a.2 prevention of lng roll
Transcription
a.2 prevention of lng roll
PREVENTION OF LNG ROLL-OVER IN AN LNG TANK PREVENTION DU RETOURNEMENT DU GNL DANS LE RESERVOIR A GNL Moriyoshi Tamura Yasuhisa Nakamura Hiroyuki Iwamoto Toho Gas Co., Ltd. Nagoya, Aichi, Japan ABSTRACT When new LNG (in cargo) enters a tank partially filled with another LNG (inventory), and the two LNGs are not fully mixed, an LNG stratified layer appears in the tank because of the slightly different densities of the two LNGs. LNG stratification in a tank causes a roll-over phenomenon that produces an almost spontaneous boil-off of large amounts of methane vapor. It is important to fully mix cargo and inventory to prevent LNG stratification in order to prevent LNG roll-over in the tank. Recently, the number of countries of LNG origin from which LNG is imported by Toho Gas Co., Ltd. has been increasing. We need to review our standards for the prevention of LNG roll-over in tanks. Thus, we study the prevention of LNG roll-over in an LNG tank experimentally and numerically. LNG entering a tank through a feed nozzle at the tank bottom (bottom feed) was simulated by using a replica tank 1/60 the scale of our tank (tank capacity; 75,000m3) at CHITA LNG terminal. The mixing of two miscible liquids was observed by LASER visualization, and the density profiles obtained numerically were compared with those measured in the tank. The mixing mechanism of light LNG entering a tank by bottom feed, which previously was unclear, was revealed. In light LNG bottom feed filling, stratification always occurs, but the interface between the two LNG layers reaches the exit of the feed nozzle. In heavy LNG filling, the lower exit of the nozzle allows the stratification to occur more easily. RESUME Quand un nouveau chargement de GNL est injecté dans un réservoir déjà partiellement rempli de GNL (stock), le mélange ne s'effectue pas complètement. Des couches stratifiées de GNL se forment dans le réservoir à cause de légères différences de densité. La stratification du GNL dans le réservoir provoque un phénomène de retournement qui se traduit par la perte par évaporation de grandes quantités de vapeur de méthane. Il est important de mélanger complètement le chargement et le stock et d'éviter la stratification du GNL pour empêcher le retournement du GNL dans le réservoir. A.2–1 Depuis quelques années, Toho Gas Co., Ltd. importe du GNL d'un nombre croissant de pays. Nous devons revoir nos normes pour éviter le retournement du GNL dans les réservoirs, et étudions donc ce phénomène sur le plan expérimental et numérique. L'injection de GNL dans un réservoir par une tuyère d'injection à la base du réservoir (alimentation par la base) a été simulée en utilisant un réservoir au 1/60e de sa grande nature (capacité du réservoir: 75.000 m3) au terminal GNL de CHITA. Le mélange des deux liquides miscibles a été observé par visualisation LASER, et les profils de densité obtenus numériquement ont été comparés à ceux mesurés dans le réservoir. Le mécanisme de mélange à l'injection du GNL léger dans un réservoir par alimentation par la base, jusque là peu clair, a ainsi été mis à jour. A l'injection de GNL léger par la base, il y a toujours stratification, et l'interface entre les deux densités de GNL atteint la sortie de la tuyère d'alimentation en dernier. A l'injection de GNL lourd, la sortie inférieure de la tuyère facilite la stratification. 1. INTRODUCTION Roll-over, which occurred in several storage tanks for liquefied natural gas (LNG) over two decades ago, is unfavorable. LNG storage tanks are ordinarily operated at ambient pressure. If roll-over occurs in a tank, a spontaneous boil-off of a large amount of methane vapour occurs. The consequent over-pressure in the tank sometimes becomes unexpectedly large. As a result, the boil-off of methane vapour in the tank has to be discharged into the atmosphere by the prompt opening of safety vents. Residents near LNG storage tanks are extremely sensitive to the safety of LNG storage tanks in Japan. In addition, global warming is drawing more attention. Therefore, it is extremely difficult to discharge methane vapour, which is combustible and has an influence on global warming, into the atmosphere. From this point of view, even now it is important to prevent roll-over in LNG storage tanks. It is well known that the cause of roll-over is LNG stratification in a storage tank. When new LNG in cargo enters a tank partially filled with another LNG (inventory) and they are not fully mixed, the LNG in the tank separates because of the slightly different densities of two layers. One is lighter than the other. This phenomenon of stratified layers is called LNG stratification. If we leave aside LNG stratification, over time the density of the upper light layer gets heavier and heavier due to the continuous evaporation of the boil-off of the methane vapour, which is a lighter component of LNG than other hydrocarbons. The bottom layer gets lighter and lighter because the layer temperature increases as a result of heat transfer from tank walls. Finally, densities of both layers become same and the two layers suddenly begin to mix. The sudden mixing in the tank causes the heat accumulated in the bottom layer to produce a large amount of boil-off gas. This is roll-over in the LNG tank. There are several techniques for mixing LNG in a tank in order to prevent roll-over after LNG stratification appears. For example, circulating LNG in the tank. Preventing LNG stratification during the tank filling operation is an important fundamental procedure for preventing LNG roll-over in the tank. A.2–2 The suitable technique for filling a tank is selected according to the following conditions; the density difference between the cargo LNG and the inventory LNG, the volume of inventory and whether the density of the cargo LNG is lighter than that of the inventory LNG. When the density of cargo LNG is heavier than that of inventory (heavy LNG filling), there are two techniques which take the difference in density into account. If the difference in LNG density is large, cargo LNG is fed by a nozzle at the top of the tank on downward angle (top feed). If the difference in LNG density is small, cargo LNG is fed by a nozzle at the bottom of tank, at an angle of about 45 degree to the horizontal plane (bottom feed). Bottom feed is preferable to top feed because in top feed, the boil-off of methane vapour increases more than that with bottom feed. When the density of cargo LNG is lighter than that of inventory LNG (light LNG feed), the bottom feed is only used because the possibility of LNG stratification is high through top feed. In Japan, information on technical standards for the prevention of LNG stratification is exchanged among companies which have their own LNG storage tanks. Though we don’t have a standard for our own tanks, we operate LNG feeding by adopting the technical standards applied to tanks of similar scale to ours. Recently, the number of countries of LNG origin from which LNG is imported by TOHO GAS CO. LTD has been increasing as the amount of city gas we produced has increased. Therefore the density range of LNG which is fed into storage tank has become wider. We need to review our operations for the prevention of LNG roll-over in tanks. Thus, we study the prevention of LNG stratification in an LNG tank experimentally and numerically. Light and heavy LNG filling into a tank through a feed nozzle at the bottom of the tank (bottom feed) was simulated by using a replica tank 1/60 the scale of our present tank (tank capacity; 75,000m3) at CHITA LNG terminal. For light LNG filling, the mixing processes that take place in a storage tank during the filling operation were numerically analysed. The nature of the mixing process, which was not previously clear, was revealed. 2. PHYSICAL MODEL FOR A TANK FILLING OPERATION Since the cargo LNG fluid, whose density differs significantly from that of the inventory, enters a tank with considerable momentum, the mixing process during LNG filling is mainly governed by both inertial and gravitational forces. The diffusion due to the concentration difference of these fluids and the natural convection by heat transfer from the tank walls are not influenced. Therefore these effects can be neglected. It is also assumed that the effect of the gas phase is ignored. In this study, the Reynolds number, Re, and the Froude number, Fr, are nondimensional governing parameters. The symbol d indicates the diameter of the nozzle into a tank; U is the mean velocity at the outlet of the nozzle; ν, the kinematic viscosity of cargo LNG; ρa, the density of the cargo; ρa, the density of the inventory; ∆ρ the difference between the two densities and g, gravitational acceleration. The Reynolds number, Re, is defined as Ud/ν and the Froude number, Fr, U/((∆ρ/ ρa)•g•d0.5. A.2–3 However, when the Reynolds number of the plume from the outlet of the nozzle is sufficiently large, the plume structure holds similarity and does not depend on the Reynolds number. In conclusion, the nondimensional parameter governing mixing during tank filling is only the Froude number. If the Froude number of an experiments is the same as that of the practical condition, real filling operations can simulate by using a replica of our LNG tank [1, 2]. 3. OPERATION CONDITION AND EXPERIMENT A schematic diagram of the experimental apparatus is shown in Figure 1. A photo is shown in Figure 2. The test liquid was clear water for the light fluid and salt water for the heavy one. Cargo fluid was first prepared in a large reservoir tank, and then transferred through a nozzle to a cylindrical replica tank, 1/60 the scale of the practical tank, by a pump at a constant flow rate. The tank is made of acrylic in order to observe the flowstream in the tank with a LASER device. The inner diameter of the tank is 94 cm; the depth is 50cm;the inner diameter of the nozzle is 8mm; the angle of the nozzle is 45 degrees to the horizontal plane. The flow rate was measured with an electro-magnetic type flow meter. The density profiles in the replica tank were measured with a conductance cell which moved automatically with a traverse device controlled by a computer. The experimental conditions were decided from the Froude number of practical conditions in order to hold fluid dynamic similarity as mentioned in the previous section. The mixing of two miscible liquids was observed by a LASER visualization technique. A small number of small tracer particles made of a polymer were dispersed in the reservoir tank. The average diameter of these particles was 30 micrometers. Their densities were almost the same as that of the liquid in the tank. Since the small particles were moving in the stream of the cargo liquid from the outlet of the nozzle during tank filling simulation, the mechanism of liquid mixing in a tank could be observed by investigating the motion of the particle. The ray of the LASER lit up the tank fluid with the tracer particles. The light reflecting on small particles was recorded with a digital video camera. The operation conditions in our practical LNG storage tank per one tank and experimental ones in this study are shown in Tables 1 and 2, respectively. The real operation conditions were selected taking into account previous operation conditions at our Chita LNG terminal and operation plans for the future. Our Chita LNG terminal has 3 unloading arms to fill LNG storage tanks from cargo. In full rate, all 3 unloading arms work, while, in 2/3 flow rate, 2 arms of 3 unloading arms work and the rest one does not. Transient density profiles are shown in Figures 3 and 4 for light fluid filling and heavy fluid filling, respectively. A.2–4 Figure 1. Schematic diagram of experimental apparatus Figure 2. Photograph of experimental apparatus Table 1. The operating conditions in a practical LNG storage tank parameter density of inventory LNG density of cargo LNG volume of feeding flow rate(full rate) flow rate(2/3 rate) initial height of inventory LNG the Froude number(full rate) the Froude number(2/3 rate) symbol unit ρa ρc V Qh Ql Hi Fr Fr kg/m3 kg/m3 m3 m3/hr m3/hr m - Light feeding 473 453 30,000 2,788 1,888 8-15 9.53 6.44 heavy feeding 457 473 30,000 2,651 1,888 5-11.5 9.96 7.08 Table 2. The experimental conditions in a replica tank parameter symbol density of inventory LNG density of cargo LNG volume of feeding flow rate(full rate) flow rate(2/3 rate) initial height of inventory LNG the Froude number(full rate) the Froude number(2/3 rate) ρa ρc V Qh Ql Hi Fr Fr A.2–5 unit Light feeding g/cm3 1.10 3 g/cm 1.00 3 cm 139,000 cm3/s 40.3 3 cm /s 27.3 cm 13.3-25 9.53 6.44 heavy feeding 1.00 1.10 139,000 44.3 31.5 8.3-19.2 9.96 7.08 50 ρc c=1.00[g/cm 3] ρaa=1.09[g/cm 3] Hi=19.2[cm] 3 Q=2.42[m /s] 40 :3.3 [min] :9.3 :15.6 :22.3 :29.2 :36.3 :43.8 :51.5 30 20 10 Light fluid filling 0 1 1.02 1.04 1.06 1.08 1.1 ρ [g/cm3] Figure 3. Transient density profiles in light fluid filling 50 ρ c=1.09[g/cm3] Heavy fluid filling ρ a=1.00[g/cm3] 3 40 Hi=19.3[cm] Q=2.66[m /s] :3.4[min] :9.1 :15.1 :21.4 :28.0 :34.9 :42.2 :49.6 :57.5 30 20 10 0 1 1.02 1.04 1.06 1.08 1.1 ρ [g/cm3] Figure 4. Transient density profiles in heavy fluid filling 4. NUMERICAL ANALYSIS IN LIGHT FLUID FILLING Germeles numerically investigated light LNG filling by bottom feed [3]. He assumed that the physical values were uniform in the horizontal plane, and that the velocity and density profiles of the plume from the nozzle were Gaussian functions. The ordinary differential equation system, which was easy to solve, was derived from the equations of conservation of volume, mass and momentum. His theory agreed on the whole with the experimental results. However, the numerical density profiles near the bottom wall did not agree with the experimental ones. We modified Germeles’ method to fit numerical calculations to the experiment results. Though the wall effect to the plume near the A.2–6 bottom wall of the storage tank is neglected in Germeles’ theory, we estimated the momentum and decreased it slightly to consider this effect. In addition, we substituted the value of the density profiles calculated in the following equation. ρ i s the calculated value by Germeles’ method; ρm is the modified value to fit experiment results; z is the vertical distance from the tank bottom; and z0 and n are fitting parameters. In a full rate case, z0=0.25, n=0.075, in a 2/3 rate case, z0=0.25, n=0.01. ρm= (z/z0)n The results of the calculation described above are shown in Figure 5 with the experimental ones. These are in very good agreement. 50 25.0 :Experiment :Calculation 19.2 Z [cm] Hi= 13.3[cm] 25 3 Q= 40.3 [cm /s] Light fluid filling 0 1 1.05 3 ρ [g/cm ] 1.1 Figure 5. Comparison of calculations and experiment results 5. MIXING MECHANISM From observations by LASER visualization, the density profiles measured and calculations, the entrainment of plume is important to mix new fluid and old fluid of inventory for both cases, light fluid filling and heavy one. The mixing mechanism for light fluid filling is shown in Figure 6. In light fluid filling, the mixed light fluid and the unmixed heavy one separate and two layers always appear as soon as the filling operation starts. Plume from the nozzle mixes with the heavy liquid as entrainment and penetrates the interface between the light mixed layer and the heavy unmixed one due to the plume’s momentum and buoyancy. Mixing between the two layers does not occur at the interface. Over time, the interface moves downward during filling. Finally, it reaches the outlet of the nozzle and stops moving. Therefore, an unmixed layer under the outlet of the nozzle remains after the filling operation in all cases of light fluid filling. The mechanism for stratification that occurs in heavy fluid filling is shown in Figure 7. As well known, the plume from the nozzle mixes with the light fluid and streams upward initially due to its momentum. However, as the vertical momentum decreases and force of gravity becomes greater than the inertia, the plume streams downward. If plume does not reach the liquid surface, LNG stratification occurs. A.2–7 There are a lot of works on heavy fluid filling because LNG stratification occurs more easily than in light fluid filling [1, 2]. However, the mechanism of mixing for light fluid filling has not been investigated fully. We therefore study the effect of filling rate on mixing in a storage tank, which is important for real operations. The numerical results of the density profiles at the end of filling are shown in Figure 8 whose parameter is the Froude number. As the Froude number increases, the difference in the density in the storage tank increases. Profiles of momentum, m and flow rate q in the vertical direction are shown in Figures 9 and 10, respectively in order to explain it. As the Froude number increases, the momentum in the vertical direction becomes large and the flow rate increases. Increasing the Froude number means that the buoyancy effect to inertia force becomes large. Therefore, when the filling rate is down in the light fluid filling, the buoyancy effect on the plume inertia force becomes large and the entrainment volume to promote mixing between the carge and the inventory increases. We found that a lower filling rate causes LNG stratification to occur less. Figure 6. Mixing mechanism in light fluid filling Figure 7. Mixing mechanism in heavy fluid filling 25 2.0 Fr=1.0 20 10.0 15 10 Hi=8 [m] Light fluid filling 5 0 460 480 ρ [kg/m 3] Figure 8. The effect of the Froude number on the density profile at the end of the filling operation A.2–8 1 1 10.0 2.0 Fr=1.0 z/Hf z/Hf 10.0 2.0 0.5 0.5 Fr=1.0 Light fluid filling Hi=8 [m] Hi=8 [m] Light fluid filling 0 10 m 0 20 10 20 30 40 50 q Figure 9. The effect of the Froude number on the vertical momentum profile at the end of the filling operation Figure 10. The effect of the Froude number on the vertical flowrate profile at the end of the filling operation 6. STANDARD OF STRATIFICATION The density of the cargo, the density of the inventory and the volume of the inventory are measured before the filling operation. We investigate the possibility of stratification with the technical standards shown in Figures 11 and 12. Figure 11 is for the light fluid filling and 12 is for the heavy one. The solid lines indicate the results of this work, the broken lines show our previous standard in which the size of the tank is similar to that of our tank. The region under the line represents safety and that over the lines represents dangerous zone. In heavy fluid filling, when the reflection point exits in density profile measured, we observed that stratification can occur from experimental observation, whereas, in light fluid filling, the heavy layer under the outlet of the nozzle always exists and two-layer separation usually occurs. Depending on the density profile, the lower heavy layer is relatively fast diminishing by mixing of natural convection after the filling operation and finally vanishes. In this case, roll-over does not occur. We judged the safety and danger by the density profile in which the lower heavy layer vanished as a standard. In the light fluid filling, the safety region of this work is small when compared with the previous standard. However, in the heavy fluid filling it is large. The reason is the difference in the location of the nozzle outlet. The position of the nozzle outlet in this work is 2.54m from the bottom of the tank vertically and that of the previous standard tank is 1.53m. In light fluid filling, when the lower unmixed layer is thin, Stratification occurs less because the lower unmixed layer can vanish more easily due to natural convection after the filling operation. Density profiles of this work and the previous one at the end of the light fluid filling are shown in Figure 13. The unmixed layer under the outlet of the nozzle in the previous standard is lower than that of this work. Therefore in this work, the dangerous region increased slightly due to the thicker unmixed layer. In the A.2–9 heavy fluid filling, the higher position of the nozzle outlet, the easier it is for the plume to reach the liquid surface and stratification occurs less. As a result, the dangerous region decreases. As described above, in the light filling, although a lower rate easily creates two different fluid mixes, the safety region increases. In the heavy fluid filling, a lower rate causes the plume to reach the liquid surface less and safety region decreases. 20 This study Previous standard 2/3 rate Stratification 10 Full rate Half rate mixed 0 Light fluid filling 2 4 (ρ c – ρ a) 6 / 8 ρ a [%] Figure 11. Stratification standard for light fluid filling 20 :This study :Previous standard :Mixed case(Previous data) :Stratification(Previous data) Stratification 10 2/3 rate Full rate Half rate mixed 0 Heavy fluid filling 2 4 (ρ c – ρ a) / 6 ρ a [%] 8 Figure 12. Stratification standard for heavy fluid filling A.2–10 :Tohogas tank :tank of previous standard 20 ρ a =473 [kg/m 3] 3 ρ c =453 [kg/m ] Hi =8 [m] 3 Q =2,788 [m /h] 10 Light fluid filling nozzle outlet 0 450 460 ρ 470 3 [kg/m ] 480 Figure 13. Difference of density profiles at the end of light fluid filling between this work and the previous standard 7. CONCLUSION We investigated about LNG stratification in our tank experimentally and numerically. By using replica tank of our tank whose its capacity is 70,000m3, at the Chita LNG terminal in Aichi Prefecture, transient density profiles were measured and mixing phenomenon was observed with a LASER visualization technique. Consequently, a new technical standard for preventing LNG stratification was obtained for our tank. In addition, the mixing mechanism during filling operation and the effect of a lower rate on stratification during light fluid filling were revealed. REFERENCES CITED 1. Smith K. A. et al., Advances in Cryogenic Engineering, 20, (1975), 124. 2. Kusakari K. et al., Ishikawajima Harima Giho, 22, 5, (1981), 1 3. Germeles A. E., J. Fluid Mech., 71, 3, (1975), 601. A.2–11