Numerical simulation and experimental study on ventilation system for powerhouses of deep underground hydropower stations

Applied Thermal Engineering 105 (2016) 151–158

Contents lists available at ScienceDirect

Applied Thermal Engineering journal homepage: www.elsevier.com/locate/apthermeng

Research Paper

Numerical simulation and experimental study on ventilation system for powerhouses of deep underground hydropower stations Yicai Liu, Shouchuan Wang, Yangbing Deng, Weiwu Ma ⇑, Ying Ma School of Energy Science and Engineering, Central South University, Changsha 410083, China

h i g h l i g h t s
A reformed RNG k–e model was proposed and applied in the simulation study.
The arch crest air supply system was taken as an independent research object.
We verify the reliability of simulation by a comparison with experimental results.
Some feasible optimization advices were given to realize evenness of cooling effect.

a r t i c l e

i n f o

Article history: Received 9 August 2015 Revised 16 March 2016 Accepted 17 May 2016 Available online 27 May 2016 Keywords: Underground hydropower station Ventilation systems Simulations Experimental study Powerhouse Arch crest

a b s t r a c t Maintaining a steady and proper indoor thermal and humidity environment in deep underground powerhouses has a significant effect on human safety and the steady operation of power-generation systems. A quick and accurate numerical method to predict and evaluate the ventilation system is seriously needed. After reviewing and summarizing the current researches, as well as research on the heat or cold source conditions of underground hydropower stations, a reformed RNG k–e model is proposed, and then applied to simulation study conducted on the ventilation system of powerhouses in Xiluodu. The actual temperature and velocity of ventilated air in different areas were also measured and compared with the simulation results, some notable differences were further discussed to improve the model. On the basis of simulation results, some feasible optimization advices for the ventilation system were given. In conclusion, the accuracy and reliability of CFD for prediction and evaluation of ventilation system in big space deep-underground is verified. Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction Deep underground hydropower stations consist of main buildings or facilities, such as water-derivation tunnels, surge shafts, pressure conduit, main powerhouse, and tailrace tunnel, all of which are located deep underground. A generator floor of underground power stations has tall building envelopes called powerhouse, with a huge cross-sectional area and up to thousands of square meters in floor area, the powerhouse is considered a ‘‘large space” in the ventilation system design. Given that the walls of the buildings mainly use native rock as their support structure, the heat or mass transfer from rocks and groundwater seepage can affect the characteristics of supplied air like temperature and humidity. Furthermore, the meters and instruments demand a steady atmospheric working condition. The heat and mass transfer

⇑ Corresponding author. E-mail address: [email protected] (W. Ma). http://dx.doi.org/10.1016/j.applthermaleng.2016.05.101 1359-4311/Ó 2016 Elsevier Ltd. All rights reserved.

during ventilation must therefore be studied, given their importance in the adjustment of ventilation systems, steady operation, and the health of workers. In most cases, arch crest system and outlets in the surrounding walls are adopted as air supply and exhaust systems for ventilation system design of powerhouses, respectively. After cooling process, the cold air is mechanically ventilated into the powerhouses through the air distributors in the arch crest, and finally returned through the exhaust systems. In which, the backflow phenomenon tends to occur in some areas and results in dead zones of temperature and humidity control.

2. Brief review of current researches In recent years, whether or not the even distribution of ventilated air can be ensured is an important standard for valuating ventilation systems in large spaces. Such standard has gradually led to the formation of a valuation system that consists

152

Y. Liu et al. / Applied Thermal Engineering 105 (2016) 151–158

Nomenclature u u Z h m Nu p Pr Re t X, Y, Z

vectorial velocity, m/s velocity, m/s compressibility factor of gas convective heat transfer coefficient, W/m2 K mass, kg Nusselt number pressure, Pa Prandtl number Reynolds number temperature, °C rectangular coordinate, m

of professional indexes, such as age of air, effective draft temperature, and air diffusion performance index [1]. In addition, the amount of fresh air calculated from the total load can hardly meet the actual needs because of the complex conditions of cold and heat sources in underground spaces, whereas setting excessive allowance in the heat or cold load results in a waste of energy. In recent years, the reasonable design of ventilation systems in underground hydropower stations based on large space ventilation technology has been a frontier study. A reasonable design should generally consider factors, such as geological conditions in underground buildings, operation conditions, locale’s conditions of heat and cold sources (i.e., rock heat flow and groundwater), and efficiency of energy consumption [2]. At present, the major method of studying air distribution in underground hydropower stations mainly includes experimentation, physical model experiments, mathematical simulation, and some combinations of the three. Before the breakthrough of CFD technology based on computer hardware, using trace elements or gas as a tool of determining the age of air was an important method of detecting air distribution in large spaces [3]. Thus far, the element tracing method remains an important method in the experimental study of the ventilation of underground spaces. Widiatmojo et al. used tracer gas SF6 to detect dead ends in the ventilation system of the Kushiro underground coal mine and conducted numerical simulation to predict air distribution [4]. The paper conducted good detection work to examine the dispersion characteristic of air, the method has great reference value in air safety, but due to limitation of trace gas method, other characteristics of the ventilated air cannot be offered. The reduced-scale model is a convenient way to study the roof supply system for underground station powerhouses. The theories are firstly developed in 1996 by Fu [5], Angui Li et al. conducted a research project on a 1:20 reduced-scale model to investigate the air distribution in the generatrix floor of the Hohhot underground hydropower station [6]. A series of tests was performed under 48 various operating conditions. The air distribution was evaluated by calculating dimensionless temperature and dimensionless velocity. Given that the effective gravity is equal to the difference between gravity and buoyancy force, the Froude number was replaced with the Archimedes number as a similar dimensionless parameter in this study. This study has taken the buoyancy into consideration. But in the model, the actual distribution of heat source conditions and heat flux from the wall and was neglected. In recent years, the technology of using CFD as a tool for airflow prediction in large spaces has been greatly developed. Liu proposed a prediction model about heat and moisture environment in underground hydropower station, the model has taken the material of building envelope into consideration, but air diffusion in the entire space cannot be given [7]. Li Xiao-dong et al.

Greek symbols specific volume, m3/kg density, kg/m3 turbulent viscosity, Pa s

m q l

Subscripts c convection e efficient e dissipation rate tunnel exit surface of tunnel

developed a method based on the study of the multi-zonal model, which established the mass and energy equilibrium equations for each macro ‘‘large control volume.” The calculation results of the wall surface temperature and heat flux can be directly treated as the energy boundary conditions in the CFD model, which is regarded as a necessary auxiliary to CFD simulation [8]. But there’s no experiment to verify the model, and the complex calculation process made the model less operable and practical. 3. Mathematical model The establishment of appropriate mathematical models can have a significant effect on improving the accuracy of simulation and on reducing computational burdens. After blown into interior space of powerhouse through air distributors in the arch crest, the fresh air is fully developed in large spaces, which belong to free air jet. The buoyancy caused by temperature rises near the wall and the generator shields contribute to natural convection. In conclusion, the airflow in a powerhouse should belong to the combination of natural convection and forced convection. 3.1. Governing equations The governing equations of the air flow are formed by the mass continuity Eq. (1), energy Eq. (2) and Navier–Stokes Eq. (3), which also determine the temperature field [9,10]:

Dq þ qr  u ¼ 0; Dt

q

De ¼ K d r2 T  Pr  u; Dt

" # Dui @P @ 2 ui 1 @ q þ qf i þ l þ ðr  uÞ : ¼ @xi Dt @xj xj 3 @xi

ð1Þ ð2Þ

ð3Þ

3.2. Turbulence model The most popular turbulence models today are the twoequation models. Among them, the standard k–e models are widely used, given their good performance with fast convergence in simulations for isothermal flow. However, they produce large errors in non-isothermal and mixed convection conditions. The RNG k–e turbulence model has great advantages in predicting near-wall flow and low Reynolds number flows and has a good performance in terms of accuracy and efficiency [11]. In this case, we use the RNG k–e model to simulate the air ventilation in the powerhouse, and the transport equations for

153

Y. Liu et al. / Applied Thermal Engineering 105 (2016) 151–158

turbulent kinetic energy k and kinetic energy dissipation rate e are formulated as follows:

@ @ @ ðquj kÞ ¼ ðqkÞ þ @t @xj @xj



le

 @k þ Gk þ Gb  qe; @xj



ð4Þ



@ @ @ @e e e2 þ C e1 Gk  C e2 q þ R ðqeÞ þ ðquj eÞ ¼ ae le @t @xj @xj @xj k k

ð5Þ

qk le

ð6Þ

Ri ¼ Gk =Gb :

ð7Þ

2

Re ¼

where le ¼ ll þ lt , le means the efficient turbulent viscosity, ll means the coefficient of kinematic viscosity, lt means the coefficient of turbulent viscosity. Gk means the turbulence kinetic energy produced by the average velocity gradient, Gb means the turbulent kinetic energy generated by the buoyancy, and R is an additional term, which can be seen as an indicator of the drawing ratio. The coefficients are Ce1 = 1.42 and Ce2 = 1.68. During the assessment of the flow states in the powerhouse, the characteristics of air near arch crest inlets, fire dampers, ventilation pipes, and exhaust outlets in the wall are always the main statics employed in calculating the Reynolds number. The result shows that, the values of the Reynolds Number mostly exceed to 105 in the core areas, and as low as 30 in near wall areas. According to the RNG theory, during the elimination of scale, a closed equation set (8) is adopted:

 2  qk t^ ^ d pffiffiffiffiffiffi ¼ 1:72 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi dt el t^3  1 þ C t

t^ ¼

le ll

ð8Þ

Equation set (8) reflects how Reynolds Number changes with the turbulent viscosity. It can be seen that the turbulent viscosity has a obvious decisive effect on the Reynolds number or even the state of turbulent flow. When it is restricted to high Reynolds number, the solution of the set is present as:

qk2 qk2 ¼ 0:0845 e e

ð9Þ

In the most CFD softwares, Eq. (9) has been taken to describe the turbulent viscosity for standard RNG k–epsilon model. To simplify the calculation, this formation has been reserved in this paper. The fresh air was blown into through the arch crest, as the powerhouses always have a large scale cross section, the natural convection may weaken the turbulent transportation, especially on the vertical direction, the phenomenon of thermal stratification is more likely to happen (especially when the air condition system runs in heating model), which make the flow deviate from the ‘isotropic turbulence’ principle. To modify the normal RNG k–e model, the effect of bouncy and thermal stratification and low Reynolds Number flow were taken into account to formulate a fixed function for efficient turbulent viscosity and energy Prandtl number. The Ri can be seen as a criterion for intensity of turbulent. When Ri P 0, the turbulent transportation is relatively weak, to include the non-isotropic kinetic diffusion, the Prandtl number of energy can be presented as:

 Pr ¼

0:85 þ 0:4ðRi  1Þ; Ri P 1 0:85;

Ri  1

qk2 qk2 ¼ 0:0845f 1 f 2 e e

" f 1 ¼ exp

0:30

#

ð0:20 þ Re=65:25Þ2:5 þ 0:06

expð1:00=ReÞ

8 1expð30RiÞ > ; Ri P 0 < ð1:37  0:37Gb =e þ 1:6Gk =eÞ f 2 ¼ 1:0; 2 6 Ri  0 > : 1:283ð1  0:239 Ri Sk =eÞ; Ri  2

ð11Þ

ð12Þ

where function f1 is used to modify the turbulent viscosity when the Re is lower than 200. Function f2 has taken the coupling effect of shear stress and thermal stratification on the Reynolds stress and heat flux in account [12,13]. 3.3. Heat transfer between rocks and airflow Given that the powerhouse is deeply buried, the main supporting structure of the buildings is made from native rocks, and the heat convection between rock and airflow cannot be neglected in ventilation. The surface heat transfer coefficient is effectively determined with the velocity of airflow, and an experimental equation has been presented by Zhang et al. as Eq. (10) [14]:

hc ¼ 0:08u2 þ 1:8363u þ 10:573:

ð13Þ

4. Numerical simulation 4.1. Research object

C t  100

le ¼ ll þ lt ¼ C 0l

le ¼ ll þ lt ¼ C 0l

To verify the reliability of the model, we applied it to a simulation study on Xiluodu hydropower station. As one of the largest underground hydropower stations in the world, Xiluodu hydropower station on Jinsha River has the largest underground caverns, in which the powerhouse is buried exceeded 400 m deep. The situation of heat (cold) sources is complex and highly representative in the similar large underground engineering projects. What is more, like most projects, Xiluodu adopted the arch crest supply system (roof air supply system) and sidewall air return system for mechanical ventilation. Thus, the research has great reference value for ventilation system design in other underground macro-engineering projects. The focus of this study is the right bank powerhouse with generator unit rankings 10F–18F, which is shown in Fig. 1. The

ð10Þ

And to better predict the low Reynolds number flow, the efficient turbulent viscosity was presented as:

Fig. 1. Graph of the whole generator floor.

154

Y. Liu et al. / Applied Thermal Engineering 105 (2016) 151–158

plant is 364.5 m long along the positive Z axis direction, which includes the following sections in turn: secondary erecting bay, generator unit rankings 10F–18F, and major erecting bay. The building structure (i.e., cross sections along the XOY plane) can be seen in Fig. 2, which shows that the arch crest, arch crest inlets and outlets, and main building of powerhouse are all positioned symmetrically. The stairwells down to the generatrix floor are near the right side wall. After being cooled by the native reservoir water and mechanical refrigeration, the cold air is mechanically ventilated into the powerhouse through the arch crest supply system, in which the cooling air is delivered along the Z axis in the positive direction through two pipes positioned on the right and left sides (i.e., the right side of the positive Z axis is called the upstream side, whereas the left side is called the downstream side). In accordance with most project practices, the design of the ventilation was conducted on the basis of the heat balance principle. The heat generation of different areas in Xiluodu is presented in Table 1. Approximately 30% of the heat generated on the generator floor diffuses through the water system, and the rest diffuses through ventilation and air conditioning system. After being cooled by the native reservoir water and two-stage mechanical refrigeration, the temperature of the ventilated air is approximately 18 °C. The result shows that the ventilation volume for the powerhouse is 800,000 m3/h when all the nine generators are operating at full load. The temperature in the volute layer is approximately 25.8 °C then, which can fully satisfy the conditions for safe operation.

temperature areas. The meshes of the heat source, such as the walls and surfaces of the generators, are subdivided into smaller sizes to ensure the precision of the simulation prediction for the boundary layer. In accordance to Ref. [8], the 3D structures not exceeding 500 mm are not meshed to lighten the computation load and save calculation resources in this study. The arch crest and supply pipes in it are seen as whole ‘‘large control volume,” and the characteristics of air obtained from the simulation are taken as the boundary conditions for the ventilation inlets of the flow field in the powerhouse. Some simplifications for the simulation model are adopted as follows: (1) The whole near-wall flow field adopts the standard wall surface function. (2) The variation of the outside atmospheric pressure is ignored. (3) The average value of the heat dissipation rate of the generator shields are taken (i.e., the rate usually varies with the height from the ground elevation changes, but the differences can be neglected). (4) The ventilation system operates at a steady condition. (5) The variation of the specific heat capacity of air is ignored. The numerical calculation is conducted in a normal atmospheric pressure environment via the PISO (Pressure Implicit with Splitting of Operators) algorithm and the improved single-precision model of the turbulence model. The energy equations are active, and the influence of gravity is considered to improve accuracy [15–17].

4.2. Settings and simplifications

As the pipes in arch crest of the powerhouse are large relatively, and the boundary layer is comparably thin. Thus, the entry speed value is set as the average speed, that is, 3.2 m/s, which is calculated according to the volume speed of supplied air. The arch crest supply system can be seen as a whole flow field with 86 outlets (i.e., air distributors for the powerhouse), when the cooled air is piped through the arch crest system, heat exchange occurs between the air and atmospheric environment outside the pipes, the latter is surrounded by a native rock structure, which has considerable heat storage capacity and negligible temperature variation. According to some relative Refs. [16,17], the boundary condition for rock should be set as a ‘‘wall” with an effect thickness and distal fixed temperature, and the heat flux from the rock ranges between 0.83 W/m2 and 7.74 W/m2 (at cooling working conditions) according to some comparative simulation works of underground hydropower stations. The results of former simulation in arch crest supply system are adopted as the boundary conditions of the inlets for the interior flow field of the powerhouse. Given the steady operation of the generators, the boundary condition of the generator shields is set as the ‘‘wall” with a steady heat flux, and the value is set at approximately 500 W/m2 according to the Report on Ventilation and Air Conditioning Systems of Underground Powerhouse of Hydropower Station on Jinsha River. Based on the continuity of flow, the characteristics of the air distribution at different areas were all subjected under statistical analysis. The error data were also eliminated. Finally, the trends of the air temperature and velocity were plotted.

The interior space of the powerhouse is a complex threedimensional (3D) body, containing velocity variations and uneven

Fig. 2. Z-axis view of the powerhouse. Table 1 Heat generation statics of Xiluodu. No.

Area

Heat generation (109 J/h)

1 2 3 4 5 6 7 8 9 10 11

Generator floor Electrical dissection Turbine floor Spiral casing floor Tail pipe layer corridor Generatrix floor Main transformed cavern GIS floor Cable shaft Secondary plant Total

6.23 2.12 0.80 0.62 0.22 5.80 1.92 0.60 3.16 1.05 24.9

4.3. Boundary conditions

4.4. Model adjustment To examine the accuracy of the simulation, the characteristics of air in different areas were measured with tools, such as Fluke 971 hygrothermograph, infrared thermometer, and anemograph. As the data of inlets matches the measured values so well, the data of outlets can greatly reflect the accuracy of the simulation for heat sources and cooling effect of air condition. The measured values and calculation results of the characteristics of outlets in the

Y. Liu et al. / Applied Thermal Engineering 105 (2016) 151–158

vertical wall are illustrated in Figs. 3 and 4, the relevant data at the stairwell down to generatrix floor is also given. Though the measured values and the simulation ones were basically the same and had similar trends, some notable differences there exists. In the initial section corresponding to the secondary erecting bay, Fig. 3 shows that the measured velocity at the outlets in the wall were considerably higher than those in the simulation, while it is lower the at the stairwells. Fig. 4 shows that the measured temperatures were obviously lower than those in the simulation of the section, especially at the stairwell. As the stairwell is comparably large in scale, a little difference in the velocity and temperature can force great influence on heat balance. The flow at the stairwell belongs to the nature ventilation, compare to the mechanical ventilation, it has lower speed and less directionality. In fact, the cold flow coming from the generatrix floor makes it even hard to describe the temperature and velocity characteristics on the interfaces. Thus, the measured data is adapted to define the boundary conditions. By contrast, in the area from the outlet rankings 23 to 27 corresponding to section near the traffic tunnel, the velocity measured was obviously smaller than that in the simulation, in other word, the outlets in the wall exhausted more air in the simulation than they actually did. The traffic tunnel brings a considerable volume of airflow intake, and the boundary condition of the traffic tunnel was originally set as the ‘‘inlet” with an average horizontal velocity. Given the temperature values match well as shown in Fig. 4, which means the heat balance was well satisfied in simulation, thus, the velocity might have been overvalued to bring a load for air exhaustion. In fact, the velocity varies a lot on the exit surface of the traffic tunnel, as there are some pressurized ventilators positioned on the top of the tunnel, the velocity becomes even bigger as the height increases. According to the measured values, a simple fitting formula for the velocity distribution was proposed, in which the tunnel was simplified as a half cylinder with diameter of 11 m.

utunnel ¼ 0:12 cosðr=ro Þ þ 0:15 lnðh=r o  1Þ

155

Fig. 4. Temperature of the stairwells and outlets in the vertical wall.

ð14Þ

where the utunnel refers to the velocity of coming flow, r refers to the distance from the center, the h refers to he height above the ground, the ro refers to the half-diameter of the tunnel. The formula was adopted in the simulation as a UDF(User-Defined Functions) to describe the boundary conditions on the exit surface of the tunnel. The results of the modified

Fig. 5. Velocity of the stairwells and outlets in the vertical wall (modified simulation).

simulation are given in Figs. 5 and 6. In which the simulation shows great consistency with the actual data.

5. Ventilation system study and discussion 5.1. Arch crest

Fig. 3. Velocity of the stairwells and outlets in the vertical wall.

The velocity and temperature graphs of the ventilated air at different distributors can be seen in Figs. 7 and 8. Among them, the distributor rankings #1–#6, #7–#36, and #36–#43 corresponded to the secondary erecting bay, generator unit rankings 10F–18F, and major erecting bay, respectively. Fig. 7 shows that the velocity plots of supplied airflow in the upstream and downstream pipes have good consistency. When the air is transported, as a result of the pipeline resistance, the flow velocity at the distributors drops from 4.5 m/s (in the secondary

156

Y. Liu et al. / Applied Thermal Engineering 105 (2016) 151–158

Fig. 6. Temperature of the stairwells and outlets in the vertical wall (modified simulation).

Fig. 8. Temperature trend of the distributors in the arch crest.

Fig. 9. Top view of the secondary erecting bay.

Fig. 7. Velocity trend of the distributors in the arch crest.

erecting bay) to 2.3 m/s (in the major erecting bay). Fig. 8 shows that, as the velocity drops, a remarkable temperature rise occurs from 18.2 °C (in the secondary erecting bay) to 23.8 °C (in the major erecting bay), which represents a difference of approximately 5.6 °C. Thus, the velocity drops, and the temperature of air increases because of the pipe resistance and cold loss in the transportation. Finally, a large difference exists in the air temperature and velocity between the two heads of the arch crest ventilation system, thereby leading to a large supplied cold capacity difference between the secondary and major erecting bays. 5.2. Interior space of the powerhouse As the workers and relative facilities are all positioned on the generator floor, the diffusion characters of velocities and tempera-

tures in the interior space of the powerhouse are worth studying in detail. The results of former simulation are adopted as the boundary conditions of the inlets for the interior flow field of the powerhouse. The secondary erecting bay contains a generator unit, which can directly reflect the cooling effect for the generator unit. Thus, the characteristics of air distribution in the secondary erecting bay are further discussed in this study, whereas the rest of the powerhouse is not analyzed in detail because of lack of space. The top view of secondary erecting bay is shown in Fig. 9. The secondary erecting bay has a 49.5 m-long roof consisting of 11 beams. The arch crest system over this section of the generator floor has 10 air distributors and 6 exhaust outlets. The vertical wall has 5 exhaust outlets. The temperature outline graph of the airflow at #1 distributor (Z = 11.25 m, XOY plane) is shown in Fig. 10. This graph illustrates that the temperature distribution is not symmetrical, and the air near the upstream wall (right side) is significantly cooler than that near the downstream wall (left side). The right-side space has a considerably large cold airflow field. The velocity of the coming air at the two distributors does not vary

Y. Liu et al. / Applied Thermal Engineering 105 (2016) 151–158

considerably. However, the only difference is that the stairwells are near the right side, which forms a suction effect. Thus, the flow resistance of the right side is considerably small. The velocity graph of the airflow at generator 10F (Z = 31.5 m, XOY plane) is shown in Fig. 11. This graph illustrates that the maximum speed of the airflow occurs in the middle area, which reaches 3.68 m/s. This observation shows that the flow fields formed by the air delivered from the two sides of the distributors enhance each other. The velocity decreases further from the middle area, the minimum speed occurs in areas near the floor, wall, and shield of the generator, and the temperature is relatively high. Cycle flow areas exist in the both corners formed by the arch crest and vertical walls, which may lead into a dead zone of temperature adjustment. Fig. 12 illustrates the temperature distribution on the center sections (X–Y and Y–Z planes) of generator 10F, and all the sampling points are at a height of 2 m. The temperature drops signifi-

157

cantly as the distance from the center increases. At approximately 700 mm from the shield, the air temperature becomes relatively steady, reaching approximately 21.2 °C. In general, the temperature on the X–Y plane is lower than that on the Y–Z plane, indicate that the Y–Z direction has a better cooling effect. Cause the cold air vertically emerges from the distributors at nearly the same speed, thus, the differences in temperature are mainly caused by horizontal reasons, such as the suction effect formed by the air outlets in the wall, as the air is exhausted horizontally, thereby enhancing the intensity of the cold airflow in the X–Y plane direction. 5.3. The whole generator floor The temperature of the generator floor was measured following the line connecting the centers of the top faces of all the generators (Y = 2 m), and the measured values are all marked as shown in Fig. 13. Fig. 13 shows that the measured values of the temperature

Fig. 10. Temperature outline at distributor #1 (Z = 11.25 m).

Fig. 12. Temperature distribution at the center section of generator 10F (Z = 31.5 m, Y = 2 m).

Fig. 11. Velocity contour at generator 10F (Z = 31.5 m).

Fig. 13. Temperature of the line through the generator centers.

158

Y. Liu et al. / Applied Thermal Engineering 105 (2016) 151–158

and simulation results have a great consistency, particularly in the middle section of the generator floor. A slight difference occured in the section between Generators 10F and 11F, which showed that the measured temperature was relatively low but did not vary considerably. The result showed that CFD could be reliable in predicting the temperature distribution of large underground spaces.

6. Conclusion The research shows that the CFD software can be applied to predict ventilation in underground hydropower stations. Moreover, the modified RNG k–e two-equation model has an advantage in striking a balance between efficiency and accuracy during the simulation of airflow in large spaces. The accuracy can be greatly enhanced if the boundary condition are well set, especially the ones of the outlets and inlets with lager scale, like exit surfaces of stairwells and tunnels. In the Xiluodu, the difference in the temperature and velocity at the two heads of the arch crest system exceeds to 5.6 °C and 2.2 m/s, respectively. For most mechanical pressurization ventilation which adopt roof air supply and sidewall air return, the uniformity of the air characteristics in the arch crest ventilation system is an important factor for the evenness of the conditioning effect. To ensure this effect, the thermal insulation of the pipes in the arch crest should be improved by preventing cold loss. Pressurization devices can also be positioned to keep the airflow steady in the transportation pipes. In the simulation and experimental analysis on the secondary bay, the temperature and velocity distribution was given in detail, which shows the model has great performance in dead zone detection and peak temperature prediction. The differences in cooling effect for the generator at X–Y plane and Y–Z plane shows that a reasonable arrangement of the positions of the outlets and distributors can greatly enhance ventilation. Acknowledgements This study is supported by the National Natural Science Foundation of P.R. China (Project No: 51276201) and Research and Development Projects of Hunan Province of P.R. China (2015JC3047).

References [1] W.K. Chow, L.T. Wong, W.Y. Fung, Field measurement of the air flow characteristics of big mechanically ventilated spaces, Build. Environ. 31 (1996) 541–550. [2] Y.B. Deng, W.G. Zhao, P. Chen, F. Dai, The research on air conditioning technology of Xiluodu hydropower station, Technol. Hydropower Station 05 (2014). 83–85+89. [Chinese Version]. [3] W.K. Chowa, W.Y. Fungb, L.T. Wonga, Preliminary studies on a new method for assessing ventilation in large spaces, Build. Environ. 37 (2002) 145–152. [4] A. Widiatmojo, K. Sasakia, Y. Sugaia, Y. Suzukib, H. Tanakab, K. Uchidab, H. Matsumoto, Assessment of air dispersion characteristic in underground mine ventilation: field measurement and numerical evaluation, Process Saf. Environ. Prot. 93 (2015) 173–183. [5] X.Z. Fu, Research of ceiling air supply in underground water powerhouses, Heat. Ventilating Air Cond. 1 (1996) 59–62 [Chinese Version]. [6] A.G. Li, Z.J. Liu, J.F. Zhang, J.M. Wang, Reduced-scale model study of ventilation for large space of generatrix floor in HOHHOT underground hydropower station, Energy Build. 43 (2011) 1003–1101. [7] X.C. Liu, Formation mechanism of heat and moisture environment in underground hydropower station and its energy-saving control strategy [Doctoral Thesis], College of Urban Construction and Environment Engineering of Chongqing University, Chongqing, China, 2014 [Chinese Version]. [8] X.D. Li, J. Gao, P.S. Gao, Simulation of temperature distribution in a large space with CFD based on a multi-zone model, J. Harbin Inst. Technol. 37 (2005) 683– 686 [Chinese Version]. [9] I. Diego, S. Torno, J. Torao, M. Menéndez, A practical use of CFD for ventilation of underground works, Tunneling Underground Space Technol. 26 (2011) 189– 200. [10] A. de Jesus Freire, J.L. Coelho Alexandre, V.B. Silva, N.D. Couto, A. Rouboa, Compact buried pipes system analysis for indoor air conditioning, Appl. Therm. Eng. 51 (2013) 1124–1134. [11] W. Feng, M.N. Wang, S. Heb, Y.Y. Deng, Computational study of effects of traffic force on the ventilation in highway curved tunnels, Tunneling Underground Space Technol. 26 (2011) 481–489. [12] M.S. Hossain, W. Rodi, A turbulence model for buoyant flows and its application to vertical buoyant jets and plumes in turbulent buoyant jets and plumes, HMT-Series, vol. 6, Pergamon Press, Oxford, 1982. [13] J. Gao, X. Zhang, J.L. Zhao, The reformed RNG model for convection of indoor air, Build. Sci. 10 (2010) 324–331 [Chinese Version]. [14] L. Zhang, Y.M. Chen, G.F. Tang, C.J. Ren, G.F. Zhang, The simulation research and statics analysis on heat exchange of wall, Chinese National Academic Conference on HVAC & Refrigeration of 2000, 2000, p. 4. [15] C. Huang, Y. Song, X. Luo, Application of the gebhart-block model for predicting vertical temperature distribution in a large space building with natural ventilation, in: Proceedings of the Sixth International Conference for Enhanced Building Operations. Shenzhen, 2006, pp. 6–9. [16] P. Hollmuller, B. Lachal, Cooling and preheating with buried pipe systems: monitoring, simulation and economic aspects, Energy Build. 33 (2001) 509– 518. [17] A. Pfitsch, M. Bruene, B. Steiling, M. Killing-Heinze, B. Agnew, M. Irving, M. Irving, J. Lockhart, Air flow measurements in the underground section of a UK light rail system, Appl. Therm. Eng. 32 (2012) 22–30.