Construction of System for Evaluation of Environmental ... · Another is the Weather Information System that supplies ... Construction of System for Evaluation of Environmental ...

P-4b-226

1

1. PurposeOn emergency case at nuclear power station, the prediction of environmental activity concentration

and dose rate is very important for decision of countermeasure of radiation protection. These predictions are ableto be figured by simulation models. TEPCO has own system because we need to estimate the impact of theradioactivity inside the nuclear power site.

TEPCO has been used the system with a two-dimensional segmented plume model since 1988. Thissystem becomes aged, so we reconstruct the new system with three-dimensional simulation models. We canobtain from this system environmental dose information in real-time after an emission of radioactive materials.The new system includes a prognostic model to predict local wind field around the site. We would have helpfulinformation to decide when we should release radioactive gas from PCV(Primary Containment Vessel) ventbefore emission.

2. Outline of the systemTEPCO plans to implement the new emergency response system DIANA(Dose Information Analysis

at Nuclear Accident) in three nuclear power stations and the head office. Prototype of the new system wasdeveloped in FY1997 and it is working at the head office. The new operational systems are implemented in twonuclear power stations of Fukushima Daiichi Nuclear Power Station(1F) and Fukushima Daini Nuclear PowerStation(2F) in FY1999. The final system configuration of our plan is shown in Figure 2-1. Two data servers arecooperative with the new system. One is the Data Processing System that supplies on-site data every ten minutes.Another is the Weather Information System that supplies GPV data twice a day. The new operational system atthe head office is to be connected to each system at nuclear power station. The results of calculation at thesenuclear power stations can be reproduced at the head office.

DIANA is working 24 hours every day except while the system is shutdown for maintenance. Twofunctions in normal operation are shown in Table 2-1. When the system receives onsite data, the observed valueat stack monitor is compared with preset threshold to determine if simulation calculation should startautomatically. When the system receives GPV data, wind field forecast calculation starts automatically over thewide region around the site.

Due to the forecasted wind field, we can decide when we would release radioactive gas from PCVvent to the environment while radioactive materials are confined in the containment vessel.

After radioactive materials are emitted from stack into the atmosphere, we can estimate anenvironmental impact in real-time with onsite input data. Additional functions are available to estimateenvironmental impact caused by non-monitored emission from any place inside the site.

To accomplish these functions effectively, DIANA is composed by automated data acquisition sub-systems, numerical simulation models, and user interface. The data flow diagram for the major functions of real-time dose estimation and prediction is shown in Figure 2-2. Model description is in the next chapter.

The new system supports three different domains for calculation, which are 10km*10km*0.5km,25km*25km*1km, and 50km*50km*1km. Though we are mainly intend to obtain dose information in narrowregion on accident, we have the option to make wide range simulation for post accident re-calculation.

The output data items are air concentrations, air absorbed dose rate, surface deposition, thyroid dose,internal effective dose equivalent and external dose equivalent caused by the radioactive rare gas, radioactiveiodine, and radioactive particular nuclides emitted into the atmosphere at the nuclear accidents.

DIANA can be used in exercises for emergency as well as in accident. We usually use an academicdata prepared for the exercises. As to meteorological data, real online data at the exercise can be used to givestaffs opportunity how to deal effectively with a tense situation of undefined consequence.Figure 2-1 DIANA system configuration and related systems

Construction of System for Evaluation ofEnvironmental Emergency Dose

N.Suzuki1, K.Sugai1, K.Hayashi1,M.Suzuki2, H.Suwa2, Y.Kato2,

F.H.Liu3, and S.Kodama3

1Tokyo Electric Power Company, Chiyoda-ku, Tokyo, 100-0011, Japan2Japan NUS Company, Minato-ku Tokyo 108-0022 Japan

3CRC Research Institute, Inc., Koto-ku Tokyo 136-8581 Japan

P-4b-226

2

Table 2-1 Functions that DIANA performs automatically in normal operation

Figure 2-2. Data flow diagram for real-time dose estimation and predictive dose estimation

３．ModelsThe simulation models in the DIANA system are three-dimensional diagnostic wind field model,

prognostic wind field model, PIC(Particle-In-Cell) diffusion model and cell dose Model.

Nuclear Observed on-site data Computer EWS EWSPower Meteorological station data TCP/IP DIANA TCP/IP WeatherStation Stack monitor processing (on-site system (GPV) Information 1F, 2F, KK(*) Monitoring post system data) System

TCP/IP (on-site data)

TEPCO (*) EWS EWSHead Office 1F:Fukushima Daiichi Nuclear Power Station DIANA TCP/IP Weather

2F:Fukushima Daini Nuclear Power Station system (GPV) InformationKK:Kashiwazaki Kariwa Nuclear Power Station System

Japan GPV : Grid Point Value GPVMeteorological RSM : Regional Spectral Model (RSM : upper)Business (RSM : surface)Support Center

Input Model OutputGPV data : twice a day Prognostic wind field model Forecasted hourly wind field 51hours from 9:00 Wind field forecast for 51hours Domain:150km x 150km 51hours from 21:00 twice a day upon GPV input Mesh : 3km x 3kmOnsite stack monitor count data : Comparison with threshold Signal to activate DIANA in accident mode every 10 minutes Survey of rare gas count Automatic detection of rare gas emission

to campare with preset threshold Automatic start of dose calculation

P-4b-226

3

3.1 Diagnostic wind field modelA diagnostic wind field model was developed based on MEDIC/MATHEW(1). MEDIC/MATHEW

was originally developed by LLNL and applied in ARAC(2) since 1976, which gives services for US DOEfacilities and DOD facilities. The vertical structure of atmosphere is modeled as three layers, namely surfacelayer, boundary layer and free atmosphere. Observed wind speed data and wind direction data are interpolatedand extrapolated in each layer to get first guess wind fields U0 (U0,V0,W0) in the calculation domain. Thenminimal modification is done by variation method(3) to get mass-consistent wind field U(U,V,W). U iscalculated from U0 by following equation.

xUU

∂∂+= λ

α 21

0 21

yVV

∂∂+= λ

α 21

0 21

zWW

∂∂+= λ

α 22

0 21

∂

∂+∂∂+

∂∂−=

∂∂

+

∂∂+

∂∂

zW

yV

xU

zyx0002

12

2

22

21

2

2

2

2

2αλααλλ

with the boundary conditions, λ=0 at the free boundary surfaces, and

　 0

212

2

22

21

2

2

2

2

2 Unz

ny

nx

n zyx ⋅−=∂∂⋅⋅

+

∂∂⋅+

∂∂⋅ αλ

ααλλ at terrain surface,

where α1 /α2 is parameter of modification as vertical wind speed over horizontal wind speed, and n(nx, ny, nz) is normal unit vector to the terrain surface.

Main advantage of this model is that the model requires only observed data of wind velocity andatmospheric stability available at the sites. In addition it is very fast to calculate the wind field.

3.2 Prognostic wind field modelLocal wind circulation assessment and prediction system (LOCALS) is used in calculating the wind

fields for evaluating dose information. The basic equations of the model are base on the model of Kikuchi etal.(4), Kimura and Arakawa(5). The initial condition and the boundary condition of the wind model areinterpolated from GPV of Japan Meteorological Agency. The outline of GPV is shown in Table 3-1.

Table 3-1 Specification of GPV data

The followings are the governing equations in LOCALS. They are written in a terrain-following coordinate

system z* defined by GT

GT zz

zzzz−−=*

Coordinates:　Cartesian coordinates in horizontal and z* terrain following coordinates in vertical direction2）Dynamic equation: Boussinesq hydrostatic hypothesis3）Turbulence model: Mellor-Yamada closure model level 2.04）Constant flux layer: Monin-Obukhov similarity theory5）Upper boundary condition: Klemp and Durran radiation condition6）Pressure: Hydrostatic equation7）Surface temperature diurnal variation: force restore method Equations of motion:

−

Θ+Θ−=+++

*

* +

*

* *

2

zuK

zhz

xz

zzzgh

xhfhv

zhuw

yhuv

xhuu

thu

mTG

T

T

∂∂

∂∂

∂∂θ

∂π∂

∂∂

∂∂

∂∂

∂∂

Data Mesh resolution Pressure level Item Forecast time CommentsRSM surface Origin : Latitude N90゜Longitude E0゜ surface U, V, T, T-Td , 51 h Distribution from JMBSC

Latitude x longitude = 12' x 15' R, Cld 9:00JST ~ 12:00 - 15:00~ 20km x 20km 21:00JST~ 00:00 - 03:00

(time interval : 1h) (Twice a day)RSM upper Origin : Latitude N90゜Longitude E0゜ surface Ps 51h Distribution from JMBSC

Latitude x longitude = 24' x 30' 925hPa z, U, V, T, T-Td 9:00JST~ 12:00 - 15:00~ 40km x 40km 850hPa z, U, V, T, T-Td 21:00JST~ 00:00 - 03:00

700hPa z, U, V, T, T-Td, ω (Time interval : 3 h) (Twice a day)500hPa z, U, V, T, T-Td

z : Geopotential height of pressure level, U and V : Components of horizontal wind, T : Air temperature T-Td : Difference between air temperature and dew point, Ps : Air pressure at the sea level, ω: Vertical velocity, R : precipitation intensity, Cld : Cloud amount

P-4b-226

4

−

Θ+Θ−−=+++

*

* +

*

* *

2

zvK

zhz

yz

zzzgh

yhfhu

zhvw

yhvv

xhuv

thv

mTG

T

T

∂∂

∂∂

∂∂θ

∂π∂

∂∂

∂∂

∂∂

∂∂

Thermodynamic equation:

∂ θ∂

∂ θ∂

∂ θ∂

∂ θ∂

∂∂

∂θ∂

ht

hux

hvy

hwz

zh z

Kz

Th+ + + =

** * *

2

Continuity equation: ∂

∂∂∂

∂∂

hux

hvy

hwz

+ + =**

0

Hydrostatic equation: ∂ π

∂θ

zhz

g

T*=

Θ2

List of symbols and contents ｘ，ｙ， z* ：coordinate in east-west, south-north and vertical direction ｕ，ｖ， w* ：wind component in x, y, z* direction ｆ ：Coriolis parameter θ ：deviation of potential temperature Θ ：average potential temperature π ：deviation of pressure in Exner function formation ｇ ：gravity acceleration zG ：elevation of terrain zT ：top of the model ｈ ：depth of atmosphere in the model（ zT ― zG ）

Km ：coefficient of momentum diffusion Kh ：coefficient of heat diffusion

dt

dzw ** ≡ ; Km , Kh is modified by the assumption of Mellor-Yamada closure model.

3.3 Diffusion modelThe diffusion model was developed based on ADPIC(6) in ARAC. Basic equation below is derived

from diffusion equation and continuity equation for incompressive atmosphere. 0

=

∂∂

−+

∂∂

−+

∂∂

−+

zC

CKWC

zyC

CK

VCyx

CCKUC

xtC zyx

∂∂

∂∂

∂∂

∂∂

with boundary conditions of a steady concentration flux at free boundary, and a zero concentration flux at terrain surface,where C, U(U,V,W), and K(Kx,Ky,Kz) are air concentration of emitted material, wind field and diffusioncoefficient, respectively. Diffusion coefficient is derived by Taylor’s relation from empirical diffusion parametersdescribed in the Japanese Meteorology Guideline to estimate air concentration at an emergency.

This equation is solved by a PIC methodology where distribution of emitted material is representedby numbers of mathematical particles. Each particle has its own radioactivity, age, size, and other characteristics.So we can easily take into account some physical process in diffusion calculation, such as radioactive decay, drydeposition, wet deposition, and gravitational fall of particular matter.

3.4 Dose modelAn external dose model is cell dose model. In this model, an air absorbed dose rate D(x,y,0) at a

terrain surface is calculated as summation of air absorbed dose rate due to each cell in the calculation domain.An external dose is a time integration of air absorbed dose rate. Conversion factor from air absorbed dose rate toexternal dose is assumed to be 1.0 because main contribution is by gamma ray. An air absorbed dose rate D(i,j,k)due to a cell of unit radioactive concentration is derived from numerical integration of the following formulaover the cell volume, and tabulated in the system.

D(x,y,0) =∑χ(i,j,k) D(i,j,k)

whereD i j k K E a r r B r x y z dx dy dz( , , ) exp[ ] / ( ) ( ) ( ' , ' , ' ) ' ' '= ⋅ ⋅ − ⋅∫∫∫ ⋅1 4 2µ µ π µ χ

P-4b-226

5

D(x,y,0) : surface air absorbed dose rate at (x,y,0) χ(i,j,k) : mean air concentration of in a cell with relative indices (i,j,k) to (x,y,0) D(i,j,k) : surface air absorbed dose rate due to a cell (i,j,k) with unit air concentration K1: Conversion factor to absorbed dose rate E: effective gamma ray energy μa: absorption coefficient to air μ: total absorption coefficient to air r : distance from point (x‘,y’,z‘) in the cell to the point to be calculated B(μr) : buildup factor of gamma ray to the air χ(x’,y‘,z’) : unit air concentration in the cell(i,j,k) . We referred the values for K1,μa,μand B(μr) from Taki et.al.(7) . An internal dose is calculated from the time integrated surface air concentration based on the ICRP publ.30.Conversion factor from air concentration to internal dose is tabulated in the system.

4. Model evaluationModel evaluation study was performed separately for a combination of diagnostic wind field model

and PIC diffusion model and for prognostic wind field model.

4.1 Evaluation of diagnostic wind field model and PIC diffusion modelWe made an evaluation of a combination of diagnostic wind field model and PIC diffusion model by

two ways. First we compared calculated air concentrations and dose rates by the models with the analyticalconcentrations and numerically integrated air absorbed dose rates for gaussian dispersion equation under steadywind field and flat terrain. The coincidences are quite well for both the air concentrations and air absorbed doserates. We could confirm that our computer codes are written correctly.

Next we compared calculated air concentrations by the models with measured values in field tracerexperiments. We selected four cases of the field experiments for the model evaluation , which are Hanford 67U62 , Hanford 67 V7, SONGS 25 and SONGS 27. A series of Hanford 67 experiments were conducted through1967 to 1973 at the Hanford site in Washington state, USA. A series of SONGS field experiments wereperformed through 1976 to 1977 inside the San Onofre Nuclear Generation Station in California state, USA.Outline of these experiments is summarized in Table 4-1.

The scatter plot diagrams of the calculated concentration over observed concentration are shownthrough Figure 4-1 to Figure 4-4. Black dots represent the ratio of calculated concentration over measuredconcentration. In Figure 4-1 and in Figure 4-3, we also plotted the concentration ratio as white circles in thecases of flat terrain model calculation. In every case, the scatter plots diagrams showed that results are quitegood. As an exception, large underestimation of factor 50 by the model occurred at the downwind distance of12.6km in Hanford experiments. We also confirmed that if terrain is modeled as non-flat, coincidence of modelresults with field observations gets better than in flat terrain case.

Table 4-1 Outline of field experimentItem Hanford 67 U62 Hanford 67 V7 SONGS 25 SONGS 27

Date of Experiment 1968/6/7 13:00 - 13:30 1973/9/25 12:01 - 12:31 1977/1/22 10:45 - 11:45 1977/1/22 12:45 - 13:45Tracer ZnS ZnS SF6 SF6Emission time 30min Continuous emission 30min Continuous emission 1 hour Continuous emission 1 hour Continuous emissionEmission Height GL 26m GL 26m GL 2m GL 2mSampling points 15 points, GL 1.5m 15 points, GL 1.5m 15 points, GL 0.6m 15 points, GL 0.6mDownwind Distance 0.8 km - 3points 0.4 km - 3points 300m - 5 points 300m - 5 points

1.6 km - 3points 0.8 km - 3points 700m - 10points 700m - 10points 3.2 km - 3points 1.2 km - 3points 7.0 km - 3points 1.6 km - 3points 12.6 km - 3points 3.2 km - 3points

Meteorological Obs. Wind Velocity GL 2.1m, 15.2m, 30.5m, 45.7m, GL 2.1m, 15.2m, 30.5m, 45.7m, GL 10m, 40m at Bluff Tower GL 10m, 40m at Bluff Tower

61.0m, 91.4m, 122m 61.0m, 91.4m, 122m GL 10m, 40m at Inland tower GL 10m, 40m at Inland tower at the emission point at the emission point GL 30ft, 120ft at Beach tower GL 30ft, 120ft at Beach tower

GL 10m at Beach Suppl. tower GL 10m at Beach Suppl. towerGL 10m at Hill mast - 1,2,3 GL 10m at Hill mast - 1,2,3

Temperature Profile at the emission point Profile at the emission point Temperature Difference at Temperature Difference at Bluff tower, Inland tower, Bluff tower, Inland tower, & Beach tower & Beach tower

Region of Simulation 25km x 25km x 350m 25km x 25km x 350m 11.25km x 11.25km x 280m 11.25km x 11.25km x 280m( mesh size ) (500m x 500m x 17.5m) (500m x 500m x 17.5m) ( 22.5m x 22.5m x 14m) ( 22.5m x 22.5m x 14m)

P-4b-226

6

Figure 4-1 Scatter plot of concentrations Figure 4-2 Scatter plot of concentration Hanford U62 Hanford V7

Figure 4-3 Scatter plot of concentrations Figure 4-4 Scatter plot of concentration　　　　 SONGS 25 SONGS 27

4.2 Compare the results of numerical model with the solution of analysisHere, we compare the numerical model results with the analysis results in the condition of virtual bell-

shaped terrain. The situation of the calculation is shown as follows.

Grid interval: dx=dy=5000mArea: 60×60 in east-west and south-north direction, with 5000m depthInitial condition: U=10 (m/s) in all layerspotential temperature = Zg ⋅+γθ Z: height(m) γ:0.004（K/m）　 gθ =290(K)Surface: non-slip condition at surface without incoming and outgoing radiation, non Coriolis effect and nonviscosity.Boundary condition: cyclic boundary in horizontal direction; radiation condition in upper boundaryIntegration time: 10 hours from initial time.Terrain: bell-shaped mountain.

P-4b-226

7

( ) 23

21

+

=

WR

HZg ( ) ( )2

02

0 YYXXR −+−=

0=<X, Y=<300(km), 0X = 0Y =150(km), W=20(km), H=500(m)The vertical current is compared within the cross section at center line of the mountain. The analysis

results are solved and provided by Mr. S.Takahashi of Japan Meteorological Agency. As shown in Figure 4-5 thesimulation results are agree well with the analysis results.

Figure 4-5 Distribution of vertical velocity in center cross section. The upper shows analysis result, the bottom shows simulation result

4.3 Simulation caseThe wind forecasting system in DIANA covers a 150km square area, where 1F is located in the center.

We calculate an area with 50meshes in both x and y direction at 3km intervals. A mass conservation model isused to interpolate wind filed from the area above to three areas as follows. They are 1) large area with a 50kmsquare at 1000m interval both in x, y directions; 2) the middle area with 25km square at 500m intervals; and 3)10km square at 200m intervals in x, y directions. The initial condition and boundary condition of LOCALS ismade from RSM GPV of 9:00JST August 1, 1999. The 51hours forecast was done about wind, temperature andso on. Distribution of surface wind at 0:00 August 2 1999 is shown in Figure 4-6. The wind direction is southerlyon the sea and near the shore. And it is SSW at the land. The 6 hours’ results are compared with the observationdata obtained from site. Figure 4-7 shows good coincidence between calculated and observed wind at thesurface. We made a diffusion calculation and dose calculation for hypothetical rare gas emission event from astack. The distribution of surface air absorbed dose rate (shown in Figure 4-8) shows that the plume flows out toNNE due to the predicted wind direction of SSW at higher level.

P-4b-226

8

Figure 4-6 Surface distribution of surface wind at 00:00 August 2, 1999 in Fukushima Prefecture.

Figure 4-7 Wind time series of calculation(broken line) Figure 4-8 Distribution of predicted air

and observation values(solid line) at 10m level absorbed dose rate at 02:00 August 2 from 22:00 August 1 to 3:00 August 2. due to assumed emission (Simulation initial time is 9:00 August 1 )

5．Results

We evaluated the integrated models of diagnostic wind field model and PIC diffusion model by twoways. The first is the comparison of calculated values by the models with the solutions of gaussian dispersionequation under steady wind field and flat terrain. Both the air concentrations and air absorbed dose ratescalculated by the models coincides with analytical concentrations and numerically computed air absorbed doserates. The second is the comparison of calculated concentrations by the models with observed values in fieldtracer experiments. The scatter plots diagrams showed that results are quite good within about 10km ofdownwind distances. If terrain is modeled as flat, coincidence of results gets worse.

We evaluated the prognostic wind field model by comparison of model results with analyticalsolutions under bell-shaped terrain. The comparison showed that the model represents analytical solution verygood.

The prototype system was made to integrate all the models. We made calculation by the prototypesystem for 1F with the site data and GPV data on August 2, 1999. The prognostic wind field was calculated fromGPV data distributed one day before. The interpolated wind speeds and wind direction at the wind observationtower in 1F are qualitatively in good agreement with the observed wind speed and wind direction.

6．ConclusionWe developed an emergency dose estimation system based on the three-dimensional diagnostic wind

field model, prognostic wind field model, PIC diffusion model and cell dose model. The prototype system wasimplemented in the head office of TEPCO in 1997. Current operational systems in three nuclear power stations

P-4b-226

9

and the head office will be replaced with the new systems. Based on the wind field forecast by the new system,we will be able to decide when we should release pressurized gas from the PCV vent at the severe accident.After the emission occurred from the stack, this system will calculate environmental dose every ten minutesautomatically. Predictive dose can be calculated on demand. We hope to make effective countermeasure timelyto the potential radioactive disaster at a severe accident by utilizing the information from the new system.

REFERENCES

1. C. A. Sherman, A Mass-Consistent Model for Wind Fields over Complex Terrain.2. J. Appl. Meteor. , 17, 312-319 (1978)3. T. J. Sullivan, et. al., Atmospheric Release Advisory Capability : Real-Time Modeling of Airborne4. Hazardous Materials. Bull. Amer. Meteor. Soc., 74912, 2343-2361(1993)5. Y. Sasaki, An Some Basic Formalisms in Numerical Variation Analysis. Mon.6. Weather. Rev., 98, 875, (1970)7. Y. Kikuchi, S.Arakawa, F.Kimura, K.Shirasaki and Y.Nagano, Numerical study on the effects of

mountains on the land and sea breeze circulation in the Kanto District.8. J. Meteor. Soc. Japan, 59,723-738 (1981)9. F.Kimura, and S.Arakawa, A numerical experiment of the nocturnal low level jet over the Kanto Plain. J.

Meteor.Soc.Japan, 61,848-861 (1983).10. R. Lange, ADPIC – A Three Dimensional Particle-In-Cell Model for the Dispersal of Atmospheric11. Pollutants and Its Comparison to Regional Tracer Studies. J. Appl. Meteor., 17, 320-329, (1978)12. M. Taki, et. al., Isopleths of Surface Air Concentration and Surface Air Absorbed Dose rate due to13. a Radioactive Cloud released from a Stack(II). JAERI-M 90-206 (1990)