Journal Pre-proof

A design on sustainable hybrid energy systems by multi-objective optimization for

aquaculture industry

Nhut Tien Nguyen, Ryuji Matsuhashi

PII: S0960-1481(20)31590-1
DOI: https://doi.org/10.1016/j.renene.2020.10.024
Reference: RENE 14314

To appear in: Renewable Energy

Received Date: 30 June 2020

Revised Date: 15 September 2020
Accepted Date: 5 October 2020

Please cite this article as: Nguyen NT, Matsuhashi R, A design on sustainable hybrid energy systems
by multi-objective optimization for aquaculture industry, Renewable Energy (2020), doi: https://
doi.org/10.1016/j.renene.2020.10.024.

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition
of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of
record. This version will undergo additional copyediting, typesetting and review before it is published
in its final form, but we are providing this version to give early visibility of the article. Please note that,
during the production process, errors may be discovered which could affect the content, and all legal
disclaimers that apply to the journal pertain.

© 2020 Published by Elsevier Ltd.

Credit Author Statement

1. Nguyen Nhut Tien: Conceptualization, Methodology, Software, Validation, Formal

analysis, Investigation, Writing - Original Draft, Visualization.
2. Ryuji Matsuhashi: Writing - Review & Editing, Supervision, Project
administration.
3. Vo Tran Thi Bich Chau: Resources, Data curation

of
ro
-p
re
lP
na
ur
Jo
 1

A Design on Sustainable Hybrid Energy

Systems by Multi-objective Optimization

for Aquaculture Industry

of
Nhut Tien Nguyen and Ryuji Matsuhashi

ro
1
2 Abstract— This paper presents an optimal design for sustainable hybrid energy systems for the
-p
3 aquaculture sector, which inherently requires intensive energy. The designed system is energized by
re
4 renewable resources to produce pure oxygen in situ through water electrolysis for oxygenation
lP

5 according to the changes of dissolved oxygen of species under culture. Moreover, the by-product
na

6 hydrogen from the electrolysis process is used to generate backup power for an eventual power

failure. The mathematical models of the system were developed for simulation and optimization to
ur

8 assess the performance of the system regarding technical, economic, and environmental aspects as
Jo

9 multi-objective functions in autonomous mode as well as on-grid mode. The merits of the proposed

10 system are demonstrated at a shrimp farm. Furthermore, the optimal results and their sensitivity

11 analysis showed that the sustainable hybrid energy system operating in grid-connected mode, which

12 possesses such attractive features as producing onsite pure oxygen for oxygenation and utilizing the

13 by-product hydrogen for generating backup power, could bring significant benefits for farmers

14 thanks to a notable reduction in the annualized cost of the system as well as CO2 emission in

This work was supported in part by the Can Tho University Improvement Project VN14-P6 through the Japanese Official Development Assistance (ODA)
Loan.
Nhut Tien Nguyen is with the Department of Electrical Engineering and Information Systems, the University of Tokyo, Tokyo 113-8654, Japan, and also with
the Department of Electrical Engineering, Can Tho University, Can Tho 900000, Vietnam (e-mail: [email protected]).
Ryuji Matsuhashi is with the Department of Electrical Engineering and Information Systems, the University of Tokyo, Tokyo 113-8654, Japan (e-mail:
[email protected]).
 2

1 comparison with the conventional system, which is powered by the national grid to run common

2 paddlewheel aerators for oxygenation.

3
4 Keywords— aeration, onsite pure oxygen, electrolyzer, renewable energy, fuel cell, by-product

5 hydrogen

6 I. INTRODUCTION

7 The aquaculture sector continues dominating aquatic food production globally and remains a growing

of
8 industry that provides a vital source of food as well as income for numerous countries, particularly in Asia

ro
9 [1]. Aquafarming production systems are diverse, ranging from extensive to intensive models. The

10

11
-p
extensive systems have the lowest stocking densities, produce low yields below 50 kg/ha/year from one or

two crops in a year, while the intensive systems typically produce yields from 5,000 to 15,000 kg/ha/year
re
12 [2]. However, the fast growth of aquaculture production has been accomplished in part through high
lP

13 intensity, which entails dramatically high energy demand.

na

14 One of the major problems farming aquatic animals is the expenses of maintaining water quality. In fact,
ur

15 the concentration of dissolved oxygen (DO) in cultivating ponds is one of the critical parameters to assure

16 the health and survivability of species under culture. Aquatic animals reared in ponds, which are quite
Jo

17 different from species living in rivers and seas, require a high amount of DO in their pond, especially in

18 intensive aquaculture systems. In order to prevent DO inadequacy and increase yields, the aeration systems

19 are employed to improve oxygen content [3], [4].

20 In general, there are three popular types of aerators, such as splasher, bubbler, and pure oxygen systems

21 [5]. Bubbler and splasher aerators are driven by motors and pumps powered by a national power grid or

22 diesel fuel in remote areas to increase the oxygen transfer rate between the air and the water pond whilst the

23 pure oxygen contact system improves DO concentration by releasing oxygen bubbles into the water. The

24 two common types of aerators utilized for aeration systems are paddlewheel and propeller-aspirator-pump

25 aerators since their oxygen transfer and water circulation are efficient [6].
 3

1 Aerators are often run continuously throughout the day, even when the DO level in the pond is saturated,

2 resulting in off-gassing DO, wasted energy, and high operation costs [3]. Therefore, to maintain suitable

3 DO and reduce power consumption, some investigations focusing on controlling aerators have been

4 considered. For instance, reference [7] showed the benefits of reducing energy consumption when aerators

5 are controlled in intermittent mode. Furthermore, the automatic control proposed in [8] and intelligent

6 control described in [9], [10] also effectively decrease the energy requirements of aeration technologies.

7 Besides, configurations of mechanical aerators have been modified to improve their aeration efficiency

of
8 and energy cost [11]. In addition, a study of different configurations of impeller aerators developed by [12]

ro
9 tried to achieve the highest aeration efficiency. On the one hand, the optimal rotation speed is taken into

10 account to evaluate the nexus between energy requirement and oxygen transfer for widely-used surface

11
-p
aerators such as paddlewheel aerators in [4], [13] and spiral aerators in [14]. On the other hand, the new
re
12 generation aerators, namely impeller in [15] and new tube aeration device in [16], have been applied in
lP

13 aquaculture and enhanced oxygenation energy-saving technology with high performance.

na

14 Moreover, to achieve energy-saving and emission reduction, renewable energy resources have been

harnessed for the aquaculture industry. The fish farm in [17], as well as the shrimp farm in [18], are
ur

15

16 supplied by photovoltaic (PV) power. Likewise, wind energy is solely employed in [19] or combined with a
Jo

17 PV system in [20] to power a farm. Besides, the utilization of power generated from fuel cells [21] and

18 biomass waste [22] has become a trend for the aquaculture sector.

19 Despite the improvements for aerators and the employment of green energy for aquaculture, the demerits

20 of conventional aeration systems are low aeration efficiency and high energy consumption due to

21 atmospheric oxygenation, which contains 21% oxygen. Using pure oxygen instead of oxygen in the

22 atmosphere for aeration systems might reduce energy demand and raise yield; nevertheless, the operation

23 cost of the system is expensive due to additional transporting costs from suppliers to customers. Utilizing

24 pure oxygen produced from water electrolysis for aeration system in aquaculture is mentioned in a few

25 works in [23], [24]. Apart from that, there is no study in which a sustainable hybrid energy system is
 4

1 proposed to produce onsite oxygen and employs by-product hydrogen to generate additional power.

2 This paper proposes an optimal design on sustainable hybrid energy systems for aquaculture, utilizing

3 renewable energy to provide pure oxygen onsite from the electrolysis process for oxygenation according to

4 the DO demand of species under culture. Moreover, the self-produced hydrogen through water electrolysis

5 is stored for backup power. The whole system is optimized by multi-objective functions, such as the

6 annualized cost of the system with the loss of power supply probability in stand-alone operation mode and

7 with carbon dioxide emission in grid-connected operation mode. Consequently, the sensitivity analysis of

of
8 optimal solutions from the proposed system is carried out by Monte Carlo simulation to make comparisons

ro
9 with a conventional system run by widely-used paddlewheel aerators in terms of the annualized cost of the

10 system and the emission of CO2.

11
-p
The rest of the paper is organized as follows. The proposed system description and its power
re
12 management strategy are illustrated in Section II. Section III presents the objective functions and
lP

13 mathematical model of the components. This is followed by the proposed multi-objective optimization and
na

14 sensitivity analysis for the hybrid energy system in Section IV. Case study and analysis of numerical results

are discussed in Section V. Finally, concluding remarks are given in Section VI.
ur

15
Jo

16 II. PROPOSED SYSTEM DESCRIPTION AND POWER MANAGEMENT STRATEGY

17 A. System Configuration

18 The sustainable hybrid energy system for aquaculture farms is illustrated in Fig. 1, which depicts the

19 designed system in constructional details. The components of the system are divided into power resources,

20 power loads, and cultured ponds. The power resources consist of wind turbines, photovoltaic (PV) arrays,

21 proton-exchange membrane (PEM) fuel cells, and the national grid as backup power in on-grid operation

22 mode. Moreover, the main power load of the system is the alkaline electrolyzer, while the baseload

23 involves illuminating loads, water treatment systems, and water pumps.

24
 5

Wind Power Solar Power Fuel Cell

~ = =
PCC ~ ~ ~

of
ro
National Grid Hydrogen Tank
~
Illuminating Load Water Pump
-p =
re
Water Treatment System
Compressor
lP

Compressor
na
ur

Electrolyzer
Jo

Cultured Ponds Microbubble System Oxygen Tank

Fig. 1. Sustainable hybrid energy system for aquaculture farms.

1 B. System Operation
 6

1 Aeration for aquaculture ponds continuously runs all day long to sustain good water quality and gain

2 biomass, although it is required more at night than in the daytime. During the day, the primary power

3 extracted from solar panels and wind turbines is converted to AC and fed into the system. The baseload is

4 supplied by stable power because the water treatment system needs a stable power to produce clean water

5 for water electrolysis and cultured ponds. Moreover, in order to provide clean water for the system, a low-

6 cost water treatment technology, which is detailed in [25], is adopted. The point-of-use disinfection

7 technology utilizing the ultraviolet tube with a cost under 50 USD could produce a high flow rate of treated

of
8 water at 5 L/min [25], which could provide sufficient and continuous amount of clean water for the

ro
9 electrolysis process. The remaining power is used to run the alkaline electrolyzer for producing oxygen in

10 situ.

11
-p
The pure oxygen from the electrolyzer is compressed via a short-term oxygen tank to the microbubble
re
12 system at the bottom of ponds that generates oxygen microbubbles for cultivated species. The aeration
lP

13 system, which is oxygenated by pure oxygen microbubbles with high oxygen absorption efficiency instead
na

14 of oxygen from the air (21%), requires less energy demand and releases fewer emissions compared to the

conventional aeration system since the compressing and the stripping of a significant volume of nitrogen
ur

15

16 from the air is eluded.

Jo

17 The fluctuation of DO in cultured ponds is caused by total respiration of sediment, plankton, and species

18 under culture, photosynthesis production, and exchanges with the atmosphere. The DO requirement during

19 daytime in the pond is low due to the oxygen supplement from photosynthesis, corresponding with low

20 power demand from the electrolyzer. Consequently, the electrolyzer is controlled to change its input power

21 between 20% and 100% of its capacity according to the amount of DO in cultured ponds [26]. As the

22 proposed system is appropriately controlled, the energy demand is reduced, leading to low operation cost.

23 Besides, the surplus electricity is sold to the national grid via point of common coupling (PCC) based on

24 feed-in-tariffs (FITs) policy.

25 Besides pumping, lighting loads, and wastewater treatment, aeration becomes really important at night
 7

1 because the DO level drops significantly during the night due to the ceased activity of photosynthesis and

2 respiration of organisms. For this reason, it is necessary to keep sufficient DO concentration to protect

3 cultured species from low-oxygen mortality. Doubtlessly, in such conditions, wind turbine systems alone

4 might not produce enough energy for the power demand of the whole system, especially the electrolyzer.

5 Due to a high amount of oxygen demand or at daylight hours when renewable resources cannot satisfy the

6 power load demand caused by the volatility of these resources, the deficiency in power is inevitable. In

7 order to compensate the energy mismatch of the system in island operation mode, when the imbalance

of
8 occurs, the by-product hydrogen from the electrolyzer, which is compressed and stored in a hydrogen tank,

ro
9 is fed into the PEM fuel cell to regenerate electric power to supply the electrolyzer and cover the baseload.

10 However, if the power demand is higher than the capacity of the fuel cell, the system is connected to the

11
-p
national grid to purchase electricity, ensuring stable operation of the system.
re
C. Power Management Strategy
lP

12

13 The proposed strategy of power management for the system is necessary for managing the power
na

14 distribution amongst the energy resources, energy storage devices, and the power loads to adapt to the
ur

15 volatile behaviors of renewable resources and load demands. The strategy is designed to meet the power
Jo

16 load demand by power generated from renewable resources as well as fuel cells, while the national grid is

Start

On-Grid
Pgs ( t ) = Pim ( t )
Y N
Pim ( t ) ≥ 0
Off-Grid

N
Pd ( t ) = Pim ( t ) PFC ( t ) = 0 CH 2 ( t ) ≥ CHmin
2
PFC ( t ) = 0 Pgp ( t ) = − Pim ( t )

N N
mHdumped
2
>0 CH 2 ( t ) ≤ CHmax
2
− Pim ( t ) ≤ PFC
rated
(t ) PFC ( t ) = PFC
rated
(t ) Pgp ( t ) = − Pim ( t ) − PFC
rated
(t )

Y Y

mHdumped
2
=Fig.
0 PFC ( t ) =management
2. Flowchart of power − Pim ( t ) strategy.
 8

1 considered as a backup system in on-grid mode. The power flow in the system is strictly controlled by

2 comparing the generated power to the power demand, as shown in Fig. 2.

3 The imbalance between power generated from renewable resources Pgen (W) and power load demand Pload

4 (W) at an instant time is expressed by:

Pload ( t )
Pim ( t ) = Pgen ( t ) − (1)
ηinv

5 where Pim is the imbalance power (W) and ηinv presents the inverter efficiency.

6 The working strategy of the hybrid system is determined by the imbalance of renewable power. If the

of
7 generated power by wind turbines and PV arrays is higher than the load demand, the excessive power is

ro
8 sold to the national grid in on-grid mode or dumped in autonomous mode. The sold and the dumped power

9 are calculated as follows:

-p
re
Pgs ( t ) = Pim ( t ) (2)
lP

Pd ( t ) = Pim ( t ) (3)
na

10 where Pgs and Pd are the power sold and dumped (W), respectively.

11 Besides, the fuel cell stops due to the surplus of electricity. The control system checks the hydrogen
ur

12 level in the storage tank. If the hydrogen level is lower than the maximum level of hydrogen CHmax (kg) in the
Jo

13 storage tank, the hydrogen flow produced from the electrolyzer is fed directly to the tank until the tank

14 reaches its capacity and the surplus hydrogen mHdumped (kg) is dumped to the temporary tank, being sold for
2

15 different applications.

16 When renewable energy might not meet load demand, the fuel cell is harnessed. Nevertheless, the

17 amount of unmet power is purchased directly from the national grid, with reliability being assumed 100%

18 when the fuel cell cannot work due to a lower level of hydrogen compared to the minimum level of

19 hydrogen CHmin in the storage tank. On the contrary, the fuel cell produces enough power to meet the
2

20 demand. Nonetheless, if the power demand is higher than the capacity of the fuel cell PFCrated (W), the power
 9

1 shortage is provided by the national grid to prevent the loss of power supply. The purchased power from

2 the national grid might be obtained by:

Pgp ( t ) = − Pim ( t ) − PFC ( t ) (4)

3 where Pgp is the power bought from the national grid (W) and PFC presents the output power of the fuel cells

4 (W).

5 III. PROBLEM FORMULATION

6 A. Objective Functions

of
7 1) Annualized Cost of System (ACS)

ro
8 The annualized cost of the system, which is composed of the annualized cost of components in the

9
-p
system CACC , the total cost of purchasing electricity CGP , the total cost of selling electricity CGS , and the total
re
10 selling cost of dumped hydrogen CHS , is computed by the following equation:
lP

ACS = CACC + CGP − CGS − CHS (5)

na

11 The annualized cost of components in the system is calculated as follows:

ur

C ACC = C NPC ⋅ CRF ( i, N ) (6)

Jo

12 where CNPC is the net present cost or life-cycle cost involving the capital cost, the cost of maintenance, the

13 cost of replacement, the depreciation tax benefit, and the salvage value incurred over the project lifetime N

14 (years) [27]. CRF ( i, N ) denotes the capital recovery factor used to convert the present value into a series of

15 equal annual cash flows, having the mathematical formulation expressed as:

i
CRF ( i, N ) = (7)
1 − (1 + i )
−N

16 where i refers to the interest rate (%).

17 The cost of power exchange between the sustainable energy system and the national grid, including the

18 cost of purchasing electricity from the grid and the cost of selling surplus electricity to the grid in a year, is

19 defined as follows:
 10

T
CGP = ∑ Pgp ( t ) ⋅ c gp (8)
t =1

T
CGS = ∑ Pgs ( t ) ⋅ c gs (9)
t =1

1 where T denotes the number of hours in a year (8760 hours), cgp and cgs are the purchasing and selling

2 electricity price ($/kWh), respectively. In stand-alone operation mode, the cost of power exchange is

3 neglected.

4 The by-product hydrogen is utilized for power generation; nevertheless, the surplus hydrogen, which is

of
5 dumped due to the limited capacity of the hydrogen tank, is sold to fertilizer factories. The cost of selling

ro
6 dumped hydrogen is given by:

T
C HS = ∑ mHdumped
2
t =1
-p
( t ) ⋅ chs (10)
re
7 where chs is the selling price of hydrogen ($/kg).
lP

8 The first objective function, which is an economic evaluation, is to minimize the ACS .

9 2) Loss of Power Supply Probability (LPSP)

na

10 The second objective function LPSP , which is defined as the ratio between the total unmet load and the
ur

11 total power load demand, is applied to minimize the loss of power supply probability according to equation
Jo

12 [28]:

∑P unmet (t )
LPSP = t =1
T (11)
∑ Pload ( t )
t =1

13 where LPSP is the loss of power supply probability used to evaluate the reliability of the system in islanded

14 operation mode due to the stochastic nature of renewable energy resources and Punmet is the unmet load at

15 each hour (W), which is given by:

0, ( P (t ) < P (t ))
Punmet ( t ) = 
load total
(12)
 Pload ( t ) − Ptotal ( t ) , ( P (t ) > P (t ))
load total

16 where Ptotal is the generated power from available sources (W).

 11

1 3) Carbon Dioxide Emission (EMI)

2 The total amount of CO2 emission from the national grid in case of grid-connected operation mode is

3 minimized by the third objective function and expressed as:

T
EMI = ∑ Pgp ( t ) ⋅ egrid (13)
t =1

4 where EMI is the amount of CO2 emission of the system during a year (t-CO2) and egrid is the grid emission

5 factor (g-CO2/kWh).

6 B. Mathematical Model of the Components in the Sustainable Hybrid Energy System

of
7 1) The External Uncertainty Models

ro
8 As a matter of fact, the meteorological factors, such as solar irradiance, ambient temperature, and wind

9
-p
speed, are sensitive to a particular topography due to their intermittent nature. As a result, the generated
re
10 power from such renewable energy resources as solar and wind power is variable and uncertain; however,
lP

11 these power resources have stochastic behavior, which could be characterized by a probabilistic distribution
na

12 function (PDF). Furthermore, to evaluate the influence degree of the interannual variations of the

13 meteorological variables on the optimization outcomes, the sensitivity analysis is conducted by Monte
ur

14 Carlo simulation based on scenarios in which the uncertainties fluctuate around the original value.
Jo

15 The stochastic uncertainty model representing deviations from the original series follows PDFs. The

16 interannual variations of average daily irradiation, average temperature, and average wind speed of a whole

17 year are described by a Gaussian PDF with mean value and standard deviation [29].

18 The uncertain scenarios having hourly series of meteorological variables in a year are generated

19 according to the random samples following the Gaussian distribution, the hourly values of each original

20 series, and the annual mean values of uncertain variables in several years, as shown in the following

21 equation [29]:

N k ( µ j , σ 2j )
θ j (t ) = ϕ j (t ) ⋅
µj
(1 ≤ k ≤ N samp ) (14)
 12

1 where θ j ( t ) is the uncertain parameter, ϕ j ( t ) denotes such time-dependent phenomena as solar irradiance

2 ( j = G ) , air temperature ( j = Ta ) , and wind speed ( j = v ) , and N samp is the number of discrete samples followed

3 by a Gaussian PDF N ( µ , σ 2 ) with annual mean µ and standard deviation σ .

4 2) Solar Photovoltaic Model (PV)

5 The output power generated hourly from the PV system PPV (W) is dependent on ambient factors (solar

6 radiation, air temperature, and wind speed) as well as the module’s performance. The mathematical

7 presentation employed to compute the output power of the system is provided by [30]:

of
 µ 9.5 ⋅ µ PV ⋅ (TNOCT − 20 ) ⋅ (1 − η PV STC
) ⋅ G t  ⋅ A ⋅ G t
PPV ( t ) = η PV ⋅ (Ta ( t ) − TSTC ) + g ( ) g ( )

ro
STC
⋅ 1 + STC
PV
(15)
 η PV 800 ⋅η PV ⋅ ( 5.7 + 3.8 ⋅ν ( t ) )
PV

STC

8 -p
where Gg refers to the global solar radiation on the tilted surface (W/m2), ηPV
STC
presents the module efficiency
re
9 at standard test condition (STC) (%), µ PV indicates the temperature coefficient of the output power (%/°C),
lP

10 TSTC refers to the STC temperature (25 °C), TNOCT denotes the nominal operating cell temperature (°C), and

2
na

11 APV represents PV module area (m ).

12 3) Wind Turbine Model (WT)

ur

13 The hourly generated power by the WT system is expressed as a function of the available wind velocity
Jo

14 at a specific hub height of the wind turbine by the following equation [31]:

0 ( v < vi or v > vo )

rated v ( t ) − vi
3 3

PWT ( t ) = ηWT ⋅ PWT ⋅ 3 ( vi ≤ v ≤ vr ) (16)
 vr − vi3
η ⋅ P rated ( vr ≤ v ≤ vo )
 WT WT

15 where PWT denotes the power produced by WT system (W), ηWT is the efficiency of the WT system, PWTrated is the

16 rated power of WT (W), and vr , vi , and vo are the rated, cut-in, and cut-out wind speed (m/s), respectively.

17 The wind speed at the desired hub height of the WT is converted from the wind speed measured at a

18 reference hub height as follows [32]:

 13

ln ( h z0 )
v ( t ) = vref ( t ) ⋅
ln ( href z0 )
(17)

1 where v is the wind speed at hub height (m/s), vref indicates the measured wind speed at reference height

2 (m/s), h denotes the WT hub height (m), href stands for the reference height (m/s), and z0 presents the surface

3 roughness length (m).

4 4) Fuel Cell Model (FC)

5 The amount of hydrogen consumed by FC depends on its output power, as shown in the following

6 expression [33]:

of
ro
α FC ⋅ PFCrated
+ β FC ⋅ PFC ( t )

( PFC ≤ Peff ⋅ PFC )
max rated

mHcons
2

( t ) = 
-p   PFC ( t ) max 
α FC ⋅ PFC + β FC ⋅ PFC ( t ) ⋅ 1 + Feff ⋅  rated − Peff  
rated
 (18)
   PFC  
re
 max rated
( Peff ⋅ PFC < PFC < PFC )
rated
lP

7 where mHcons is the hourly fuel consumption of FC (kg/h), α FC and β FC are the coefficients of a hydrogen
2
na

8 consumption curve, Peffmax is the power of maximum efficiency, and Feff is the correction factor considering the

9 high consumption above Peffmax . These values are associated with the FC efficiency ranging between 31% and
ur

10 46% [33].
Jo

11 5) Hydrogen Tank Model (HT)

12 The level of hydrogen in the storage tank is written as follows [34]:

CH2 ( t ) = CH2 ( t − 1) + mHprod

2
( t ) − mHcons2 ( t ) ηtank (19)
13 where CH ( t ) and CH ( t − 1) refer to the states of the hydrogen mass in the storage tank (kg) at the time t and t − 1
2 2

14 , respectively. mHprod is the hydrogen mass flow produced from the electrolyzer hourly (kg/h), and ηtank is the
2

15 storage efficiency.

16 6) Power Load Model (PL)

17 The primary power load of the proposed system is consumed by the electrolyzer, while a small portion of

18 power goes to the baseload. The energy required to operate the electrolyzer depends on the losses of
 14

1 oxygen caused by organisms as well as the oxygen produced through photosynthesis activity in ponds. The

2 changing rate of the DO level in a pond dDO dt (mgL-1h-1) comprising photosynthesis production PO (mgL- 2

1 -1
3 h ), total respiration of organisms RO (mgL-1h-1), and exchange with the atmosphere EO (mgL-1h-1), which
2 2

4 is disregarded due to the negligible wind speed on the pond surface, is expressed as:

dDO
= PO2 ( t ) − RO2 ( t ) ± EO2 ( t ) (20)
dt

5 The amount of oxygen produced by photosynthesis depending on ambient factors (intensity of

6 photosynthetically active solar radiation (PAR) and water temperature) and concentration of chlorophyll-a

of
7 is given as [35]:

ro
(1−α ⋅R ( t ) )
PO2 ( t ) = 9.6 ⋅ α par ⋅ RS ( t ) ⋅1.036( w ( ) ) ⋅ e par S ⋅ Chla
T t − 20

8
-p (21)

where α par indicates the ratio of PAR to broadband solar radiation of investigated sites (m2/kW), RS refers to
re
9 the broadband solar radiation (kW/m2), Tw denotes water temperature of the pond (°C), and Chla presents the
lP

10 concentration of chlorophyll-a in the pond (mg/L).

na

11 The total respiration rate in the pond involving the respiration rate of aquatic species under culture Rspecies
ur

12 (mgL-1h-1), namely fish or shrimp, sediment respiration due to decaying organic matter Rsediment (mgL-1h-1),
Jo

13 and water respiration caused by phytoplankton Rwater (mgL-1h-1) is provided by the following equation:

RO2 ( t ) = Rspecies ( t ) + Rsediment ( t ) + Rwater ( t ) (22)

14 The average value of water respiration rate ranges between 0.01 and 0.86 mgL-1h-1 while the sediment

15 respiration rate is ranging from 0.3 to 0.56 mgL-1h-1 [36]. Regarding the cultured species respiration rate,

16 the shrimp oxygen consumption rate is a function of water temperature Tw (°C), salinity S in the pond (ppt),

17 wet body weight mshrimp (g), and stocking density of shrimps Stshrimps (shrimps/m2) [37],[38] whilst the channel

18 catfish respiration rate depends on the average fish weight m fish (g) and pond water temperature [39]. The

19 expressions for calculating the respiration rate of shrimp and fish are provided by the following equations:
 15

Rshrimp ( t ) = f (Tw , S , mshrimp , Stshrimps ) (23)

R fish ( t ) = −0.999 − 9.57 ⋅ 10 −4 ⋅ m fish + 6 ⋅ 10 −7 ⋅ m 2fish + 0.0327 ⋅ Tw − 8.7 ⋅10 −6 ⋅ Tw2 + 3 ⋅10 −7 ⋅ m fish ⋅ Tw (24)
-1 -1
1 where Rshrimp and R fish are respiration rates of shrimp and fish (mgL h ), respectively.

2 The total demand of oxygen for a pond TOD (kg/h) is given by:

( )
TOD ( t ) = Ro2 ( t ) − Po2 ( t ) ⋅ V pond ⋅10 −3 (25)
3 where Vpond denotes the pond volume (m3) and 10 −3 refers to the conversion factor (kg/g).

4 The electric power acquired to run the electrolyzer to keep stable DO in the pond according to the

of
5 fluctuation of DO is computed by [24]:

ro
2 ⋅ nH 2 ⋅ F ⋅ U c ⋅ QO2 ( t )
PELZ ( t ) = (26)
-p
80.69 ⋅η F

where PELZ is the power required by electrolyzer (W), n H indicates the number of electrons transferred per
re
6 2
lP

7 hydrogen molecule, F denotes the Faraday’s number (96485 C/mol), Uc represents the cell’s actual voltage

8 (V), ηF refers to Faradic efficiency, and QO is the total oxygen flow rate produced from electrolyzer (Nm3/h).
na

9 QO2 has a relationship with TOD according to the following equation [40]:
ur

QO2 ( t ) = 0.6999 ⋅ TOD ( t ) ηµ (27)

Jo

10 where ηµ is the oxygen absorption efficiency generated by the microbubble system.

11 In order to compare the features of the proposed system with the conventional system run by mechanical

12 aerators in terms of ACS and EMI , the electrical power requirement for mechanical aerators is taken into

13 account, which is calculated by the following expression [41]:

TOD
PMA = (28)
AE
14 where PMA is electric power demand of the mechanical aerator (W) and AE refers to the aeration efficiency of

15 the aerator (kg/kWh). The actual oxygen-transfer efficiency of a mechanical aerator is about 50% of the

16 standard aeration efficiency [42].

 16

1 IV. PROPOSED MULTI-OBJECTIVE OPTIMIZATION AND SENSITIVITY ANALYSIS FOR THE DESIGNED SYSTEM

2 A. Multi-objective Optimization for the Hybrid System

3 1) Constraints

4 The optimization problem attempts to achieve the optimal values of six decision variables including tilt

5 angle βt (°), azimuth angle γ a (°), PV peak capacity PPVpeak (Wp), WT rated capacity PWTrated (Wr), FC rated capacity

rated
6 PFC (Wr), and the mass of HT mtank (kg). The constraints of the variables should be met:

0 ≤ β t ≤ 90o

of
−180o ≤ γ a ≤ 180o
0 ≤ PPVpeak ≤ PPV

ro
max

(29)
0 ≤ PWT
rated
≤ PWT
max

0 ≤ PFC
rated
-p
≤ PFC
max

0 ≤ mtank ≤ mtank
max
re
7 where PPVmax , PWTmax , PFCmax , and mtank
max
denote the maximum values of PV capacity, WT capacity, FC capacity, and
lP

8 the mass of HT, respectively.

The PV system is of the maximum value of PV capacity according to the capacity potential, which
na

10 depends on the packing factor, the available land of the investigated site, and the power over the module
ur

11 unit area [43]. Similarly, the maximum WT capacity is set to the same value as PV capacity. Besides, the
Jo

12 hydrogen storage is constrained based on its minimum and maximum capacity given by:

CHmin2 ≤ CH2 ( t ) ≤ CHmax

2
(30)
13 where CHmin and CHmax are the minimum and maximum levels of hydrogen in the storage tank, respectively.
2 2

14 2) Implementation of Non-dominated Sorted Genetic Algorithm (NSGA-II)

15 The optimization simulations are developed in the Matlab environment. NSGA-II, which is a robust

16 multi-objective evolutionary algorithm based on Pareto-optimal solutions, is utilized to find the optimal

17 system configuration [44]. A fast non-dominated sorting technique is employed in NSGA-II to sort the

18 population into various non-dominated levels, with each solution being assigned fitness, which is equal to

19 its non- dominated level [45]. Furthermore, the obtained Pareto-optimal solutions in this algorithm
 17

1 comprise a set of optimal solutions called the Pareto front.

2 The necessary steps illustrated in Fig. 3 for solving nonlinear optimal configuration of the system are

3 presented as follows:

4 Step 1: Input the primary data: annual meteorological data of the investigated location (hourly solar

5 radiation, ambient temperature, and wind speed), technical as well as economic data for components in the

6 system, emission parameters, biophysical parameters of the cultured pond, and parameters of the NSGA-II

7 algorithm.

of
ro
-p
re
lP
na
ur
Jo
 18

Start

- Meteorological data: solar radiation, air temperature, and wind speed

- Technical and economical data for components in the system
- Emission parameters of the system
- Cultured pond parameters
- NSGA-II parameters

Initialize population
( β ,γ
t a , PPVpeak , PWT
rated rated
, PFC , mtank )

PV WT FC HT PL

of
model model model model model

ro
Off-grid -p
Select operation
mode
On-grid
re
Compute objectives Compute objectives
f1 = ACS f1 = ACS
lP

f 2 = LPSP f 2 = EMI
na

Evaluate objective Constraint of

functions decision variables
ur
Jo

No Maximum number
of generations?

Yes

Generate pareto
front

End

Fig. 3. Flowchart of NSGA-II for multi-objective optimization.

1 Step 2: Produce the initial population based on decision variables (azimuth angle, tilt angle, the capacity
 19

1 of PV, wind turbine, fuel cell, and hydrogen tank), which are gene codes of each individual in the

2 population with size N pop . Then apply genetic manipulation, such as crossover and mutation mechanisms, to

3 produce the offspring population.

4 Step 3: Compute the energy output for operation in one year through the combination between the input

5 data and system models such as PV, wind turbine, fuel cell, hydrogen tank, and power load.

6 Step 4: Select operation modes and calculate objective functions such as ACS and LPSP in island mode and

7 ACS and EMI in grid-connected mode.

of
8 Step 5: Implement NSGA-II to assess objectives according to objective functions and constraints.

ro
9 Step 6: Update the initial population and repeat step 2 if the maximal number of iterations is not satisfied.

10 -p
Step 7: Sort obtained solutions and generate Pareto set for sorted non-dominated solutions when the

termination criteria are met.

re
11

12 The population size, together with the number of generations of the algorithm, is set to 200. Furthermore,
lP

13 the crossover and mutation rates of decision variables are set as 0.9 and 0.1, respectively.
na

14 B. Sensitivity Analysis by Monte Carlo Simulation

ur

15 It is necessary to carry out sensitivity analysis for optimization outcomes due to uncertainties in the input
Jo

16 variables. Nonetheless, the Monte Carlo analysis method, which often applies to assumptions for system

17 cost, is very useful when input values expose uncertainties [46]. Hence, Monte Carlo simulation is

18 conducted in this paper since it is the most suitable method for complicated systems having a variety of

19 uncertainties [47].

20 To evaluate the effect of uncertain parameters on the fitness of the system, the procedures of sensitivity

21 analysis by Monte Carlo simulation after performing deterministic optimization in Fig. 4 are shown below:

22 Step 1: Collect data of the system for Monte Carlo simulation.

23 Step 2: Perform statistical analysis study from historical and given data to obtain mean values and

24 standard deviations for several years.

 20

1 Step 3: Generate uncertainty scenarios with a total of N samp random samples from Gaussian PDFs.

2 Step 4: Calculate the output power for a one-year operation by combining the generated scenarios and

3 optimized decision variables through PV, WT, FC, HT, and PL models.

4 Step 5: Select the operation modes and compute the fitness corresponding to each mode.

5 Step 6: Keep recording results and repeat step 3 if the maximal number of iterations is not reached.

Start

of
Historical data (solar radiation, temperature, and wind speed) and given data of the system

ro
Generate random samples of uncertain parameters by PDFs

Solar radiation
θG ( t )
Wind speed
θv ( t )
-p Air temperature
θTa ( t )
Optimized
decision variables
re
lP

PV WT FC HT PL
model model model model model
na

Off-grid Select operation On-grid

mode
ur

Compute objectives Compute objectives

f1 = ACS f1 = ACS
f 2 = LPSP f 2 = EMI
Jo

Record results

No Maximum number
of samples?

Yes

Show results

End

Fig. 4. Framework of scenario generation by Monte Carlo simulation for sensitivity analysis.
6 Step 7: Extract results for sensitivity analysis when the stopping criteria are met.
 21

1 V. RESULTS AND ANALYSIS OF CASE STUDY

2 A. Case Data Introduction and Parameter Settings

3 The proposed sustainable hybrid energy system and its optimal design are applied to a shrimp farm

4 located at Vinh Chau, Vietnam, which is situated 9.42° north latitude, 106.7° east longitude, and -8 m

5 elevation. The investigated shrimp farm consists of three grow-out ponds, one pond for storing water, and

6 another pond for water treatment. Each grow-out pond is of a depth of 1.2 m and a surface area of 1400 m2.

of
ro
-p
re
lP
na
ur
Jo
 22

1 According to Köppen-Geiger climatic chart [48], the location of the shrimp farm has the equatorial

2 savannah climate with dry winter (Aw), with an apparent seasonal variation in precipitation between dry

3 and rainy seasons in which the former starting from November to April and the latter lasting from May to

4 October. The meteorological data, which includes solar irradiance, air temperature, and wind velocity

5 captured by a sample per hour precision in a typical year (8760 hours), is collected from the local weather

6 station for simulation optimization and shown in Fig. 5. The solar irradiance is somewhat constant during

of
ro
-p
re
lP
na
ur
Jo
 23

1 the year with a daily mean value of 4.8 kWh/m2 [49]. The average value of the air temperature is about 27

2 °C. Moreover, the wind speed in this region is abundant at 80 m height, with its value averaging at 7 m/s

3 [50].

4 The rearing period of shrimp at the implemented farm lasts about three months. The primary power at the

5 farm during the cultivating period is consumed by electrolyzer, which is employed to produce pure oxygen

6 for aerating the cultured ponds, at 80%, leaving the baseload at 20%.

7 TABLE I

of
1200
1000

ro
800
600
400
200
-p
re
0
0 1000 2000 3000 4000 5000 6000 7000 8000 9000
lP
na

31
30
29
ur

28
27
26
Jo

25
24
23
0 1000 2000 3000 4000 5000 6000 7000 8000 9000

18
16
14
12
10
8
6
4
2
0
0 1000 2000 3000 4000 5000 6000 7000 8000 9000

Fig. 5. Annual solar irradiance (a), air temperature (b), and wind speed at 80 m height (c).
 24

1 BIOPHYSICAL CHARACTERISTICS OF GROW-OUT PONDS

Quantity Value Refs
Number of ponds 3
Surface area of grow-out pond (m2) 1400
Depth of water (m) 1.2
Salinity (ppt) 1
Average water temperature (°C) 30
Air pressure (atm) 1
Average shrimp weight (g) 30
Stocking density (shrimps/m2) 200
Average water respiration rate (mgL-1h-1) 0.4 [36]
Average sediment respiration rate (mgL-1h-1) 0.43 [36]
Oxygen absorption efficiency of bubble system (%) 95 [51]
Ratio of PAR to broadband solar radiation (m2/W) 0.002 [35]
Average chlorophyll-a concentration (mg/L) 0.1 [52]

of
ro
-p
re
lP
na
ur
Jo

Fig. 6. The relationship between oxygen production and consumption rate with power demand at the farm.
2 The total oxygen requirement in grow-out ponds, which have biophysical characteristics presented in

3 Table 1, determines the operation mode of the electrolyzer. The power consumption of the electrolyzer

4 changes during the day according to losses of oxygen due to the respiration of organisms in ponds and

5 oxygen production by photosynthesis, reaching the highest value at night when the photosynthesis activity

6 stops as illustrated in Fig. 6. Furthermore, the power consumption of two compressors for oxygen and

7 hydrogen tanks is included in the electrolyzer consumption since the operation of two compressors depends

8 on the flow rates of oxygen and hydrogen produced from the electrolyzer. Thus, the load profile of the

9 system, followed by the pattern of the electrolyzer, demands more electricity at night.

10 A summary of the design parameters of the models of components in the hybrid system for simulation
 25

1 optimization is given in Table 2.

2 TABLE II
3 PARAMETERS OF THE MODELS OF COMPONENTS IN THE SYSTEM
Component Parameter Value Refs
PV system Standard test condition efficiency (%) 17
(Polycrystalline silicon type) Temperature coefficient of open circuit (%/°C) -0.32
Voltage at point of maximum power (V) 37.8
Standard test condition temperature (°C) 25
Nominal operation cell temperature (°C) 45
Area of module (m2) 1.94
Maximum power (W) 330
Wind turbine system Cut-in wind speed (m/s) 2.5
Rated wind speed (m/s) 10

of
Cut-out wind speed (m/s) 25
Hub height (m) 69
α FC (kg/kWh)

ro
Fuel cell 0.004 [33]
β FC (kg/kWh) 0.05 [33]

Feff (years)
-p
Output power at maximum efficiency (%) 20
1
[33]
[33]
re
Hydrogen tank Storage efficiency (%) 95 [53]
Electrolyzer Average cell voltage (V) 1.9 [54]
Faradic efficiency (%) 95 [54]
lP

Inverter Efficiency (%) 98

4 The specifications of components in the system at the shrimp farm for techno-economic optimization are
na

5 described in Table 3. The project lifetime of the system is assumed 20 years based on the longest lifespan
ur

6 of the component in the system. Therefore, some components in the system, such as electrolyzer, fuel cell,
Jo

7 and inverter, have to be replaced several times over the lifetime period according to the lifespan of these

8 components, resulting in additional cost in the overall expense of the system.

10

11

12

13 TABLE III
14 PARAMETER SETTINGS FOR SIMULATION OPTIMIZATION
Parameter Value Refs
Initial cost of PV ($/kW) 1000 [55]
Initial cost of wind turbine ($/kW) 1980 [55]
Initial cost of electrolyzer ($/kW) 1200 [56]
 26

1 Initial cost of fuel cell ($/kW) 3000 [57]

2 Initial cost of hydrogen tank ($/kg) 1000 [53]
Initial cost of inverter ($/kW) 90 [58]
3 Initial cost of mechanical aerator ($/kW) 440 [41]
4
Purchase price of electricity from grid ($/kWh) 0.0804
Sale price of electricity to grid ($/kWh) 0.0935 [59]
5 Sale price of hydrogen ($/kg) 2 [60]
CO2 emission factor from grid (g/kWh) 636 [61]
6
Project lifetime (years) 20
7 PV system lifetime (years) 20 [62]
Wind turbine lifetime (years) 20 [63]
8 Electrolyzer lifetime (years) 10 [56]
9 Fuel cell lifetime (years) 5 [57]
Hydrogen tank lifetime (years) 20 [53]
10 Inverter lifetime (years) 15 [64]
Mechanical aerator lifetime (years) 4 [41]

of
11
Tax rate (%) 10 [65]
12 Interest rate (%) 6 [65]

ro
Maintenance rate of PV system (%) 1.5 [55]
13
Maintenance rate of wind turbine (%) 1.8 [55]
14

15
-p
Maintenance rate of electrolyzer (%)
Maintenance rate of fuel cell (%)
Maintenance rate of hydrogen tank (%)
2.5
5.8
1
[56]
[57]
[53]
re
16 Maintenance rate of mechanical aerator (%) 10 [41]
17
lP

18
19
20
na

21
22
23 B. Results and Analysis
ur

24 1) Optimization Results in Stand-Alone Mode

Jo

25 The relationship between LPSP and ACS in stand-alone operation mode, which is powered by PV arrays

26 and WT system with and without energy storage, is depicted in Fig. 7. This figure presents the annualized

27 cost of the system at various reliability levels. The LPSP value decreases dramatically with the rise of ACS .

28 The lowest value of LPSP is 0 indicating the highest reliability whilst the value of 1 shows that the load

29 demand is never satisfied.

 27

LPSP (%)

of
ro
-p
re
Fig. 7. Pareto fronts in stand-alone mode.
lP

1 As shown in Fig. 7, the system without energy storage has the minimum value of LPSP at about 28%,

2 while this value of the system supported by energy storage is at about 20%. On the one hand, the reliability
na

3 of the system supported by energy storage, i.e., fuel cell and hydrogen storage, is improved slightly, at
ur

4 about 8% due to the low efficiency of FC. The reason for this is that the hydrogen fuel, which is directly
Jo

5 converted into electricity by FC for backup power, is the by-product through water electrolysis. Apart from

6 that, the production flow rate of hydrogen depends on the operation of the electrolyzer, which is used to

7 produce the main-product oxygen according to DO changes in cultured ponds to provide aeration for

8 cultivated species. On the other hand, the system cost corresponding to LPSPmin (20%) is about 416,620

9 USD, which is considerably higher than the system without energy storage (273,933 USD) due to the

10 additional expense of the energy storage system, particularly the four-time replacement of FC during the

11 project lifetime.

12 Moreover, the process of simulation optimization is explained specifically by analyzing the relationship

13 between the available capacity of power resources and LPSP , as shown in Fig. 8. The rise in the capacity of
 28

1 the PV system and the WT system significantly improves the reliability of the system. However, when the

2 value of LPSP is lower than 30%, LPSP decreases fractionally since the capacity of PV and WT system

of
ro
-p
re
lP
na

Fig. 8. The relationship between capacity of PV and WT with LPSP.

3 approaches the limit. The maximum available value of WT is higher than that of PV (1,000 kWr and 704
ur

4 kWp, respectively) because the system demands more power at night than the daytime. Also, the moderate
Jo

5 wind velocity at the farm might not produce sufficient energy for the system at night; hence, the reliability

6 of the system could not be improved dramatically even the WTs reach the maximal rated capacity.

7 Besides, when the energy storage is introduced in the system, the relationship between the capacity of PV

8 and WT with FC is presented in Fig. 9. The capacity of the FC is affected by the capacity of the PV system

9 and the WT system. The low capacity of the power resources often coincides with the small capacity of the

10 FC. As shown in Fig. 9, although the capacity of the power resources initially increases dramatically, the

11 capacity of the FC is still very low since the hydrogen is produced at a low rate with the low capacity of

12 power resources. After the capacity of the power resources reaches 100 kWp for the PV system and 200

13 kWr for the WT system, the FC capacity starts to increase. Afterward, the capacity of the FC grows
 29

1 substantially due to the increase in the capacity of the PV and WT systems. Furthermore, the FC capacity

2 cannot rise beyond 70 kWr because the PV and WT systems reach the maximum available values.

of
ro
-p
re
lP
na
ur

Fig. 9. The relationship between capacity of PV and WT with FC.

Jo

5 Similarly, the influence of FC capacity on values of LPSP and ACS is illustrated in Fig. 10. The utilization

6 of FC to provide backup power generated from by-product hydrogen could improve the reliability of the

7 hybrid system. It can be seen in Fig. 10 that the LPSP initially decreases dramatically, although the capacity

8 of FC is still very low. The cause of this decline is the increase in the capacity of renewable energy

9 resources. Nevertheless, when the renewable resources close to their upper boundary, the value of LPSP

10 continues to decline gradually with the sharp rise of FC capacity, reaching 20% with the maximal values of

11 ACS and FC capacity at about 416,620 USD and 67 kWr, respectively. The system cost witnesses a

12 dramatic increase due to the high capital cost of FC and associated HT together with the four-time

13 replacement cost of FC.

 30

LPSP (%)

of
ro
-p
re
Fig. 10. The relationship between FC capacity and LPSP with ACS.
lP

1 Undoubtedly, the power system having the minimum value of LPSP at 20% could not meet the security of

the proposed system. Likewise, the previous study carried out by the authors’ own in [24] showed that the
na

3 hybrid energy system in islanded operation mode comprising of the PV system, WT system, and battery
ur

4 with the support of backup power from a diesel generator could satisfy the power demand at the shrimp
Jo

5 farm. Although the system in the islanded mode in [24] could provide the high reliability for the shrimp

6 farm with low gas emissions, the system cost is considerably higher than that of the grid-connected system,

7 which could produce revenues by selling not only hydrogen but also the surplus electricity to the national

8 grid. By contrast, the islanded system could only produce profits from selling hydrogen while the excess

9 energy is stored in the battery to supply power for the system anytime when renewable power is not

10 available.

11 In general, the designed system for the aquaculture industry necessitates stable, sufficient, and secure

12 power resources, which could supply power continuously without lengthy blackout; therefore, the proposed

13 system energized by renewable resources and supported by FC as primary backup must be connected to the
 31

1 national power system to satisfy the power shortage.

2 2) Optimization Results in Grid-Connected Mode

3 When the proposed system operates in connection with the national power grid, the carbon dioxide

4 emission from the grid is considered as the objective function EMI besides ACS . The Pareto front optimized

of
ro
-p
re
lP
na
ur
Jo

Fig. 11. Pareto front for on-grid operation mode.

5 in on-grid mode with the total support from the national grid as the secondary backup power system is

6 presented in Fig. 11. The increase in the ACS leads to a significant decrease in CO2 emission, reaching the

7 minimal value at 124.32 tCO2 with the ACS at 152,386 USD. Nevertheless, the ACS has the negative values

8 indicating that the proposed system is able to procure revenues from selling both surplus electric power to

9 the national grid and dumped hydrogen to local fertilizer factories.

10 The optimal solutions from the Pareto front are divided into two groups. The group with orange solutions

11 in Fig. 11 has ACS values lower than ACS of the conventional system (CS) run by paddlewheel aerators. On

12 the contrary, the costs of the group with purple solutions are higher than those of the CS.
 32

1 A comparison amongst two solutions of interest from orange solutions and the CS powered by the

2 national grid is illustrated in Table 4. Solution 1 (Sol 1) corresponds to the lowest value of ACS and the

3 highest amount of CO2 emission among orange solutions, while Sol 2 is the intermediate solution, of which

4 values are nearly equal to the average values of all orange solutions in terms of ACS and EMI . However, in

5 real applications, designers could choose one of the orange solutions based on their preferences.

6 The capacity of PV and WT of Sol 1 reaches the maximum available values while the FC is not

7 employed for backup power, which is purchased from the national grid totally. Thus, besides selling

of
8 surplus electric power to the national grid, all of the dumped hydrogen from the electrolysis process is

ro
9 collected and sold to fertilizer plants, producing revenues for farmers about 9,112 USD in a year. By

10 contrast, the amount of emission released from the grid is high (231.35 tCO2) compared to Sol 2 (208.01
-p
11 tCO2).
re
12 TABLE IV
lP

13 CHARACTERISTICS OF TWO SOLUTIONS OF INTEREST AND CONVENTIONAL AERATION SYSTEM

Sol 1 Sol 2 CS
na

PV capacity (kWp) 998.65 998.62 -

Wind turbine capacity (kWr) 999.09 998.96 -
Fuel cell capacity (kWr) 0.06 12.30 -
ur

Hydrogen tank capacity (kg) 0.51 12.25 -

ACS (USD) -9,112 7,160 35,090
EMI (tCO2) 231.35 208.01 261.96
Jo

14 With the utilization of FC as backup power in Sol 2, the amount of CO2 emission declines considerably

15 in comparison with Sol 1 and CS (23.34 tCO2 and 53.95 tCO2, respectively). The utilization of FC often

16 coincides with HT. In combination, the extra cost associated with such configuration is inevitable.

17 Although the system cost of Sol 2 is considerably higher than Sol 1, it is still significantly lower than CS

18 (7,160 USD and 35,090 USD, respectively); hence, farmers could save about 27,930 USD per year.

19 3) Sensitivity of Deterministic Solution

20 The influence of uncertainties of meteorological variables on optimal solutions is assessed by Monte

21 Carlo simulation based on uncertain scenarios. A number of 1000 random samples are generated according

22 to Gaussian PDFs for iteratively simulating the proposed system.

 33

1 The probabilistic simulation is performed over the intermediate Sol 2. The PDFs illustrated in Fig. 12

2 present the variability of Sol 2 for the ACS and the EMI . The normal curves shown in Fig. 12 best fit the

3 PDFs.

90
80
70
60
50
40
30
20

of
10
0
-15 -10 -5 0 5 10 15 20 25 30

ro
90
80
70
-p
re
60
50
40
lP

30
20
10
0
na

190 195 200 205 210 215 220 225

Fig. 12. PDFs of the results of ACS (a) and EMI

ur

(b).
Jo

4 The average values, standard deviations, minimum, and maximum values of the ACS and the EMI are

5 given in Table 5. The values of ACS range between -12,273 USD and 28,677 USD, which mean that farmers

6 could get revenues or pay high annual cost according to the uncertainties of meteorology. Nonetheless, the

7 maximum values of ACS with substantially low probability are lower than ACS of the CS. Furthermore, the

8 amount of CO2 emission fluctuates between 193.87 tCO2 and 222.84 tCO2, which is significantly lower

9 than EMI of the CS.

10 TABLE V
11 RESULTS OF SENSITIVITY ANALYSIS FOR SOLUTION 2
Mean Standard deviation Min Max
ACS (USD) 7,457 6,712 -12,273 28,677
EMI (tCO2) 208.26 4.60 193.87 222.84
 34

1 VI. CONCLUSION

2 An optimal design on a sustainable hybrid energy system for the aquaculture industry is proposed in this

3 work. The designed system is quite different from the traditional systems for aquaculture in the

4 configuration as well as operation. In order to assess performance and validate results, the simulation and

5 optimization models were developed. The optimal results of the proposed system were compared with the

6 CS regarding technical, economic, and environmental aspects. An excellent avouchment found in optimal

7 outcomes and sensitivity analysis shows that the proposed system connecting to the national grid is the best

of
8 solution instead of operating in islanded mode due to strictly-needed reliability. Moreover, when properly

ro
9 designed and operated, the proposed system energized by green energy resources could produce pure

10 -p
oxygen in situ for oxygenation based on the DO demand of species under culture and the by-product

hydrogen through the electrolysis process to generate backup power presents technical, economic, and
re
11

12 environmental benefits in comparison with the CS. In other words, farmers could save operation costs from
lP

13 reducing energy demand and contribute to environmental protection by limiting the release of CO2
na

14 emission for their system in the aquaculture industry. Moreover, the outcomes from the paper can be
ur

15 guidelines to instruct optimal design and operation for aquaculture farms not only in Vietnam but also in
Jo

16 other countries.

17 REFERENCES
18
19 [1] F. A. O. U. Nations, "Global fish economy," in GLOBEFISH Highlights October 2019 ISSUE, with Jan. – Jun. 2019 Statistics: A quarterly update on
20 world seafood markets. Rome, Italy: Food & Agriculture Org., 2020, pp. 11-12.
21 [2] A. W. Fast, "CHAPTER 14 - PENAEID GROWOUT SYSTEMS: AN OVERVIEW," in Marine Shrimp Culture, A. W. Fast and L. J. Lester Eds.
22 Amsterdam: Elsevier, 1992, pp. 345-353.
23 [3] C. E. Boyd, "Pond water aeration systems," Aquacultural Engineering, vol. 18, no. 1, pp. 9-40, 1998/07/01/ 1998.
24 [4] J. M. Moore and C. E. Boyd, "Design of small paddle wheel aerators," Aquacultural Engineering, vol. 11, no. 1, pp. 55-69, Jan 1992.
25 [5] C. E. Boyd, E. L. Torrans, and C. S. Tucker, "Dissolved oxygen and aeration in ictalurid catfish aquaculture," Journal of the World Aquaculture
26 Society, vol. 49, no. 1, pp. 7-70, Feb 2018.
27 [6] C. E. Boyd, E. Jory Darryl, G. W. Chamberlain, and A. Global Aquaculture, "Aerated pond management," in Operating Procedures for Shrimp
28 Farming: Global Shrimp OP Survey Results and Recommendations, 1st ed. St. Louis, MO, USA: Global Aquaculture Alliance, 2006, pp. 68-75.
29 [7] D. Zhu, X. Cheng, D. J. Sample, and M. N. Yazdi, "Effect of intermittent aeration mode on nitrogen concentration in the water column and sediment
30 pore water of aquaculture ponds," Journal of Environmental Sciences, vol. 90, pp. 331-342, 2020.
31 [8] R. H. Hoagland, D. B. Rouse, D. Teichert-Coddington, and C. E. Boyd, "Evaluation of automated aeration control in shrimp ponds," Journal of
32 Applied Aquaculture, vol. 11, no. 3, pp. 45-55, Sep 2001.
33 [9] F. C. Cruz, K. Mahmudov, A. Marouchos, and A. Bilton, "A Feasibility Study on the Benefits of Feedback Aerator Control in Precision Aquaculture
34 Applications for the Developing World," in International Design Engineering Technical Conferences and Computers and Information in Engineering
35 Conference, 2019, vol. 59193: American Society of Mechanical Engineers, p. V02BT03A004.
36 [10] H. Deng, L. Peng, J. Zhang, C. Tang, H. Fang, and H. Liu, "An intelligent aerator algorithm inspired-by deep learning," Mathematical biosciences
37 and engineering: MBE, vol. 16, no. 4, pp. 2990-3002, 2019.
38 [11] S. Thakre, L. Bhuyar, and S. Deshmukh, "Effect of different configurations of mechanical aerators on oxygen transfer and aeration efficiency with
39 respect to power consumption," International Journal of Aerospace and Mechanical Engineering, vol. 2, no. 2, pp. 100-108, Feb 2008.
 35

1 [12] M. Adel, M. R. Shaalan, R. M. Kamal, and D. S. El Monayeri, "A comparative study of impeller aerators configurations," Alexandria Engineering
2 Journal, vol. 58, no. 4, pp. 1431-1438, 2019.
3 [13] S. M. Roy, S. Moulick, C. K. Mukherjee, and B. C. Mal, "Effect of rotational speeds of paddle wheel aerator on aeration cost," American Research
4 Thoughts, vol. 2, no. 1, pp. 3069-3087, Nov 2015.
5 [14] S. M. Roy, S. Moulick, and B. C. Mal, "Design characteristics of spiral aerator," Journal of the World Aquaculture Society, vol. 48, no. 6, pp. 898-
6 908, Dec 2017, doi: doi:10.1111/jwas.12410.
7 [15] J. Marappan et al., "Assessment of the new generation aeration systems efficiency and water current flow rate, its relation to the cost economics at
8 varying salinities for Penaeus vannamei culture," Aquaculture Research, 2020.
9 [16] C. Zhang, B. Song, J. Shan, Q. Ni, F. Wu, and S. Wang, "Design and optimization of a new tube aeration device," Aquaculture International, pp. 1-
10 15, 2020.
11 [17] F. Hahn and R. Pérez, "Design and evaluation of a dissolved oxygen controller for solar powered fish tanks," British Journal of Applied Science &
12 Technology, vol. 5, no. 2, pp. 173-188, Jan 2015.
13 [18] P. E. Campana, L. Wästhage, W. Nookuea, Y. Tan, and J. Yan, "Optimization and assessment of floating and floating-tracking PV systems integrated
14 in on-and off-grid hybrid energy systems," Solar Energy, vol. 177, pp. 782-795, Jan 2019.
15 [19] K. Mahmudov, A. Mahmoud, S. Sur, F. C. Cruz, and A. M. Bilton, "Feasibility of a wind-powered aeration system for small-scale aquaculture in
16 developing countries," Energy for Sustainable Development, vol. 51, pp. 40-49, Aug 2019.
17 [20] W. Nookuea, P. E. Campana, and J. Yan, "Evaluation of solar PV and wind alternatives for self renewable energy supply: case study of shrimp
18 cultivation," Energy Procedia, vol. 88, pp. 462-469, Jun 2016.
19 [21] D. M. Atia, F. H. Fahmy, N. M. Ahmed, and H. T. Dorrah, "A new control and design of PEM fuel cell system powered diffused air aeration system,"
20 TELKOMNIKA Indonesian Journal of Electrical Engineering, vol. 10, no. 2, pp. 291-302, Apr 2012.
21 [22] Y. Shiratori et al., "Biogas power generation with SOFC to demonstrate energy circulation suitable for Mekong Delta, Vietnam," Fuel Cells, vol. 19,
22

of
no. 4, pp. 346-353, Apr 2019.
23 [23] N. N. Tien, R. Matsuhashi, and V. T. T. B. Chau, "A sustainable energy model for shrimp farms in the Mekong Delta," Energy Procedia, vol. 157, pp.
24 926-938, Jan 2019.
25 [24] N. T. Nguyen and R. Matsuhashi, "An Optimal Design on Sustainable Energy Systems for Shrimp Farms," IEEE Access, vol. 7, pp. 165543-165558,

ro
26 2019.
27 [25] S. A. Brownell et al., "Assessment of a low-cost, point-of-use, ultraviolet water disinfection technology," Journal of water and health, vol. 6, no. 1,
28 pp. 53-65, 2008.
29
30
31
[26]

[27]
-p
S. Pascuzzi, A. Anifantis, I. Blanco, and G. Scarascia Mugnozza, "Electrolyzer performance analysis of an integrated hydrogen power system for
greenhouse heating. A case study," Sustainability, vol. 8, no. 7, pp. 629-643, Jul 2016.
M. Campbell, P. Aschenbrenner, J. Blunden, E. Smeloff, and S. Wright, "The drivers of the levelized cost of electricity for utility-scale
re
32 photovoltaics," White Paper: SunPower Corporation, 2008.
33 [28] B. Ai, H. Yang, H. Shen, and X. Liao, "Computer-aided design of PV/wind hybrid system," Renewable energy, vol. 28, no. 10, pp. 1491-1512, 2003.
34 [29] R. Dufo-Lopez, I. R. Cristobal-Monreal, and J. M. Yusta, "Stochastic-heuristic methodology for the optimisation of components and control variables
lP

35 of PV-wind-diesel-battery stand-alone systems," Renewable Energy, vol. 99, pp. 919-935, 2016.
36 [30] J. A. Duffie and W. A. Beckman, "Design of photovoltaic systems," in Solar Engineering of Thermal Processes, 4th ed. Hoboken, NJ, USA: John
37 Wiley & Sons, 2013, pp. 745-773.
38 [31] R. Chedid, H. Akiki, and S. Rahman, "A decision support technique for the design of hybrid solar-wind power systems," IEEE Transactions on
39 Energy Conversion, vol. 13, no. 1, pp. 76-83, Mar 1998.
na

40 [32] J. F. Manwell, J. G. McGowan, and A. L. Rogers, "Wind characteristics and resources," in Wind Energy Explained, 2nd ed. Chichester, United
41 Kingdom: John Wiley & Sons, 2010, pp. 23-89.
42 [33] R. Dufo-Lopez and J. L. Bernal-Agustín, "Multi-objective design of PV–wind–diesel–hydrogen–battery systems," Renewable energy, vol. 33, no. 12,
43 pp. 2559-2572, 2008.
ur

44 [34] A. K. Kaviani, G. Riahy, and S. M. Kouhsari, "Optimal design of a reliable hydrogen-based stand-alone wind/PV generating system, considering
45 component outages," Renewable energy, vol. 34, no. 11, pp. 2380-2390, 2009.
46 [35] Z. Xu and Y. Xu, "A deterministic model for predicting hourly dissolved oxygen change: development and application to a shallow eutrophic lake,"
Jo

47 Water, vol. 8, no. 2, pp. 41-55, Jan 2016.

48 [36] A. W. Fast and C. E. Boyd, "Water circulation, aeration and other management practices," in Marine Shrimp Culture, A. W. Fast and L. J. Lester
49 Eds., 1st ed. Amsterdam, The Netherlands: Elsevier, 1992, pp. 457-495.
50 [37] C. Bett and L. Vinatea, "Combined effect of body weight, temperature and salinity on shrimp Litopenaeus vannamei oxygen consumption rate,"
51 Brazilian Journal of Oceanography, vol. 57, no. 4, pp. 305-314, Dec 2009.
52 [38] L. Vinatea, W. Muedas, and R. Arantes, "The impact of oxygen consumption by the shrimp Litopenaeus vannamei according to body weight,
53 temperature, salinity and stocking density on pond aeration: a simulation," Acta Scientiarum. Biological Sciences, vol. 33, no. 2, 2011.
54 [39] J. W. Andrews and Y. Matsuda, "The influence of various culture conditions on the oxygen consumption of channel catfish," Transactions of the
55 American Fisheries Society, vol. 104, no. 2, pp. 322-327, 1975.
56 [40] J. Ivy, "Summary of electrolytic hydrogen production: milestone completion report," National Renewable Energy Lab., Golden, CO, USA,
57 NREL/MP-560-36734, Sep 2004.
58 [41] A. Kumar, S. Moulick, and B. C. Mal, "Selection of aerators for intensive aquacultural pond," Aquacultural Engineering, vol. 56, pp. 71-78, Sep
59 2013.
60 [42] C. E. Boyd and C. S. Tucker, "Aeration," in Pond Aquaculture Water Quality Management. Boston, MA: Springer US, 1998, pp. 306-353.
61 [43] S. Chen et al., "The potential of photovoltaics to power the belt and road initiative," Joule, vol. 3, pp. 1-18, Aug 2019.
62 [44] K. Deb, "Elitist multi-objective evolutionary algorithms," in Multi-Objective Optimization Using Evolutionary Algorithms, 1st ed. Chichester, United
63 Kingdom: Wiley, 2009, pp. 239-288.
64 [45] Y. Katsigiannis, P. Georgilakis, and E. Karapidakis, "Multi-objective genetic algorithm solution to the optimum economic and environmental
65 performance problem of small autonomous hybrid power systems with renewables," IET Renewable Power Generation, vol. 4, no. 5, pp. 404-419,
66 2010.
67 [46] P. Vithayasrichareon and I. F. MacGill, "A Monte Carlo based decision-support tool for assessing generation portfolios in future carbon constrained
68 electricity industries," Energy policy, vol. 41, pp. 374-392, 2012.
69 [47] M. Perninge and C. Hamon, "A stochastic optimal power flow problem with stability constraints—Part II: The optimization problem," IEEE
70 Transactions on Power Systems, vol. 28, no. 2, pp. 1849-1857, 2012.
71 [48] M. Kottek, J. Grieser, C. Beck, B. Rudolf, and F. Rubel, "World map of the Köppen-Geiger climate classification updated," Meteorologische
72 Zeitschrift, vol. 15, no. 3, pp. 259-263, 2006.
73 [49] J. Polo et al., "Solar resources and power potential mapping in Vietnam using satellite-derived and GIS-based information," Energy Conversion and
74 Management, vol. 98, pp. 348-358, Jul 2015.
75 [50] S. Joost and A. Eric, "Wind Energy Potential Vietnam," Netherlands Enterprise Agency, The Netherlands, RVO-015-1901-RP-INT, 2018.
 36

1 [51] K. Muroyama, K. Imai, Y. Oka, and J. i. Hayashi, "Mass transfer properties in a bubble column associated with micro-bubble dispersions," Chemical
2 Engineering Science, vol. 100, pp. 464-473, Aug 2013.
3 [52] C. E. Boyd and C. S. Tucker, "Ecology of aquaculture ponds," in Pond aquaculture water quality management: Springer, 1998, pp. 8-86.
4 [53] Z. Huang et al., "Modeling and multi-objective optimization of a stand-alone PV-hydrogen-retired EV battery hybrid energy system," Energy
5 conversion and management, vol. 181, pp. 80-92, 2019.
6 [54] J. Wanjiku, M. Khan, P. Barendse, and A. Sebitosi, "Analytical sizing of an electrolyser for a small scale wind electrolysis plant," in 2010 IEEE
7 International Energy Conference, 2010: IEEE, pp. 10-15.
8 [55] N. Q. Khanh, "Analysis of future generation capacity scenarios for Vietnam," in "Green Innovation and Development Centre (GreenID), Vietnam,"
9 GreenID., Hanoi, Vietnam, Oct 2017.
10 [56] A. Buttler and H. Spliethoff, "Current status of water electrolysis for energy storage, grid balancing and sector coupling via power-to-gas and power-
11 to-liquids: A review," Renewable and Sustainable Energy Reviews, vol. 82, pp. 2440-2454, Feb 2018.
12 [57] S. M. Hakimi, A. Hasankhani, M. Shafie-khah, and J. P. Catalão, "Optimal sizing and siting of smart microgrid components under high renewables
13 penetration considering demand response," IET Renewable Power Generation, vol. 13, no. 10, pp. 1809-1822, 2019.
14 [58] R. Fu, R. M. Margolis, and D. J. Feldman, "US solar photovoltaic system cost benchmark: Q1 2018," National Renewable Energy Lab., Golden, CO,
15 USA, NREL/TP-6A20-72399, Nov 2018.
16 [59] Danmark and Energistyrelsen, "Overview of Vietnam energy sector," in Vietnam Energy Outlook Report 2017, 1st ed. Hanoi, Vietnam: Danish Energy
17 Agency, 2018, pp. 11-50.
18 [60] J. Armijo and C. Philibert, "Flexible production of green hydrogen and ammonia from variable solar and wind energy: Case study of Chile and
19 Argentina," International Journal of Hydrogen Energy, vol. 45, no. 3, pp. 1541-1558, 2020.
20 [61] W. Christine, "Situation analysis of the Vietnamese electricity sector," Arepo Consult., Berlin, Germany, Mar 2016.
21 [62] S. Mandal, B. K. Das, and N. Hoque, "Optimum sizing of a stand-alone hybrid energy system for rural electrification in Bangladesh," Journal of
22

of
Cleaner Production, vol. 200, pp. 12-27, 2018.
23 [63] S. Singh, M. Singh, and S. C. Kaushik, "Feasibility study of an islanded microgrid in rural area consisting of PV, wind, biomass and battery energy
24 storage system," Energy Conversion and Management, vol. 128, pp. 178-190, 2016.
25 [64] Z. Movahediyan and A. Askarzadeh, "Multi-objective optimization framework of a photovoltaic-diesel generator hybrid energy system considering

ro
26 operating reserve," Sustainable Cities and Society, vol. 41, pp. 1-12, Aug 2018.
27 [65] T. C. Nguyen, A. T. Chuc, and L. N. Dang, "Green Finance in Viet Nam," in Handbook of Green Finance: Energy Security and Sustainable
28 Development, J. D. Sachs, W. T. Woo, N. Yoshino, and F. Taghizadeh-Hesary Eds., 1st ed. Singapore: Springer Singapore, 2019, pp. 675-705.
29 -p
re
lP
na
ur
Jo
 Highlights

- Green technology is applied to the aquaculture industry.

- Pure oxygen produced onsite for oxygenation leads to higher yield and energy saving.

- By-product hydrogen is used for backup power.

- Optimal planning and design for an energy system help farmers save operation costs and

of
reduce CO2.

ro
- Useful guidelines to instruct optimum design and operation for aquaculture farms.

-p
re
lP
na
ur
Jo
Declaration of Interest Statement

The manuscript has been read and approved by all authors. The order of authorship
has been agreed upon among co-authors, and the criteria for authorship as defined in the
“Uniform requirements for manuscripts” have been met. This manuscript has not been
published previously and is not under consideration for publication elsewhere. We have no
conflicts of interest to disclose.

of
Nguyen Nhut Tien

ro
On behalf of all authors

-p
re
lP
na
ur
Jo