Energy 215 (2021) 118956

Contents lists available at ScienceDirect

Energy
journal homepage: www.elsevier.com/locate/energy

System impact of heat exchanger pressure loss in ORCs for smelter off-
gas waste heat recovery
Monika Nikolaisen a, *, Trond Andresen a
a
SINTEF Energy Research, Department of Gas Technology, Sem Sælands Vei 11, 7034, Trondheim, Norway

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

Article history: Applying Rankine cycles to smelter off-gas could increase the required off-gas fan power in an order of
Received 9 February 2020 magnitude equivalent to the power production. Predicting the fan power is not straightforward since it is
Received in revised form affected in two contradictory ways: 1) the heat recovery heat exchanger creates additional off-gas
10 September 2020
pressure loss, increasing fan power; 2) off-gas cooling reduces pressure loss in the off-gas handling
Accepted 27 September 2020
system downstream of the cycle, reducing fan power. The purpose of our study is to analyze the effect of
Available online 6 October 2020
fan power on optimum system performance. While additional fan power can be calculated based on heat
exchanger pressure loss, the reduction in fan power depends on the total pressure loss downstream of
Keywords:
Waste heat recovery
the cycle, which is unknown. As an alternative to calculating fan power reduction, we account for the off-
Aluminum smelter off-gas gas cooling effect by including only parts of the fan power caused by heat exchanger pressure loss.
ORC optimization Results from three cases show that both heat exchanger and cycle performance strongly depend on the
Heat exchanger pressure loss potential for downstream pressure loss reduction. Thus, the total pressure loss in the downstream off-gas
Hydrocarbon working fluid handling system has a significant impact on the optimum heat exchanger and cycle performance, and
should be accounted for during system design.
© 2020 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/).

1. Introduction industry owing to the low quality of rejected heat and a lack of
nearby demand for heat [5]. Further research is necessary to pave
The aluminum industry is one of the most energy intensive the way for cost-efficient and large-scale surplus heat utilization in
industries in the world [1], and accounts for around 3.5% of global the aluminum industry.
electricity consumption and 1% of global CO2 emissions [2,3]. In An option for more widespread utilization of surplus heat in the
Norway, aluminum production represents about 17% of all inland aluminum industry is conversion into electric power, or "heat-to-
electricity consumption [4]. Still, the aluminum industry is ex- power" conversion. Heat-to-power conversion can be realized us-
pected to consume even more energy in the future, as the demand ing Rankine cycles, also known as organic Rankine cycles (ORCs)
for aluminum is projected to increase 2e3 times by 2050 [2]. The when utilizing an organic working fluid. Research on Rankine cy-
industry has targeted a 50% reduction in emissions by 2050, cles covers a broad range of application areas and research topics. A
meaning that, in practice, the specific emissions have to be reduced few examples include power production from engine waste heat
by at least 75% [2]. [8,9], geothermal heat [10,11] and solar heat [12], and evaluation of
Around half of the energy input to aluminum smelters is lost to expander technology [13e15], cycle architecture [16,17], and
the surroundings as surplus heat [5e7]. Recovering this surplus thermo-economic analysis [10,18,19]. Our article focuses on the
heat could significantly improve energy efficiency and reduce particular challenges concerning application of Rankine cycles for
emissions in the aluminum industry. The two most efficient ways of surplus heat-to-power conversion in the aluminum industry.
utilizing surplus heat are direct re-use at the same temperature However, results could also be valid for other industries with
level and upgrading to a higher temperature level using a heat similar heat source characteristics and boundary conditions.
pump. However, such reutilization is limited in the aluminum There are several potential surplus heat sources that can be
recovered at primary aluminium production facilities. Our study
focuses on surplus heat originating from the electrolytic reduction
* Corresponding author. process. This process takes place in a series of aluminum electrol-
E-mail addresses: monika.nikolaisen@sintef.no (M. Nikolaisen), trond. ysis cells, as illustrated in Fig. 1. In this example, the electrolysis
andresen@sintef.no (T. Andresen).

https://doi.org/10.1016/j.energy.2020.118956
0360-5442/© 2020 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
M. Nikolaisen and T. Andresen Energy 215 (2021) 118956

Nomenclature DTmin pinch point temperature difference (K)

DW_ change in power (kW)
A heat transfer surface area (m2)
Ac cross-sectional flow area (m2) Subscript
dh hydraulic diameter (m) cond condenser
h enthalpy (kJ/kg) evap evaporator
HRHE heat recovery heat exchanger () exp expander
L length (m) gas electrolysis cell off-gas
m_ mass flow (kg/s) gen generator
h efficiency () ind indirect fluid
p pressure (Pa) is isentropic
P perimeter (m) mech mechanical
Q_ heat duty (kW) rec recuperator
s entropy (kJ/kgK) sink heat sink
U overall heat transfer coefficient (W/m2K) source heat source
W_ power (kW) tot total
Dp pressure loss (Pa) wf working fluid

Wet scrubber
and stack
Fan Gas
treatment
Off-gas duct system
center

Aluminium
electrolysis cell

Fig. 1. Sketch of a generic primary aluminum production facility.

cells are organized in clusters of eight, and a duct system transports total energy content in the gas could be significant and in some
the off-gas from the cells to a gas treatment center, fan, wet cases larger than that dispersed through the sidewalls [22]. Utili-
scrubber and stack. The fan recovers the total off-gas pressure loss zation of the off-gas also has several practical advantages. First of
through the system. Surplus heat from the electrolysis cells mainly all, the off-gas is considered the most easily accessible heat source
takes the form of heat dissipation through the steel shell of the at aluminum facilities since it is already collected in ducts and
electrolysis cells, and as heat contained in the warm off-gas exiting chimneys [7,22]. Moreover, off-gas utilization requires no or only
the cells. minimal structural modifications to the electrolysis cells, and does
Utilization of surplus heat dispersed through the sidewalls of not have a significant influence on the thermal energy balance of
aluminum electrolysis cells has been investigated in several the cells [27,28]. We focus on off-gas surplus heat utilization due to
research articles [1,20,21]. This heat source has a relatively high its practical advantages and significant energy content. However,
exergy content, i.e. potential for power production, owing to the other heat sources at aluminum plants are also important to
temperature level of the rejected heat, which typically ranges be- investigate to maximize surplus heat recovery.
tween 200 and 350
C [1,22]. However, recovering heat from the Conversion of surplus heat to power is challenging in general
electrolysis sidewalls is challenging because it can affect the ther- because of low cost-efficiencies [1]. Surplus heat sources with low
mal energy balance of the electrolysis process itself, the sidewalls specific energy density,1 such as aluminum smelter off-gas, intro-
are difficult to access and the smelter environment is chemically duce additional challenges; the heat recovery heat exchanger will
reactive [1]. Moreover, utilizing the heat rejected through the create additional off-gas pressure loss and thereby contribute to
sidewalls would require new infrastructure in connection to a large additional fan power consumption. Due to the large volume flow of
number of electrolysis cells, as evident from Fig. 1. the gas, the additional fan power could be significant compared to
Recovering the heat contained in the electrolysis off-gas is an the power production, and in some cases equal in magnitude to the
interesting alternative to utilization of heat rejected from the power output [7]. Thus, accounting for fan power during power
sidewalls. The off-gas mainly consists of air infiltrated into the cells cycle optimization is critical to system performance. As can be seen
and CO2 released in the electrolytic reduction process [5]. Air in the following, previous research has been insufficient in ac-
infiltration is necessary for operational reasons and to limit fluoride counting for the off-gas fan power during Rankine cycle optimi-
emissions, and significantly reduces the temperature of the off-gas zation. Even when the change in fan power has been calculated, the
[23]. The off-gas typically exits the cells at a temperature in the
range 150e180
C, depending on ambient temperature and other
factors [24e26]. Since the off-gas has a relatively low temperature, 1
Aluminum electrolysis off-gas is similar to air in thermodynamic properties at
the specific potential for power production is limited. However, the approximately atmospheric pressure and a temperature in the range of 150e180
C.

2
M. Nikolaisen and T. Andresen Energy 215 (2021) 118956

analyses are limited by fixed heat recovery heat exchanger pa- gas fan. Thus, the results cannot be generalized to plants with
rameters. Note that auxiliary power on the heat sink side could also different pressure loss in the gas collection and treatment system.
be of a significant magnitude, especially in the case of air cooling. Yu et al. [7] accounted for the change in off-gas fan power during
However, our study involves a water heat sink with a relatively Rankine cycle optimization. They formulated an equation for net
small ratio of pumping power to net power. power that included the additional fan power due to off-gas heat
Research on off-gas heat-to-power conversion has covered exchanger pressure loss, as well as the reduction in fan power due
different topics, such as optimization of Rankine cycle working fluid to off-gas cooling. They optimized the Rankine cycle using a multi-
and operating conditions [29,30]. The effect of an off-gas cooling objective optimization approach with the purpose of finding an
limit has also been investigated, which is necessary to avoid optimal compromise between net power, exergy efficiency and
corrosion problems caused by condensation of acidic components heat transfer surface area. The resulting net power was 1.2 MW,
[31]. Ladam et al. [31] investigated the effect of such a cooling limit accounting for 2.4 MW additional fan power due to off-gas heat
on Rankine cycles that recover heat from the source both directly exchanger pressure loss, and 1.3 MW reduction in fan power due to
and indirectly through an indirect heat transfer loop. Their results off-gas cooling. Although the study accounted for the change in fan
showed that the cooling limit had less impact on indirect cycles power during optimization, the analysis involved several simplifi-
than direct cycles, but the latter achieved higher power output. cations, such as fixed heat exchanger geometries (i.e. no heat
Other studies have analyzed the potential for implementation of exchanger geometry optimization) and a fixed off-gas pressure loss
Rankine cycles at existing aluminum plants in the Nordic countries. and heat transfer coefficient. Thus, the optimization tool had
For instance, Yu et al. [7] investigated both Rankine cycles and limited freedom to optimize performance, and there may be a
combined heat and power cycles at Alcoa’s plant in Fjardaal, Ice- higher net power with a more optimal compromise between power
land. They found that the Rankine cycles achieved the highest production and fan power. Furthermore, the results are only valid
exergy efficiency, whereas the combined heat and power cycles for the specific evaluated aluminum production facility. In any case,
achieved the highest energy efficiency. Another example is it is worth noting that the net increase in fan power, 1.1 MW, was
Børgund [32], who evaluated implementation of Rankine cycles at approximately the same as the net power, demonstrating the
Hydro’s plant in Øvre Årdal, Norway. One of her findings was that importance of accounting for fan power during cycle optimization.
Rankine cycles were more suitable than other power cycles, such as While a few studies have accounted for the change in off-gas fan
Stirling engines, steam cycles and Kalina cycles. A similar conclu- power in Rankine cycle analysis, the performance of the heat re-
sion was reached by Kolasin  ski [33], who concluded that the covery heat exchanger has not been optimized. Furthermore, no in-
Rankine cycle was a promising technology for heat-to-power con- depth analysis has been made of how fan power affects system
version in the metal smelting industry. performance under different boundary conditions. To close this
The design of heat exchangers for off-gas heat recovery is also an research gap, our study evaluates the effect of fan power on opti-
important research topic, since such heat exchangers are prone to mum system performance by optimizing both heat exchangers and
fouling and abrasion [34]. Fouling of heavy dust particles is critical cycle operating conditions simultaneously. We do not evaluate a
to performance and should be avoided, as it reduces heat transfer specific nominal pressure loss in the downstream gas collection
coefficient, increases pressure drop and requires expensive main- and treatment system. Instead, we evaluate different downstream
tenance [28]. We have assumed that fouling does not affect the conditions, and thus different potential for fan power reduction, by
performance of the heat recovery heat exchanger based on the including only parts of the fan power caused by heat exchanger
assumption that fouling can be avoided using circular channels pressure loss. The underlying assumption is that the part of fan
without surface enhancements [24,35]. power from heat exchanger pressure loss that is not included can
Our scope is to investigate the effect of off-gas fan power on be compensated for by an equivalent reduction in fan power from
optimum system performance, which is not a trivial task. As off-gas cooling. We analyze the two extreme cases, which include
mentioned, fan power increases due to off-gas pressure loss in the none and all of the fan power from heat exchanger pressure loss,
heat recovery heat exchanger. At the same time, off-gas cooling respectively. The optimal solution will lie somewhere between
contributes to a reduction in fan power, due to reduced volume flow these two extremities, depending on conditions in the downstream
in the downstream gas collection and treatment system [7,32]. The off-gas handling system. In addition to the two extreme cases, we
magnitude of the fan power reduction is not only dependent on the analyze a case that includes half of the fan power from heat
reduction in off-gas temperature and volume flow, but also on the exchanger pressure loss, representing a more practical solution.
total pressure loss in the downstream gas collection and treatment The analysis may contribute to improving the energy recovery
system, since fan power is proportional to both volume flow and potential in the smelting industry by enabling a better under-
pressure loss. Another important effect is that power production standing of how Rankine cycles and heat exchangers should be
can increase with higher off-gas pressure loss through the heat designed for optimum performance.
recovery heat exchanger, due to the positive correlation between
heat exchanger pressure loss and heat transfer coefficient [36]. 2. Methods
Some research articles have calculated the change in fan power
when implementing a Rankine cycle for heat-to-power conversion 2.1. Case description
from aluminum smelter off-gas. For example, Børgund [32] found
that Rankine cycle implementation resulted in a net decrease in fan Fig. 2 shows a sketch of a generic aluminum production facility
power compared to the baseline scenario. However, the cycle she in a scenario with heat-to-power conversion from the electrolysis
evaluated was not optimized, but rather simulated based on a off-gas. Several Rankine cycles are placed throughout the plant to
number of fixed parameters, such as an off-gas heat exchanger recover surplus heat close to the cells at a high off-gas temperature.
pressure loss of 1000 Pa and a pinch point temperature difference For simplicity, the system boundary of the analysis only encom-
of 15
C in the heat recovery heat exchanger. Børgund stressed that passes one Rankine cycle, recovering surplus heat from a cluster of
the calculated effect on fan power should only be seen as a pre- eight electrolysis cells. Due to the methodology used, considering
liminary indication due to the simplifications made. Furthermore, more or all the Rankine cycles would have no effect on optimization
the analysis is only valid for the specific evaluated aluminum plant, results.
with a nominal pressure difference of about 5000 Pa across the off- The off-gas exits the cells at a temperature of 150
C and is
3
M. Nikolaisen and T. Andresen Energy 215 (2021) 118956

Wet scrubber
and stack
Fan Gas
treatment System boundary
Off-gas duct system
center 80 °C
150 °C
RC RC RC RC RC RC

Rankine cycle Aluminium

electrolysis cell

RC RC RC RC RC RC

Fig. 2. Sketch of a generic primary aluminum production facility with heat-to-power conversion from off-gas.

collected in a duct system that transports the off-gas to the gas Table 2
treatment center, fan, wet scrubber and stack. The fan recovers the Heat sink specifications.

total off-gas pressure loss through the system. When there is no Heat sink Unit Value
surplus heat recovery, we refer to this pressure loss as the nominal Fluid e Water
pressure loss. Implementing a Rankine cycle will affect the nominal Inlet temperature (
C) 10
pressure loss and thereby the fan power in two ways: 1) additional Mass flow (kg/s) 23
pressure loss through the heat recovery heat exchanger, increasing Pressure (bar) 5.0
fan power, and 2) reduction in pressure loss due to off-gas cooling,
reducing fan power. Both effects are accounted for in our study.
Table 1 and Table 2 show the heat source and heat sink speci- adapted from a direct cycle to an indirect cycle, and a heat source
fications, respectively. The off-gas heat source is modelled as air, fan has been added. The heat exchangers modelled by Hagen et al.
which is assumed to provide sufficient thermodynamic accuracy. [36] were only geometrically described on the working fluid sides,
We have imposed a lower cooling limit of 80
C on the off gas to with the exception of the recuperator, which was described on both
avoid condensation of acidic components. The acid dew point de- sides. In our model, all heat exchangers are geometrically described
pends on many factors and is difficult to determine, and 80
C on both sides. The modifications to the model and assumptions
represents a typical conservative limit [37,38]. The heat sink is made by Hagen et al. [36] are summarized in Table 3. REPROP 9 was
cooling water at 10
C, which is a representative annual mean used for calculating thermodynamic properties [39] in both
temperature in the Nordic countries. The heat sink mass flow was models.
set to 23 kg/s, which was the optimal value resulting from pre- The cycle model contains several heat exchangers, as well as
liminary system optimizations, as described in Section 3 Results pumps, an expander, and a heat source fan. The heat exchanger
and discussion. model is described in Section 2.2 Heat exchanger model. The
models for the expander, pumps, and fan require the inlet state and

2.2. Rankine cycle model

The modelled Rankine cycle is sketched in Fig. 3. Heat is Heat source fan Off-gas heat
recovered indirectly from the heat source through a heat recovery source
heat exchanger (HRHE). A heat source fan is located downstream of Heat recovery
the HRHE to calculate the additional fan power caused by off-gas heat exchanger
heat exchanger pressure loss. The indirect heat transfer loop uses Indirect
pressurized water as heat transfer fluid, which is commonly used in fluid pump Indirect loop
industrial applications of aluminum off-gas energy recovery
[23,26]. An indirect system solution was chosen since the working
fluid is a flammable hydrocarbon, which receives heat from the
Evaporator
indirect fluid through the evaporator and produces power in a Working fluid
Rankine cycle. Excess heat is transferred to the heat sink in the Expander
condenser. The model is generic with respect to subcritical and
transcritical operation, which for all simulations in this work is
Recuperator

determined by the solver manipulating the variable expander inlet

pressure.
The cycle model has been developed based on the model
described by Hagen et al. [36]. In our work, their model has been
Working
fluid pump
Table 1
Heat source specifications.

Heat source Unit Value Condenser

Fluid e Air Heat sink
Inlet temperature (
C) 150
Lower cooling limit (
C) 80
Heat sink pump
Mass flow (kg/s) 12
Pressure (bar) 1.0
Fig. 3. Sketch of the modelled Rankine cycle.

4
M. Nikolaisen and T. Andresen Energy 215 (2021) 118956

Table 3
Modifications to the method used by Hagen et al. [36].

Hagen et al. [36] Our method

Heat source Pressurized water at 140
C. Electrolysis off-gas at 150
C.

Process Direct Rankine cycle without heat source Indirect Rankine cycle with heat source fan,
auxiliary equipment. pressurized water as heat transfer fluid.
Heat exchangers Geometries described on working fluid sides only. Geometries described on both hot and cold sides.
Working fluid side surface area used in calculations Hot and cold side average surface area used in
of heat exchanger area. calculations of heat exchanger area.
Working fluid Propane, propene, n-butane, mixture between propene Propene.
and n-butane, R-134a.

outlet pressure to be specified and are modelled with constant

_ net ¼ W_ exp  W_  W_ _
isentropic and mechanical efficiencies according to Eq. (1) through W wf pump sink pump  W ind pump
(4). From the saturated liquid state at the condenser outlet, the  DW_ (5)
fan;net
components are calculated in sequence, and fluid states are upda-
ted along the way. Two equality constraints are imposed to ensure
DW_ fan;net refers to the net change in fan power caused by imple-
that the pressure and temperature on each side of a stream split
menting a Rankine cycle, which is given by DW _
point are identical. Two inequality constraints requiring dry vapour fan;net ¼
at expander inlet and outlet are also included. Optimization vari- DW_ fan;HRHE  DW_ fan;cooling . The value of DW_ fan;HRHE is the addi-
ables, the objective function and additional constraints are tional fan power caused by heat exchanger pressure loss, and the
described in Table 6. value of DW_
fan;cooling is the reduction in fan power caused by off-gas
Pump and fan model:
cooling. Table 4 gives an overview of the two effects on fan power.
While the value of DW _ can be calculated based on off-gas
fan;HRHE
hðpout ; sin Þ  hin heat exchanger pressure loss, the value of DW_
his ¼ (1) fan;cooling depends on
hout  hin unknown factors outside the system boundary. To evaluate
different boundary conditions, we investigate three cases that
include different parts of the fan power caused by heat exchanger
_ pump ¼ ð1 = h _ out  hin Þ
mech Þmðh
W (2) pressure loss:

Expander model: Case 1. :

 Fan power is not included in the calculation of net power, i.e. D

hin  hout _
his ¼ (3) W ¼0
fan;net
hin  hðpout ; sin Þ
 For this to hold, the fan power reduction from off-gas cooling
would have to equal the fan power from heat exchanger pres-
sure loss
_ exp ¼ h mðh
W gen _ in  hout Þ (4)

Hagen et al. [36] compared the performance of five different Case 2. :

hydrocarbon working fluids. They allowed the optimization solver
to choose between subcritical and transcritical process design for  Half of the fan power from heat exchanger pressure loss
each fluid, and found that four of the five investigated fluids is included in the calculation of net power. i.e.
resulted in transcritical configuration under optimal conditions. DW_ ¼ 0:5,DW
fan;net
_
fan;HRHE
The only fluid that resulted in subcritical configuration under  This case could represent a scenario where fan power reduction
optimal conditions was n-butane, which achieved lower power equals half the fan power from heat exchanger pressure loss
output than the other fluids for the same values of total heat
transfer area. The poor performance of the subcritical fluid owed in
part to relatively low heat transfer coefficients. Based on the results Case 3. :
from Hagen et al. [36] and the similar conditions in our work, we
have chosen propene as the working fluid in this study.  The total fan power from heat exchanger pressure loss is
The Rankine cycle net power is given by Eq. (5): included in the calculation of net power, i.e.
DW_ fan;net ¼ DW_ fan;HRHE
 This case disregards the reduction in fan power, and as such
Table 4
represents the strictest boundary conditions
Description of the two different effects on fan power.

Effect on Description Value

fan power
2.3. Heat exchanger model
DW_ fan;HRHE Additional fan power caused by off-gas Known
pressure loss in the heat recovery heat Function of pressure loss in
exchanger heat recovery heat The heat exchanger model is based on the generic heat
exchanger exchanger model described by Hagen et al. [36], which is a
DW_ fan;cooling Reduction in fan power caused by off- Unknown simplified heat exchanger model that is able to account for heat
gas cooling and reduced pressure loss Function of pressure loss transfer mechanisms and calculate heat transfer area without
outside the system boundary outside system boundary
specifying the heat exchanger type. The model simply assumes that
5
M. Nikolaisen and T. Andresen Energy 215 (2021) 118956

Table 5 optimization variables and constraints are given in Table 6. Based

Heat transfer and pressure drop correlations. on the value of optimized variables and boundary conditions, all
Flow Heat transfer Pressure loss operating conditions in the system can be calculated. Note that
Single-phase Gnielinski [40] Selander [41]
expander inlet pressure is a free optimization variable, which
Two-phase Boyko and Friedel [43] with means that the solver can choose between subcritical and tran-
Kruzhilin [42] single-phase formulation scritical process designs.
(condensation) by Selander [41] Each case was subjected to the same constraint on a total heat
transfer surface area of 750 m2 to ensure a fair comparison between
the different cases. Note that this means the heat exchangers were
not constrained by a minimum pinch point temperature difference.
With a constraint on total surface area, the solver will search for the
the hot and cold side fluids flow counter-currently through chan-
heat exchanger geometry parameters that yield the optimal heat
nels, and only specifies the hydraulic channel diameters, cross-
transfer area distribution between the four heat exchangers, as well
sectional flow areas and channel length, e.g. dh , Ac and L. The hy-
as the optimal trade-offs between overall heat transfer coefficients
draulic diameters and flow areas differ on the hot and cold sides,
and pressure losses. A heat transfer area of 750 m2 was set in this
whereas the channel length is the same. The heat transfer area on
particular case to represent a relatively large system with small
each side of the heat exchanger is given by Acold=hot ¼ Phot=cold , L,
temperature differences, utilizing most of the available potential in
where P refers to the perimeter, given by P ¼ ð4 ,Ac Þ= dh.
the heat source. Furthermore, the maximum off-gas heat exchanger
Based on these geometry parameters and fluid state points in
pressure loss was set to 5000 Pa to avoid unlimited pressure loss in
one end of the heat exchanger, the heat exchanger model calculates
Case 1, where fan power is not included in the objective function.
heat duty, overall heat transfer coefficient, pressure loss and heat
Observe that heat exchanger length is an optimization variable,
transfer area. The calculations are performed by solving three dif-
meaning that the optimizer can choose a recuperator length of 0 if
ferential equations in a set of equidistant numerical integration
this is optimal. Thus, a recuperator is only included if it results in a
steps, as described by Hagen et al. [36]. Local fluid properties, heat
higher net power.
transfer coefficients and pressure gradients are evaluated for each
Fixed parameters are given in Table 7. A number of parameters
integration step.
were fixed to simplify the optimization problem, including the
In the absence of in-depth economic functions per component,
cross-sectional flow area on the heat sink side of the condenser,
the sum of all heat exchanger areas could be considered as an initial
fixed to 100 cm2, and the working fluid outlet pressure at the
representation of system costs. At initial design stages, it is not
condenser outlet, fixed to 10 bar. Heat exchanger hydraulic di-
obvious whether the hot or cold side of the heat exchangers has the
ameters were also fixed during optimization. It was assumed that
highest cost, and therefore the average area is used in the calcu-
sufficient heat exchanger optimization flexibility is achieved by
lation of total area, i.e. AHX ¼ ðAcold þ Ahot Þ=2. The total heat
optimizing heat exchanger length and cross-sectional flow areas.
transfer surface area in the Rankine cycle is given by Atot ¼ AHRHE þ
This assumption is evaluated in Section 3 Results and discussion.
Aevap þ Acond þ Arec .
Heat transfer and pressure loss correlations are given in Table 5,
and are the same as those applied by Hagen et al. [36]. As Hagen 3. Results and discussion
et al. [36] pointed out, these correlations could be considered
suitable to apply in the generic heat exchanger model, even for Table 8 shows the main results from system optimization. Note
fluids not considered during development of the correlations. Note that an overall system optimization was performed for each of the
that we have not reported a correlation for boiling since all opti- cases. Observe first Case 1, where fan power was not included
mizations resulted in transcritical operation. during optimization, i.e. DW_ ¼ 0. For this reason, the opti-
fan;net
mizer makes no effort to minimize off-gas pressure loss, which
2.4. System optimization reaches the upper limit of 5000 Pa. As a result, fan power from off-
gas heat exchanger pressure loss becomes relatively high,
An overall system optimization was performed for each of the amounting to almost half of the power produced in the expander.
three cases, which involves maximizing net power by optimizing This is likely an unfeasible solution, demonstrating the need for
cycle operating conditions and heat exchanger geometries simul- including fan power during optimization, or alternatively to set a
taneously. The system is optimized using the gradient-based con- more practical upper limit on maximum off-gas pressure loss.
strained optimization solver NLPQL [44]. The objective function, However, it is unlikely that setting such a limit results in the most

Table 6
Objective function, optimization variables and constraints.

Symbol Description

Objective function W_ net Net power

Process optimization variables m_ wf Working fluid mass flow
m_ ind Indirect fluid mass flow
hin;exp Expander inlet enthalpy
pin;exp Expander inlet pressure
Tout;HRHE;ind Indirect fluid outlet temperature from the HRHE
Heat exchanger optimization variables L Channel length
Ac;hot Hot side cross-sectional flow area
Ac;cold Cold side cross-sectional flow area
Constraints Atot Total heat transfer surface area
DpHRHE;max Max HRHE off-gas pressure loss

6
M. Nikolaisen and T. Andresen Energy 215 (2021) 118956

Table 7 the cases only involve different formulations of fan power from
Fixed parameters. heat exchanger pressure loss. The change in expander and pumping
Parameter Unit Value power indicates a coupling between cycle and heat exchanger
Pumps Isentropic efficiency e 0.70
performance. Table 9 shows the optimized process and heat re-
Motor efficiency e 0.95 covery heat exchanger parameters, which demonstrate the
Expander Isentropic efficiency e 0.85 connection between heat exchanger and process parameters. For
Generator efficiency e 0.95 example, off-gas pressure loss, overall heat transfer coefficient and
Fan Isentropic efficiency e 0.90
heat duty in the HRHE decrease from Case 1 to 3, while the surface
Motor efficiency e 0.95
Heat exchangers Total heat transfer surface area (m2) 750 area and minimum temperature difference increase. At the same
HRHE Hydraulic diameter, cold side (mm) 10 time, expander inlet enthalpy and pressure decrease, explaining
Hydraulic diameter, hot side (mm) 60 the reduction in expander power. Since working fluid outlet pres-
Max HRHE off-gas pressure loss (Pa) 5000
sure from the condenser is fixed, a lower expander inlet pressure is
Evaporator Hydraulic diameter, cold side (mm) 10
Hydraulic diameter, hot side (mm) 20
coupled to a smaller pressure lift in the working fluid pump and
Condenser Hydraulic diameter, hot side (mm) 20 therefore lowers the pumping power. Since these parameters are all
Hydraulic diameter, cold side (mm) 20 the result of a complex overall system optimization, it is difficult to
Cold side cross-sectional flow area (cm2) 100 identify a direct link between the different parameters. However, it
Working fluid outlet pressure (bar) 10
is obvious that there is a connection between heat exchanger and
Recuperator Hydraulic diameter, hot side (mm) 20
Hydraulic diameter, cold side (mm) 10 cycle performance, and that optimum heat exchanger and cycle
design both depend on plant-specific factors outside the system
boundary.
As mentioned, the surface area of the heat recovery heat
optimal design of the heat recovery heat exchanger. Instead, we
exchanger increases from Case 1 to Case 3. This is likely caused by
should allow the optimizer to choose the most optimal off-gas
the reduction in off-gas pressure loss and overall heat transfer co-
pressure loss based on the effect on fan power, as done in Case 2.
efficient, which can be observed in Table 9. Observe also that the
In Case 2, half of the fan power from off-gas heat exchanger
_ HRHE length decreases from Case 1 to Case 3, even though the
pressure loss was included during optimization, i.e. DW ¼ fan;net surface area increases. This is because the optimizer instead
0:5,DW _
fan;HRHE . With this formulation of fan power, the optimizer chooses to increase cross-sectional flow areas, which more than
has an incentive to reduce off-gas pressure loss, yielding a signifi- doubles on the off-gas side from Case 1 to Case 3. This indicates that
cantly lower pressure loss and fan power than in Case 1. Net power optimizing only heat exchanger length would not provide a suffi-
is lower than in Case 1, not only because we have included fan cient degree of heat exchanger design freedom.
power in the calculation of net power, but also because expander The cross-sectional flow area on the hot side of the heat re-
power is lower. covery heat exchanger is between 100 and 200 times larger than
Case 3 accounts for the entire fan power from heat exchanger the flow area on the cold side. This is a relatively high ratio, but
pressure loss during optimization. This represents the strictest could be possible in novel heat exchanger concepts for smelter off-
boundary conditions, where a potential reduction in fan power is gas heat recovery [45].
disregarded. As a result, the optimizer chooses an even lower off- Table 10 shows the optimized evaporator, condenser and recu-
gas pressure loss and equivalent fan power than in Case 2. This perator parameters. Observe that the heat transfer surface areas of
case results in the lowest net power, both because of a high net fan these heat exchangers decrease from Case 1 to Case 3, which is
power and a low expander power. coupled to the increasing area requirement of the heat recovery
Case 1 and Case 3 account for no and all of the fan power from heat exchanger and a fixed total area. Note also that the optimized
off-gas pressure loss, respectively. In practice, the optimal system recuperator length is above zero, implying that recuperation
solution will lie somewhere between Case 1 and Case 3, depending resulted in higher net power.
on the potential for reduction in fan power. We need more Since several process and geometry parameters were optimized
knowledge of processes downstream of the system boundary to simultaneously, there is a possibility that the global optimum so-
calculate the exact fan power reduction. However, we can identify lution has not been found. To simplify the optimization problem,
two important factors that will impact the fan power reduction. The
first is the value of pressure loss through the main processes such as
the gas treatment center, wet scrubber and stack. If this pressure
Table 9
loss is high, off-gas cooling has the potential to yield a higher fan Optimized process and heat recovery heat exchanger parameters.
power reduction. The other factor is the design of the duct system
and the position of the heat recovery heat exchanger in the system; Unit Case 1 Case 2 Case 3

placing this unit close to the electrolysis cells will give a higher Process
potential for pressure loss reduction in the ducts and vice versa. m_ wf (kg/s) 2.0 2.0 2.0
Investigating these factors is an interesting path for future work. m_ ind (kg/s) 2.8 3.2 3.3
hin;exp (kJ/kg) 740 720 720
Expander and pumping power differ in Cases 1e3, even though
pin;exp (bar) 51 47 45
Tin;HRHE;cold (
C) 65 64 65
HRHE
L (m) 35 22 18
Table 8
Ac;hot (cm2) 4200 7100 9000
Main results from system optimization.
Ac;cold (cm2) 39 50 53
Cases W_ net W_ exp W_ pumps DW_ f an;net DW_ f an;HRHE DpHRHE AHRHE (m2) 510 560 570
(kW) (kW) (kW) (kW) (kW) (Pa) Q_ (kW) 850 850 820
U (W/m2K) 120 80 68
Case 1 118 147 29.2 0.00 71.1 5000
Dphot (Pa) 5000 1200 680
Case 2 99.3 135 26.8 8.75 17.5 1230
Dpcold (kPa) 23 12 9
Case 3 93.1 129 25.7 9.72 9.72 680
DTmin (
C) 13 16 18

7
M. Nikolaisen and T. Andresen Energy 215 (2021) 118956

we chose to fix some of the parameters that varied little during Table 11

preliminary optimizations. These parameters include the working Optimized condenser working fluid outlet pressure and heat sink mass flow
resulting from preliminary optimizations.
fluid pressure at the condenser outlet ðpcond;wf Þ and the heat sink
mass flow (m_ sink ). Table 11 shows the optimized values of pcond;wf Cases pcond;wf m_ sink
(bar) (kg/s)
and m_ sink resulting from the preliminary optimizations, demon-
strating a relatively small variation from Case 1 to Case 3, with Case 1 10.5 21.9
Case 2 10.4 22.7
average values of 10 bar and 23 kg/s, respectively (which were set
Case 3 10.5 22.4
as the fixed values). Note that the fixed values deviate slightly from Fixed value 10 23
the average values in Table 11, owing to results from two cases that
were left out.
During the preliminary optimizations, we also optimized the Table 12
hydraulic diameter on the off-gas side of the heat recovery heat Optimized hydraulic diameter on off-gas side of the
exchanger (dh;gas Þ. Table 12 shows the optimized hydraulic di- heat recovery heat exchanger resulting from pre-
liminary optimizations.
ameters in Case 1e3, and Table 13 shows the change in main system
optimization results when adding dh;gas , pcond;wf and m_ sink as opti- Cases dh;gas
mization variables. Table 12 demonstrates an increasing trend in (mm)

hydraulic diameter from Case 1 to Case 3, but Table 13 shows a Case 1 55.9
small impact on net power (only 0.2%) and other optimization re- Case 2 60.7
Case 3 66.7
sults. In fact, net power is reduced in two out of three cases, indi-
Fixed value 60
cating reduced probability of finding the global optimum with an
excessive number of optimization variables. These observations
also indicate that optimizing hydraulic diameters is not critical to
heat exchanger performance, and supports the assumption that Table 13
sufficient heat exchanger optimization flexibility can be achieved Change in system optimization result when adding condenser working fluid outlet
pressure, heat sink mass flow and off-gas side hydraulic diameter as optimization
by optimizing only the cross-sectional flow-area and heat
variables.
exchanger length.
Cases W_ net W_ exp W_ pumps DW_ f an;net DW_ f an;HRHE DpHRHE

Case 1 0.2% 1.1% 4.7% 0.0% 0.0% 0.0%

4. Conclusions Case 2 0.2% 0.6% 4.0% 0.5% 1.1% 1.6%
Case 3 0.2% 1.4% 5.8% 1.3% 1.3% 1.5%
We have optimized Rankine cycle and heat exchangers for po-
wer production from aluminum smelter off-gas. The objective was
to investigate the effect of fan power on optimum system perfor- heat exchanger pressure loss. In this way, we could evaluate
mance. We developed three cases that accounted for the reduction different conditions in the downstream gas collection and treat-
in fan power by including different parts of the fan power caused by ment system.
Table 10 Even though each case was optimized with the same constraint
Optimized evaporator, condenser and recuperator parameters. on total heat exchanger surface area, they resulted in significantly
Unit Case 1 Case 2 Case 3
different cycle and heat exchanger performance. When fan power
was neglected during system optimization, off-gas pressure loss
Evaporator
reached the upper limit and fan power became unfeasibly high.
L (m) 77 68 65
Ac;hot (cm2) 28 27 27 However, a high off-gas pressure loss resulted in an efficient heat
Ac;cold (cm2) 39 35 33 recovery heat exchanger and a high expander power. On the other
Aevap (m2) 81 65 59 hand, when we accounted for the entire fan power from heat
Q_ (kW) 850 850 820 exchanger pressure loss, optimization resulted in a relatively low
U (W/m2K) 2300 2700 2900 off-gas pressure loss and equivalent fan power. This restricted the
Dphot (kPa) 40 48 47 performance of the heat recovery heat exchanger and resulted in a
Dpcold (kPa) 86 88 89
DTmin (
C) 0.7 1.1 1.1
relatively low expander power. In practice, the optimum solution
Condenser depends on the exact value of fan power reduction, which is gov-
L (m) 48 48 43 erned by the total off-gas pressure loss outside the system
Ac;hot (cm2) 150 120 120 boundary. Thus, the optimal system solution is unknown, but lies
Ac;cold (cm2) 100 100 100 somewhere in the range between the two extremities that account
Acond (m2) 120 110 98
for either none or all of the fan power from heat exchanger pressure
Q_ (kW) 720 730 700
loss.
U (W/m2K) 1300 1400 1400
Dphot (kPa) 17 24 24 Our analysis also shows that the optimum heat recovery heat
Dpcold (kPa) 140 140 130 exchanger design varies considerably in the different cases,
DTmin (
C) 2.4 2.6 3.0 implying that the optimum design depends on the potential for
Recuperator downstream fan power reduction. For example, if the heat recovery
L (m) 16 11 11
Ac;hot (cm2) 180 150 150
heat exchanger is placed close to the electrolysis cells, the potential
Ac;cold (cm2) 27 25 23 for pressure loss reduction in the ducts is higher and the heat re-
Arec (m2) 38 22 22 covery heat exchanger can be designed with a higher pressure loss.
Q_ (kW) 110 84 79 The results also show that there is a strong connection between
U (W/m2K) 460 520 520 heat exchanger and cycle performance. Thus, downstream condi-
Dphot (kPa) 4.3 3.8 3.9 tions strongly impact both heat exchanger and cycle performance,
Dpcold (kPa) 17 13 15
(
C)
and should be accounted for during system design. An interesting
DTmin 3.0 4.7 4.2

8
M. Nikolaisen and T. Andresen Energy 215 (2021) 118956

path for future work is to consider the effect of plant-specific off- system. Energy 2017;141:1084e96.
[14] Song P, Wei M, Zhang Y, Sun L, Emhardt S, Zhuge W. The impact of a bilateral
gas handling conditions on system performance.
symmetric discharge structure on the performance of a scroll expander for
ORC power generation system. Energy 2018;158:458e70.
Credit author statement [15] Declaye S, Quoilin S, Guillaume L, Lemort V. Experimental study on an open-
drive scroll expander integrated into an ORC (Organic Rankine Cycle) system
with R245fa as working fluid. Energy 2013;55:173e83.
Monika Nikolaisen: Conceptualization, Methodology, Software, [16] Manente G, Lazzaretto A, Bonamico E. Design guidelines for the choice be-
Investigation, Writing - original draft, Writing - review & editing, tween single and dual pressure layouts in organic Rankine cycle (ORC) sys-
tems. Energy 2017;123:413e31.
Visualization, Project administration Trond Andresen: Conceptu-
[17] Rohde D, Hagen BAL, Trædal S, Andresen T. Comparison of rankine cycle and
alization, Methodology, Validation, Writing - review & editing, trilateral flash cycle for power production from low temperature heat sources.
Visualization, Supervision, Project administration, Funding Refrigeration Science and Technology 2018.
acquisition [18] Braimakis K, Karellas S. Integrated thermoeconomic optimization of standard
and regenerative ORC for different heat source types and capacities. Energy
2017;121:570e98.
Declaration of competing interest [19] Elsido C, Mian A, Martelli E. A systematic methodology for the techno-
economic optimization of Organic Rankine Cycles. Energy Procedia
2017;129:26e33.
The authors declare the following financial interests/personal [20] Barzi YM, Assadi M, Parham K. A waste heat recovery system development
relationships which may be considered as potential competing and analysis using ORC for the energy efficiency improvement in aluminium
interests: The authors are employed by the research institute SIN- electrolysis cells. Int J Energy Res 2018;42(4):1511e23.
[21] Barzi YM, Assadi M. Heat transfer and thermal balance analysis of an
TEF Energy Research, wherein their work has been funded by the aluminum electrolysis cell side lines: a heat recovery capability and feasibility
HighEFF and COPRO projects, which target improvement of in- study. In: ASME international mechanical engineering congress and exposi-
dustrial energy efficiency. tion, proceedings (IMECE); 2013.
[22] Zhao R, Nowicki C, Gosselin L, Duchesne C. Energy and exergy inventory in
aluminum smelter from a thermal integration point-of-view. Int J Energy Res
Acknowledgements 2016;40(10):1321e38.
[23] Sørhuus A, Wedde G. Pot gas heat recovery and emission control. In: Light
metals. Cham: Springer; 2016. B. G., D. M., and T. G., Editors.
This publication has been funded by the research projects [24] Qassab HAA, Mohd SSAA, Wedde G, Sørhuus A. HEX retrofit enables smelter
HighEFF and COPRO. HighEFF is a Center for an Energy Efficient and capacity expansion. In: Suarez CE, editor. Light metals 2012. Cham: Springer
Competitive Industry for the Future, a 8-year Research Center un- International Publishing; 2016. p. 815e20.
[25] de Gromard A, Lim C, Bouhabila EH, Cloutier B, Frainais M. Development on
der the FME-scheme (Center for Environment-friendly Energy electrolytic cell gas cooling. In: Grandfield J, editor. Light metals 2014. Cham:
Research, grant no. 257632/E20). COPRO was a four-year compe- Springer International Publishing; 2016. p. 623e8.
tence building project within industrial surplus-heat-to-power [26] Sørhuus AK, Ose S, Nilsen BM. Possible use of 25 MW thermal energy
recovered from the potgas at alba line 4. In: Hyland M, editor. Light metals
conversion (EnergiX grant no. 255016/E20). The authors grate- 2015. Cham: Springer International Publishing; 2016. p. 631e6.
fully acknowledge the financial support from The Research Council [27] Nowicki C, Gosselin L, Duchesne C. Waste heat integration potential assess-
of Norway and user partners of HighEFF and COPRO. ment through exergy analysis in an aluminum production facility. Energy
Technology. In: Carbon dioxide management and other technologies. Hobo-
The authors acknowledge Nancy Eik-Nes, Stian Trædal and ken, NJ, USA: John Wiley & Sons Inc; 2012. 2012.
Brede Hagen for their inputs on language and presentation of the [28] Fleer M. Heat recovery from the exhaust gas of aluminum reduction cells.
results. Iceland: Reykjavík University; 2010.
[29] Castelli AF, Elsido C, Scaccabarozzi R, Nord LO, Martelli E. Optimization of
organic rankine cycles for waste heat recovery from aluminum production
References plants. Frontiers in Energy Research 2019;7(44).
[30] Wang Z, Zhou N, Luo L, Guo J. Thermodynamic analysis of power generation
[1] Cascella F, Gaboury S, Sorin M, Teyssedou A. Proof of concept to recover system based on waste heat from aluminium reduction cell. Light Metals;
thermal wastes from aluminum electrolysis cells using Stirling engines. En- 2010.
ergy Convers Manag 2018;172:497e506. [31] Ladam Y, Børgund M, Næss E. Influence of heat source cooling limitation on
[2] Cullen JM, Allwood JM. Mapping the global flow of aluminum: from liquid ORC system layout and working fluid selection: the case og aluminium in-
aluminum to end-use goods. Environ Sci Technol 2013;47(7):3057e64. dustry. In: TMS light metals; 2014.
[3] Milford RL, Allwood JM, Cullen JM. Assessing the potential of yield improve- [32] Børgund MA. Power production from low temperature aluminium electrolysis
ments, through process scrap reduction, for energy and CO2 abatement in the cell off-gases. In: Department of energy and process engineering. Norwegian
steel and aluminium sectors. Resour Conserv Recycl 2011;55(12):1185e95. University of Science and Technology; 2009.
[4] Produksjon og forbruk av energi. energibalanse. [Accessed 2020-01-28]; [33] Kolasin  ski P. In: The application of ORC systems for waste heat recovery in
Available from: https://www.ssb.no/energi-og-industri/statistikker/ metal smelting industry; 2019. ORC2019.
energibalanse. [34] Bouhabila EH, Næss E, Einejord VK, Kristjansson K. An innovative compact
[5] Nowicki C, Gosselin L. An overview of opportunities for waste heat recovery heat exchanger solution for aluminum off-gas cooling and heat recovery. In:
and thermal integration in the primary aluminum industry. JOM 2012;64(8): Sadler BA, editor. Light metals 2013. Cham: Springer International Publishing;
990e6. 2016. p. 793e7.
[6] Ladam Y, Solheim A, Segatz M, Lorentsen O-A. Heat recovery from aluminium [35] Sorhuus A, Rye K, Nyland G. Increased energy efficiency and reduced HF
reduction cells. Light Met 2011:393e8. 2011. emission with new heat exchanger. Light metals. 2010. p. 249e54.
[7] Yu M, Gudjonsdottir MS, Valdimarsson P, Saevarsdottir G. Waste heat re- [36] Hagen BAL, Nikolaisen M, Andresen T. A novel methodology for Rankine cycle
covery from aluminum production. In: Minerals, metals and materials series; analysis with generic heat exchanger models Applied Thermal Engineering.
2018. p. 165e78. 2020.
[8] Yang F, Zhang H, Yu Z, Wang E, Meng F, Liu H, Wang J. Parametric optimization [37] Solheim A, Senanu S. Recycling of the flue gas from aluminium electrolysis
and heat transfer analysis of a dual loop ORC (organic Rankine cycle) system cells. In: Light metals 2020. Cham: Springer International Publishing; 2020.
for CNG engine waste heat recovery. Energy 2017;118:753e75. [38] Aarhaug TA, Ratvik AP. Aluminium primary production off-gas composition
[9] Guillaume L, Lemort V. Comparison of different ORC typologies for heavy-duty and emissions: an overview. JOM 2019;71(9):2966e77.
trucks by means of a thermo-economic optimization. Energy 2019;182: [39] Lemmon EW, Huber ML, McLinden MO. NIST standard reference database 23:
706e28. reference fluid thermodynamic and transport properties-REFPROP. 2013.
[10] Usman M, Imran M, Yang Y, Lee DH, Park BS. Thermo-economic comparison of Version 9.1.
air-cooled and cooling tower based Organic Rankine Cycle (ORC) with R245fa [40] Gnielinski V. New equations for heat and mass transfer in turbulent pipe and
and R1233zde as candidate working fluids for different geographical climate channel flow, vol. 16; 1976. p. 359e68. April.
conditions. Energy 2017;123:353e66. [41] Selander WN. Explicit formulas for the computation of friction factors in
[11] Mohammadzadeh Bina S, Jalilinasrabady S, Fujii H. Energy, economic and turbulent pipe flow. Chalk River, Ontario, Canada: Chalk River Nuclear Labs;
environmental (3E) aspects of internal heat exchanger for ORC geothermal 1978.
power plants. Energy 2017;140:1096e106. [42] Boyko LD, Kruzhilin GN. Heat transfer and hydraulic resistance during
[12] Orosz MS, Mueller A, Quolin S, Hemond H. Small Scale Solar ORC system for condensation of steam in a horizontal tube and in a bundle of tubes. Int J Heat
distributed power. SolarPaces 2009. Mass Tran 1967;10(3):361e73.
[13] Pantano F, Capata R. Expander selection for an on board ORC energy recovery [43] Friedel L. Improved friction pressure drop correlations for horizontal and

9
M. Nikolaisen and T. Andresen Energy 215 (2021) 118956

vertical two-phase pipe flow, vol. 18; 1979. p. 485e91. from challenging industrial off-gas e model-based investigation of a novel
[44] Schittkowski K. NLPQL: a fortran subroutine solving constrained nonlinear heat recovery concept. In: IIR international rankine 2020 conference -heating.
programming problems. Ann Oper Res 1986;5(1):485e500. Glasgow, UK: Cooling and Power Generation; 2020.
[45] Skjervold VT, Skaugen G, Andresen T, Nekså P. Enabling power production

10