Annals of Nuclear Energy 159 (2021) 108331

Contents lists available at ScienceDirect

Annals of Nuclear Energy

journal homepage: www.elsevier.com/locate/anucene

PWR core loading pattern optimization with adaptive genetic algorithm

Chol So ⇑, Il-Mun Ho, Jong-Suk Chae, Kwang-Hak Hong
Faculty of Energy Science, Kim Il Sung University, Pyongyang, Democratic People’s Republic of Korea

a r t i c l e i n f o a b s t r a c t

Article history: In this paper, a new adaptive genetic algorithm (AGA) is proposed for the loading pattern optimization
Received 17 November 2020 (LPO) of the pressurized water reactor (PWR) core. The tournament selection, the two-point crossover
Received in revised form 9 March 2021 and the mutation based on randomly swapping positions between two fuel assemblies (FAs) are applied.
Accepted 8 April 2021
New calculation formulae are introduced to adjust effectively the crossover and mutation probabilities
Available online 24 April 2021
according to the fitness value of the individual. The proposed algorithm is implemented in the LPO for
the first core of the 1000 MWe PWR. The objective function is to minimize the maximum radial power
Keywords:
peaking factor (RPPF) at the equilibrium of Xe under the constraint condition for the cycle length. The
Fuel management
Loading pattern optimization
maximum RPPF of the obtained LP is decreased than that of the loading pattern (LP) by the designer.
Adaptive genetic algorithm The results show that the proposed AGA is effective to improve the convergence rate of genetic algo-
PWR rithms (GAs) in the LPO.
Ó 2021 Elsevier Ltd. All rights reserved.

1. Introduction algorithm (GSA) (Mahmoudi et al., 2016), imperialist competitive

algorithm (ICA) (Akbari et al., 2018).
In the operation of the nuclear reactor, the loading pattern opti- GAs are stochastic optimization methods based on the biologi-
mization (LPO) plays a very important role. Its goal is to determine cal evolutionary theory of natural selection and survival of the fit-
the loading patterns (LPs) for producing full power with adequate test founded by Darwin. GAs have been widely applied to the LPO
safety margins (Levine, 1987). and yielded outstanding results. However, due to the low local
The LPO problem is a complex combinatorial optimization search capability, the performance of the conventional GA (CGA)
problem, and the principal characteristics of this problem are is not so good as those of other optimization methods such as
non-linearity, multimodality, discrete solutions with nonconvex PSO (Khoshahval et al., 2011), TS and ant system (AS) (François
functions, disconnected feasible regions and high dimensionality et al., 2013), HS (Aghaie et al., 2013) and GSA. Therefore, in the
(Stevens et al., 1995). Due to these characteristics of the LPO prob- use of GAs, it is very important to improve the local search capabil-
lem, achieving the optimal LP is very difficult and efficient opti- ity and increase the convergence rate of the algorithm.
mization methods are needed. Several techniques have been applied to increase the conver-
Since classical approaches such as gradient-based methods are gence rate of GAs in the LPO. Considering that the diversity of off-
not appropriated to the optimization of the core pattern spring in a population has a great impact on efficient search in a
(Meneses et al., 2009), several stochastic optimization methods design space, distributed genetic algorithm (DGA) was developed
have been applied to optimize LP such as simulated annealing (Yamamoto and Hashimoto, 2002). In DGA, candidate solutions
(SA) (Stevens et al., 1995), tabu search (TS) (Lin et al., 1998; Hill are distributed in the form of independent islands and evolved in
and Parks, 2015), ant colony optimization (ACO) (Machado and parallel. Migrations of some candidates between islands are peri-
Schirru, 2002; De Lima et al., 2008; Lin and Lin, 2012), genetic algo- odically performed to prevent the early convergence. Because
rithms (GAs) (Zhao et al., 1998; DeChaine, 1995; Yilmaz et al., diversity in overall population is adequately controlled by the
2006), particle swarm optimization (PSO) (Khoshahval et al., number of island and migration frequency, the performance of
2011), Cellular Automata (Fadaei and Setayeshi, 2009), harmony DGA is better than that of CGA. In order to improve the conver-
search (HS) (Aghaie et al., 2013; Nazari et al., 2013), evolutionary gence of GAs, the technique to eliminate most of invalid genotypes
harmony search (EHS) (Poursalehi, 2015), gravitational search was used and the worth technique was applied in both crossover
and mutation so that various fuel assemblies can be placed only
in certain core positions (Alim et al., 2008). Hybrid mutation oper-
⇑ Corresponding author. ators were proposed for integer coded GA (ICGA) to solve LP prob-
E-mail address: [email protected] (C. So).
lem in the pressurized water reactor (PWR) (Norouzi et al., 2011).

https://doi.org/10.1016/j.anucene.2021.108331
0306-4549/Ó 2021 Elsevier Ltd. All rights reserved.
C. So, Il-Mun Ho, Jong-Suk Chae et al. Annals of Nuclear Energy 159 (2021) 108331

They used these hybrid mutation operators to improve the compu- eighth core configuration for the first cycle presented by the
tation efficiency by shortening the long execution time of ICGA. designer is shown in Fig. 1. Characteristics and configurations of
The idea of coupling GA and Hopfield neural network (HNN) was each type of FA loaded in the first core are shown in Table 2 and
proposed to speed up the convergence of GAs by improving the Fig. 2.
quality of the initial population, and applied to the burnable poison
placement problem (Khoshahval and Fadaei, 2012). Based on com- 2.2. Calculations of the neutronic parameters
paring the effectiveness of GA and SA in the LPO for PWR, a hybrid
GA which switches to SA whenever the solution becomes stagnant In general, due to safety and economical requirements, LPs sat-
was developed (Zameer et al., 2014). To evaluate the effects on the isfying some constraints are searched in the fuel management. The
genetic variance of the population, different mutation strategies minimization of the radial power peaking factor (RPPF) and the
were applied and the results were compared (Israeli and Gilad, maximization of the effective multiplication factor or the critical
2018). The results showed that adaptive mutation strategies are boron concentration are widely used as constraints. The critical
highly beneficial to the optimization process and provide addi- boron concentration and RPPF of the core corresponding to the
tional and useful control measures over the stochastic optimiza- given LP are obtained via the core physics calculation by the core
tion process. simulation code.
In this work, a new adaptive GA (AGA) is proposed and imple- For the LP analysis of PWR, we have developed the core simula-
mented for the LPO of a PWR core. In GAs for the LPO, the adaptive tion code, TPSQ/AGANGC. AGANGC, the Adaptive Genetic Algo-
probabilities were mainly employed in the mutation process, and rithm Nodal Green’s function Code, is based on solving two
have been not applied to the crossover process. In addition, most dimensional (2D) and three dimensional (3D) multi-group neutron
of the adaptive probabilities employed in the mutation were rela- diffusion equations by the nodal Green’s function method on Neu-
tive to the generation number. The advantage of AGA is that the mann boundary condition (Hu and Zhao, 1998). TPSQ, the Trans-
crossover and mutation probabilities are adjusted according to mission Probability code for SQuare assemblies, is a lattice
the fitness of the individual to improve the convergence. However, physics code, which performs homogeneous calculations for fuel
the original AGA has the disadvantage that premature convergence and non-fuel cells and the fuel assembly by the collision probabil-
can occur easily. In this paper, new calculation formulae are intro- ity method and the transmission probability method. The multi-
duced to adjust effectively the crossover and the mutation proba- group constants at different burn-up, different fuel and coolant
bilities. The proposed algorithm is applied to optimize the first temperatures and different boron concentrations for nine types
cycle pattern for the core of a PWR with an electrical power rating of fuel assemblies calculated by TPSQ are input to AGANGC code.
of 1000 MW, and its performance is compared with those of CGA
and the original AGA. The optimization results show that the pro- 2.3. Adaptive genetic algorithm model
posed AGA is very effective to speed up the convergence rate of
GAs for the LPO. To solve the LPO problem by GA, the LP must be represented as
a string of symbols being called the chromosome. Chromosomes in
2. Material and method this work are encoded by using an integer vector. Because the first
core of the reactor has a one-eighth symmetric shape, the encoding
2.1. Reactor core description. is performed for this symmetric configuration. In the vector, inte-
gers signify the kind of FAs and positions of these integers repre-
The core consists of 177 of rectangular fuel assemblies and has sent positions of FAs in the core. The position order of FA begins
one-eighth symmetric shape. Nine types of fuel assemblies (FAs) at the center of the core, and is set to increase from down to up
with different enrichments and different numbers of burnable poi- and from left to right. The example of the encoded LP is shown
son rods are different are loaded into the first core. Each FA is made in Fig. 3.
of a 16
16 square pin cells and has 5 guide tubes. A summary of The initial population is created by including the individual
the reactor design specifications are given in Table 1. The one- which the FAs of Table 2 are randomly arranged. To guarantee
the diversity of the initial population, the new generated individual
is compared with already generated individuals; if this individual
Table 1 is not repeated, it is included into the initial population. When
Design data for the 1000 MWe PWR.
the FAs are arranged, quantities of each type of FA in the core are
Parameter Value preserved.
Reactor core operating conditions
Reactor nominal thermal power (MW) 2815
Pressurizer operating pressure (MPa) 15.82
Core active length (mm) 3800
Coolant inlet temperature (℃) 296
Flow area of the core (m2) 4.165
Coolant flow rate (t/h) 5.346
103
Equivalent diameter of the core (m) 3.124
Number of fuel rods in core 41,772
Fuel rod
Fuel pellet material Enriched UO2
Cladding material Zircalloy-4
Fuel pellet diameter (mm) 8.26
Clad outer diameter (mm) 9.7
Clad inner diameter (mm) 8.43
Fuel rod pitch (mm) 12.85
Guide tube
Tube material Zircalloy-4
Outer diameter (mm) 24.934
Inner diameter (mm) 22.9
Fig.1. The one-eighth core configuration presented by the designer.

2
C. So, Il-Mun Ho, Jong-Suk Chae et al. Annals of Nuclear Energy 159 (2021) 108331

Table 2
Various FAs in the first core.
235
FA Quantity of U Enrichment Number of fuel rods per Number of burnable poison rods per Gd2O3 concentration in burnable posion rod
type FA (%) FA FA (%)
A 45 1.28 236
B 20 2.34 236
B1 8 2.34/1.28 176/52 8 4
B2 16 2.34 232 4 4
C 12 2.84/2.34 184/52
C1 32 2.84/2.34 176/52 8 4
D 12 3.34/2.84 184/52
D1 8 3.34/2.84 176/52 8 4
D2 24 3.34/2.84 128/100 8 4

Fig. 2. Configurations of FAs.

Once the creation of the initial population is completed, the crossover, two points are randomly selected in the parents and
evolution starts with this initial population by applying genetic the genes placed between these points are exchanged.
operators including selection, crossover and mutation. After the crossover, the mutation operator is applied to intro-
The tournament selection method with the tournament size 2 is duce genetic diversity to the population. The mutation is per-
performed as the selection operator. The elitist strategy is also used formed by swapping two randomly selected genes. To avoid the
to speed up the convergence of GA, which survives the best indi- production of the duplicate individuals, this process will be
vidual in each generation. repeated several times.
Because crossover and mutation operations are carried out in a
2.3.1. Crossover and mutation random way, the FAs with the same identification number can be
The parent chromosomes are combined by crossovers and a located more or less than their quantity in Table 2. In order to pre-
new population is produced. The new population has many charac- serve the quantity of each type of FA, one of FAs with the increased
teristics of old population. It is believed that high potential parents number is randomly selected and replaced by the FA with the
will generate better offspring chromosomes (Amjady and Shirzadi, decreased number. This process is continued until quantities of
2008). Two-point crossover was implemented in our AGA. In this each type of FA on the current LP are equal to those in Table 2.
3
C. So, Il-Mun Ho, Jong-Suk Chae et al. Annals of Nuclear Energy 159 (2021) 108331

Fig. 3. The example of the encoded LP.

2.3.2. Probabilities of crossover and mutation where P c;max and Pm;max are the maximum values of crossover and
Srinivas and Patnaik (1994) proposed an AGA, which the cross- mutation probabilities, respectively. As shown in Eqs. (3) and (4),
over and mutation probabilities are adaptively set according to the the crossover and mutation probabilities do not decrease to zero
fitness of the individual by the following equations: even when the fitness of the individual is equal with the maximum
8  fitness value in the population. These probabilities allow the popu-
> 0
< k1 f max f ; f 0 P f lation to continue to evolve without stagnation. The number of gen-
Pc ¼ f max f
ð1Þ erations is used as the convergence criterion.
>
: 0

k2 ; f <f
8 
2.3.3. Fitness function
>
< k3 f max f ; f P f In GAs, the fitness value of each of individual is calculated by
Pm ¼ f max f
ð2Þ the fitness function and the optimal solution is searched on the
>
: 
k4 ; f <f basis of fitness values. Therefore, the fitness function significantly
affects on the convergence rate of GAs. The fitness function is made
where k1 , k2 , k3 and k4 are the constants belonging to (0, 1). In gen- from the objective function. One of the objective functions being
0
eral, k1 ¼ k2 ¼ 1:0 and k3 ¼ k4 ¼ 0:5. f is the better fitness value widely used in the fuel management is the maximum RPPF mini-

between parents involved crossover operation, f max and f are the mization. It means that the fuel assemblies should be arranged in
best and the average of fitness value in the population, respectively, the core in such a manner that the neutron flux is flattened as high
and f is the fitness value of the individual to be undergone muta- possible (Driscoll et al., 1990).
tion. From Eqs. (1) and (2), for individuals whose the fitness is In our study, the objective function for the LPO is to minimize
higher than the average fitness of the population, small crossover the maximum RPPF at the equilibrium of Xe under the constraint
and mutation probabilities are assigned to retain the good individ- condition which the cycle length must be satisfied. So the fitness
ual in the population carried over to the next generation. On the function is set by:
contrary, if an individual’ fitness is lower than the average fitness, fit ¼ w1
f 1 þ w2
f 2 ð5Þ
high crossover and mutation probabilities are assigned to remove
the individual. Therefore, AGA can be provided optimal and adap- where w1 and w2 are penalty factors. f 1 and f 2 , the penalty function,
tive crossover and mutation probabilities unlike CGA. However, as are,
shown in Eqs. (1) and (2), the crossover and mutation probabilities
become zero when the fitness of the individual is equal with the f 1 ¼ 1=F xy

maximum fitness value in the population. It may cause the popula- 1; C B P C B;0 ð6Þ
f2 ¼
tion to become stagnant, thus occurring premature convergence 0; C B < C B;0
easily. In order to overcome this drawback and enhance the perfor-
mance of AGA, we introduce the following equations to adjust effec- where F xy is the maximum RPPF at the equilibrium of Xe, C B;0 the
tively the crossover and mutation probabilities, P c and Pm : limitation on the boron concentration and C B the boron concentra-
8  
tion for the given LP at the end of cycle (EOC). F xy and C B are
>
< Pc;max  Pc;max expð0:382Þ f 0 f
 ; f Pf
0 obtained via the 2D diffusion-burnup calculation by TPSQ/AGANGC
Pc ¼ f max f
ð3Þ code. In this work, w1 and w2 are taken as 1.0 and 0.2, respectively.
>
: 0

Pc;max ; f <f These penalty factors were empirically determined through many
runs to result in good solution. These values can vary depending
8 
>

on the problem.
< Pm;max  Pm;max expð0:618Þ f f
 ; f Pf Finding the optimal LP is performed by applying above men-
Pm ¼ f max f
ð4Þ
>
:  tioned genetic operators and the fitness function. The flowchart
Pm;max ; f <f of the LPO using AGA is shown in Fig. 4.
4
C. So, Il-Mun Ho, Jong-Suk Chae et al. Annals of Nuclear Energy 159 (2021) 108331

Table 3
Start
The optimization results by the proposed AGA.

LP CB, ppm Fxy

Initialize population
2D 3D 2D 3D
1 16.43 18.11 1.287 1.294
Evaluation of fitness value Core physics calculation 2 21.86 20.84 1.283 1.291
3 18.33 20.12 1.269 1.272
Selection 4 22.69 24.51 1.275 1.300

Calculation of the crossover probability

No
rand<Pc?

Yes
Crossover

Calculation of the mutation probability

No
rand<Pm?

Yes
Mutation
Fig. 5. The one-eighth core configuration corresponding to the optimal LP found by
the proposed AGA.

No Is the population
size satisfied? The changes of the RPPF according to the changes of the burn-
up in the optimal LP are shown with those in the reference LP in
Yes
Fig. 6. As shown in Fig. 6, the RPPF of the initial equilibrium of
No Is the criterion
Xe for the optimal LP was decreased about 0.024 than that for
satisfied? the reference LP. The cycle burn-up of the proposed optimal LP is
13873.4 MWd/tU, which was increased than that of the reference
Yes
LP.
End To compare the performance of the proposed AGA with those of
CGA and the original AGA, 50 experiments of each algorithm were
Fig. 4. Flowchart of the LPO using AGA. carried out. The crossover probability and the mutation probability
in CGA were selected 0.6 and 0.4 respectively via several prelimi-
nary experiments. The mean and standard deviation of the maxi-
mum RPPF of the best LPs found after each experiment in each
3. Optimization result algorithm are shown in Table 4. The mean and standard deviation
of the best fitness function evaluation in each generation also are
The LPO for the first core of the 1000 MWe PWR was imple- shown in Fig. 7.
mented by employing the above described AGA. The constraint
condition for the LPO is that the cycle length must be equal to or
longer than that of the reference LP proposed by the designer. Both
the population size and the maximum generation number were
selected as 100. The core physics calculations for the reference LP
was performed before the optimization. The cycle burn-up of the
reference LP is 13764.2 MWd/tU. C B at the burn-up of
13750 MWd/tU and F xy of the reference LP obtained through 2D
and 3D core physics calculations are 11.05 ppm and 10.31 ppm,
1.301 and 1.281, respectively, therefore, C B;0 was set to
11.05 ppm. The good results by the proposed AGA were achieved
with Pc,max = 0.9 and Pm,max = 0.4. The 2D and 3D calculation results
of the obtained LPs are shown in Table 3.
As shown in Table 3, in all of obtained LPs, maximum RPPFs are
not exceeding 1.30 while cycle lengths are satisfied the constraint
condition. Considering 3D calculation results for the reference LP, it
can be known that the third LP is the optimal LP fitting for the opti-
mization target. In the proposed LP, the maximum RPPF was
decreased about 0.01 than that in the reference LP.
The one-eighth core configuration corresponding to the optimal
LP found by the proposed AGA is compared with the reference Fig. 6. The changes of the RPPF according to the changes of the burn-up in the
arrangement in Fig. 5. optimal LP and the reference LP.

5
C. So, Il-Mun Ho, Jong-Suk Chae et al. Annals of Nuclear Energy 159 (2021) 108331

Table 4 CRediT authorship contribution statement

Mean and standard deviation of the maximum RPPF.

Algorithm Mean Standard deviation Chol So: Conceptualization, Investigation, Methodology, Project
The proposed AGA 1.293 0.011 administration, Writing - original draft, Supervision, Writing -
The original AGA 1.312 0.016 review & editing. Il-Mun Ho: Investigation, Formal analysis, Vali-
CGA 1.364 0.022 dation, Resources. Jong-Suk Chae: Visualization, Writing - review
& editing. Kwang-Hak Hong: Data curation, Software.

Declaration of Competing Interest

The authors declare that they have no known competing finan-

cial interests or personal relationships that could have appeared
to influence the work reported in this paper.

References

Levine, S., 1987. In-core fuel management of four reactor types. In: Ronen, Y. (Ed.),
Handbook of Nuclear Reactors Calculations. CRC Press.
Meneses, A.A.M., Machado, M.D., Schirru, R., 2009. Particle swarm optimization
applied to the nuclear reload problem of a pressurized water reactor. Prog. Nucl.
Energy 51, 319–326.
Stevens, J.G., Smith, K.S., Rempe, K.R., 1995. Optimization of pressurized water
reactor shuffling by simulated annealingwith heuristics. Nucl. Sci. Eng. 121, 67–
88.
Lin, C., Yang, J., Lin, K., Wang, Z., 1998. Pressurized water reactor loading pattern
design using the simple tabu search. Nucl. Sci. Eng. 129, 61–71.
Hill, N.J., Parks, G.T., 2015. Pressurized water reactor in-core fuel management by
Fig. 7. Comparison of the average fitness values with different optimization tabu search. Ann. Nucl. Energy 75, 64–71.
algorithms. Machado, L., Schirru, R., 2002. The Ant-Q algorithm applied to the nuclear reload
problem. Ann. Nucl. Energy 29, 1455–1470.
De Lima, A.M.M., Schirru, R., Da Silva, F.C., Medeiros, J.A.C.C., 2008. A nuclear reactor
As shown in Table 4, the mean and standard deviation of the core fuel reload optimization using artificial antcolony connective networks.
maximum RPPF obtained by the proposed AGA is smallest. As Ann. Nucl. Energy 35 (9), 1606–1612.
shown in Fig. 7, the average fitness values in the proposed AGA Lin, C., Lin, B.F., 2012. Automatic pressurized water reactor loading pattern design
using ant colony algorithms. Ann. Nucl. Energy 43, 91–98.
increase much faster than those in other algorithms. The error bars Zhao, J., Knight, B., Ephraim, N., Soper, A., 1998. FuelGen: a genetic algorithm-based
in Fig. 7 show that the standard deviation of the results in the pro- system for fuel loading pattern design in nuclear power reactors. Expert Syst.
posed AGA is smallest throughout the search process. The results in Appl. 14, 461–470.
DeChaine, M.D., 1995. Stochastic Fuel Management Optimization Using Genetic
Table 4 and Fig. 7 clearly show that the proposed AGA reach the
Algorithms and heuristic rules Ph.D. Thesis. The Pennsylvania State
optimal solution or the neighborhoods of the optimal solution University.
much faster than CGA and the original AGA. Yilmaz, S., Ivanov, K., Levine, S., Mahgerefteh, M., 2006. Application of genetic
algorithms to optimize burnable poison placement pressurized water reactors.
Ann. Nucl. Energy 33, 446–456.
4. Conclusion Fadaei, A.H., Setayeshi, S., 2009. A new optimization method based on cellular
automata for VVER-1000 nuclear reactor loading pattern. Ann. Nucl. Energy 36,
659–667.
Finding the optimal LP is very important to supply the efficient Aghaie, M., Nazari, T., Zolfaghari, A., Minuchehr, A., Shirani, A., 2013. Investigation of
operation and the adequate safety margins of PWR. GAs have been PWR core optimization using harmony search algorithms. Ann. Nucl. Energy 57,
1–15.
widely applied to solve the LPO problem. The main disadvantage of Nazari, T., Aghaie, M., Zolfaghari, A., Minuchehr, A., Norouzi, A., 2013. WWER core
CGAs is the low local searching capability, resulting in the early pattern enhancement using adaptive improved harmony search. Nucl. Eng. Des.
convergence. To overcome this disadvantage, a new AGA for the 254, 23–32.
Poursalehi, N., 2015. Development of a new approach, evolutionary harmony search
LPO has been proposed in this paper. The tournament selection,
algorithm, for the LPO problem. Prog. Nucl. Energy 81, 78–90.
the two-point crossover and the mutation based on the randomly Mahmoudi, S.M., Aghaie, M., Bahonar, M., Poursalehi, N., 2016. A novel optimization
swapping positions between two FAs were applied. New calcula- method, Gravitational Search Algorithm (GSA), for PWR core optimization. Ann.
Nucl. Energy 95, 23–34.
tion formulae were introduced to adjust effectively the crossover
Akbari, R., Abbasi, M., Faghihi, F., Mirvakili, S.M., Mokhtari, J., 2018. A novel multi-
and the mutation probabilities. To verify the effectiveness of the objective optimization method, imperialist competitive algorithm, for fuel
proposed AGA, the first core of the 1000 MWe PWR has been cho- loading pattern of nuclear reactors. Prog. Nucl. Energy 108, 391–397.
sen. The objective function is to minimize the maximum RPPF with Khoshahval, F., Minuchehr, H., Zolfaghari, A., 2011. Performance evaluation of
PSO and GA in PWR core loading pattern optimization. Nucl. Eng. Des. 241,
the cycle length that is equal with or longer than that of the refer- 799–808.
ence LP. The critical boron concentration and the maximum RPPF François, J.L., Ortiz-Servin, J.J., Martín-del-Campo, C., Castillo, A., Esquivel-Estrada, J.,
required for the calculation of the fitness function are calculated 2013. Comparison of metaheuristic optimization techniques for BWR fuel
reloads pattern design. Ann. Nucl. Energy 51, 189–195.
using the TPSQ/AGANGC code. In the optimization, a LP satisfying Yamamoto, A., Hashimoto, H., 2002. Application of distributed genetic algorithm for
the optimization target was found. The maximum RPPF of the pro- in-core fuel optimization problems under parallel computational environment.
posed LP was reduced than that of the reference LP. The compar- J. Nucl. Sci. Technol. 39, 1281–1288.
Alim, F., Ivanov, K., Levine, S.H., 2008. New genetic algorithms (GA) to optimize PWR
ison of performances with CGA and the original AGA showed that reactors. Ann. Nucl. Energy 35, 121–131.
the new AGA can greatly improve the convergence rate of GAs in Norouzi, A., Zolfaghari, A., Minuchehr, A.H., Khoshahval, F., 2011. An enhanced
the LPO. integer coded genetic algorithm to optimize PWRs. Prog. Nucl. Energy 53, 449–
456.
Khoshahval, F., Fadaei, A.H., 2012. Application of a hybrid method based on the
Funding combination of GA and H.N.N for burnable poison placement. Ann. Nucl. Energy
47, 62–68.
Zameer, A., Mirza, S.M., Mirza, N.M., 2014. Core loading pattern optimization of a
None. typical two-loop 300 MWe PWR using Simulated Annealing (SA), novel

6
C. So, Il-Mun Ho, Jong-Suk Chae et al. Annals of Nuclear Energy 159 (2021) 108331

crossover Genetic Algorithms (GA) and hybrid GA(SA) schemes. Ann. Nucl. Amjady, N., Shirzadi, A., 2008. Unit commitment using a new integer coded genetic
Energy 65, 122–131. algorithm. Euro. Trans. Electr., Power Published online in Wiley Inter Science.
Israeli, E., Gilad, E., 2018. Novel genetic algorithm for loading pattern optimization Srinivas, M., Patnaik, L.M., 1994. Adaptive probabilities of crossover and mutation in
based on core physics heuristics. Ann. Nucl. Energy 118, 35–48. genetic algorithms. IEEE Trans. Syst. Man Cybern. 24 (4), 656–667.
Hu, Y.M., Zhao, X.F., 1998. Advanced nodal Green’s function method on Neumann Driscoll, M.J., Downar, T.J., Pilat, E.E., 1990. The Linear Reactivity Model for Nuclear
boundary condition. J. Tsinghua Univ. (Sci. Technol.) 38 (4), 17–21 (in Chinese). Fuel Management. American Nuclear Society.