Journal of Energy Storage 6 (2016) 105–115

Contents lists available at ScienceDirect

Journal of Energy Storage

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

Sodium nickel chloride battery steady-state regime model for

stationary electrical energy storage
Sebastian Dambone Sessaa,* , Giorgio Crugnolab , Marco Todeschinib , Stefano Zinb ,
Roberto Benatoa
a
Department of Industrial Engineering, University of Padova, Via Gradenigo 6/A, 35131 Padova, Italy
b
Fiamm Sonick, Viale Europa 75, 36075 Montecchio Maggiore (VI), Italy

A R T I C L E I N F O A B S T R A C T

Article history: The purpose of this paper is presenting a reliable modelling of sodium-nickel chloride battery in order to
Received 26 September 2015 have a powerful tool which is able to foresee the steady state battery behaviour in both discharge and
Received in revised form 18 March 2016 charge operations. The proposed modelling approach allows representing both constant current
Accepted 19 March 2016
operations and variable charge or variable discharge current ones, but it does not allow passing
Available online 25 March 2016
instantaneously from one battery mode to another one. This method is based on experimental measures.
All the main modelling steps are described and a comparison between the model results and real battery
Keywords:
measures, with the same conditions, is presented. The very good agreement between measures and
Sodium nickel chloride batteries
Stationary electrical energy storage
model confirms the robustness of the approach for steady state applications. The paper proposes to adopt
Battery modelling a set of standard battery measures from which it is possible to infer a simple but very precise modelling
structure.
ã 2016 Elsevier Ltd. All rights reserved.

1. Introduction
In the paper, the steady state modelling of a NaNiCl2 storage
cell, part of Na-beta battery family, is described. This is one of the
most promising technologies, in the short term, for load levelling,
voltage regulation, time shifting and power fluctuation mitigation
of the renewable energy sources in High Voltage networks [1].
Some installations of NaNiCl2 battery for large-scale stationary
storage purpose in the high voltage network are described in Refs.
[2,3].
The most common battery modelling approaches available in
literature [4–8] do not consider neither the cell resistance
dependence on the current rate and DoD, nor the cell voltage
dependence on DoD and temperature contemporaneously, which
are typical NaNiCl2 battery behaviours. The existing Sodium Nickel
Chloride battery models [9,10] are very complicated because they
use several parameters, such as different resistances and
capacitances, whose exact values have to be obtained. This is
due to the fact that these approaches are conceived for electrical
vehicle applications [11,12]. The chief characteristic of this model is
its seeming simplicity, since the NaNiCl2 cell is represented by a
voltage source with a non-linearly variable resistance in series,
* Corresponding author.
E-mail address: sebastian.dambonesessa@unipd.it (S. Dambone Sessa).
which depends upon Depth of Discharge (DoD) and current rates.
The aim of this paper is demonstrating that only one variable
resistance is sufficient to represent with very good precision the
stationary working of NaNiCl2 batteries, by starting from the cell
measures of the:

complete charge/discharge Open Circuit Voltage (OCV) as a
function of Depth of Discharge (DoD);

internal resistance for different current rates as a function DoD
(the resistance dependence on the temperature has been
experimentally demonstrated to be negligible. This is due to
the fact that the cell temperature is maintained in the optimum
temperature range of 260  C–300  C by means of some heaters.
This guarantees an optimal and almost constant ionic conduc-
tivity);

temperature for different current rates and DoD.
Therefore, the described model does not consider transient
phenomena during the cell operation, but it is a steady-state
model. This choice is suitable for studying and simulating the effect
of electrochemical storage in the electrical network with the
purpose of stationary applications, but it is not suitable to
represent the behaviour of a battery installed on a vehicle. Paper
[9] provides a very good approach to represent vehicular
applications of NaNiCl2 batteries. Nevertheless, the model can
be considered a “dynamic model” as well. This is due to the fact

http://dx.doi.org/10.1016/j.est.2016.03.005
2352-152X/ ã 2016 Elsevier Ltd. All rights reserved.
106 S. Dambone Sessa et al. / Journal of Energy Storage 6 (2016) 105–115
Table 1
Nominal NaNiCl2 cell electric characteristics [2,3].
Average open circuit discharge voltage
End-of-discharge voltage
Capacity [Ah]
Typical discharge regime
Electrochemical Efficiency [%]
Square section side [mm]
Length [mm]
Weight [kg]
Volume [m3]
Freeze-thaw
Volumetric energy density [kWh/m3] or [Wh/l]
Gravimetric energy density [Wh/kg]
Operating temperature [ C]
that the cell internal resistance varies in a non-linear way during
the charge/discharge operation, so that the cell behaviour during
charge and discharge continuously changes. Therefore, the model
has to reproduce “dynamically” these variations.
The chosen modelling structure allows adapting beyond to one
cell also to a NaNiCl2 module (240 cell series connected) or units
(64 modules parallel connected) [13] very easily. In order to test
the model, the simulation responses have been compared with real
cell charge/discharge measurements with a very good agreement.
The model is wholly based on experimental data which are DoD,
current rate and temperature dependent.
1.1. Cell characteristics
The basic cell structure consists of an anode part, made by
molten sodium (Na), a solid ceramic electrolyte, namely beta-
alumina (b00 -Al2O3) [14,15] and a cathode part made by nickel
chloride (NiCl2) with the addition of metallic doping substances
[16]. One of the main peculiarities of this technology is that its solid
electrolyte presents a very low ionic resistivity in the temperature
interval between 250  C and 300 C, with an operating temperature
equal to 260  C [17,18].
Fig. 1. Discharge reaction scheme of NaNiCl2 storage cell doped with iron.
2.6
2.33 V
38
from 3 to 5 h (current from C/3 to C/5 A)
90
36
236
0.680
3  104
no limitations
280
140
260
The main parameters of the cell are reported in Table 1.
2. Theory and calculation
2.1. Basic NaNiCl2 cell discharge electrochemistry
By measuring voltage and current during the discharge
operation of one NaNiCl2 cell, it is possible to infer an equivalent
electrical circuit which is able to simulate the cell behaviour.
In particular, if the cell is supplying a load (see Fig. 1), a
difference between the OCV of the cell and the On-Load one (OLV)
at different DoD can be observed. The cell OCV in discharge mode is
bigger than the OLV. Consequently, this voltage drop can be
modelled as an electrical non-linear resistance. In order to
understand the cell internal resistance and OCV variations during
the discharge process, the basic NaNiCl2 cell electrochemistry has
to be analysed. By considering the scheme in Fig. 1, once the switch
s is closed, the sodium anode releases electrons and, as a
consequence, sodium positive ions are formed in the anode zone.
Consequently, an electric current flows through the circuit and
the sodium ions pass through the beta-alumina to the cathode
zone of the cell. In the latter zone, the positive ions and the
 S. Dambone Sessa et al. / Journal of Energy Storage 6 (2016) 105–115 107

Fig. 2. (a) Measured NaNiCl2 cell discharge OCV at the cell nominal temperature of 260  C; (b) Cell resistance variation as a function of DoD and discharge current rates.

released electrons react with the cathode nickel-chloride with the presents values which are suitable for an optimal discharge
following reaction [19]: performance. Moreover Fig. 2b shows that besides the DoD, the cell
resistance varies as a function of the discharge current rates as
NiCl2 + 2Na+ + 2e ! 2NaCl + Ni
well.
This reaction takes place close to the ceramic electrolyte, but The cell resistance in discharge mode was calculated by
during the discharge process the reaction front, shown in Fig. 1, measuring the voltage drop which occurs by passing from the
progressively comes close to the current collector, and the cell open circuit operation to the on-load one, for 100 DoD intervals
propagation distance of Na+ ions increases consequently. This at a constant discharge current rate, by means of the following
determines a rise in the cell electrical internal resistance, which is equation:
therefore proportional to the DoD. In order to limit this rise in the
rd = (OCV(DoD, I)  OLV(DoD, I))/I
internal resistance, doping metallic additives are added in the
cathode. The chosen doping metals have to react with Na+ ions in where I is the discharge current. This calculation was repeated for
the same way of Ni. For example, a typical NaNiCl2 cell cathode different discharge current rates.
additive is iron, whose chemical reaction with Na+ and electrons is Furthermore, it is worth considering that the discharge current
similar to the Ni reaction: rate fluctuations involve a variation in the cell temperature as well,
which does not affect the cell internal resistance, but which has to
FeCl2 + 2Na+ + 2e ! 2NaCl + Fe
be considered for the calculation of OCV by means of Eq. (1). A set
Differently from Ni, this iron reaction is produced for a lower of temperature measures for different discharge current rates is
OCV, i.e. 2.35 V. The cell OCV voltage decreases, during the shown in Fig. 3. In the following only the relationship between the
discharge process, from the fully charged cell voltage, which is
2.6 V, to the fully discharged one, which is 2.33 V (see Fig. 2a), with
the following temperature (q) dependence:
OCV(#) = OCV*  216106(#  260) (1)
*
where OCV must be the open circuit voltage at the nominal
operating temperature. Eq. (1) is an empirical formula derived
from experimental evidences. In this context, when the voltage
decrease to 2.35 V, Na+ ions start to react with iron, so that the front
of this new reaction starts close to the electrolyte. Therefore, the
ions propagation distance is reduced, and the cell internal
resistance too. Whereas the iron reaction continues, the reaction
front moves towards the current collector, and the resistance
progressively increases. So that it is possible, over a given
resistance increasing, to add a second metal additive to lower
the reaction front once again and to consequently lower the
internal resistance.
Fig. 2b shows the measured NaNiCl2 cell internal resistance as a
function of the DoD for different discharge current rates. The two
peaks in the graph, at 75% and 85% of DoD respectively, represent
the starting of the metal additive reactions, which involve a rapid
decrease of the cell resistance. For the typical NaNiCl2 cell
discharge currents, which are from C/3 to C/5, the cell resistance
Fig. 3. Cell measured temperature variations as a function of current rates.
108 S. Dambone Sessa et al. / Journal of Energy Storage 6 (2016) 105–115
Fig. 4. Complete NaNiCl2 discharge mode modelling scheme.
temperature variations and the internal resistance ones (which in
turn depends on DoD and on discharge current rate) is considered.
The heaters influence on the cell temperature was neglected
because the cell temperature depends only marginally from
heaters in steady state stationary applications. Indeed the heat
produced by charging and discharging cycles is sufficient to
maintain the operating temperature and no further heaters are
required [9]. Heaters are in operation only when the battery is idle
and the temperature falls below 260  C.
2.1.1. Simulink NaNiCl2 cell discharge modelling
Simulink was the chosen software for NaNiCl2 cell modelling
because it allows implementing Matlab functions inside the
program. By starting from the measured data of cell internal
resistance and OCV, the first step is creating a controlled real
voltage supply whose no-load voltage is equal to the NaNiCl2 OCV
(see Fig. 4). The voltage supply internal resistance is equal to the
cell measured one. As it has been explained in Section 2.1, the cell
internal resistance varies as a function of DoD and of the discharge
current rates. Since in the Simulink library there is not a resistance
block able to simulate this kind of variation, it is necessary to
implement a set of interpolation functions in order to represent
correctly the cell internal resistance.
2.1.2. Modelling of the variable cell resistance
The resistance measurements shown in Fig. 2b can be
represented in matrix form. The elements rdm of this matrix are
equal to the measured discharge resistance values at different
discharge current rates cr and DoD. The matrix number of rows is
equal to the DoD measure intervals DoD_m, the number of columns
is equal to the discharge current rates in correspondence of which
the cell resistance is measured. The cell discharge resistance
rd(DoD_m, cr) in matrix form is represented in Fig. 5.
Once the resistance matrix is created, it is necessary to
implement a function able to calculate the instantaneous cell
DoD during the discharge simulation. Therefore, the current i(t)
supplied by a controlled voltage source of the Simulink library has
to be integrated by means of a Simulink limited integral block, the
 S. Dambone Sessa et al. / Journal of Energy Storage 6 (2016) 105–115 109

output of which is the internal cell resistance as a function of the

discharge current and DoD, i.e.
r(i(t)*,DoD(t))
It is worth highlighting that the increase of the number of
columns of the rd matrix allows the user to increment the precision
in the resistance evaluation. For the comparison reported in
Section 3, the resistance measures have been performed for
9 different current rates from 0 to C.

2.1.3. Setting of the Simulink controlled voltage source

By considering the OCV measurement as a function of DoD in
Fig. 2a it can be implemented in the shape of the vector OCV
(DoD_m), the elements of which are the OCV measurements OCVm
Fig. 5. Cell resistance measures in the shape of a matrix.
for different DoD measurement intervals DoD_m, as it is shown in
Fig. 6.
By interpolating OCV(DoD_m) with DoD(t), with “interpol3”
limit of which is the maximum cell charge capacity Mcc (see Fig. 4). function in Fig. 3, it is immediate to obtain the cell OCV as a
By comparing the limited integral output with Mcc, which is function of DoD(t), which is the OCV of the controlled voltage
assumed equal to 38 Ah, and by taking into account that source. In order to take into account the dependence of OCV on the
DoD = (1  SoC), it is immediate to deduce the normalized DoD temperature, the cell temperature variation as a function of DoD
variation as a function of time: and discharge current rate c_r (see Fig. 3), can be represented in
0 1 matrix form, as it has been explained for the cell resistance. The
Zt structure of temperature matrix q (DoD_m, c_r) is shown in Fig. 7.
BMcc  iðtÞdtC
B C With the same procedure described for the cell resistance
B C
B 0 C matrix, the matrix #(Dod_m, c_r), the current i(t)*and DoD(t) are
DoDðtÞ ¼ 1  B C ð2Þ
B Mcc C interpolated by means of “interpol2” in Fig. 4, the output of which
B C
@ A is the cell temperature as a function of the discharge current and
DoD, i.e.
It is worth noting that it is not possible to use in the model the #(i(t)*, DoD(t))
current i(t) as it is measured by the Simulink current sensor. In fact,
this current is never zeroed during the simulation, so that it is not Finally, by connecting #(i(t)*,DoD(t)) and OCV(DoD(t)) with a
possible to represent the end of the cell discharge. Therefore, the function block in which Eq. (1) is implemented (see Fig. 4), the OCV
current i(t) has to be calculated as the derivative i(t)* of the current of the controlled voltage source can be obtained, i.e.
integral. OCV(#, DoD(t))
0 t 1
Z Hence the discharge voltage V _d of the Simulink controlled
d@ iðtÞA voltage source is:
iðtÞ ¼ 0
V d ¼ OCVð#; DoDðtÞÞ  fr½iðtÞ ; DoDðtÞ  iðtÞ g ð3Þ
dt
For a constant current discharge, before of integrating i(t), it has The cell discharge voltage is the voltage of source, and the cell
to be limited at the desired current value by means of a Simulink discharge current is the derivative of the integrated source current
saturation block. In this way it is not necessary to create a constant i(t)*.
current load.
By considering that the DoD_m intervals in the resistance 2.2. Basic NaNiCl2 cell charge electrochemistry
matrix are different from DoD(t), the resistance matrix r(Dod_m,
c_r), the current i(t)*and DoD(t) have to be interpolated by means of By measuring voltage and current during the charge operation
an interpolation function indicated with Interpol1 in Fig. 4, the of one NaNiCl2 storage cell, it is possible to infer an equivalent
electrical circuit which is able to simulate the cell behaviour.

Fig. 6. OCV as a function of DoD measure intervals in the shape of a vector. Fig. 7. Cell temperature measures in the shape of a matrix.
110 S. Dambone Sessa et al. / Journal of Energy Storage 6 (2016) 105–115

Fig. 8. Charge reaction scheme of NaNiCl2 storage cell.

In particular, if the cell is powered by an external energy source
(see Fig. 8), a difference between the open circuit voltage (OCV) of
the cell and the on-load one (OLV) at different state of charge (SoC)
can be observed. The cell OCV in charge mode is lower than the
OLV. So that, this voltage difference can be modelled as a charge
electrical resistance rc by considering the following equation:
OLV = OCV + rcI
By considering the scheme in Fig. 8, once the switch s is closed,
the NiCl2 cathode releases electrons and, as a consequence, sodium
positive ions are formed, with the following reaction:
2NaCl2 + Ni ! NiCl2 + 2Na+ + 2e
These sodium ions pass thorough the ceramic electrolyte to the
anode zone of the cell. The electrons flow through the electric
circuit, and sodium is formed in the cell anode part, by following
the reaction:
2Na+ + 2e ! 2Na.
The measured NaNiCl2 cell OCV is shown in Fig. 9a.
The charge cell resistance rc can be derived from Eq. (4) by
performing a set of cell OLV measurements and by taking into
(4)
account the measures of Fig. 9a, by means of the following
expression:
rc = (OLV(DoD)  OCV(DoD))/I
where I is the charge current.
The rc behaviour in charge mode is shown in Fig. 9b.
It is worth considering that the charge current rate and
temperature variations do not affect the cell charge resistance
behaviour.

Fig. 9. (a) Measured NaNiCl2 cell OCV; (b) NaNiCl2 cell charge resistance as a function of SoC.
 S. Dambone Sessa et al. / Journal of Energy Storage 6 (2016) 105–115 111

2.2.1. Modelling of the cell charge resistance as a function of SoC

In order to calculate the cell SoC during the simulation, the
current i(t) absorbed by a controlled voltage source of the Simulink
library has to be integrated by means of a Simulink limited integral
block, the limit of which is the maximum cell charge capacity Mcc.
By comparing the limited integral output with Mcc, which is
assumed equal to 38 Ah and taking into account that SoC = (1 
DoD) it is immediate to derive the normalized SoC variation as a
function of time:
0 t 1
Z
B iðtÞdt  Mcc C
B C
B C
B0 C
SoCðtÞ ¼ 1  B C ð5Þ
B Mcc C
B C
@ A

Fig. 10. Cell charge resistance for different SoC in the shape of a vector.
Analogously to the cell discharge modelling, the cell charge one
begins with the implementation of a controlled real voltage source,
taking into account the measured data of cell internal resistance
and OCV. The OCV of the source is equal to the NaNiCl2 OCV. The
voltage supply internal resistance is equal to the cell measured one.
By adding a Simulink saturation block before of integrating i(t),
a constant current cell charge can be simulated. In this way it is not
necessary to create a constant current load in Simulink.
It is worth noting that the current i(t) is negative because it
enters into the voltage supply.
By means of a Matlab function block, the measurement curve of
Fig. 8b is implemented in the shape of a vector rc (SoC_m) whose
elements are the measured charge resistance rcm for different SoC
measure intervals SoC_m (see Fig. 10).

Fig. 11. NaNiCl2 charge mode modelling scheme.

112 S. Dambone Sessa et al. / Journal of Energy Storage 6 (2016) 105–115
Fig. 12. Cell OCV in charge mode for different SoC in the shape of a vector.
Because of the SoC(t) is obviously different from the SoC
measure intervals of rc(SoC_m), by interpolating SoC(t) and
rc(SoC_m) by means of “interpol4” function in Fig. 11 the cell
charge resistance as a function of SoC(t) can be obtained, i.e. rc(SoC
(t)).
2.2.2. Control of the Simulink controlled voltage source
By means of a Matlab function block, the measurement curve of
Fig. 9a is implemented in the shape of a vector OCV(SoC) whose
elements are the measured OCV for different SoC measure intervals
(see Fig. 12).
By interpolating OCV(SoC) with SoC(t) by means of “Interpol5”
function in Fig. 11, the OCV as a function of SoC(t), can be obtained,
i.e.
OCV(SoC(t))
Once OCV(SoC(t)) and the internal resistance rc(SoC(t)) of the
cell are calculated, the charge voltage V_c of the Simulink
controlled voltage source has to be:
V c ¼ OCV þ ðrc ðSoCðtÞÞ  iðtÞ Þ
The dependence of V_c on the temperature has not been
considered because, as it is mentioned in Section 2.2, the
temperature variation does not affect the charge resistance.
Fig. 13. a) Comparison between the model battery voltage results and the measured ones for a complete discharge with a current of 2.38 A (0.075  C); b) Comparison
between the model battery voltage results and the measured ones for a complete discharge with a current of 38 A (1.187  C).
As it is explained in Section 2.1.2, in order to obtain V_c, the
current i(t) has to be calculated as the derivative i(t)* of the current
integral, i.e. the output of the limited integration block. In order to
simulate the cell charge, it is sufficient to apply a voltage source at
the controlled voltage source terminals. The cell charge voltage is
the controlled voltage source one, and the cell discharge current is
the derivative of the integrated controlled voltage source one i(t)*.
3. Results
3.1. Comparison between battery charge/discharge measures and
model responses
In order to validate the above presented cell model, a
comparison between the model results and a series of battery
measures is presented. The charge and discharge cell models have
been joined into a unique model, by means of logic functions which
enable the charge or the discharge model on the basis of the
current direction. It is worth highlighting that the following
comparisons are referred to a NaNiCl2 module (Fiamm SoNick
ST523 module) [13], i.e. a battery consisting of 240 cells in series,
with incorporated battery management system (BMS). Therefore, a
BMS model has been added to the previous model structure. In
particular, in the following comparisons, the BMS current
limitation in charge mode is shown. This BMS function avoids
that the battery voltage increases too much, up to 640 V by limiting
the charge current in turn.
It is worth remembering that the battery self-consumption due
to the heater actions was not considered in the model. This
simplification is licit because the exothermic cell reactions
themselves guarantee the battery operating temperature in steady
state operations. The heaters are involved when the battery
temperature drops below 260  C, therefore it implies a little heater
power consumption, which can be neglected. Overall, although the
cells have 100% coulombic efficiency, and therefore, have no self-
discharge [19], the base operating temperature of 260  C requires
that even in a very well insulated battery box utilizing evacuated
thermal insulation, some heat loss is unavoidable. A 20 kWh
ð6Þ
battery would have a heat loss of around 80 W and this can be
thought of as self-discharge. It is not possible to classify this as a
percentage of capacity since it depends upon the operating cycle
but it varies from 10% per day if the battery is not used at all
 S. Dambone Sessa et al. / Journal of Energy Storage 6 (2016) 105–115 113

Fig. 14. (a) Comparison between the model battery voltage results and the measured one for a 5 A charge current (0.156  C) from 0% to 95% of the SoC; (b) Comparison
between the model battery current results and the measured one for a 5 A (0.156  C) charge current from 0% to 95% of the SoC.

(exhibiting a natural cooling of around 3  C/h [9]) to zero for an order to highlight the limits of the model, a discharge transient has
intensively used battery in which the temperature never falls to been simulated and compared with real measures. During a
260  C. In order to increase the model accuracy and to consider the NaNiCl2 module discharge, by opening and closing alternately the
battery stand-by mode or a very low current charge or discharge circuit, the discharge voltage was measured. The discharge current
for a short time, the heaters influence should be taken into account. magnitude was 0.312  C (10 A).
The modular structure of the model allows the user to implement a Fig. 16 shows the comparison between model results and
block which considers this aspect. measures during the above mentioned transient operation. The
In Fig. 13 two comparisons between the model battery voltage voltage error between the two curves is small, but the model is not
results and the measured ones for a complete discharge, with a able to follow correctly the battery voltage behaviour. Therefore, in
constant current of 0.075  C (2.38 A) (see Fig. 13a) and 1.187  C order to represent battery transient operations, the model needs to
(38 A) (see Fig. 13b) are shown. The maximum voltage difference be further developed, with the inclusion of new elements to the
between the two curves of Fig. 13a is 1.05% and the discharge time modelling structure.
error is zero. The maximum voltage difference between the two
curves of Fig. 13b) is 2.04% and the discharge time error is zero.
In Fig. 14a and b a comparison between the model battery
voltage and current results and the measured ones for a charge
from 0% to 95% of the SoC, with a constant current of 0.156  C (5 A),
is shown. In particular in Fig. 14b) it is possible to see the BMS
effect, which cuts the current in order to limit the battery voltage
increase.
The maximum voltage difference between the two curves of
Fig. 14a) is 4.6% and the maximum error between the two current
curves of Fig. 14b) is 5.2%. The model was tested for discharge
variable currents too. In Fig. 15 a comparison between the
measured battery voltage and the model results for a variable
discharge which progressively varies from 0.156  C to 1.093  C
(from 5 to 35 A, from 0 to 80% of DoD) is shown. The figure
represents the battery behaviour if a continuously variable load
(which represents a hypothetical grid current requirement) is
applied to it. In order to verify the reliability of the model results,
variations of the battery current and voltage during the discharge
process were considered simultaneously and compared with real
measurements. Therefore, the focus of the comparison was to
analyze the voltage variation on the discharge current one. DoD
varies from 0% to 80% during the discharge, and it was not reported
in the figure because it is not relevant in this analysis. The
maximum difference between the curves is 4.4%. Fig. 15. Voltamperometric curves representing the comparison between the
Even though the focus of the present modelling approach is measured battery on-load voltage and the model results for a variable discharge
representing the steady state behaviour of NaNiCl2 batteries, in current which varies from 0.156  C to 1.093  C (from 5 to 35 A, from 0 to 80% of
DoD).
114 S. Dambone Sessa et al. / Journal of Energy Storage 6 (2016) 105–115
Fig. 16. Comparison between the measured on-load battery voltage and the model results for transient discharge conditions.
4. Conclusions and discussion
In this paper an experimental data-based model has been
developed with a very good agreement between real discharge-
charge measures and model results. The present approach is
suitable for representing the battery energy intensive application
on the electrical network, which requires battery charging or
discharging in long intervals. NaNiCl2 battery rated charge/
discharge time (from 3 to 5 h) fits very well with this application.
The model was tested for both constant and variable charge/
discharge currents. The differences between measures and model
results are basically due to the fact that it is very difficult to have in
the model exactly the same current curves of real discharge/charge
measures. Moreover, the power consumption of the heaters has
been neglected; this simplification is acceptable in consideration
of the fact that the model errors are small.
A comparison between battery measures and model results in
transient conditions has been presented. In this case the present
approach does not follow perfectly the battery voltage behaviour.
This implies that the set of experimental measures which are used
to infer the steady state battery model is not sufficient to correctly
represent the transient behaviour of NaNiCl2 batteries. The authors
hope to further develop the proposed approach to create a
completely general battery model, with the same “simple”
structure.
In conclusion, the present NaNiCl2 model appears an useful tool
in order to foresee the effect of the energy storage installation in
the electrical network, by adding a power conversion system [20]
and electrical network models to the battery model.
Besides, the presented modelling approach could be easily
accounted for Na-beta family batteries, e.g. sodium-sulphur
batteries [21–23].
Acknowledgement
S.D.S gratefully acknowledges CA.RI.PA.RO Foundation for the
financial support to the PhD scholarship.
References
[1] Grigorii L. Soloveichik, Battery technologies for large-scale stationary energy
storage, Annu. Rev. Chem. Biomol. Eng. 2 (2011) 503–527.
[2] R. Benato, N. Cosciani, G. Crugnola, S. Dambone Sessa, G. Lodi, C. Parmeggiani,
M. Todeschini, Sodium nickel chloride battery technology for large-scale
stationary storage in the high voltage network, J. Power Sources 293 (2015)
127–136, doi:http://dx.doi.org/10.1016/j.jpowsour.2015.05.037.
[3] R. Benato, N. Cosciani, S. Dambone Sessa, G. Lodi, C. Parmeggiani, M.
Todeschini: “La tecnologia sodio-cloruro di nichel (Na-NiCl2) per l’accumulo
elettrochimico stazionario sulla rete di trasmissione”, L’Energia Elettrica, Vol.
91, N 4, pp. 71–84, 2014.
[4] G. Pebriyanti, A lithium-ion battery modeling for a HIL-battery simulator,
2013 International Conference on Computer, Control, Informatics and Its
Applications, Jakarta (2013).
[5] K. Bae, S. Choi, J. Kim, C. Won, Y. Jung, LiFePO4 dynamic battery modeling for
battery simulator, 2014 IEEE ICIT, Busan, Korea (2014).
[6] R.C. Kroeze, P.T. Krein, Electrical battery model for use in dynamic electric
vehicle simulations, 2008 IEEE PES, Rhodes (2014).
[7] B. Wu, B. Chen, Study the Performance of Battery Models for Hybrid Electric
Vehicles, 2014 IEEE/ASM (MESA), Senigallia, 2014.
[8] A. Damiano, M. Claudia, M. Ignazio, Experimental validation of a dynamic
energy model of a battery electric vehicle, 4th International Conference on
Renewable Energy Research and Applications, Palermo, Italy (2015).
[9] T.M. O’sullivan, C.M. Bingham, R.E. Clark, Zebra battery technologies for the all
electric smart car, SPEEDAM International Symposium on Power Electronics,
Electrical Drives, Automation and Motion (2006).
[10] E. Micolano, M. Broglia, Sviluppo di modelli di sistemi di accumulo di tipo
tradizionale ed avanzato per impieghi nella generazione distribuita al fine
della loro rappresentazione in sistemi complessi, CESI Report A5053120, 2005.
[11] C. Capasso, V. Sepe, O. Veneri, M. Montanari, L. Poletti, Experimentation with a
ZEBRA plus EDLC based hybrid storage system for urban means of transport,
Electrical Systems for Aircraft, Railway, Ship Propulsion and Road Vehicles
(ESARS), 2015 International Conference on, vol. no., 3–5 March, 2015, pp. 1–6.
 S. Dambone Sessa et al. / Journal of Energy Storage 6 (2016) 105–115
[12] O Veneri, C. Capasso, S. Patalano, Experimental study on the performance of a
ZEBRA battery based propulsion system for urban commercial vehicles,
Applied Energy, Available online 24 February 2016, ISSN 0306-2619.
[13] R. Manzoni, M. Metzger, G. Crugnola, ‘ZEBRA Electric Energy Storage System:
From R&D to Market’ HTE hi.tech.expo, Milan, 25-28 November 2008.
[14] X. Lu, G. Xia, J.P. Lemmon, Z. Yang, Advanced materials for sodium-beta
alumina batteries: status, challenges and perspectives, J. Power Sources 195
(2010) 2431–2442.
[15] M. Rahmane, C. Iacovangelo, Materials for Advanced Sodium Metal Halide
Batteries, Ceramic Leadership Summit, Baltimore, MD (USA), 21–22 June
(2010).
[16] H. Bohm, G. Beyermann, ZEBRA batteries, enhanced power by doping, J. Power
Sources 84 (1999) 270–274.
[17] X. Lu, G. Li, J.Y. Kim, J.P. Lemmon, V.L. Sprenkle, Z. Yang, The effects of
temperature on the electrochemical performance of sodium nickel chloride
batteries, J. Power Sources 215 (2012) 288–295.
[18] J.L. Sudworth, R.C. Galloway, Secondary batteries—high temperature systems/
sodium–nickel chloride, in: Juergen Garche, Chris Dyer, Patrick Moseley,
115
Zempachi Ogumi, David Rand, Bruno Scrosati (Eds.), Encyclopaedia of
Electrochemical Power Sources, 42009, pp. 312–323.
[19] J.L. Sudworth, The sodium/nickel chloride (ZEBRA) battery, J. Power Sources
100 (2001) 149–163.
[20] M. Andriollo, R. Benato, M. Bressan, S. Dambone Sessa, F. Palone, R.M. Polito,
Review of power conversion and conditioning systems for stationary
electrochemical storage, Energies 8 (2015), doi:http://dx.doi.org/10.3390/
en8020960.
[21] J. Sudworth, R. Tilley, The Sodium/Sulfur Battery, Chapman and Hall, London,
1985.
[22] M. Andriollo, R. Benato, S. Dambone Sessa, N. Di Pietro, N. Hirai, Y. Nakanishi, E.
Senatore, Energy intensive electrochemical storage in Italy: 34.8 MW sodium–
sulphur secondary cells, J. Energy Storage 5 (February) (2016) 146–155, doi:
http://dx.doi.org/10.1016/j.est.2015.12.003.
[23] A. Bito, Overview of the sodium–sulfur battery for the IEEE stationary battery
committee, IEEE Pow. Eng. Soc. General Meeting (2005).