Structural model is the basis for the establishment of three-dimensional model, which reflects the formation of spatial framework and plays constrained role in phase model and property model. Structural model consists of interval model and fault model, which based mainly on structural surface owing to lack of faults. In the need of subsequent cap quality evaluation, mudstones can be divided into 4 layers based on structural interpretation of Huangliu formation. Flat grid spacing is set to 100 m * 100 m with 105 layers of vertical grid. Specific longitudinal layer is divided as follows: Four mudstones are successively divided into 20, 10, 10, 15 layers; the lower part of the reservoirs is divided into 50 layers with 2.2 m single average thickness.

During the first section of Huangliu formation (T30 - T31), the center of basin is conducive to the development of gravity flow. With transgression intensified, the scope of the deposition significantly expanded, the center of basin received a large reservoir system owing to surrounding active source system. Seismic facies have a transition from imbricate structure to chaotic characteristics, reflecting that the west of low delta directly connected to the submarine fan in the east. Main waterways have obviously undercut phenomenon in three-dimensional seismic profile, indicating rapid strong erosion with similar shape of submarine fan at home and abroad.

West of Dongfang a structure mainly developed delta front and fore-delta, up-leaning lithologic traps was formed by HST overlying mudstone and low fore-delta mudstone. During the first section of Huangliu formation, offshore beach and sand bars developed from Hainan provenance in the east. Inverted layered reservoir of III gas group in the thin edge of DFA well indicates hydrodynamic orientation. It’s predicted that higher permeability reservoirs may be in higher parts and thicker sand dam body.

On the basis of sedimentary facies and its digital boundary [10] [11] [12] , a deterministic sedimentary facies model will be established (Figure 1) as the restraint frame of lithology model and property model.

Petro-physical studied on four wells shows that: P-impedance properties can better distinguish between sandstone and mudstone. Mudstone shows high impedance, mainly concentrates in more than 8,000,000 kg/m³ * m/s region, while sandstone shows low impedance, the main concentrates in an area of less than 8,000,000 kg/m³ * m/s. Therefore, wave impedance inversion can basically describe the variation between different wells. On the basis of the structural model, wave impedance inversion model will be obtained from wave impedance inversion results to grid model (Figure 2).

Three general ways of obtaining P-wave velocity: pre-stack simultaneous inversion, stack velocity from seismic data processing and then calculate longitudinal wave velocity through DIX formula, by using other seismic attributes with better correlation. Pre-stack simultaneous inversion method is not applicable in this area because of all from post-stack data. In this study, we use two methods to establish P-wave velocity model: One is obtained by DIX formula, which input discrete point of velocity, then sampled in the geological model and interpolated to form a layer velocity model [12] [13] [14] [15] [16] . Through this method, the difference between processing velocity and Sonic calculation velocity is obvious. Another method is to establish the velocity model by using relationships between various seismic attributes and velocity. Statistical analysis through speed and impedance data of drilled wells (Figure 3), it’s easy to find that two properties have good correlation with 0.98 R-squared value, which indicates this method is feasible. So P-wave velocity model can be obtained from the above wave impedance data volume.

Compared with these two methods, we can find that it’s difficult to explain partial low-speed or high-speed stratigraphy owing to taking no consideration of time-depth conversion and velocity prediction by pressure, as well as limitation of velocity resolution. Thus, changes in the processing velocity is relatively stable, in terms of velocity obtained by wave impedance can reflect high and low

Combined with acoustic travel time curves and MTD data, pore pressure prediction model is summarized to represent the relationship between P-wave velocity difference and pressure coefficient [20] [21] . During seismic attribute modeling process, two models based on seismic processing velocity and P-wave velocity(interval velocity) are established. Combined with pressure prediction model and two velocity models, pore pressure prediction model can be obtained respectively by discriminant function.

According to the analysis of prediction error, pressure predicted by wave impedance has high accuracy with actual pressure [22] [23] [24] [25] [26] . The consideration of seismic data imaging, restriction of velocity resolution as well as partial low-speed or high-speed stratigraphy lead to pressure anomaly is difficult to get. Pore pressure by processing velocity can reflect overall vertical pressure trends. The pore pressure from impedance conversion can closely reflect the pressure characteristics. Therefore, comprehensive analysis shows that accuracy of pore pressure is much higher by wave impedance model. Pressure prediction model is recommended to use impedance conversion velocity model (Figure 5).

Different from pore pressure prediction models, solid model of pore pressure is obtained from interpolated trend control. Based on pressure coefficient of a single well, solid model of pore pressure is updated from time to time by newly drilled wells with seismic attributes as lateral trends. For Dongfang district, solid model of pore pressure is applied to data of good quality, more accurate velocity, pressure coefficient and interval velocity associated with good relativity.

Figure 6(a) and Figure 6(b) are contour maps of reservoir and cap pressure coefficient through pressure prediction model constraints ,among which 6(a) is contour maps of reservoir pressure coefficient, 6(b) is contour map of cap pressure coefficient. As can be seen from the figures, the reservoir pressure coefficient is concentrated from 1.91 to 1.93, the average is about 1.93, partial pressure of the pressure coefficient is 1.94. Changes of cap pressure coefficient are relatively large, concentrated from 1.7 to 1.8, the average is 1.78, partial pressure of is 1.81, which can reflect the cap sealing capacity. Meanwhile different pressure coefficients and structures of reservoir and cap have different requirements of specific gravity of slurry.