Upload
others
View
5
Download
0
Embed Size (px)
Citation preview
Using a Combination of Weighting Factors, Genetic Algorithm
and Ant Colony Methods to Speed up the Reloading Pattern Optimization
of VVER-1000 Reactors
Yashar Rahmania , Ali Pazirandeh
b, Mohammad B. Ghofrani
c, Mostafa Sadighi
d
aDepartment of Physics, Faculty of Engineering, Sari Branch,
Islamic Azad University, Sari, Iran
bDepartment of Nuclear Engineering, Science and Research Branch,
Islamic Azad University, Tehran, Iran
cDepartment of Energy Engineering, Sharif University of Technology,
Tehran 8639-11365, Iran dOfogh Consulting Engineers, No.3, Separ Alley, Africa Blvd.,
Postal code: 1518716713, Tehran, Iran
ABSTRACT This paper discusses a new and innovative method, which significantly speeds up the reloading
pattern optimization of VVER-1000 reactors. To reduce the range of possible searchable loading
patterns and consequently speed up calculations for the reloading pattern optimization of the
BUSHEHR VVER-1000 reactor (second cycle), first, according to the values of weighting factors of
the 16 applicable fuel assemblies (in this cycle), these assemblies are categorized into 6 limited groups.
Then, the optimum arrangement was obtained in the 6-group state by using the genetic algorithm
method. Finally, the position of each subset fuel assembly in the 6 compacted groups was determined
by using the weighting factor method and ant colony algorithm. This way, the re-transformation process
of the optimum arrangement was performed from the 6-group to the real 16-group state. It should be
noted that the arrangement obtained from the re-transformation process was completely in compliance
with the optimum reloading pattern achieved from the 16-group genetic algorithm calculations, which
verifies the methods applied in this research and also the accuracy of the re-transformation calculations.
By using the method discussed in this paper, identical optimum arrangement can be achieved in nearly
half of the required runtime for conventional stochastic optimization methods. Of course, the proposed
method is not limited to use the genetic algorithm and ant colony in the optimization process; it can
accelerate the reloading pattern optimization of the reactor using other stochastic algorithms. In this
research, to increase the accuracy of the objective parameters computations in each of the arrangements,
thermo-hydraulic calculations were made and the effectiveness of temperature feedbacks in the cell and
neutronic calculations was discussed. Furthermore, all the structural constraints are observed in this
method. As no research or data have ever been reported on the reloading pattern optimization of
BUSHEHR VVER-1000 reactor in the second cycle, it was impossible to make comparisons.
Therefore, by using the coupling of WIMSD5-B, CITATION-LDI2 and WERL codes, the time-
dependent thermo-neutronic computations were made in the second cycle to ensure the safety and
desirability of the proposed pattern during the cycle.
Keywords— Reloading pattern optimization, VVER-1000 reactor, genetic algorithm, weighting
factor method, ant colony algorithm
INTRODUCTION
In the calculation of the optimum loading pattern for a nuclear reactor, several parameters, including
the multiplication factor maximization, the raising of the desired safety threshold during the cycle
length and the flattening of the radial power peaking factor should be taken into account.
Several studies have been conducted to calculate the reloading pattern of VVER-1000 reactors; among
them, one can refer to:
1. Genetic algorithms [1,2,3]
2. Particle swarm optimization [4,5]
3. Neural network & simulated annealing [6]
4. Cellular automata & simulated annealing [7]
5. Cellular automata [8]
6. Perturbation theory [9]
7. Artificial Bee Colony [10]
However, most of these studies have been conducted in the first or equilibrium cycles, which have
less fuel variety compared with the second cycle of BUSHEHR VVER-1000 reactor.
As this paper aimed to achieve a method for the reloading pattern optimization of VVER-1000
reactors in transient cycles, first, the fuel compositions were calculated for each remaining assembly
from the first cycle. After determining the desired cycle length (300 days), a quick estimation was made
for specifying the type and quantity of the loadable fresh assemblies to provide the required excess
reactivity for passing the cycle length with the reused fuel assemblies remaining from the previous
cycle.
After this stage, an approach was designed using a combination of the weighting factor method and
the genetic algorithm to quickly optimize the reloading pattern of the BUSHEHR VVER-1000 reactor
(in the second cycle).
The coupling of WIMSD5-[11] and CITATION-LDI2[12] codes with the thermo-hydraulic
calculations were used to calculate the effective multiplication factor and the radial power peaking
factor distribution of the reactor core. As there were no similar studies or reported data for reloading
pattern optimization of VVER-1000 reactor in the second cycle, it was not possible to compare the
results; however, in this paper, in order to ensure the desirability of the proposed arrangement, the
thermo-neutronic behavior of the reactor core (with this arrangement) was simulated during the cycle.
The evaluation of the results indicated that the proposed arrangement enjoys a desirable safety margin.
GENERAL INTRODUCTION TO THE MAIN CHARACTERISTICS OF THE BUSHEHR VVER-1000
REACTOR
The BUSHEHR VVER-1000 reactor falls into the category of pressurized light water reactors. The
core of this reactor is composed of 163 hexahedral fuel assemblies [13]. Ten groups of control rod
assemblies are used to control and maintain the reactor’s safety. In the second operational cycle, 16
types of fuel assemblies are used including 12 types that remain from the first cycle (109 fuel
assemblies). Chromium diboride has been used as a burnable absorber in two of the 16 types of
assemblies. Table.1 describes the main characteristics and working conditions of the BUSHEHR
VVER-1000 reactor [13].
Fig.1 also outlines the general layout of the core of BUSHEHR VVER-1000 reactor and the
arrangement of the fuel assemblies in the first operational cycle [13].
TABLE 1. THE MAIN TECHNICAL CHARACTERISTICS OF VVER-1000 (V-446) REACTOR
3000 Reactor nominal thermal power, MW
30 Coolant heating in the reactor, Co
4 Number of loops, pieces.
4.14 Flow area of the core,m2
1.44 Maximum allowable radial power
peaking factor
Hexahedral
prism
Fuel assembly geometry
3.53 Fuel height in the cold state, m
pellets
UO2
Fuel
)( 232
4
TiOODy
CB + Absorbing material
0.236 Pitch between fuel assemblies, m
31075.12 −× Pitch between the fuel rods, m
Fig.1 The arrangement of the VVER-1000 reactor core in the first cycle
BURNUP CALCULATION PROCESS
Because the purpose of this study is to describe the process of the optimum reloading pattern
proposition in the second cycle of BUSHEHR VVER-1000 reactor, burnup calculations should be
conducted to determine the fuel compositions of the reused fuel assemblies.
Therefore, a coupling of neutronic and thermo-hydraulic calculations was used. To do this, the
physical group constants of the fuel assemblies and reflectors were calculated using the WIMSD5-B
code.
Furthermore, to obtain the time-dependent changes in the fuel composition and calculate the rate of
burnup in each fuel assembly, the computational capabilities of the WIMSD5-B code were utilized. By
inserting the physical group constants obtained from the WIMSD5-B code into the input file of the
CITATION-LDI2 code and also defining the geometry of the reactor core, the effective multiplication
factor and the three-dimensional distribution of the reactor’s thermal power were calculated.
In this study, a thermo-hydraulic software was designed using the Enveloped Pin method [14],[15].
Furthermore, the Dittus-Boelter [15], Ross-Stoute [15],[16] and Lee-Kesler [17] models were used in
the calculations of the heat transfer coefficient of coolant, gap conductance coefficient and gap
pressure, respectively. In addition, to estimate the concentration of the released gaseous fission products
into the gap space of fuel, the Weisman [18] model was used.
By using the results of the thermo-hydraulic calculations, the temperature and density of the fuel, clad
and coolant elements (in each fuel assembly) were applied to the neutronic calculations and thus, a
continuous sequence of neutronic and thermo-hydraulic calculations was created. Fig.2 provides a
schematic description of the applied computational flowchart.
Fig.2 Schematic description of the applied computational flowchart.
THE CALCULATION OF THE TYPE AND QUANTITY OF THE LOADABLE FRESH FUEL ASSEMBLIES IN
THE SECOND CYCLE
The estimation of the rate of fuel depletion in the 163 assemblies of the reactor core (in the first
cycle), specified that 109 fuel assemblies of the first cycle can be reused in the second cycle and 54
fresh fuel assemblies together with those 109 assemblies should be used to complete reloading of the
reactor core in the second cycle. The calculation in this section aims to determine the type and quantity
of these fresh fuel assemblies, which is be used in the reloading of the second cycle. Generally, the
designer proposes 11 types of fuel assemblies in VVER-1000 reactors, which can be used in all of the
operating cycles to complete the reloading of reactor core. In Table.2 the characteristics of these fuel
assemblies has been discussed.
Therefore, to complete the arrangement of the reactor core in the second cycle, some fresh fuel
assemblies should be selected among these 11 types of assemblies.
As the positions of the fuel assemblies in the reactor core are divided into six and twelfth-fold
symmetric cells (Fig.5), these 54 fresh fuel assemblies should be made using 6 and 12 fuel packages.
Studying the types of these 109 reused assemblies shows that 6 sets of the whole 9 symmetric six-
fold cells in the reactor core must be reserved for these types of assemblies. Therefore, to achieve the
whole of the possible states of loading 54 fresh fuel assemblies, we have to consider the states, in which
not more than three sets of six-fold assemblies are used.
It should be noted that, 54 required fresh fuel assemblies could be made in four different states.
Table.3 describes these 4 different states.
TABLE .2THE SPECIFICATIONS FOR DIFFERENT TYPES OF
APPLICABLE FUEL ASSEMBLIES IN VVER-1000 REACTORS
Quantity of fuel rods
(enrichment U-235,%
weight)
Burnable absorber rod
(BAR) characteristics
FA type U-235,
% weight
Fuel rod
Type 1
Fuel rod
Type 2
existence Number content
natural,
g/cm3 1.6% 1.6 311
(1.6)
-
2.4% 2.4 311
(1.6)
-
2.4%
B0.02
2.4 311
(1.6)
+ 18 0.02
2.4%
B0.036
2.4 311
(1.6)
+ 18 0.036
3.62% 3.62 245
(3.7)
66
(3.3)
-
3.62%
B0.02
3.62 245
(3.7)
66
(3.3)
+ 18 0.02
3.62%
B0.036
3.62 245
(3.7)
66
(3.3)
+ 18 0.036
4.02% 4.02 245
(4.1)
66
(3.7)
-
4.02%
B0.02
4.02 245
(4.1)
66
(3.7)
+ 18 0.02
4.02%
B0.036
4.02 245
(4.1)
66
(3.7)
+ 18 0.036
4.02%
B0.050
4.02 245
(4.1)
66
(3.7)
+ 18 0.050
TABLE .3 THE ARRANGEMENT OF 54 FRESH FUEL ASSEMBLIES IN THE REACTOR CORE
According to the reasons mentioned before, states III and IV never occur. Therefore, among the
different loading states of 54 fresh fuel assemblies, only states I and II should be considered, which is
illustrated in Fig.3
Fig.3 Allowed placements for fresh fuel assemblies
After creating all the possible states for loading of fresh fuel assemblies, it was realized that only
5005 states could occur. To avoid spending too long time for testing the desirability of each possible
state, ranking is performed for each of these possible states using the weighting factor method. Based
on that, a loading state would be searched which has a desirable effective multiplication factor at the
end of a 300-day cycle.
APPLYING WEIGHTING FACTORS FOR EACH POSSIBLE STATE OF FRESH ASSEMBLIES LOADING
To search for the appropriate loading state of the fresh fuel assemblies, the time-dependent
calculations should be made for estimating the fuel composition and subsequently the core’s excess
reactivity at each 5005 possible loading states (at the end of the cycle).
To do so, WIMSD5-B code was used for the time-dependant calculation of fuel consumption in each
assembly .Therefore, at each of the possible state, first, the fuel composition of the fresh assemblies (at
that state) and the reused 109 assemblies were averaged; then, designing a virtual fuel assembly was
begun based on the averaged fuel composition.
To perform the calculations of fuel consumption in the virtual fuel assembly, the thermal power was
considered about 18.4049 megawatt (163
3000MW ) and the cycle length was considered to be 300 days.
After calculating the fuel composition of each state and obtaining the physical group coefficient of the
virtual assembly at the end of the cycle (using WIMSD5-B code), the data were linked to the
CITATION-LDI2 code through which the multiplication factor of the reactor core was calculated for
this possible state.
In order to calculate the effective multiplication factor using the CITATION-LDI2 code, the fuel
composition of the virtual assembly was assumed for all the 163 fuel assemblies within the reactor core.
In this research to overcome the phenomena such as samarium locking, it was supposed that the
effective multiplication factor of the reactor should be considered a little greater than the critical value
(approximately 1.02) at the end of the 300-day cycle.
Making time-dependant calculations for fuel depletion in each 5005 possible state would take a lot of
time. Therefore, to avoid spending a long time on calculations, an innovative approach based upon the
weighting factor method was used.
In this method, first, the fuel composition of each 11 types of the fresh fuel assemblies was calculated
at the end of the 300-day cycle. Next, the fuel-composition-dependent weighting factors were
calculated for each of these assemblies (at the end of the cycle).
With respect to the type and quantity of fresh fuel assemblies at each of these 5005 possible states,
the equivalent weighting factor was calculated.
Strictly speaking, the equivalent weighting coefficient (for each of the 5005 possible states) was
calculated through averaging the weighing factors of the fuel assemblies in that state.
This way, instead of making time-consuming calculations for estimating the fuel depletion in each of
the possible states, the weighting factors of each state were estimated through making calculations for
only 11 types of the fresh fuel assemblies. Later, the ascending grading of these 5005 states was
performed through the weighting factors of each possible loading state, as the state with minimum
weighting factor is placed in the first row and the state with the maximum weighting factor is placed in
the last row of the 5005 states.
It was evident that the possible states with high weighting factors, due to their higher fuel enrichment,
enjoyed higher effective multiplication factor than the states with lower weighting factor. Therefore, the
Bisection computational method was used in this section.
In this regard, first, fuel depletion calculations were made for each of the upper and lower boundaries of
the possible states. Next, calculating the average weighting factors in both domains, the state with the
lowest difference between its weighting factor and the averaged weighting factor was searched. Then
the effective multiplication factor was calculated in that middle state. If the effective multiplication
factor was lower than the desired quantity, calculations of the interval division should be continued at
the lower half of the states set which has higher weighting factors as compared with the earlier state;
otherwise, the upper half would be selected. According to the Bisection method, the span of possible
states would be halved in the following steps so that the searching process would reach the desired
result.
The important point which was considered as the second condition of selection in these calculations
was the choice of the possible state with the lower effective multiplication factor at the beginning of the
cycle within the sets of states with an equal multiplication factor at the end of the cycle. This is because
lower effective multiplication factor at the start of a cycle, equivalently, leads to more stability and
safety in the reactor’s core against positive reactivity feedbacks induced by boric acid dilution.
Furthermore, it would enjoy a higher safety margin as far as the rest of the thermo-neutronic parameters
are concerned.
Fig.4 schematically describes the computational flowchart applied in this section.
Meanwhile, Table.4 describes the results of the calculations and the offer made for the type and
quantity of the fresh fuel assemblies applicable in completing the BUSHEHR VVER-1000 reactor
reloading pattern in the second cycle.
Fig.4 A schematic description of the applied computational flowchart to
determine the type and quantity of the fresh fuel assemblies in the second cycle
TABLE .4 TYPES AND QUANTITIES OF THE FRESH FUEL ASSEMBLIES APPLICABLE IN THE SECOND CYCLE
Fuel assembly
type
Quantity
1.6% 0
2.4% 0
2.4%B0.02 0
2.4%B0.036 0
3.62% 12
3.62%B0.02 6
3.62%B0.036 12
4.02% 24
4.02%B0.02 0
4.02%B0.036 0
4.02%B0.050 0
USING THE COMBINATION OF WEIGHTING FACTOR METHOD AND GENETIC ALGORITHM FOR THE
RELOADING PATTERN OPTIMIZATION OF THE VVER-1000 REACTOR
After estimating the types and quantities of the loadable fuel assemblies in the second cycle, the fuel-
composition-dependent weighting factors of each assembly were calculated.
Then, according to the weighting factor of each type of assemblies, the assemblies with close
weighting factors were categorized in the compacted groups, so that the number of varieties of the fuel
assemblies was reduced to six.
Afterwards, the fuel composition of the subset assemblies was averaged in each of the compressed
groups. Then, as a sample of that group, a virtual assembly with the averaged fuel composition was
created.
Besides, the optimum arrangement at the six-group state was obtained using the genetic algorithm.
Then, the re-transformation of the optimal arrangement from 6-group state into the real 16-group was
performed using the radial distribution of its maximum power peaking factor and the estimated
weighting factor. Fig.8 schematically describes the computational flowchart applied in this section.
Calculating the Weighting Factor in Each Type of Fuel Assemblies
To calculate the weighting factor of each of the 16 types of fuel assemblies based on their fuel
compositions, the WIMSD5-B and CITATION-LDI2 codes will be coupled with thermo-hydraulic
computations to calculate the produced thermal power of each fuel assembly per equal neuron flux.
This weighting factor (factor C) is obtained by dividing the produced thermal power of each fuel
assembly by the thermal power of a typical assembly with a fuel composition averaged from the entire
fuel assemblies (Formula 1).
ave
i
iP
PC = (1)
Where Ci is the weighting factor value, i is the type of fuel assemblies, Pi is the produced thermal
power in each type of fuel assembly and Pave is the produced power of the fuel assembly with the
averaged enrichment.
In order to calculate this factor, we assumed an equal neutron flux in numerator and denominator of
this formula; as a result, neutron flux parameter was removed from both the numerator and
denominator of this correlation.
Table.5 shows the calculated weighting factor values that are based on the fuel compositions of each
the 16 type of fuel assemblies.
TABLE .5 THE WEIGHTING FACTOR BASED ON THE
FUEL COMPOSITION OF EACH ASSEMBLY (“C” FACTOR)
“C” factor Fuel
assembly
type
number
0.8267205 24-2 1
0.8266882 24-2 2
0.8250063 24-2 3
0.8250063 24-2 4
0.8268361 24-2 5
0.8475394 24B20-2 6
0.8551277 24B36-2 7
0.8531355 24B36-2 8
1.185350 36-2 9
1.196011 36-2 10
1.183665 36-2 11
1.205054 36B36-2 12
1.231149 36-1 13
1.128669 36B20-1
14
1.074982 36B36-1
15
1.359755 40-1 16
1 FA
(average enrichment)
The calculated values of weighting factors in the 16 types of fuel assemblies show that this
parameter’s values for some assemblies are similar (Table.5). Thus, these assemblies are classified into
fewer groups based on their weighting factor values. For this purpose, fuel assemblies of 1 to 8 are
assigned to group one; 9 to 12 to group two; and 13, 14, 15, and 16 to groups three, four, five, and six,
respectively .Table.6 indicates the C factor values, and the quantity of assemblies in each of the limited
six groups.
TABLE .6 THE NUMBER AND THE VALUES OF C FACTOR
FOR ASSEMBLIES IN EACH OF THE SIX LIMITED GROUPS
The types of fuel
in sextet groups
Burnup C factor Quantity
1 (group 2.4%)
reused fuel
12.7-14.56 0.8403266 67
2 (group 3.6%)
reused fuel
9.6-12.17 1.201491 42
3( group 3.6%)
fresh fuel
0 1.231149 12
4 ( group 3.6%B0.02)
fresh fuel
0 1.128669 6
5( group 3.6%B0.036)
fresh fuel
0 1.074982 12
6( group 4.02%)
fresh fuel
0 1.359755 24
Description of the reloading pattern optimization of BUSHEHR VVER-1000 reactor using the
genetic algorithm
Having done the coding process and defined the arrangement of the reactor core as a chromosome
with 19 genomes, the random generation of the initial population of the chromosomes was performed in
the following step. It should be noted that in the present research, the initial population consists of 20
chromosomes. After evaluating the initial population and selecting 10 chromosomes as the initial
parents, producing the next generation was carried out using the mutation and crossover operators. In
this research, the number of offsprings for each generation was considered to be 7 times bigger than the
number of the parents’ population (10).
In the present paper, the optimization calculations were made in two states using two different fitness
functions. In the first state, the optimality condition was considered as per the access to an arrangement
with flattened power distribution. In the second state, the definition of the fitness function was offered
in order to achieve an arrangement with the minimum “maximum power peaking factor” and maximum
“effective multiplication factor”.
The chromosomes of each generation were evaluated using the determined fitness function.
Meanwhile, nine superior chromosomes and one of the worst chromosomes were selected.
It should be noted that the reason for choosing a bad chromosome as one of the 10 parents selected in
each generation was to prevent the trapping of the search process in a series of local optimum results.
In addition, in order to prevent the loss of the desired chromosomes of the earlier generation and the
probable reduction of optimality during the process of searching, the selecting of superior chromosomes
in each generation is performed among the offsprings of that generation and the selected chromosomes
of the previous generation.
Therefore, this process is performed successively in each of the generations to achieve convergence.
Finally, to determine the appropriate time to finish the calculations, the fitness function values of the
selected parents of the last three generations are compared. In case of coincidence or negligible
differences among them, the calculations will be ended.
Coding of the VVER-1000 Arrangements
The hexagonal structure of the core of VVER-1000 reactors has a one-twelfth symmetry. In addition,
as the fuel assemblies of this reactor are hexagonal, the 163 positions of the reactor core are divided
into 9 twelve-fold sets, 9 six-fold sets, and one central cell.
Hence, the core of the VVER-1000 reactor is divided into 19 positions. By knowing the type of the
fuel assemblies in these 19 positions, it is possible to generate the arrangement of the reactor core. Fig.5
describes the classifying of the reactor core positions into six and twelve-fold cells.
Fig.5 The classifying of the reactor core positions
into 6- and 12-piece cells
Therefore, the geometrical structure of the reactor core is coded as a 19-genome chromosome. Fig.6
describes this chromosome and quantity of the relevant cells.
Fig.6 the coding of a reloading pattern as a 19-genome chromosome
The Evaluation and Selection of Chromosomes
In this research, the calculation of reloading pattern optimization was made in two states.
To estimate the quantity and type of the fresh fuel assemblies, the calculation was made in a way to
keep the criticality at an optimal level during the cycle. Therefore, in the first state, the optimization
was conducted so as to achieve an arrangement by flattened thermal power distribution.
When the criticality of the reactor is assured during the cycle (at a desirable level), the reactor core
with a flattened power distribution, results in more reduced fluctuations caused by the boron dilution
phenomenon and temperature feedbacks.
Therefore, to evaluate the generated chromosomes in each generation, chromosomes with a flatter
power distribution (minimum SPPE factor) were selected.
∑=
−=163
1
)1(m
mk PPFSPPE
(2)
where PPFm is the power peaking factor in each of the assemblies and SPPEk indicates the flattening
level of the radial power peaking factor distribution of each chromosome.
In the second state, the evaluation of chromosomes was carried out as per the access to a reloading
pattern with the maximum effective multiplication factor and the safe maximum power peaking factor.
To do so, the optimization condition was considered in maximizing the value of the fitness function
offered in the following formula.
)()1(),( maxmax PPFPPFYKXPPFKF uleffeff −+−= (3)
Where effk is the multiplication factor of the reactor and
maxPPF and ulPPF are the maximum power
peaking factor of the reactor and the greatest possible value of this factor (5), respectively. X and Y are
the weighting coefficients, which have been considered as 268 and 27, respectively.
It is noteworthy that the second state calculations were only performed to ensure the desired and rapid
performance of the research’s proposed method at multi-objective optimization states.
3. Using Crossover and Mutation Operators to Generate Offsprings of Each Generation
After the processes of selecting the parent chromosomes of the next generation, offspring
chromosomes are generated using operators such as crossover and mutation.
In this research, the population of the offspring chromosomes was considered to be 7 times bigger
than the parent chromosomes and the probabilities of the mutation (Pm)and crossover (Pc) operators
were considered as 0.7 and 0.3, respectively. Therefore, in a 70-piece population of the offspring
chromosomes, 49 chromosomes are generated through the mutation operator and the 21 remaining
chromosomes are generated through the crossover operator.
In this research, a one-point mutation operator and a four-point crossover operator were used. Fig.7
schematically describes the procedure of these operators. In the process of generating offspring
chromosomes, it is necessary to observe the structural constraints including matching quantity of each
type of fuel assembly applied in each arrangement with its permissible quantity and prohibition of
placing the fuel assemblies with burnable poison at the specified positions for entering control rods.
Therefore, each offspring chromosome caused by these operators was examined. If the constraints are
not matched with the generated chromosomes, they are eliminated and the operators are used
continuously to achieve the population, which is matched with the constraints.
As a result, using these operators in each generation, the offspring chromosomes are generated and
this cyclic process is carried out repeatedly until optimum result is achieved.
Fig.7 The performance of operators
a) one-point mutation; b) four-point crossover
To explain more clearly, Fig.8 schematically describes the computational flowchart applied in this
section.
Fig.8 A schematic description of the computational flowchart
applied in the optimization section using the genetic algorithm
RE-TRANSFORMATION PROCESS
Re-transformation of the optimum 6-group Arrangement to the real 16-group pattern by using
weighting factor concept
The process of re-transformation is based on calculating the radial power-peaking-factor distribution of
the proposed six-group arrangement. Then, based on the power peaking factor for the fuel assemblies in
groups one and two, the fuel assemblies will be positioned based on their C weighting factors while
observing the structural constraints.
It is performed in such a way that the assemblies with the highest C factor are located in places with the
lowest power peaking factors. This process will continue in this way until the last fuel assembly with
the lowest C factor is located in the place with the highest power peaking factor.
Thus, the units in groups one and two are divided into their subset fuel assemblies, and the optimum 6-
group arrangement is converted to 16 groups.
Re-transformation of the optimum few group patterns to the real mode by using ant colony
algorithm
As a second solution, the re-transforming process has been done by using the ant colony algorithm.
In this method, by using the η parameter (in the algorithm), the structural constraints can be easily
considered during the loading pattern optimization.
In order to escape from the local optimum solutions, the random values (between 0-1) were generated
for each position of the condensed groups.
In continuation, if the value of the random number in each position was lesser than 0.2, then the
uniform pheromone was assigned for the possible states of that position and subsequently the random
type of fuel assembly was selected for it.
To prevent from retreating of the optimization process, the value of the objective function of the
mentioned ant was compared with its previous value and in order of superiority of the previous state, it
is remained.
In this section, the initial population consists of 10 ants which have been generated randomly from the
6-group optimum pattern.
Fig.9 schematically describes the computational flowchart applied in this section.
Fig.9 A schematic description of the computational Flowchart applied
in the optimization section using the ant colony algorithm (in re-transforming process)
RESULTS
Figs.10 to 12 compares the rate of changes of the fitness function for the first state optimization
model (minimizing SPPE factor) in the 6 and 16-group calculations.
051015202530354045505560657075808590
0 1 2 3 4 5 6 7 8 9 10
Generation
Fitness function (SPPE)
Fig.10 Changes in the rate of first fitness function (SPPE)
during the generation 1 to10 in the 16-group state
11.9511.975
1212.02512.0512.07512.1
12.12512.1512.17512.2
12.22512.2512.27512.3
12.32512.3512.37512.4
11 12 13 14 15 16 17 18 19 20 21 22
Generation
Fitness function (SPPE)
Fig.11 Changes in the rate of first fitness function (SPPE) during
the generation 11 to 21 in the 16-group state
0
10
20
30
40
50
60
70
80
90
100
0 1 2 3 4 5 6 7 8 9 10 11 12 13
Generation
Fitness Function (SPPE)
Fig.12 Changes in the rate of first fitness function (SPPE)
in the 6-group state
By making the optimization calculations at the 16-group state, the optimum reloading pattern is
obtained after 21 generations. In spite of the convergence of the 6-group genetic algorithm calculations
in the 12th generation, this method reached the optimum arrangement equivalent to the 16-group state
from the 8th generation on.
Figures 13 and 14 describe the changes of the fitness function in the second re-transformation model(
by using ant colony algorithm).
In the second mode of optimization (using formula no.3),by making the optimization calculations at
the 16-group state, the optimization process is converged after 26 generations and in the 6-group
genetic algorithm calculations this method reached the optimum arrangement equivalent to the 16-
group state in the 11th generation .
146
146.2146.4
146.6146.8147
147.2
147.4
147.6147.8148
148.2
148.4148.6
148.8
0 1 2 3 4 5 6 7 8 9 10 11
Iteration
Averaged fitness function
Fig.13 the changes of the averaged fitness function of ants in each iteration
148.35
148.4
148.45
148.5
148.55
148.6
148.65
148.7
148.75
148.8
148.85
148.9
148.95
0 1 2 3 4 5 6 7 8 9 10 11
Iteration
The fitness value's of best Ant
Fig.14 Changes of the fitness function of the best ant in each iteration
It is observed that the number of evaluations in whole of 10 iterations of ant colony algorithm is much
lesser than the required evaluations for 2 generations of the genetic algorithm method (140 genomes).
Figs.15-16 shows the optimum reloading patterns in 6 and 16-group states of the second optimization
mode. It should be noted that, after re-transformation computations in the 6-group optimum
arrangement (by using the ant colony algorithm), exactly the same 16-group optimum arrangement was
obtained.
Fig.15 The proposed optimum reloading pattern of
the BUSHEHR VVER-1000 reactor (6-groups) in second mode of optimization
Fig. 16. The optimum reloading pattern of the VVER-1000 reactor in 16-group in the second mode of
optimization (obtained from the re-transformation process by ant colony algorithm and also the 16-
group genetic algorithm calculations)
As in the earlier state, it is observed that for an equal value of the fitness function, the 6-group
calculations enjoy a higher convergence velocity as compared with the 16-group model.
Table.7 describes the value of objective parameters in 16 and 6-group optimum reloading patterns (in
first mode of optimization).
TABLE .7 A DESCRIPTION OF THE BASIC PARAMETERS OF
THE PROPOSED OPTIMUM ARRANGEMENTS IN 6 AND 16-GROUP COMPUTATIONS
Optimization type SPPE Keff PPFmax
16 group 11.897610 1.1547 1.1539
6 group 12.502290 1.1552 1.1815
Figs.17-20 shows the optimum reloading patterns and distribution of the radial power peaking factor
of the optimum arrangements in 6 and 16-group computations(in first mode of optimization). It should
be noted that, after re-transformation computations in the 6-group optimum arrangement (by using
weighting factor method), exactly the same 16-group optimum arrangement was obtained.
Fig.17 The proposed optimum reloading pattern of
the BUSHEHR VVER-1000 (V-446) reactor (6-groups)
Fig.18 The radial power-peaking-factor distribution
of the proposed optimum reloading pattern (6-groups)
Fig. 19. The optimum reloading pattern of the VVER-1000 reactor in 16-group (obtained from the
retransformation process and also the 16-group genetic algorithm calculations)
Fig.20 The radial power-peaking-factor distribution of the proposed optimum reloading pattern (16-
groups)
Fig.21 describes the rates of change in the maximum power peaking factor of the proposed pattern
during the second cycle.
Fig. 21 The time dependent changes of maximum power peaking factor
in the proposed reloading pattern (second cycle)
DISCUSSION AND CONCLUSION
As Figs 10-14 show, using the proposed method in this research, identical optimum arrangements
can be achieved through the usual optimization method by using the genetic algorithm but with fewer
generations and, consequently, within a shorter run time.
Of course, the proposed method is not limited to use the genetic algorithm in the optimization
process; it can accelerate the reloading pattern optimization of the reactor using other stochastic
algorithms.
In addition, by comparing the re-transformed 16-group arrangement from the 6-group state and the
optimum pattern obtained from the 16-group state, it is observed that there is a complete conformity
between these two arrangements. This shows that the proposed method and also the applied
computational process in the re-transformation section are correct.
In this research, to increase the accuracy of the objective parameters computations in each of the
arrangements, thermo-hydraulic calculations were made and the effectiveness of temperature feedbacks
in the cell and neutronic calculations was discussed. Furthermore, all the structural constraints are
observed in this method, while in the other studies [3to10], the constraints were not considered.
As no research or data have ever been reported on the reloading pattern optimization of VVER-1000
reactors in the second cycle, it was impossible to make comparisons. Therefore, the time-dependent
thermo-neutronic computations in the second cycle were made to ensure the safety and desirability of
the proposed pattern during the cycle. As shown in Fig.21, the arrangement enjoys a desired safety
margin during the second cycle.
REFERENCES
[1] C. Guler, S. Levine, K. Ivanov, J. Svarny, V. Krysl, P. Mikolas, J. Sustek, “Development of the
VVER core loading optimization system,” Ann.Nucl.Energy., vol. 31, 2004, pp.747-772 .
[2] F. Alim, K. Ivanov, S. H. Levine, “New genetic algorithms (GA) to optimize PWR reactors Part III:
The Haling power depletion method for in-core fuel management analysis,” Ann.Nucl.Energy., vol. 35,
2008, pp.121-131.
[3] N. Mohseni, M. Boroushaki, M. B. Ghofrani, M. H. Raji, “Application of minimum plutonium
criteria at EOC to optimize the fuel loading pattern in VVER-1000 reactors,” Ann.Nucl.Energy., vol.35,
2008, pp.269–276 .
[4] D. Babazadeh, M. Boroushaki, C. Lucas, “Optimization of fuel core loading pattern design in a
VVER nuclear power reactors using Particle Swarm Optimization,” Ann.Nucl.Energy., vol.36, 2009,
pp.923–930 .
[5] M. Khoshahval, A. Zolfaghari, H. Minuchehr, M. Sadighi, A. Norouzi, “PWR fuel management
optimization using continuous particle swarm intelligence,” Ann.Nucl.Energy .,vol. 37, 2010,
pp.1263-1271.
[6] A. H. Fadaei, S. Setayeshi, “LONSA as a tool for loading pattern optimization for VVER-1000
using synergy of a neural network and simulated annealing,” Ann.Nucl.Energy., vol.35, 2008, pp.1968–
1973.
[7] A. H. Fadaei, S. Setayeshi, S. Kia, “An optimization method based on combination of cellular
automata and simulated annealing for VVER-1000 NPP loading pattern,” Nucl.Eng.Des., vol. 239,
2009, pp.2800–2808 .
[8] A. H. Fadaei, S. Setayeshi,” A new optimization method based on cellular automata for VVER-
1000 nuclear reactor loading pattern,” Ann.Nucl.Energy., vol.36, 2009, pp.659–667 .
[9] M. Hosseini, N. Vosoughi, “Development of a VVER-1000 core loading pattern optimization
program based on perturbation theory,” Ann.Nucl.Energy., vol.39, 2012, pp.35-41.
[10] O. Safarzadeh, A. Zolfaghar, A. Norouzi, H. Minuchehr,” Loading pattern optimization of PWR
reactors using Artificial Bee Colony,” Ann.Nucl.Energy., vol. 38, 2011, pp.2218-2226 .
[11] M. J. Roth,J. D. Macdougall, P. B. Kemshell, “Winfrith Improved Multigroup Scheme Code
System (WIMSD5-B manual),” NEA-1507 , Atomic Energy Establishment ,1997.
[12] T.B.Fowler, D.R.Vondy, G.W.Cunningham, “CITATION-LDI2: Nuclear Reactor Core Analysis
Code System,” CCC-643ORNL, Oak Ridge National Laboratory, 1999.
[13] Atomenergoproekt, “BUSHEHR VVER-1000 Reactor Final Safety Analysis Report (Chapter
4),”49.BU.10.0.OO.FSAR.RDR001, Ministry of Russian Federation of Atomic Energy, 2003.
[14] Y. Rahmani, ”Safety Analysis of VVER-1000 Type Reactors Against Pump Failure Accident in 2
and 4 Loops,” in Proc. Int. Conf. Safety assurance of NPP with WWER, GIDROPRESS, Podolsk,
Russia, May. 17-20, 2011.
[15] N. E. Todreas and M. S. Kazimi, Nuclear System 1, p.332-338, Taylor and Francis, Massachusetts,
United states, 1993.
[16] Y. Rahmani, M. Rahgoshay, ”Study of the Role of Gap Conductance Coefficient of Fuel on
Increasing Safety in VVER-1000 Reactors,” Proc. Int. Conf. Safety assurance of NPP with WWER,
GIDROPRESS, Podolsk, Russia, May. 17-20, 2011.
[17] R. E. Sontag and C. Borgnakke, Fundamentals of Thermodynamics, 5rd ed., John Wiley and
Sons, New York, United States ,1997.
[18] J. Weisman, P. E. Macdonald, A. I. Miller, H. Ferrari, “Fission Gas Release from UO2 Fuel Rods
with Time Varying Power Histories,” Trans. Am. Nucl. Soc, vol.12,1969, pp. 900 .