Dynamic Simulation and Optimization of Integrated
Dynamic Simulation and Optimization of Integrated
OPEN ACCESS
Article
Dynamic simulation and optimization of integrated
clean energy water systems
Highlights
Two nuclear hybrid energy
systems are evaluated to
deliver clean energy and
water
A penalty function is
introduced to account for
mismatch of energy
demand and supply
Article
Dynamic simulation and optimization of integrated
clean energy water systems
Stephen Hills,1 Seth Dana,1 and Hailei Wang1,2,*
SUMMARY
Nuclear hybrid energy systems (NHES) are a viable option to provide clean power
by combining renewable energy sources such as wind and solar. This study ana-
lyzes two types of NHES that use small modular reactors (SMR) and wind turbines
to produce clean energy and water. The first system uses freeze desalination (FD)
and the second system uses reverse osmosis (RO) to produce clean water. Both
systems are optimized using net present value at two case locations. The FD sys-
tem can better meet the energy demand using the stored thermal energy to
boost the power during peak hours, which allows less capital investment on its
design compared to the RO system. However, the results from the two cases
reveal that the RO system can be more economic when water price is more
than $1.50/m3. A sensitivity analysis also identified the critical system parameters
on the net present value of the systems.
INTRODUCTION
More than half of the world’s electrical energy needs are currently covered by conventional fossil fuel power
plants whose efficiency for a single cycle power plant is significantly lower than that of other plants, such as
combined cycle power plants (CCPPs). CCPPs combine two or more thermodynamic cycles, such as a gas
turbine with a heat recovery steam generator, enabling an attractive higher efficiency (up to 60%) and
providing a reduction in pollutant emissions (Watson, 1996). Coupled with low-cost shale gas, CCPPs
have become more and more common over the last decade. At the same time, renewable energies such
as wind and solar have seen tremendous growth. With further penetration of renewable energy, it poses
a great challenge for traditional power plants as their demands fluctuate considerable within hours. Using
the solar energy in CA, US, as an example, the phenomenon has become known as the ‘‘duck curve’’ as
shown in Figure 1. During the day when plenty of solar energy is available, it depresses the power gener-
ation from traditional power plants. After sunset, the ramp rate required for meeting the energy needs puts
tremendous stress on traditional power generation systems (Jones-Albertus, 2017). This requires power
plants to be more flexible in meeting the power demands.
To keep the efficiency of the power plant as high as possible and to decrease the thermo-mechanical fa-
tigue, it is desirable to maintain the power generation at full capacity (Benato et al., 2016, pp. 76–85).
The challenge is to store the energy that the power plant generates during off-peak hours to then be
used during peak hours to reduce the stress on the power plant during peak hours. This requires a dynamic
control strategy that can respond to the variable working conditions to store as much energy as possible
during off-peak hours and then release the stored energy at the optimal time and rate during peak hours to
reduce the stress on the power plant.
One way to increase the reliability of systems using renewable energy and decrease the influence of energy
fluctuations is to use a hybrid energy system (HES). Hybrid energy systems combine two or more forms of 1Department of Mechanical
renewable energy sources, as well as energy storage, to improve system performance and energy reli- and Aerospace Engineering,
ability, and to overcome limitations inherent in single energy sources. The operator of hybrid energy sys- Utah State University, 4130
Old Main Hill, Logan, UT
tems decides on the optimal scheduling of its resources and trading power with the main grid in an optimal
84321, USA
way (Taghizadeh et al., 2020). A more specific type of HES is the Nuclear Hybrid Energy System (NHES). 2Lead contact
NHESs have the potential to produce commodities such as electricity and hydrogen and allow for electricity
*Correspondence:
grid load following (Pinsky et al., 2020). A benefit of using an NHES is that the reactor can continuously op- [email protected]
erate at full capacity, so the fluctuation in power output is not as large as compared to a system that only https://doi.org/10.1016/j.isci.
relies on wind or solar power. Also, this system can potentially provide carbon-free energy production. 2022.104015
Figure 1. The ‘Duck Curve’ published by CAISO in 2013 shows how solar energy can cause the net load on the
power grid to fluctuate throughout the day
Ongoing research has looked into the use of NHES and the apprehensions of the public (Suman, 2018, pp.
166–177), the use of small modular nuclear reactors (SMRs) in system design (Bragg-Sitton, 2021, pp. 323–
356), and adding other power sources, such as solar, along with water desalination to improve economic
and system efficiency (Wang et al., 2021).
In this study, two NHESs that can meet the power demand of a given region cost effectively while producing
clean water have been designed and optimized. The first NHES uses freeze desalination (FD) to both store
thermal energy to boost power production during peak hours and produce clean water. This optimized sys-
tem is compared to the second NHES that uses excess power to run a Reverse Osmosis (RO) system to
generate clean water. The sizing of the components and the number of components is optimized to avoid
over-producing while meeting the power demands. The schematic diagrams of both NHES designs are
shown in Figure 2. For the FD system, when there is excess power from the SMRs and wind turbines (usually
during low power demand periods), the refrigeration cycle is turned on to produce ice out of the seawater
(or salt water in general). When the power demand exceeds the total power outputs from the SMRs and
wind turbines (i.e., peak energy demand periods), the stored ice is used to lower the steam cycle
condensing temperature/pressure through the power-boosting condenser, and increase the steam cycle
efficiency and power output. After utilizing its latent heat, the thawed water becomes a source of clean wa-
ter. In contrast, the RO system is a simpler system. It produces clean water when there is excess power from
the SMRs and wind turbines. However, it does not provide the energy storage and power boosting func-
tions. In comparing these two systems, the Net Present Value (NPV) is used to see which system is more
advantageous for a given power demand, air temperatures and wind profiles.
Recent research has been done on the dynamic modeling of combined-cycle power plants, specifically
those with a gas turbine unit coupled to a heat recovery steam generator (Casella and Colonna, 2012,
pp. 91–111; Casella and Pretolani, 2006; Casella et al., 2011, pp. 9549–9554). The Modelica language has
been used to model the dynamics of these power plants and investigate the stresses on them during start
up. Work has also been done using Modelica to optimize the control of a combined-cycle power plant to
reduce the stresses during fluctuations in power demand. This is of growing importance as the fluctuation
of power demand for power plants continues to increase.
One of the main challenges in using a hybrid energy system effectively is optimizing it for size and cost.
Much research has been done and is continuing to be done on optimization techniques for HES (Sana-
jaoba, 2019, pp. 655–666; Duman and Guler, 2018, pp. 107–126; Gharibi and Askarzadeh, 2019, pp.
25428–25441). When using Modelica to model a system, there are multiple choices of software available
for optimization. One option is to load the model in MATLAB and use it for optimization (Tica et al., 2012,
pp. 256–265). Another software that has been used to optimize a Modelica model is CasADi (Shitahun et al.,
2013, pp. 446–451). Some other software that is designed specifically for Modelica and for dynamic opti-
mization is Optimica and JModelica.org (Akesson et al., 2010, pp. 1737–1749). Another important aspect
when modeling a system using Modelica is to make sure that it is as simplified as possible to reduce the
compilation time. It is important to have a powerful solver, especially when optimizing a complex dynamic
system as it can dramatically decrease the time needed to run the model (Cafferkey and Provan, 2015, pp.
210–215). One such possible solver, developed by the Idaho National Laboratory, is RAVEN. It can perform
adaptive samplings of large input spaces and can create reduced order models representative of the input
space (Rouxelin et al., 2020).
RAVEN stands for Risk Analysis and Virtual Environment. It is a multipurpose probabilistic and uncertainty
quantification framework, capable of communicating with any system code, implemented with an inte-
grated validation methodology involving several different metrics (Narcisi et al., 2019, pp. 419–432). It is
a useful tool for analyzing systems as it can integrate the system code with various plugins that are available
to RAVEN to perform system optimization and economic analysis. One such plugin for economic analysis in
RAVEN is TEAL. RAVEN communicates with other software code, such as OpenModelica, through XML
files. RAVEN can be used to run multiple iterations of a simulation with the prescribed variables being
altered to see how the system responds. Outputs from the simulation can be saved and plots can be gener-
ated by RAVEN to visualize the results. The outputs from the simulations can also be used by other plugins,
such as TEAL, to perform economic analysis. TEAL stands for Tool for Economic AnaLysis. It is a RAVEN
plugin that is used to deploy economic analysis for RAVEN workflows. TEAL can compute economic metrics
such as NPV (Net Present Value).
advantages over RO, such as higher salt rejection (CDI, MD), higher recovery of water (MD), fewer pre-treat-
ment stages (MD, FO) and the ability to use low-grade energy (MD, FO). Currently, stand-alone technolo-
gies cannot compete with RO until certain challenges are addressed, such as high energy consumption
(Skuse et al., 2021).
Ice-based Freeze Desalination (FD) is a technique to recover fresh water from seawater that has been
receiving increased interest over the years (Sahu et al., 2020). It involves the cooling of saline water below
freezing whereas the salt remains in the solution. The ice crystals can subsequently be separated from the
concentrated brine, washed and melted to obtain clean water (Shi et al., 1990, pp. 316–328). Some of the
advantages of FD is that the energy required to freeze water is only 1/7 of its latent heat of vaporization,
which means FD can be more effective on energy utilization than the distillation process. Also, corrosion
and scaling can be negligible during the heat exchange because of the low operating temperature. An
added benefit is that there is no discharge of hazardous chemicals into the environment (Xie et al., 2018,
pp. 293–300). Although FD is a promising desalination technique and the process could potentially be
coupled with energy storage, effective separation produced ice and brine in order to achieve acceptable
salinity level has been challenging. Reverse Osmosis is an effective approach toward molecular separation
and is widely used commercially for desalination of seawater (Vatanpour and Zoqi, 2017, pp. 1478–1489). It
works by using a partially permeable membrane to push pressurized water through, while preventing larger
particles, such as salt, through the membrane. An advantage of reverse osmosis over some of the other
desalination techniques is higher energy efficient, which leads to lower levelized cost of water (Okampo
and Nwulu, 2021). Some of its drawbacks include membrane degradation because of fouling or corrosion.
However, research is currently being done to improve the membrane, allowing for enhanced desalination,
anti-fouling and durability properties (Liu et al., 2021).
MODELING
In this work, OpenModelica has also been chosen as the dynamic modeling tool to simulate the processes
of the two NHES systems (for detailed information of the FD and RO systems, please refer to a recent pub-
lication (Hills et al., 2021)). The OpenModelica models leverage some of the components already devel-
oped in the open-sourced Modelica libraries and ThermoPower library. Compared to other dynamic
modeling tools such as Matlab/Simulink, Modelica as an object-oriented programming language allows
the modelers to focus on the physical process modeling with little concern about the optimal/efficient so-
lution algorithms. Figure 3 shows the overall OpenModelica model of the FD system, including the steam
cycle, the refrigeration cycle, the wind turbines, the ice/water tanks and the cooling tower, along with the
inputs and connections for those components. There is a legend at the top that shows what the icon is for
each of the major parts in the system.
The NHES was modeled using OpenModelica, making use of some of the component models already in
the Modelica and ThermoPower libraries. The overall OpenModelica model of the NHES is shown in Fig-
ure 3. It shows the major parts of the system: the steam cycle, the refrigeration cycle, the wind turbines, the
ice/water tanks and the cooling tower. The legend at the top shows what the icon is for each of the major
parts in the system. The steam cycle and refrigeration cycle shown in Figure 3 can be expanded to display
the individual components that make up both cycles. Note that an in-depth explanation on how the com-
ponents in the system were modeled can be found in a recent publication (Hills et al., 2021).
Figure 4 shows the OpenModelica model for the steam cycle with its major components. The cycle is simpli-
fied compared to a more complex real-life steam cycle it represents. It has a power-boosting condenser
(PBC) as shown to further cool down the cooling tower return water. The turbine has a lumped efficiency
equivalent to a system with regeneration. Inputs to the steam cycle are the mass flow rate, thermal power
from the reactor, the return water mass flow rate and temperature from the cooling tower and the ice water
mass flow rate from the ice/water tank. Outputs from the steam cycle are the water outlet temperatures
from the PBC and main condenser, as well as the cycle efficiency and power output.
Figure 5 shows the refrigeration cycle model in OpenModelica. As with the steam cycle, the refrigeration
cycle model is a simplified version of the real-life system. The main inputs for the cycle are the water tem-
perature and mass flow rate to the condenser from the cooling tower, the ice water temperature and mass
flow rate from the tank and the power available to run the cycle. The outputs from the cycle are the water
temperature going back to the cooling tower and the amount/fraction of ice produced going into the tank.
The equations used for modeling the components, along with assumptions and simplifications, and the
logic used for the controllers are described in the following sections.
‘‘The Economics of Small Modular Reactors’’ was the basis for calculating the recurring costs from opera-
tion and maintenance (The Economics of Small Modular Reactors, 2017). It gives estimates for the fixed
($135/kW-yr) and variable ($3/MWh) O&M, as well as an estimate for the cost of fuel ($8.5/MWh), with
the fuel disposal cost included.
costs and power costs to run the system (Seawater Desalination Power Consumption, 2011). The cost
model used for the RO system lumps together the capital costs and operating costs into one large single
cost, which is based on the levelized cost of water. The overall cost for a 15 MW system with a 20-year
lifespan was set at $225 million and increases with a scale factor of 0.5 as the size of the RO system
increases.
S is the sum of the costs, k is the current rate for electricity, y is the power produced and yp is the predicted
power production. This approach to calculating the penalty was selected so that the cost would be sym-
metric about the demand and increase greatly the farther the system was from meeting the demand.
This helped to ensure that the system would be better fit to meet the power demand.
The systems were evaluated using OpenModelicav1.14.1 (64-bit) with OMSimulatorv2.1.0. RAVEN v. 2.0
was used to run multiple cases for each of the systems to find the optimal system sizing for highest net pre-
sent value.
Figure 10. Results from freeze desalination system in Salt Lake City
(A) Power for Refrigeration Cycle.
(B) Ice in tank.
(C) Efficiency of steam cycle.
(D) Penalty energy throughout the day.
Optimization process
Five design parameters have been identified for the FD system as they are associated with the size of its key
subsystems, including 1) number of SMRs; 2) number of wind turbines; 3) size of the compressor for the
refrigeration cycle; 4) size of the ice storage tanks; 5) size of the power boosting condenser. For the RO sys-
tem, in addition to the number of SMRs and number of wind turbines, the size of the RO system is included
as one of the three design parameters for the optimization.
The optimization process to find the optimal sizing of the system to meet the power demand with the highest NPV
is shown in Figure 6. It is broken down into two main parts. The first part is when the Monte Carlo sampler is run in
RAVEN to explore the design space for the design parameters of each system. The sampler runs multiple cases of
the OpenModelica model, changing the following variables (depending on the system): (1) Number of SMRs; (2)
number of wind turbines; (3) size of the compressor for the refrigeration cycle; (4) size of the tanks; (5) size of the
power boosting condenser; and (6) size of the RO system. It produces the following outputs, which will be used in
the second part: (1) The component sizes for the simulation; (2) the amount of drinkable water produced; (3) the
amount of power generated; and (4) the penalty charges. The second part of the optimization process uses the
outputs from the sampled simulations to perform an economic analysis using TEAL. TEAL then calculates the net
present value (NPV) for the sampled systems. These results are used to refine the sampling space. This process will
be repeated multiple times to reach an optimal system sizing. Then the NPV of the optimized system can be
computed. The optimized systems using reverse osmosis and freeze desalination can then be compared and
the best system for the given inputs can be determined.
RESULTS
This section discusses the results from optimizing both OpenModelica models, the Freeze Desalination
(FD) system and the Reverse Osmosis (RO) system for two locations. The locations chosen were Salt
Lake City, UT and San Diego, CA.
Input profiles
Figure 7 shows the input profiles (power demand, air temperature and wind speed) used to run the models.
The data was taken from a hot summer day in July as it represents typical variable load and relatively high
power demand in the afternoon.
The power demand data for Salt Lake City came from the US Energy Information Administration (Hourly
Electric Grid Monitor, 2019). A plot of its profile for the given day is shown in Figure 7A. The data for the
air temperature came from Weather Underground (Weather History, 2021). It has records of hourly data
for air temperature. Figure 7B shows the air temperature profile for the chosen summer day in Salt Lake
City. The data for the wind speed came from a wind farm outside of Salt Lake City, in Milford, UT. This loca-
tion was chosen as it was near Salt Lake and is a windy location where wind turbines would realistically be
located. The data came from the US National Renewable Energy Laboratory wind prospector tool (Wind
Prospector, 2020). Data was not available for the year 2019 (from which the power demand and air temper-
atures come), so the data was taken from the same day in 2012 instead. Figure 7C shows the wind speed
profile for the summer day.
Optimization process
To get a good estimate of the sizing search space to use for the number of SMRs, the OpenModelica model
was run a few times with the number of SMRs varied until the power generation was close to the power
demand. Then a Monte Carlo analysis was performed with a relatively large sample space. The perfor-
mance of each of the runs was ranked by NPV and the sizing of the best runs was used to narrow the search
space for a second Monte Carlo analysis. This process was repeated with the search space decreasing for
each iteration until the optimal solution was found.
Figures 9 and 10 show how the FD system operates for the optimal case. Figure 9 shows how the system
meets the power demand, with the power also being broken down into two types, power from the SMRs
and power from the wind turbines. Notice how the system meets or exceeds the power demand from hours
0–19 and 22–24. Figure 10A shows the power available for the compressor to run the refrigeration cycle.
From about hours 3–5 the power available is a little more than the 20 MW size of the compressor. From
about hours 13–22 the compressor is turned off as there is no excess power to run the refrigeration cycle.
Figure 10B display the percent of ice in the tank throughout the day. The ice percentage builds up to about
75% by hour 13 and then decreases to zero a little after hour 19 as the ice is used up in order to boost power
output through the power-boosting condenser. Figure 10C shows the efficiency of the steam cycle and how
the use of the power boosting condenser from hours 13–22 improves it. Notice how the efficiency is
boosted when the PBC is turned on at hour 13, with the boosting decreasing from hours 19–22 as the
tank runs out of ice. Figure 10D shows the accumulation of penalty energy throughout the day. Notice
that there is a penalty from hours 3–5 when the system is producing more power than the compressor
can use for the refrigeration cycle. Also, from hours 19–22 there is more significant penalty when the system
does not meet the power demand.
Figure 11 shows the temperature of the waters in the power boosting condenser. Although the outlet tem-
perature of the water from the tank is always set to 1 below the inlet temperature of the water from the
cooling tower, the mass flow of the water from the tank is effectively zero from hours 0–13 as power boost-
ing is not needed and 22–24 as the stored latent heat in the ice/water tanks is depleted. This can be inferred
as the temperature difference between the inlet and outlet water from the cooling tower is zero over the
same time period.
years Figure 12 shows the optimization results. As shown, designs with 42 SMRs and 8 wind turbines per
SMR give relatively higher NPV.
Figures 13 and 14 show how the RO system operates for the optimal case. Figure 13 shows how the system
meets the power demand. Notice that the system oscillates around the power demand during the peak
hours from 15–21 as the wind turbines’ power output fluctuates more, which requires the larger number
of wind turbines. Figure 14A shows how much power is available to run the reverse osmosis process. Notice
that from hours 2.5–5.5 there is excess power that can’t be used by the 19 MW RO system. Figure 14B shows
the amount of clean water that has been generated throughout the day by the RO process. Most of the wa-
ter is generated in the first 12 h. The amount of penalty energy incurred by the system throughout the day is
shown in Figure 14C. Notice the penalty from over-production from hours 2.5–5.5 and then penalties from
underproduction for the rest of the day.
Input profiles
The following figure shows the input profiles (Power Demand, Air Temperature and Wind Speed) used to
run the models. The data again was taken from a hot summer day. The day was chosen as it had a relatively
high power demand for the given season. The power demand data for San Diego also came from the US
Energy Information Administration (Hourly Electric Grid Monitor, 2019). Power demand data for San Diego
was created by using the power demand for CA for the given days and scaling it to the number of house-
holds in San Diego. The power demand for San Diego for the given summer day is shown in Figure 15A. The
data for the air temperature came from Weather Underground. Figure 15B shows the air temperature pro-
files for the chosen summer day in San Diego. The data for the wind speed came from a wind farm outside of
San Diego, in Boulevard, CA. This location was chosen as it was near San Diego and a windy location where
Figure 14. Results from the optimal reverse osmosis system in salt lake city
(A) Power available to RO.
(B) Clean water generated throughout the day.
(C) Penalty Energy.
wind turbines would realistically be located. The data also came from the US National Renewable Energy
Laboratory wind prospector tool. Again, the data was not available for the year 2019 (from which the power
demand and air temperatures come), so the data was taken from the same days in 2012. Figure 15C shows
the wind speed profiles for the given summer day.
condenser of 1000 kg/s. The calculated net present value (NPV) for this case was $2.67e10 after
40 years. Figure 16 shows the results of the optimization process. Figures 17 and 18 show how the FD
system operates for the optimal case. Figure 17 shows how the system meets the power demand, with
the power also being broken down into two types, power from the SMRs and power from the wind tur-
bines. Notice how the system meets or exceeds the power demand from approximately hours 0–17.5 and
21.5–24.
Figure 18A shows the power available for the compressor to run the refrigeration cycle. From about hours
11–14 and 17–22 the compressor is turned off as there is no excess power to run the refrigeration cycle. For
the last hour of the day the power available is more than the 16 MW size of the compressor. Figure 18B show
the percent of ice in the tank. The ice percentage reaches a maximum of about 29% by hour 11 and
then decreases to zero by hour 19. Figure 18C shows the efficiency of the steam cycle and how it is boosted
by the power boosting condenser. Notice how the boosting decreases from hours 17.5–22 as the tank ran
out of ice. Figure 18D shows the accumulation of penalty energy throughout the day. Notice that most of
the penalty is from hours 18–21 when the system does not meet the power demand. There is also a little
penalty during the last hour when too much power is generated. Figure 19 shows the temperatures
of the waters in the power boosting condenser. Notice how the temperature of the water at the outlet
to the steam condenser decreases from the inlet from the cooling tower when the power boosting
condenser is on.
of the amount of penalty energy incurred by the system throughout the day. The penalty from hours 18–21
is because of not meeting the power demand. The penalty from hours 22.5–24 is because of excess power
exceeding the size of the 15 MW RO system.
DISCUSSION
This section provides some discussions and explanation of the results shown in the previous section.
A sensitivity analysis of relevant system variables was conducted first, which was then followed by
analyzing the results for both the FD and RO systems and what make one system better suited than
the other.
Sensitivity analysis
A sensitivity analysis was performed using the data for Salt Lake City. The variables that were used in the
sensitivity analysis were: SMR number, wind turbine number, tank size, compressor size, PBC max size,
RO size, steam heat transfer coefficient in main condenser, the turbine isentropic efficiency and the price
of energy during peak hours. To calculate the sensitivity of each variable, the system was first run at the
optimal solution (i.e. for FD system: 42 SMRs, 2 wind turbines per SMR, 1.8 million gallon tank, 20 MW
compressor, 1,400 kg/s max flow for PBC, 5000 W/(m2K) steam heat transfer coefficient, 0.85 isentropic ef-
ficiency of turbine and $0.22/kWh for electricity during peak hours). Then each individual variable was var-
ied by 2.5%, first by increasing, then by decreasing the variable. The NPV for each case was calculated and
the variation from the optimal solution was calculated as shown in the following equation:
NPVi NPVopt
NPV variation =
NPVopt
The sensitivity of each variable was then calculated by taken the average of the absolute value of the two
NPV variations. The results of the sensitivity analysis are shown in Figure 23. It shows that the NPV calcula-
tion is most sensitive to a change in the turbine isentropic efficiency. A 2.5% change in the value of the isen-
tropic efficiency of the turbine resulted in about a 4.5% change in the NPV as compared to the optimal
solution. This impact is so large as it affects the overall efficiency of the steam cycle for its entire life cycle.
The next closest variable was the SMR, which changed the NPV by a little less than 2%. This impact is lower
as it mostly affects the capital cost. The only other variable that was able to change the NPV by more than
1% was the RO size. The remaining variables had minimum change on the NPV. This shows how important it
is to have a good estimate of the turbine performance.
Similarly, for San Diego, the study has shown the optimal FD system resulted in a higher NPV ($2.67e10)
than the optimal RO system ($2.33e10). In this case, the RO system did a better job in meeting the power
demand with a lower amount of penalty energy. The amount of clean water generated by the RO system
was about twelve times larger than the FD system (about 1.37 million m3 to about 113,500 m3), which could
be more advantageous given San Diego’s need for a large amount of clean water. In order for the NPV of
the RO system to equal the FD system, the price of water would need to rise from $1.25 per m3 to $1.52 per
m3. In performing the same hypothetical case without penalty energy, the FD system now requires 55 SMRs
(increased from 52 of the optimal case) and 555 wind turbines (increased from 520 of the optimal case). The
NPV for this case was also slightly lowered to be 2.63e10. Although this system would have a larger initial
Figure 22. Results from the optimal reverse osmosis system in San Diego
(A) Power available to RO.
(B) Clean water generated throughout the day.
(C) Penalty Energy.
capital cost and a lower NPV based on the penalty function used, this system would be more advantageous
if the penalty for not meeting the power demand were increased.
One possible way to find a more optimal solution would be to uncouple the number of wind turbines with
the number of SMRs. For simplicity, the number of wind turbines was linked with the number of SMRs. This
made it difficult to refine the number of wind turbines when there were a large number of SMRs. It was also
noted that the NPV would continue to increase with an increasing max size of the power-boosting
condenser (PBC). One reason for this is that the cost of the PBC was assumed to be small in relation to
the other component costs. Thus, setting the limit of the PBC size in this study may have impeded the
FD system potential.
Based on this study, both the cost of water and cost of the penalty energy play important roles in deciding
which system is more economically viable. As the cost of water increases, the value of the RO system rises as
it produces more clean water. As the cost of the penalty energy increases, the FD system would be
preferred. It is also important to note that the economic models used in this study are rather simplified.
Thus, more detailed cost models for the key system components would more accurately reflect the costs
of real-life systems.
Conclusion
This article provided an analysis of the fit of two types of nuclear hybrid energy systems (NHES) for deliv-
ering clean energy and water. Both systems couple small modular reactors (SMR) with wind turbines to
provide clean power and drive either a freeze desalination (FD) or a reverse osmosis (RO) system for clean
water. The FD system was able to boost power production of the SMR using the produced ice, whereas the
RO system was able to generate more clean water with the requirement of more wind turbines to better
meet the power demand.
Both systems were modeled in OpenModelica leveraging various component models in its libraries. Sys-
tem optimization and NPV analysis were performed using RAVEN and its plugin TEAL. RAVEN performs
Monte Carlo analyses over the sample space of each design variable to find the optimal component sizes
for each system. A quadratic penalty cost function was used for unmet power and excess power to help with
finding the optimal system size.
Based on the two case studies using Salt Lake City and San Diego, the results reveal that the FD system can
better meet the power demand, thus less penalty energy. This is attributed to the power-boosting capa-
bility of the freeze desalination system. During the peak hours of power demand, the FD system increases
its power production by about 12% by operating the steam cycle more efficiently. According to the eco-
nomic figures, as long as the water price is below $1.50/m3, it is more economic for the FD system.
Thus, when water price rises above $1.50/m3, the RO system could be more advantageous as it generates
substantially more clean water. Although the aforementioned general findings are likely held for other
similar cities, the optimal design of each system (FD or RO) can change from location to location based
on the respective input datasets (demand, costs of energy and water).
synthetic time-series datasets generated from reliable data sources would be one of the focuses of the
work going forward.
NOMENCLATURE
Acronyms
CCPP Combined cycle power plant
CDI Capacitive deionization
FD Freeze desalination
FO Forward osmosis
GWe Gigawatt electrical
GWth Gigawatt thermal
HES Hybrid energy system
IRR Internal rate of return
MD Membrane distillation
NHES Nuclear hybrid energy system
NPV Net present value
PI Profitability index
PBC Power boosting condenser
SD San Diego
SLC Salt Lake City
SMR Small modular reactor
RAVEN Risk analysis and virtual environment
RO Reverse osmosis
TEAL Tool for economic analysis
Variables
a Cost for the reference size
D Driver size of the component/system
N Multiple number
R Reference size of the component/system
S Scaling factor
S Sum
y Total power produced by system [W]
yp Expected total power production by system [W]
STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:
ACKNOWLEDGMENTS
The authors would like to thank the financial support provided by Nuclear Regulatory Commission (NRC)
through the Award No. 31310019M0014.
AUTHOR CONTRIBUTIONS
SH: Modeling, software, manuscript preparation. SD: RO system modeling, manuscript preparation, edit-
ing. HW: System concepts and designs, model and result discussions, manuscript review and editing.
DECLARATION OF INTERESTS
The authors declare no competing interests.
REFERENCES
Akesson, J., Arzen, K.-E., Gafvert, M., Bergdahl, framework for the design of reverse osmosis 110712. https://doi.org/10.1016/j.rser.2021.
T., and Tummescheit, H. (2010). Modeling and desalination plants under food-energy-water 110712.
optimization with optimica and JModelica.org– nexus considerations. Desalination 503, 114937.
languages and tools for solving large-scale https://doi.org/10.1016/j.desal.2021.114937. Pawlowski, S., Huertas, R.M., Galinha, C.F.,
dynamic optimization problems. Comput. Chem. Crespo, J.G., and Velizarov, S. (2020). On
Eng. 34, 1737–1749. https://doi.org/10.1016/j. Duman, A.C., and Guler, O. (2018). Techno- operation of reverse electrodialysis (red) and
compchemeng.2009.11.011. economic analysis of off-grid PV/wind/fuel cell membrane capacitive deionisation (MCDI) with
hybrid system combinations with a comparison of natural saline streams: a critical review.
Almasi, A. (2014). How much will your compressor regularly and seasonally occupied households. Desalination 476, 114183. https://doi.org/10.
installation cost?, Retrieved from chemical Sustain. Cities Soc. 42, 107–126. https://doi.org/ 1016/j.desal.2019.114183.
processing. https://www.chemicalprocessing. 10.1016/j.scs.2018.06.029.
com/articles/2014/how-much-will-your- Pinsky, R., Sabharwall, P., Hartvigsen, J., and
compressor-installation-cost/. Gharibi, M., and Askarzadeh, A. (2019). Size and O’Brien, J. (2020). Comparative review of
power exchange optimization of a grid- hydrogen production technologies for nuclear
Bandar, K.B., Alsubei, M.D., Aljlil, S.A., Darwish, connected diesel generator-photovoltaic-fuel hybrid energy systems. Prog. Nucl. Energy 123,
N.B., and Hilal, N. (2021). Membrane distillation cell hybrid energy system considering reliability, 103317. https://doi.org/10.1016/j.pnucene.2020.
process application using a novel ceramic cost and renewability. Int. J. Hydro.Energy 44, 103317.
membrane for brackish water desalination. 25428–25441. https://doi.org/10.1016/j.ijhydene.
Desalination 500, 114906. https://doi.org/10. 2019.08.007. Quickfacts United States (2021), Retrieved from
1016/j.desal.2020.114906. United States Census Bureau:https://www.
Hills, S., Dana, S., and Wang, H. (2021). Modeling census.gov/quickfacts/fact/table/US/PST045219.
Benato, A., Bracco, S., Stoppato, A., and and simulation of nuclear hybrid energy systems
Mirandola, A. (2016). Dynamic simulation of using OpenModelica. Energy. Convers. Manag. Rouxelin, P., Alfonsi, A., Ivanov, K., and Strydom,
combined cycle power plant cycling in the 247, 114724. https://doi.org/10.1016/j. G. (2020). Energy group search engine based on
electricity market. Energ. Convers. Manag. 107, enconman.2021.114724. surrogate models constructed with the RAVEN/
76–85. https://doi.org/10.1016/j.enconman.2015. NEWT/PHISICS sequence. Nucl. Eng. Des. 356,
07.050. Hourly Electric Grid Monitor (2019). Retrieved 110356. https://doi.org/10.1016/j.nucengdes.
from U.S. Energy Information Administration: 2019.110356.
Black, G.A., Aydogan, F., and Koerner, C.L. https://www.eia.gov/electricity/gridmonitor/
(2019). Economic viability of light water small dashboard/electric_overview/US48/US48 Sahu, P., Krishnaswamy, S., and Pande, N.K.
modular nuclear reactors: general methodology (2020). Process intensification using a novel
and vendor data. Renew. Sustain.Energy Rev. IEA (2017). US Wind O&M Costs Estimated at continuous U-shaped crystallizer for freeze
103, 248–258. https://doi.org/10.1016/j.rser.2018. $48,000/MW; Falling Costs Create New Industrial desalination. Chem. Eng. Process. Process
12.041. Uses (IEA), Retrieved from Reuters Events Intensif. 153, 107970. https://doi.org/10.1016/j.
Renewables:https://www.reutersevents.com/ cep.2020.107970.
Bragg-Sitton, S.M. (2021). Hybrid Energy Systems renewables/wind-energy-update/us-wind-om-
Using Small Modular Nuclear Reactors (SMRs). costs-estimated-48000mw-falling-costs-create- Sanajaoba, S. (2019). Optimal sizing of off-grid
Handbook of Small Modular Nuclear Reactors, new-industrial-uses-iea. hybrid energy system based on minimum cost of
Second Edition (Woodhead Publishing), energy and reliability criteria using firefly
pp. 323–356. https://doi.org/10.1016/B978-0-12- Jones-Albertus, B. (2017). Confronting the duck algorithm. Sol. Energy 188, 655–666. https://doi.
823916-2.00013-8. curve: how to address over-generation of solar org/10.1016/j.solener.2019.06.049.
energy, Retrieved from Office of Energy Efficiency
Cafferkey, N., and Provan, G. (2015). An analysis & Renewable Energy. https://www.energy.gov/ Seawater Desalination Power Consumption (2011),
of performance-critical properties of modelica eere/articles/confronting-duck-curve-how- Retrieved from WateReuse Association:https://
models. IFACPapers Online 48, 210–215. https:// address-over-generation-solar-energy. watereuse.org/wp-content/uploads/2015/10/
doi.org/10.1016/j.ifacol.2015.05.122. Power_consumption_white_paper.pdf. .
Liu, M., He, Q., Guo, Z., Zhang, K., Yu, S., and
Casella, F., and Colonna, P. (2012). Dynamic Gao, C. (2021). Composite reverse osmosis Shi, Y., Liang, B., and Hartel, R.W. (1990).
modeling of IGCC power plants. Appl. Therm. membrane with a selective separation layer of Crystallization of ice from aqueous solutions in
Eng. 35, 91–111. https://doi.org/10.1016/j. double-layer structure for enhanced desalination, suspension crystallizers. In ACS
applthermaleng.2011.10.011. anti-fouling and durability properties. SymposiumSeries, 438 (ACS Publications),
Desalination 499, 114838. https://doi.org/10. pp. 316–328. https://doi.org/10.1021/bk-1990-
Casella, F., and Pretolani, F. (2006). Fast start-up 1016/j.desal.2020.114838. 0438.ch023.
of a combined-cycle power plant: a simulation
study with Modelica. Modelica Conference 4, Narcisi, V., Lorusso, P., Giannetti, F., Alfonsi, A., Shin, H., Kalista, B., Jeong, S., and Jang, A. (2019).
3–10. and Caruso, G. (2019). Uncertainty quantification Optimization of simplified freeze desalination
method for RELAP5-3D using RAVEN and with surface scrape freeze crystallizer for
Casella, F., Donida, F., and Akesson, J. (2011). application on NACIE experiments. Ann. Nucl. producing irrigation water without seeding.
Object-oriented modeling and optimal control: a Energy 127, 419–432. https://doi.org/10.1016/j. Desalination 452, 68–74. https://doi.org/10.1016/
case study in power plant start-up. IFAC Proc. Vol. anucene.2018.12.034. j.desal.2018.08.023.
44, 9549–9554. https://doi.org/10.3182/
20110828-6-IT-1002.03229. Okampo, E.J., and Nwulu, N. (2021). Shitahun, A., Ruge, V., Gebremedhin, M.,
Optimisation of renewable energy powered Bachmann, B., Eriksson, L., Andersson, J., and
Di Martino, M., Avraamindou, S., Cook, J., and reverse osmosis desalination systems: a state-of- Fritzson, P. (2013). Model-based
Pistikopoulos, E.N. (2011). An optimization the-art review. Renew. Sustain.Energy Rev. 140, dynamic optimization with OpenModelica
and CasADi. IFAC Proc. Vol. 46, 446–451. wp-content/uploads/2017/09/SMR-Start- Watson, W.J. (1996). The ‘success’ of the
https://doi.org/10.3182/20130904-4-JP-2042. Economic-Analysis-APPROVED-2017-09-14.pdf. combined cycle gas turbine. In IEEE International
00166. Conference on Opportunities andAdvances in
Tica, A., Gueguen, H., Dumur, D., Faille, D., and International Electric Power Generation,
Davelaar, F. (2012). Design of a combined cycle pp. 87–92. https://doi.org/10.1049/cp:19960125.
Skuse, C., Gallego-Schmid, A., Azapagic, A., and
Gorgojo, P. (2021). An emerging membrane- power plant model for optimization. Appl. Energy
based desalination technologies replace reverse 98, 256–265. https://doi.org/10.1016/j.apenergy. Weather History (2021), Retrieved from Weather
osmosis? Desalination 500, 114844. https://doi. 2012.03.032. Underground:https://www.wunderground.com/
org/10.1016/j.desal.2020.114844. history/daily/us/ca/san-diego/KSAN.
Utah Rates and Tariffs. Retrieved from Rocky
Suman, S. (2018). Hybrid nuclear-renewable Mountain Power (2019): https://www. Wind Prospector (2020), Retrieved from National
energy systems: a review. J. Clean. Prod. 181, rockymountainpower.net/about/rates- Renewable Energy Laboratory:https://maps.nrel.
166–177. https://doi.org/10.1016/j.jclepro.2018. regulation/utah-rates-tariffs.html. gov/wind-prospector/?
01.262. aL=0&bL=clight&cE=0&lR=0&mC=40.21244%
Vatanpour, V., and Zoqi, N. (2017). Surface 2C-91.625976&zL=4 .
Taghizadeh, M., Bahramara, S., Adabi, F., and modification of commercial seawater reverse
Nojavan, S. (2020). Optimal thermal and electrical osmosis membranes by grafting of hydrophilic Xie, C., Zhang, L., Liu, Y., Lv, Q., Ruan, G., and
operation of the hybrid energy system using monomer blended with carboxylatedmultiwalled Hosseini, S.S. (2018). A direct contact type ice
interval optimization approach. Appl. Therm. carbon nanotubes. Appl. Surf. Sci. 396, 1478– generator for seawater freezing desalination
Eng. 169, 114993. https://doi.org/10.1016/j. 1489. https://doi.org/10.1016/j.apsusc.2016.11. using LNG cold energy. Desalination 435,
applthermaleng.2020.114993. 195. 293–300. https://doi.org/10.1016/j.desal.2017.04.
002.
Wang, G., Yin, J., Lin, J., Chen, Z., and Hu, P.
TEAL (2021), Retrieved from GitHub. https://
(2021). Design and economic analysis of a novel Zhang, X., and Liu, Y. (2021). Integrated forward
github.com/idaholab/TEAL#readme.
hybrid nuclear-solar complementary power osmosis-absorption process for strontium-
system for power generation and desalination. containing water treatment: pre-concentration
The Economics of Small Modular Reactors (2017), Appl. Therm. Eng. 187, 116564. https://doi.org/ and solidification. J. Hazard. Mater. 414, 125518.
Retrieved from SMR Start:https://smrstart.org/ 10.1016/j.applthermaleng.2021.116564. https://doi.org/10.1016/j.jhazmat.2021.125518.
STAR+METHODS
RESOURCE AVAILABILITY
Lead contact
Further information and requests for resources and reagents should be directed to and will be fulfilled by
the lead contact, Hailei Wang ([email protected]).
Materials availability
This study did not use or generate any materials.
METHOD DETAILS
OpenModelica has also been chosen as the dynamic modeling tool to simulate the processes of the two
NHES systems. The OpenModelica models leverage some of the components already developed in the
open-sourced Modelica libraries and ThermoPower library.
Cost models for the key components were created using the RAVEN plugin TEAL. The costs are broken
down into two main parts, capital costs and recurring costs. Capital costs include the initial purchasing
price for the component or system. Recurring costs include the costs for operation and maintenance, as
well as revenue generated by producing things such as clean water or electrical power. To calculate the
penalty cost for over-producing or not meeting the power demand, a quadratic cost function was used.
The optimization process to find the optimal sizing of the system to meet the power demand with the high-
est NPV can be broken down into two main parts. The first part uses the Monte Carlo sampler in RAVEN to
explore the design space for the design parameters of each system. The sampler runs multiple cases of the
OpenModelica model, changing the following variables: (1) Number of SMRs; (2) number of wind turbines;
(3) size of the compressor for the refrigeration cycle; (4) size of the tanks; (5) size of the power boosting
condenser; and (6) size of the RO system. The following outputs: (1) the component sizes for the simulation;
(2) the amount of drinkable water produced; (3) the amount of power generated; and (4) the penalty
charges will be fed into the second part of the optimization process to perform an economic analysis using
TEAL. TEAL then calculates the net present value (NPV) for the sampled systems. These results are used to
refine the sampling space. This process will be repeated multiple times to reach an optimal system sizing.
Then the NPV of the optimized system can be computed. The optimized systems using reverse osmosis and
freeze desalination can then be compared and the best system for the given inputs can be determined.