It is generally believed that at steady state, a heavy fluid mixture cannot float, without motion, at the top of a light fluid mixture in a cavity. The expectation is that because of pressure diffusion, segregation occurs with the light fluid at the top and the heavy fluid at the bottom. We present, for the first time, an extensive set of measurements in 5-km vertical wells in a large hydrocarbon formation of 1-km thickness with horizontal dimensions on the order of several kilometers that show a high-density fluid mixture at the top of a light-density fluid mixture at steady state. The data in the 5-km wells show liquid in the middle, and vapor at the top and bottom. In the hydrocarbon formation, there is a gradual decrease of density with depth.A theoretical model based on the thermodynamics of irreversible processes is used to provide an interpretation of the unusual density variation vs. depth both in the hydrocarbon formation and in the long wells, as well as the unusual species distribution in the hydrocarbon formation. The results reveal that thermal diffusion (caused by geothermal temperature gradient) causes the segregation of heavy components in the subsurface fluid mixture to the cold side in the Earth (that is, the top), overriding pressure and molecular diffusion (Fickian diffusion). As a consequence of the competition of these three diffusion effects, a heavy fluid mixture can float at the top with a light fluid mixture underneath. In the past, thermal diffusion has been thought of as a second-order effect. For the fluid mixture in our work, thermal diffusion is the main phenomenon affecting the spatial density and species distribution.

The Yufutsu field is a new naturally-fractured reservoir in Japan with negligible matrix porosity and permeability. This large reservoir contains near-critical gas condensate fluids with unusual fluid distribution and production. The GOR shows a decreasing trend. This is perhaps the first report of such an observation in the literature. Another interesting and perhaps related observation is that measured data in some wells show the gas richness decrease with depth; methane increases while heptaneplus decreases with depth. Steady-state data from 5-km long wells at shut-in conditions show liquid in the middle, and gas in the top and bottom.We used the expression of total diffusion flux (molecular, pressure, and thermal diffusion) to analyze and interpret the data in the Yufutsu field. One-dimensional and two-dimensional computations were performed using the Newton method and a new robust algorithm. Predictions were made in the entire reservoir using one single PVT sample in good agreement with data. Based on this work, it is concluded that thermal diffusion significantly affects the fluid distribution in the reservoir with consequence of decreasing the GOR produced.

TX 75083-3836, U.S.A., fax 01-972-952-9435. AbstractThe Yufutsu field is a fractured granite, gas condensate reservoir, located on Hokkaido island, in the north of Japan. There is minimal primary porosity and secondary porosity consists of fractures. Fractures are categorised into four types according to size (mega, major, minor and halo fractures). Mega fractures govern the flow, while micro fractures govern gas storage. Gas migrated into the reservoir after the fractures were formed.

Increased interest in gas condensate reservoir which exhibit complexities due to the production of gas, condensates and many times an in-situ oil phase has been observed. As expected, there are definite production problems and exploitation concerns which are at issue with gas condensate reservoirs, or reservoirs which exhibit a combination of gas condensate and oil characteristics. In studying reservoirs of this type, there are three main areas which must be adequately addressed in order to develop an appropriate exploitation strategy. These areas are adequately summarized as:The characterization of the gas condensate fluidsThe coupling of the inherent phase behaviour and fluid flow in the porous mediaImplementation into a simulator for forecasting capability This work seeks to provide a summary of important characteristics which must be considered in each of these three areas. The knowledge gained from studying a number of these reservoirs over the years will then be implemented by the authors to provide an example of a possible approach which should be followed for designing production from such a reservoir. Based upon this work, it has been found that the characterization of the gas condensate fluids is strongly influenced by two main factors which include any degree of contamination by a free liquid phase in situ and hold-up of retrograde condensate in the formation resulting in excessive producing GORs. The coupling of the fluid phase behaviour and fluid flow in the rock appears to be governed in the areas of the degree of retrograde condensate accumulation, interfacial tension effects, and mobility effects. Part of this coupling also indicates the difference between critical condensate saturation and residual condensate saturations. Finally, an analysis of techniques to improve retrograde condensate behaviour is described and the necessary components of a reservoir simulator are discussed including results from such a simulation. Characterization of Gas Condensate Systems Condensate reservoirs are inherently more difficult to characterize correctly. The literature shows many differences between gas condensate reservoirs and dry gas reservoirs. Figure 1 provides a fairly typical GOR versus total flow rate response from a gas condensate reservoir. At very low flow rates, one has a high producing GOR and, beyond the certain minimum value in GOR, the trend is again upwards. It is easy to identify why this occurs, but sometimes, when faced with the possibility of having extra sampling runs and spending more time in the field some operators believe that the cost outweighs the benefit. In the same plot the response normally seen for an oil reservoir is also shown. With the oil reservoir, the sampling technique is fairly easy to specify; one must produce the well in the domain low enough so that a constant GOR is produced. Such is not the case with gas condensate reservoirs however. At low flow rates, the liquid hold-up will increase and slugging may result. The increased GOR to the left of the vertical line is due to the low flow rate not providing enough lift to transport the liquids in the wellbore. By contrast, one may be inducing liquid dropout in the reservoir at high flow rates to the right of the vertical line in Figure 1. In this case, as the pressure drops below the dewpoint, the liquid will begin to collect in the near wellbore region. In so doing, produced hydrocarbons will contain less liquid than they should and therefore the GOR will be high. Thus, "At what producing rate should a well be sampled?", the only way that one can adequately respond is if one already knows the character of the fluid and the dynamics of the production well. Since this information is not available, sampling gas condensate reservoirs in an optimal manner can sometimes be non-linear and include some trial and error. P. 545

They were more help to rrl. than they realize. I would like to thank my parents, Shoichi and Umeko Anraku, for their love. 1 l am proud that I am your son.