation of the mean value, the maximum value, the minimum value, the intermediate value and the RMS attribute employs subtraction after aligning the traces, by which the effect of travel time is weakened and the effect of elastic characteristics of reservoir is reflected. The scattering effect is stronger than the attenuation effect for the inhomogeneous reservoir; therefore, the amplitude for strata containing oil is weaker than that containing gas and water. Beneath the reservoir, the attenuation due to inhomogeneity in the reservoir with gas is far stronger than that with water and oil, and the amplitude of multi-scattered wave decreases apparently, which gives rise to negative values in the RMS difference attributes after gas/oil substitution. But in the case of water/oil substitution, the energy difference of multi-scattered wave multiplies, which generates the maximum positive values in the RMS difference attributes. We should note that the previous analysis only aims at the slightly inhomogeneous medium, and it cannot be applied to the strong one. When the reservoir contains steel balls, the intensity of the scattering effect by inhomogeneities will increase apparently. In spite of different travel time and attenuations caused by strong inhomogeneity with different fluids, the difference is covered by the strong scattering effect and cannot be presented in the amplitude difference attribute. Consequently, strong inhomogeneous medium perhaps only shows the scattering effect instead of fluid effect regardless of depth.

In this work, the thick single weak inhomogeneous strata shows a RMS difference amplitude anomalous value phenomenon which is similar to that in the work by Wang et al [21]. In essence, the thin interbedded layer is a slight inhomogeneous medium, so the essential factor of the RMS difference amplitude abnormal value phenomenon is like the slight inhomogeneity of medium with fluids. The multiple from the interbedded layer contributes to the abnormal value; in contrast, the strong scattering effect from the steel balls diminishes the abnormal value.

4 Application to field data

In order to apply theory to practice, this section analyzes the fluid substitution effect according to field data A. The overall reservoir physical property is poor in the study area, and the oil and gas are rich, which is a typical weak inhomogeneous sandstone block. In the study area there are two producing wells (filled with oil), and gas injection and water injection were carried out respectively, and the seismic data were recorded before (1994) and recorded after exploitation by 16 years (2010). After the seismic processing on the old and new original data, the horizontal slices of the RMS amplitude difference seismic attributes in the reservoir and 50 ms below it are shown in Fig. 10. Figures 10(a) and (b) show the RMS amplitude attribute slices in the reservoir and 50 ms below it respectively before and after gas injection, and Figs. 10(c) and (d) are corresponding to water injection. The arrows in Fig. 10 refer to the positions of the producing wells. According to Fig. 10, we could see that the amplitude difference in the reservoir presents positive value, while the value under the reservoir is close to 0 when developing by gas injection; the value approaches 0 in the reservoir, while the amplitude difference under the reservoir presents positive value when developing by water injection. Thus it can be seen that the conclusions obtained from the actual data are consistent with the above experimental results.

Fig. 10 RMS amplitude difference attributes horizontal slices in reservoir and about 50 ms below reservoir: (a, b) gas/oil substitution; (c, d) water/oil substitution

Figure 11 shows the continuous RMS amplitude difference values (normalized) of 101 traces surrounding the wells W1 and W2 varying with time. The time ranges from 2100 ms to 2700 ms. By analyzing the well data, we know that the time domain from 2210 ms to 2510 ms corresponds to the reservoir in the well W1 and 2190 ms to 2480 ms corresponds to the reservoir in the well W2. It can be clearly seen from Fig. 11 that the RMS amplitude difference in the reservoir presents the maximum positive value by using gas injection, and the negative value occurs after 2505 ms; while the value approaches 0 in the reservoir, and the amplitude difference presents positive value after 2485 ms by using water injection. By this reckoning, we can know that the bottoms of reservoirs are at 2505 ms and 2485 ms, corresponding to the W1 and W2, respectively. Besides, according to the maximum range above the bottom in Fig. 11(a), we can estimate that the top of the reservoir in W1 is at 2201 ms, thus the thickness of gas-filled reservoir is 304 ms; according to the zero-value range above the bottom in Fig. 11(b), we can estimate that the top of the reservoir in W2 is at 2188 ms, thus the thickness of water-filled reservoir is 297 ms. Hence, the results coincide with the logging interpretation information.

Fig. 11 Continuous RMS amplitude difference values (normalized) of 101 traces surrounding wells W1 (a) and W2 (b) varying with time. The time ranges from 2100 ms to 2700 ms

5 Conclusions

1) The response to the strong inhomogeneous medium is stronger than that to the weak one.

2) For the weak inhomogeneous medium, a significantly anomalous change occurs in the amplitude difference attribute after gas/oil substitution. The dynamic responses to the tight sandstone and thin interbedded layer in the inhomogeneous reservoirs can be monitored, which contributes to steam or gas injection. But for the strong inhomogeneous medium with many karst caves or fractures, it is hard to monitor the dynamic responses by time-lapse seismic technique after gas/oil substitution.

3) For the weak inhomogeneous medium, reservoir parameters reveal a slight variation after water/oil substitution, whereas in amplitude difference attribute profiles, the responses to the reservoir are obvious and consistent. Time-lapse seismic attribute differences involving average amplitude, maximum amplitude, minimum amplitude, midpoint amplitude and RMS amplitude, have weak anomalies in the reservoir, thus are likely to be difficult to use for monitoring fluids.

4) In the case of fluid substitution, the response of the amplitude difference below the reservoir is obvious. However, the strong seismic difference response does not always mean that fluid substitution occurs.

5) Under the circumstance of strong lateral inhomogeneity in the reservoir, mutual balance process cannot remove the difference of travel time caused by two different fluids. As a result, direct subtraction of two post stack migrated sections standing for different fluids respectively will lead to errors. However, the large difference value does not imply the real difference of amplitude.

6) The difference amplitude attribute anomalies are mainly caused by the weak inhomogeneous medium filled with fluids, but the scattering effect of strata with many karst caves and fractures will diminish the anomalies.

Acknowledgments

Wei Jian-xin made the physical model, and Li Fa-lv guided the time-lapse seismic data processing. We thank Zhao Qun and Wang Shu-jiang for constructive comments.

References

[1] LI Jing-ye, DVORKIN J. Effects of fluid changes on seismic reflections: predicting amplitudes at gas reservoir directly from amplitudes at wet reservoir [J]. Geophysics, 2012, 77(4): D129-D140.

[2] LI Jing-ye. Fluid substitution in a shaley sandstone reservoir at seismic scale [J]. Petroleum Science, 2012, 9(1): 31-37.

[3] SHEN Wei-jun, LIU Xiao-hua, LI Xi-zhe, LU Jia-liang. Water coning mechanism in Tarim fractured sandstone gas reservoirs [J]. Journal of Central South University, 2015, 22(1): 344-349.

[4] CHUAI Xiao-yu, WANG Shang-xu, SHI Pei-dong, WEI Xian, CHEN Wei. Applications of texture attribute analysis to seismic interpretation [J]. Journal of Central South University, 2014, 21(9): 3617-3626.

[5] YUAN Yong, ZHANG Jin-liang, LI Cun-lei, MENG Ning-ning, LI Yan. Seismic sedimentology of conglomeratic sandbodies in lower third member of Shahejie Formation (Palaeogene) in Shengtuo area, East China [J]. Journal of Central South University, 2014, 21(12): 4630-4639.

[6] ZHANG Hong, GAO Qian, XU Bin. Scattering wave field around a cavity with circular cross-section embedded in saturated soil using boundary element method [J]. Journal of Central South University, 2013, 20(11): 3296-3304.

[7] KORNEEV V A, GOLOSHUBIN G M, Daley T M, SILIN D M. Seismic low-frequency effects in monitoring fluid-saturated reservoirs [J]. Geophysics, 2004, 69(2): 522-532.

[8] LI Jing-ye, CHEN Xiao-hong. A rock-physical modeling method for carbonate reservoirs at seismic scale [J]. Applied Geophysics. 2013,10(1):1-13.

[9] GOODWAY W, CHEN T, DOWNTON J. Improved AVO fluid detection and lithology discrimination using Lame petrophysical parameters [C]// 67th SEG Annual International Meeting. America: Society of Exploration Geophysicists, 1997: 183-186.

[10] CASTAGNA J P, SWAN H W. Principles of AVO crossplotting [J]. The Leading Edge, 1997, 16(4): 337-342.

[11] HEDLIN K. Pore space modulus and extraction using AVO [C]// 70th SEG Annual International Meeting. America: Society of Exploration Geophysicists, 2000: 170-173.

[12] RUSSELL B H, HEDLIN K, HILTERMAN F J, LINES L R. Fluid-property discrimination with AVO: A Biot-Gassmann perspective [J]. Geophysics, 2003, 68(1): 29-39.

[13] CASTAGNA J P, SUN S, SIEGFRIED R W. Instantaneous spectral analysis: Detection of low-frequency shadows associated with hydrocarbons [J]. The leading Edge, 2003, 22(2): 129-127.

[14] SINHA S K, ROUTH P S, ANNO P D, CASTAGNA J P. Time frequency attribute of seismic data using continuous wavelet transform [C]// 73rd SEG Annual International Meeting. America: Society of Exploration Geophysicists, 2003: 1481-1484.

[15] HAN J J, van der MIRKO B. Complete ensemble empirical mode decomposition for seismic time-frequency analysis [C]// 82th SEG Annual International Meeting. America: Society of Exploration Geophysicists, 2012: 1-5.

[16] HAN J J, van der MIRKO B. Empirical mode decomposition for seismic time-frequency analysis [J]. Geophysics, 2013, 78(2): 9-19.

[17] WANDLER A, EVANS B, LINK C. AVO as a fluid indicator: A physical modeling study [J]. Geophysics, 2007, 72(1): C9-C17.

[18] CHUAI Xiao-yu, WANG Shang-xu, SHI Pei-dong, WEI Xian, CHEN Wei. Applications of texture attribute analysis to seismic interpretation [J]. Journal of Central South University, 2014, 21(9): 3617-3626.

[19] LI Jing-ye, WANG Shou-dong. Pressure monitoring on the real thin inter-layer gas reservoir by time-lapse seismic method [J]. Journal of Petroleum and Gas Exploration Research, 2011, 1(2): 51-58.

[20] LI J T,LINERC,STEWART R. Time-lapse seismic modeling for CO₂ sequestration at the Dickman Oilfield, Kansas [J]. Geophysics,2014, 79(2): B81-B95.

[21] WANG S X, LI X Y, DI B R, BOOTH D. Reservoir fluid substitution effects on seismic profile interpretation: A physical modeling experiment [J]. Geophysical Research Letters, 2010, 37: L10306.

[22] WU Ru-shan, AKI K. Scattering characteristics of elastic waves by an elastic heterogeneity [J]. Geophysics, 1985, 50(4): 582-595.

(Edited by YANG Hua)

Foundation item: Project(2013CB228600) supported by the National Basic Research Program of China

Received date: 2015-05-26; Accepted date: 2015-12-05

Corresponding author: CHEN Wei, PhD; Tel: +86-13125041733; E-mail: cwjycd@163.com