Assessment of Battery-Hybrid Diesel-Electric Locomotive Fuel Savings

Energy 173 (2019) 1154e1171
Contents lists available at ScienceDirect
Energy
journal homepage: www.elsevier.com/locate/energy
Assessment of battery-hybrid diesel-electric locomotive fuel savings

and emission reduction potentials based on a realistic mountainous
rail route
Mihael Cipek a, *, Danijel Pavkovi
c a, Zdenko Kljai
c b, Tomislav Josip Mlinari
cc
a
Faculty of Mechanical Engineering and Naval Architecture, University of Zagreb, Zagreb, Croatia
b
Ericsson Nikola Tesla d.d., Zagreb, Croatia
c
Faculty of Transport and Traffic Sciences, University of Zagreb, Zagreb, Croatia
a r t i c l e i n f o a b s t r a c t
Article history: This paper presents a hypothetical conversion of a conventional heavy haul diesel-electric locomotive to
Received 9 October 2018 its hybrid counterpart by incorporating a battery energy storage system. Starting from the basic pa-
Received in revised form rameters of a 1.6 MW diesel-electric locomotive currently found in the national railway company loco-
30 January 2019
motive fleet, the quasi-static model of the locomotive is derived and validated. The conventional
Accepted 19 February 2019
Available online 21 February 2019
locomotive model is then converted to its hybrid counterpart by adding a battery energy storage system
in parallel to the generator and equipping it with an adequate optimized energy management control
strategy. The hybrid locomotive powertrain components are also appropriately re-sized in order to meet
Keywords:
Hybrid electric locomotive
comparable traction force and power performance. Both the conventional and hybridized locomotive
Mountain rail route models are then used for the purpose of comparative analysis of main vehicle characteristics for the
Emissions mountainous railway route driving scenario, which includes realistic slope and speed limitations. The
Hybridization costs obtained simulation results are used to gain insights about the possible advantages of the proposed
Return-of-investment conversion/drivetrain hybridization in terms of feasible reduction of fuel consumption and related CO2
Advanced transportation technologies emissions, while also considering additional hybridization costs and return-of-investment period.
© 2019 Elsevier Ltd. All rights reserved.
1. Introduction and the development of electric machines at the end of the 19th
century, while the industrial production of electric locomotives
Currently, around 20% of the globally-produced fossil fuels (i.e. began in the 1930s [5]. An electrified railway system distributes the
primary energy) are consumed by the personal and freight trans- electrical energy trough the dedicated low or medium-voltage
portation sector, which contributes to its large carbon footprint and system (by means of an overhead conductor or a third rail) to the
greenhouse gases emissions share [1]. Hence, transportation elec- train locomotive, which can operate without having a primary
trification has been recognized as a promising way to make the energy source on-board [6]. In particular, electric locomotives are
overall system more efficient, cleaner, quieter, and less dependent characterized by higher power-to-weight ratios than diesel ones
on oil reserves [2], while incorporating the renewable energy [6], and are also able to recover a portion of kinetic energy trough
sources such as hydroelectric [3], wind and solar (photovoltaic) regenerative braking [7]. Apparently, due to significant capital
power-plants into the overall energy system can further reduce air infrastructure costs associated with railway line electrification (see
pollution and greenhouse gases emissions [4]. for example an Egyptian [8] and American [9] case study), many
The development of the railway system electrification started railway lines in Europe that are rather lengthy and are character-
along with the evolution of electrical energy distribution systems ized by low traffic densities are still not electrified [10]. Since en-
ergy and material costs for the electrified infrastructure represent a
significant portion of the transportation costs [11], a notable
number of diesel-powered locomotives are still needed for pas-
* Corresponding author. senger and freight transport [12]. Thus, for the time being, the share
E-mail addresses: mihael.cipek@fsb.hr (M. Cipek), danijel.pavkovic@fsb.hr
of diesel fuel alternatives such as bio-diesel, methanol and
(D. Pavkovic), zdenko.kljaic@ericsson.com (Z. Kljai
c), tmlinaric@fpz.hr
(T.J. Mlinari
c). hydrogen in transport is not likely to exceed 10% by 2040 [13].
https://doi.org/10.1016/j.energy.2019.02.144
0360-5442/© 2019 Elsevier Ltd. All rights reserved.
M. Cipek et al. / Energy 173 (2019) 1154e1171 1155
In particular, the energy security, which is a fundamental transport. Alternative energy storage systems such as flywheels
requirement for the country's economic prosperity, currently [30] and hydrostatic energy storage [27] have also been considered
strongly depends on the availability of oil and gas, mandating in rail transport, but have typically been characterized by inferior
timely implementation of energy policies related to energy supply performance when compared to batteries and supercapacitors.
diversification which minimize the risks of supply chain disruption Unsurprisingly, the aforementioned hybridization measures have
[14]. These energy policies should also be aimed towards higher also appealed to other similar sectors, such as maritime trans-
sustainability indices and energy efficiency of the transportation portation [31], and freight transportation in mining [32]. In the
sector [15]. A study presented in Ref. [16] has indicated that sig- latter case, reference [33] suggests that the diesel propulsion may
nificant savings can be obtained in the transportation sector in be completely replaced by advanced battery technologies, such as
Croatia by implementing energy efficiency improvement measures those based on lithium-ion and sodium metal halide chemistries.
and transport electrification. When considering the rail freight Having this in mind, the main hypotheses of the presented work
transport, it would be difficult to quantify the energy consumption are that: (i) it would be possible to convert a conventional 103 t and
and CO2 emissions due to different freight size, type of traction and 1.6 MW heavy-haul diesel-electric locomotive, currently being
terrain [17]. However, the railway freight energy efficiency study utilized by the national railway company [34], to a battery hybrid
presented in Ref. [18] has suggested that it would be possible to counterpart; (ii) and that it would result in notable fuel and
achieve up to 15% less diesel fuel consumption and a proportional greenhouse gases emissions for the particular scenario of heavy
CO2 emissions reduction through optimization of rail freight freight haul over the established mountainous railway route
transport and appropriate choice of heavy-haul diesel locomotive located within Lika region [35]. Due to inclusion of sufficiently-
engine operating modes during transport. sized battery energy storage system, the hybrid locomotive pow-
It has been suggested in Ref. [19] that over 70% of the existing EU ertrain components would also have to be appropriately re-sized in
transportation energy demands can be covered by transport sector order to meet comparable locomotive traction performance, while
electrification. This includes hybrid-electric power-trains in trans- also satisfying the weight-per-axle constraint of 18 t per axle, cor-
portation applications, and appropriate choice of energy storage responding to B category tracks according to domestic classifica-
systems capable of dealing with anticipated power-train load pro- tions [36], as well as international ones [37].
files characteristic for particular transportation applications [20]. The paper is organized as follows. Section 2 presents an over-
Even though road vehicles are increasingly being equipped with view of the conventional diesel-electric locomotive. The scalable
hybrid powertrains over the last two decades with the main goal of model of a diesel generator set with the model parameters ob-
increasing their fuel economy [21], and, consequently reducing tained from the publicly-available data for a similar-powered
their carbon footprint [22], rail vehicle hybridization is still in the diesel-electric locomotive is presented in Section 3, while model
early development stages. Conceptual design of heavy haul hybrid of diesel-electric locomotive along with the longitudinal dynamics
locomotives is given in Ref. [23], wherein different electrical energy of the railway train and virtual train driver model are outlined in
storage systems, such as electrochemical batteries and ultra- Section 4. The mountainous railway route and overall energy re-
capacitors, are analyzed and compared, while reference [24] pri- quirements for the particular route are given in Section 5. These
marily deals with the application of flywheel energy storage. results are used in Section 6 for the sizing of the main generator and
However, hybrid locomotives have been produced only as pro- battery energy storage system of the battery-hybrid locomotive
totypes [25] and mostly used as switcher locomotives [26]. Due to counterpart. Subsequent simulation analysis is present in Section 7,
the fact that majority of modern heavy-haul diesel-powered loco- while the comparative techno-economic aspects of the proposed
motives utilize electrical power transmission (so-called diesel- upgrade are given in Section 8. The concluding remarks and
electric locomotives [5]), their hybridization can be achieved by possible avenues for future work are given in Section 9.
basically adding a properly-sized electrical energy storage system,
such as one based on electrochemical batteries, supercapacitors or
2. Diesel-electric locomotive outline
flywheels, directly into the electrical transmission system [23]. The
benefits of utilization of electric energy storage system might
Railway transportation over non-electrified tracks is nowadays
include: (i) additional power when the main generator set could
predominantly based on diesel-electric locomotives,1 whose trac-
not provide sufficient traction power during extreme operating
tion system comprises a diesel engine coupled to the main gener-
conditions (i.e. acceleration under high load during climbing); (ii)
ator producing electric power which is conditioned through the
recovery of notable portion of regenerative braking power by
dedicated control unit in order to produce adequate torque at
means of electric energy storage system recharging; and (iii) the
traction motors (see Fig. 1) [12]. Such electrical power transmission
possibility of all-electric low-power driving as a means of air-
within diesel-electric locomotives is characterized by its voltage/
pollution reduction or in emergency scenarios such as the case of
current system, which may either be direct current (DC), alter-
diesel generator malfunction. Perceived disadvantages of addi-
nating current (AC), or a combination thereof, as shown in Ref. [5]:
tional hybridization are reflected in additional locomotive weight
and cost of energy storage system installation, and cycle life and
DC main generator and DC traction motors (DC-DC),
operating conditions constraints in the case of battery systems [5].
AC main generator and DC traction motors (AC-DC),
Different hybrid architectures for diesel engine-based rail ve-
AC main generator and AC traction motors (AC-AC).
hicles have been investigated in Ref. [27], with lithium-ion batteries
and supercapacitors recognized as the most promising energy
However, all aforementioned systems utilize a common DC link
storage technologies for diesel engine-based rail vehicle retrofitting
in order to transmit electrical power between the main generator
aimed at power-train hybridization. The extensive study presented
and traction motors [25], which may utilize traction inverters in the
in Ref. [28] has shown that battery-hybrid diesel-electric railway
case of AC traction motors [24]. This is favorable from the
traction can achieve up to 20% fuel consumption reduction in heavy
freight applications, whereas reference [29] points out to 24% fuel
consumption reduction through regenerative braking recuperation 1
In contrast, diesel-hydraulic traction system is sometimes used for freight
of kinetic energy and subsequent support of the diesel engine- trains, primarily because it is not adapted to the heating of coaches within pas-
based propulsion during acceleration in the case of passenger rail senger trains, as is the case when diesel-electric traction is used.
1156 M. Cipek et al. / Energy 173 (2019) 1154e1171
Fig. 1. Simplified schematic representation of diesel-electric locomotive power-train.
standpoint of adding a battery energy storage system for the pur- diesel-electric locomotive from GP38-2 series, also manufactured
pose of locomotive traction system hybridization. by the General Motors Corporation Electro-Motive Division (EMD)
from 1973 onwards, and powered by a 2000 HP “roots-blown” EMD
2.1. Target diesel-electric locomotive used by national company 16-645-E diesel engine. Reference [38] also presents a compre-
hensive approach to conducting the locomotive exhaust emissions
2062 series diesel-electric locomo- tests for four different fuel types, and includes a detailed technical
According to Ref. [34], HZ
description of the locomotive selected for testing, as well as its
tives are commonly used over Lika's railway routes. These loco-
engine power, fuel consumption and exhaust emissions measure-
motives were manufactured in 1970's by General Motors
ments. The summarized data of the baseline test from Ref. [38] are
Corporation under the designation G26CW, and have been subse-
converted to the SI unit measurement units, and listed in Table 1.
quently modernized in the early 2000's. These locomotives weigh
103 t, and are powered by a 1641 kW 16-cylinder two-stroke diesel-
engine (EMD 16-645-E) driving the main generator which provides 2.3. Throttle position power distribution and power conversion
electrical power for six traction electric motors coupled to the efficiency
traction axles by means of a fixed-ratio gearboxes [34].
For the selected type of diesel-electric locomotive, the train
2.2. EMD-645-E powered diesel-electric locomotive data driver operates a lever mechanism (engine governor) with 8 engine
throttle positions (so-called “Notches”). Each “Notch” corresponds
The realistic data for diesel-electric locomotive fuel consump- to constant power production at the main DC generator, whose
tion and emissions for each throttle lever position (so-called Notch excitation circuit is adjusted in order to maintain predefined diesel
position) are adopted from Ref. [38]. This reference presents a engine power (speed and torque) [12] (‘’ markers in Fig. 2). Idling
comprehensive overview of emission tests for a similarly-sized or neutral position corresponds to zero “Notch”. Fig. 2a shows the
Table 1
GP38-2 locomotive data based on field measurements.
Throttle Main engine power Pmg Main engine speed ue Electric traction power Pt Engine fuel consumption rate m_ f Emissions rates (g/s)
position (kW) (rad/s) (kW) (g/s)
HC CO NOX CO2
IDLE 7.83 27.02 0 3.1247 0.0364 0.0531 0.1836 9.7154

Notch 1 61.82 33.09 48.69 6.4259 0.0397 0.0653 0.3372 20.4256
Notch 2 283.33 40.84 208.86 16.8837 0.0481 0.1256 0.7861 53.3554
Notch 3 437.28 52.36 386.01 28.7275 0.0792 0.1194 1.4817 90.8920
Notch 4 769.19 60.11 688.62 49.6431 0.1014 0.1828 2.9736 157.3038
Notch 5 956.88 67.75 843.79 62.7469 0.1381 0.3714 3.9292 198.5396
Notch 6 1174.18 76.65 1026.93 78.1186 0.1936 1.1792 4.8725 245.8126
Notch 7 1412.21 86.50 1221.39 96.7663 0.3133 2.7028 5.9336 302.4819
Notch 8 1596.00 94.88 1363.59 114.9010 0.4031 5.0500 6.3678 356.1055
Fig. 2. Locomotive power vs. Notch lever position (a) and electric power conversion efficiency (b).
percentage of power generated at each “Notch” position listed in

Table 1, which is assumed the same for any engine size. PmgMAX
smg ¼ : (4)
Although most of the mechanical power generated by the diesel PmgMAX0
engine is used to drive the main generator in order to provide
Therefore, the Willans line technique first needs to find appro-
electrical power for traction motors, a small amount is used to
priate values of second-order polynomial approximation co-
power different auxiliaries (e.g. pumps, fans and air compressor)
efficients c0, c1 and c2 for each angular velocity ue and torque value
[5]. In this work, the overall efficiency of mechanical-to-electrical
te, and for scaling factor smg ¼ 1 in equation (3). This can be done by
power conversion is described by the nominal power percentage-
numerically solving a system of equations for each torque vs. speed
related efficiency coefficient hel(%P). This coefficient is indepen-
data point from the original engine fuel consumption map, which
dent of engine size, and can be defined as the ratio between output
could be done by using least squares method [41]. Due to insuffi-
Po(%P) and input Pi(%P) power as follows:
cient data, the EMD-645-E fuel consumption map needed to be
Po ð%PÞ reconstructed based on the fuel consumption map for a similarly-
hel ð%PÞ ¼ ; (1) sized diesel engine, adopted from Ref. [12]. This map is then
Pi ð%PÞ
adjusted to meet known fuel consumption values in the fifth col-
which implicitly includes the auxiliaries power consumption. umn in Table 1, and the resulting final speed-dependent approxi-
Fig. 2b shows value of efficiency coefficient derived from the mation coefficients for the target engine EMD-645-E are shown in
ratio between electric traction power taken from the 4th column Fig. 3a. By inserting the obtained polynomial coefficients (Fig. 3a)
and the main engine power from the 2nd column of Table 1 for each into equation (3), the fuel consumption rate may be represented as
traction system “Notch” position, while taking into account the fuel rate (Fig. 3b) or specific fuel consumption (Fig. 3c), wherein ‘þ’
relationship between Notches and the percentage of power shown markers denote individual throttle “Notch” positions, as explained
in Fig. 2a. in the previous section.
3.2. Normalized diesel engine emissions

3. Conventional locomotive diesel generator set performance
assessment and scaling
Reference [38] has presented locomotive exhaust emissions
analysis including hydrocarbons (HC), carbon-monoxide (CO),
This section presents a comprehensive scalability tool used for
carbon-dioxide (CO2), and nitrogen oxides (NOX). Values of these
resizing of locomotive engine and generator set, and scaling of fuel
emission rates for each Notch and diesel engine steady-state are
consumption maps. It also briefly outlines the diesel engine
given in Table 1. In order to use these values within the scalable
exhaust normalized emissions and provides the methodology of
diesel engine model, emission rates m_ HC;CO;NOX ;CO2 for each Notch
assessment of the locomotive power-train cost and weight.
from Table 1 are expressed as functions of power percentage and
are normalized by the fuel rate (also related to power percentage),
3.1. Willans line-based power-train scaling methodology thus resulting in the following emission coefficient ε definition:
According to data in Table 1, the main engine maximum power m_ HC;CO;NOX ;CO2 ð%PÞ
(related to the ultimate Notch position 8) of the target diesel- εHC;CO;NOX ;CO2 ð%PÞ ¼ : (5)
m_ f ð%PÞ
electric locomotive (PmgMAX ¼ 1641 kW) only slightly differs from
the benchmark locomotive engine (PmgMAX0 ¼ 1596 kW). In order to Fig. 4 shows the values of each coefficient vs. engine power
provide a comprehensive scalability tool for the engine re-sizing percentage. Generally speaking, emissions coefficients for carbon-
within the hybrid locomotive, a scalable model of the main monoxide and nitrogen oxides show notable increase with engine
generator set is proposed in this section. For this purpose, the diesel power (Fig. 4b and c), whereas in the cases of hydrocarbons and
engine steady-state fuel consumption can be parameterized by carbon-dioxide, this rising trend is either not emphasized (Fig. 4a),
using a fuel consumption map, wherein the fuel rate m_ f depends on or shows a slight decrease (Fig. 4d). For higher engine output power
the engine operating point (i.e. engine speed ue and engine torque values, the decreasing trend of carbon-dioxide emission curve
te): correlates with the rising trend of the carbon-monoxide curve,
which suggests that incomplete combustion may be the primary
m_ f ¼ f ðue ; te Þ: (2) cause of such CO2 emission discrepancy at high power settings. On
the other hand, emission curves for low power values also slightly
In order to obtain a scalable fuel consumption map, an deviate from ideal (near constant) values, which may be explained
approximation based on so-called Willans lines is proposed. In by the increase of particulate emissions (soot) in engine exhaust.
Ref. [39] the engine fuel-related power Pf (i.e. fuel rate multiplied by These generally occur at higher fuel rates when diesel engine is
fuel heating value lower bound) is approximated by a second-order operated at low output power [42].
polynomial for each engine operating speed, while the engine
torque and overall losses are scaled linearly with a scaling factor
3.3. Main diesel engine and generator weight and cost assessment
smg. This approach would be more accurate as the size of approxi-
mated engine is closer to the baseline size [40]. In particular, the
Reference [43] provides the weight of the main diesel engine
fuel rate approximation can be written as:
EMD-645. However, in order to take into account the main engine
overall assembly, including the generator and control unit, data
t2e
m_ f ðue ; te Þ ¼ c2 ðue Þ þ c1 ðue Þte þ c0 ðue Þsmg ; (3) from several generator sets rated from 400 kW to 800 kW are ob-
smg
tained from a catalogue [44]. A following linear fit between the
power and weight is obtained based on catalogue data:
where c2(ue), c1(ue) and c0(ue) are angular speed-dependent co-
efficients of the above polynomial approximation. The scaling fac- mmg ¼ a$PmgMAX b; (6)
tor smg represents the ratio between scaled engine power PmgMAX
and original engine power PmgMAX0, defined as follows:
Fig. 3. Angular speed-dependent coefficients of the polynomial approximation (a), engine fuel rate in g/s (b) and specific fuel consumption in g/kWh (c).
Fig. 4. Normalized locomotive exhaust gaseous emission for (a) hydrocarbons (HC), (b) carbon-monoxide (CO), (c) nitrogen oxides (NOX) and (d) carbon-dioxide (CO2).
where PmgMAX is the main engine maximum power in kW (corre-

sponding to Notch 8 throttle position), and the above linear fit Cmg ¼ Ce þ Cg $PmgMAX ; (7)
coefficient values are a ¼ 0.014 t/kW and b ¼ 0.45 t.
According to equation (6), weight of the 1596 kW main diesel
engine within the locomotive from Ref. [38] is 21.89 t, while the
weight of the 16-cylinder EMD engine according to Ref. [43] is 4. G26CW diesel-electric locomotive model and longitudinal
16.5 t. Therefore, it may be assumed that this difference relates to train dynamics
the weight of the main generator and electric traction control unit.
In order to calculate the total cost of the main engine and Appling the approximation procedure explained above to the
generator unit, specific costs from Ref. [45] have been used, measurement data in Table 1 for a similar diesel-electric locomotive
wherein specific engine cost is estimated to Ce ¼ 15 EUR/kW, while [38], with scalability coefficient smg ¼ PmgMAX/PmgMAX0 ¼ 1641/
the specific cost of electric generator is estimated to Cg ¼ 30 EUR/ 1596 ¼ 1.02823 (cf. equation (4)), the resulting approximate data is
kW, which yields the following cost estimate: given in Table 2. The resulting weight of the main engine and
Table 2
Approximate data for G26CW diesel-electric locomotive.
Throttle position Main engine power Pmg (kW) Electric traction power Pt (kW) Fuel rate m_ f (g/s) Emissions rate (g/s)
HC CO NOX CO2
IDLE 8.05 0 3.2131 0.0374 0.0546 0.1888 9.9810

Notch 1 63.56 50.02 6.6242 0.0409 0.0673 0.3476 21.0561
Notch 2 245.05 214.71 17.3520 0.0494 0.1290 0.8079 54.8354
Notch 3 449.62 396.92 29.3608 0.0809 0.1221 1.5143 92.8959
Notch 4 790.90 708.11 51.0394 0.1042 0.1879 3.0572 161.7280
Notch 5 983.90 867.66 64.5594 0.1420 0.3821 4.0427 204.2746
Notch 6 1207.33 1055.88 80.3542 0.1992 1.2129 5.0119 252.8473
Notch 7 1452.07 1255.92 99.7743 0.3231 2.7868 6.1181 311.8848
Notch 8 1641.00 1402.08 118.0862 0.4142 5.1896 6.5438 365.9487
generator set is mmg ¼ 22.52 t, calculated according to equation (6), b ¼ 7.5 km/h and g ¼ 44 km/h.
so the locomotive base weight is mlb ¼ ml -mmg ¼ 80.48 t (ml being Thus, the traction force is limited by the normal force due to the
103 t as discussed earlier). locomotive weight ml g as follows:
Ftmax ¼ ma ml gcosðaÞzma ml g; (10)

4.1. Traction and electric braking force of G26CW locomotive
where a is the track slope typically being within ±2.5 [35], which
The locomotive electric traction power (third column in Table 2) justifies the above approximation (cosa z 1 is valid in that case).
is supplied to traction motors. These motors, in turn, generate By combining equations (8) and (10), the maximum traction
wheel traction force. Since this relates to electrical-to-mechanical force curves for each Notch are derived and shown in Fig. 5a,
power conversion, its efficiency also needs to be accounted for. wherein the same negative values are assumed for the case of
For the sake of simplicity, this efficiency is assumed to be equal to regenerative braking. The corresponding electrical traction power
the efficiency of the generator power conversion hel (Fig. 2b). curves are shown in Fig. 5b. In the case of regenerative braking
Hence, the electrical power for each throttle position generates the (negative power values), the electrical power values are smaller
following traction force: because the corresponding conversion efficiency also needs to be
accounted for. Namely, the regenerative braking regime is only
3:6$hel ð%PÞPt effective at higher speeds (e.g. above 20 km/h) because the induced
Ft ¼ ; (8)
v electromotive force in traction motors is sufficient to generate
ample braking current. Therefore, the braking region below 20 km/
where v is the locomotive longitudinal velocity given in km/h and Pt h is characterized by reduced regenerative braking potential, which
is the electric traction power expressed in Watts (W). However, the mandates simultaneous application of mechanical brakes in that
above expression is subject to an additional limitation related to the case. Note that in the case of conventional diesel-electric locomo-
maximum traction force due to wheel vs. track adhesion charac- tive this braking power is dissipated at the dissipation resistors [5]
teristic. According to Ref. [46], the adhesion coefficient can be (cf. Fig. 1).
expressed as:
b 4.2. Longitudinal train dynamics

ma ¼ mmin þ ; (9)
vþg
Since longitudinal train dynamics depend on the locomotive
with the minimum adhesion coefficient in the above equation traction force, train braking force, track topography and curvature,
given as mmin ¼ 0.161, and velocity-related constants given as the type of locomotive and wagons and the connection between
Fig. 5. Maximum force (a) and power (b) traction maps.

them, they can generally be described by a system of differential assumed and calculated from chosen set values of curvature radii
equations [5]. In this paper the longitudinal train dynamics are ranging from R1 ¼ 250 m which corresponds to the minimum pre-
simplified to a point body mass motion. Thus-simplified longitu- scribed railway route curve radius [47] to a maximum value of
dinal dynamic model includes the aerodynamic and rolling resis- R2 ¼ 4250 m, by using the following averaging formula:
tance forces and the gravity force due to the track slope which need
Rð2
to be overcome by the total traction force Ft [12], and is given by the
1 wr0 wr0 R r0
following dynamic equation: wr ¼ dr ¼ ln 2 ¼ 4:99 (14)
R2 R1 r r0 R2 R1 R1 r0
R1
dv ma
Ft Fb ¼ ma þ ma gsina þ ðw þ wr Þ; (11)
dt 1000 k
where Fb is the force applied to the brakes (equal 0 during normal

driving conditions). The right-hand-side term (dv/dt)ma is the force 4.3. Locomotive power-train and train driver model
(in Newtons) due to acceleration of total train mass ma (expressed
in kg), while the second term corresponds to gravity force due to Fig. 6 shows the block diagram representation of the quasi-static
the track slope, where sina can be expressed as the ratio between model of a conventional diesel electric locomotive, where specific
elevation change and related traveled distance (sina ¼ (hnþ1-hn)/ sub-models are framed by dashed lines. The sub-model of longi-
(lnþ1-ln)). Finally, the third term at the right-hand-side contains wk tudinal dynamics is described above, while the power-train sub-
and wr which are the specific resistance coefficients of longitudinal model is modeled as static characteristics (maps) derived in Section
and curvature motion respectively. According to Ref. [12] the spe- 3 (cf. Figs. 3e5). The traction force block (cf. Fig. 5a) provides the
cific motion resistance is defined as: traction force Ft map output related to train velocity v and driver-
based Notch selection as inputs. The traction power map block (cf.

v þ vd 2 Fig. 5b) uses the same inputs to calculate the electrical transmission
wk ¼ wk0 þ k ; (12) power Pt, while the engine model block calculates the fuel rate and
v0
exhaust gas emissions mass flows for selected Notch values.
with wk0 ¼ 25 N/t, v0 ¼ 10 km/h, and where k is train type config- The train driver model calculates the throttle and brake com-
uration coefficient (k ¼ 0.25 N/t is assumed in this work, which is mands based on the difference between velocity reference vref and
valid for express and heavy cargo trains [12]) and vd is wind velocity the actual train velocity v (see for example a driver model from
addition in km/h (vd ¼ 12 km/h is commonly used, as suggested by Ref. [48]). The throttle path is modeled as a simple proportional
Ref. [12]). The curvature-specific resistance according to Ref. [12] term with gain value KDr ¼ 8, which ensures that the speed target
depends on the curvature radius r and track gauge. For standard following error is less than 1 km/h under steady-state conditions.
gauge this coefficient can be expressed as: The driver model output is quantized to provide integer values
related to Notch positions and is also limited (Notch is between 8
wr0 and þ8 in Fig. 6). The resulting partition corresponds to 8 positive
wr ¼ ; (13)
r r0 Notch values for normal driving (cf. Fig. 3), one zero-valued Notch
position for Idling, and 8 negative Notch values for electro-dynamic
with wr0 ¼ 6500 Nm/t and r0 ¼ 55 m. Consequently, the above braking via power dissipation resistors. In the case of deceleration,
expression needs to be calculated for every curvature encountered when electro-dynamic braking cannot maintain the train velocity
at the considered railway route. In order to simplify the problem of less than 5 km/h above the target value, additional mechanical
curvature-related resistance, a constant average curvature value is brakes need to be applied, which is also modeled as a proportional
Fig. 6. Block diagram of conventional diesel-electric locomotive quasi static model.

train driver braking command, related to the wheel vs. track maximum load for a single locomotive traveling over that route,
braking potential, expressed as the overall train weight multiplied which is typically limited below 600 t [34]. Since velocity limita-
with gravitational constant and adhesion coefficient from equation tions are known in advance (Fig. 7c), the train velocity reference vref
(9). (red plot in Fig. 8c) is generated as a gradual velocity change over a
2 km window by using the a-priori known limitation profile. Since
5. Mountainous railway route driving scenario such velocity target also prevents excessive accelerations and de-
celerations, comprehensive use of electric braking is anticipated in
This section presents the train longitudinal motion data over a that case. For the given speed target vref and the actual train velocity
mountainous railway route and the estimation of the energy re- v (blue trace in Fig. 8c), the train driver model results in throttle
covery potential of the battery-hybrid diesel-electric locomotive. (Notch) and braking (Brk) commands shown in Fig. 8b.
Please note that positive Notch values in Fig. 8b correspond to
5.1. Acquiring realistic route data the traction, whereas negative values correspond to electrical
braking operation (see discussion above). However, under certain
A rather simplified rail route longitudinal profile between conditions, electrical braking is not always sufficient, so additional
Zagreb and Split is presented in Ref. [35], which represents the friction brakes need to be applied, especially during train
official route data provided by the national railway company descending on a rather steep track slope (cf. Fig. 8a and b). On the
The rail route segment from the town of other hand, the train velocity plots in Fig. 8c show that the available
Hrvatske Zeljeznice (HZ).
train traction power is not always sufficient to follow the velocity
Ostarije to the town of Knin, which corresponds to the target route
reference over the ascending track profile, even though the train is
within Lika region, is shown in Fig. 7a. In order to acquire a more
still able to climb at lower velocities. Fig. 8c also shows that the
detailed elevation profile, the free on-line GPS Visualizer utility
train longitudinal velocity may exceed the reference (speed limit)
software [49] has been used herein, which supports exporting of
during certain descending intervals (e.g. 200 km < l < 280 km)
the route data related to route length and elevation into suitable
which is due to non-ideal characteristics of the power-train
spreadsheet format. Since this particular railway route includes
regenerative braking power output and the relatively simple vir-
bridges, viaducts and tunnels, some subsequent post-processing is
tual driver control law. Of course, in actual application, the human
required in order to correct the railroad slope data. Fig. 7b shows
train driver would have to strictly comply with the speed limit.
the final elevation profile obtained by using the GPS Visualizer
Note also that the proposed load configuration represents a worst
software. Speed limitation is taken from the available on-line
case scenario, so better reference tracking would be expected for
source [50], and additional speed limits of 20 km/h are included
the case of lighter load.
in order to account for several in-route stations, thus providing for
Fig. 9 shows the locomotive electrical transmission power and
more realistic train driving behavior. The final speed limit profile is
cumulative fuel consumption, along with electric energy con-
given in Fig. 7c, while the overall route is outlined at the map
sumption and energy recovery potential. In particular, the blue
shown in Fig. 7d.
trace in Fig. 9a shows the electrical transmission power, related to
the Notch variable (cf. Fig. 8b) according to the transmission power
5.2. Energy consumption and energy recovery potential
map (Fig. 5b), which also includes the power conversion effi-
ciencies. The cumulative fuel consumption volume (in liters) is
The considered driving scenario, shown in Fig. 8, corresponds to
shown as the red trace, and it is calculated from the power-related
a return journey over the mountain route given in Fig. 7, with seven
fuel mass flow according to the following expression:
fully-loaded 90 t cargo wagons. This roughly corresponds to the
Fig. 7. Officially-available simplified elevation profile (a), GPS Visualizer-generated elevation profile (b), speed limitation profile (c) and map outline of considered railway route (d).
Fig. 8. Proposed driving scenario elevation (a), train driver commands (b) and velocity (c) profiles.
Fig. 9. Locomotive electrical transmission power and cumulative fuel consumption (a), electric energy used (b) and energy recovery potential (c).
braking periods, followed by the integration of positive traction

ðtf power values until the accumulated energy would be depleted
1
Vf ¼ m_ f dt; (15) (spent), which would represent the traction-assist periods made
rf available by the recovered energy.
0
where rf ¼ 850 kg/m3 ¼ 850 g/L is the diesel fuel density (grams per 6. Battery-hybrid locomotive model
liter).
Fig. 9b shows the electric transmission energy requirement for This section proposes a battery-hybrid diesel-electric locomo-
the considered driving scenario, while Fig. 9c shows the electric tive configuration wherein the battery is adequately sized in order
energy recovery potential, which is subsequently used as a basis for to store the estimated braking energy recovery potential. Due to
battery energy storage system sizing. It is estimated by integrating additional battery power capability, the main diesel engine can be
the electrical transmission power during electrical traction-based somewhat downsized, while retaining the same traction
performance, thus keeping the locomotive weight within the same Table 3
limits. Li-Ion battery cell and overall hybrid locomotive battery system parameters.
N E (kWh) Qmax (Ah) Pmax (kW) m (kg)

6.1. Battery model Cell 1 0.06 15.9 0.4 0.63
Battery 15000 900 1192.5 6000 9450
Due to exceptional energy density and durability, lithium-ion
(Li-Ion) battery technology is considered for battery-hybrid loco-
motive retrofitting. The battery model is derived from the equiva- 20% of energy reserve is also included, thus resulting in 900 kWh of
lent circuit shown in Fig. 10a, which results in the following well- final battery energy storage capacity. This requires N ¼ 15000 bat-
known relationship [51]: tery cells, which are arranged into 75 parallel blocks of 200 series-
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi connected cells each (in order to obtain the desired operational DC
dSoC U 2oc ðSoCÞ 4RðSoC; iÞPbatt Uoc ðSoCÞ link voltage between 620 V and 780 V). Finally, by multiplying the
¼ ; (16) cell weight and capacity with the number of battery cells N, the
dt 2Qmax RðSoC; iÞ
overall battery weight and capacity are calculated and listed in
where 0 SoC 1 is the battery State-of-Charge, Qmax is the battery Table 3.
capacity, and Pbatt is the battery power flow. Within the proposed In order to calculate the cost of the proposed Li-Ion battery, its
battery model, the open circuit voltage Uoc and internal resistance R specific cost needs to be estimated. It depends on the aspects of
are made dependent on the SoC, while the internal resistance is also battery utilization and battery technology itself, and it has also
made dependent on the current direction (i.e. dependent on shown a strong declining trend in recent years. According to
whether the battery is being charged or discharged) [52]. Ref. [55], in 2016 the total energy installation costs for utility-scale
For the case of discharging and relatively low battery resistance, Li-Ion battery system have been between 200 USD/kWh and 1260
the above expression simplifies to the following integral-type USD/kWh which approximately corresponds to the range between
battery SoC model: 177 EUR/kWh and 1113 EUR/kWh for the current EUR to USD ex-
change ratio of 1/1.132. Reference [55] also predicts that battery
dSoC Pbatt system costs will continue to decline and are likely to be between
¼ ; (17) 77 and 574 USD/kWh in 2030. According to Ref. [56], battery system
dt Uoc ðSoCÞQmax
cost for electric vehicle (EV) applications in 2016 also ranged be-
wherein the battery SoC rate of change represents the ratio between 203 and 415 USD/kWh, with cost prediction over the next
tween instantaneous power demand and the instantaneous stored couple of years to be around or below 200 USD/kWh (177 EUR/
energy. kWh).
Open circuit voltage and internal resistance characteristics, For the purpose of battery cost modeling in this work, the cost of
adopted from Autonomie software, are shown in Fig. 10b and c for a the proposed battery is estimated to cbatt ¼ 200 EUR/kWh, thus
single Li-Ion battery cell. It is assumed that the battery is connected reflecting the declining battery system cost trends and the fact that
directly to the common DC link so the battery power in equation the presented research is aimed at future retrofitting designs of
(16) is calculated as: diesel locomotives. In particular, this lower cost estimate satisfies
both costing ranges (i.e. those related to stationary battery appli-
Pbatt ¼ Pt Pg ; (18) cations and vehicular battery systems) because future battery costs
are estimated to 200 EUR/kWh or below according to data provided
where Pg is the main generator electric power. Note also that pos- by Refs. [55,56]. Hence, the battery cost is calculated according to
itive Pbatt values correspond to battery discharging operation. the following straightforward expression:
Battery sizing has a major effect on the vehicle range and per-
formance [53]. The first row in Table 3 shows the Li-Ion cell pa- Cbatt ¼ cbatt $Ebatt ; (19)
rameters (such as cell energy E, charge capacity Qmax, power rating
Pmax and mass m), which are obtained by scaling down the Li-Ion where Ebatt is the battery energy storage capacity.
battery used in Ref. [54] with respect to the number of its cells. In
order to calculate the required number of cells for the hybrid 6.2. Approximate data for downsized EMD-645 engine
locomotive battery, the maximum value of energy of recovery po-
tential from Fig. 9c can be used, which results in approximately Using the sizing approximation procedure outlined in Section 3
500 kWh of storage capacity. However, the battery is oversized for a for the case of main engine and generator, which can be re-sized by
50% energy reserve (750 kWh) in order to ensure that its energy is using the smg scaling coefficient (see Table 2), the same procedure
sufficient to cover for different driving requirements, and also to may be applied to the battery-hybrid locomotive power-train. The
avoid deep battery discharges. Moreover, in order to account for maximum electrical traction power which is generated by the main
inevitable capacity reduction caused by battery aging, additional generator for the considered locomotive is Ptmax ¼ 1402.08 kW. This
Fig. 10. Battery equivalent circuit (a), open-circuit voltage vs. state-of-charge dependence (b), and internal resistance vs. state-of-charge and current direction (c).
Table 4
Downsized main engine and generator data.
Throttle position Main engine power Pmg (kW) Generator power Pg (kW) Fuel rate m_ f (g/s) Emissions rate (g/s)
HC CO NOX CO2
IDLE 6.43 0 2.5704 0.0299 0.0436 0.1510 7.9910

Notch 1 50.81 40.01 5.2994 0.0326 0.0538 0.2781 16.8449
Notch 2 196.00 171.77 13.8816 0.0395 0.1032 0.6463 43.8683
Notch 3 359.71 317.54 23.4887 0.0647 0.0977 1.2115 74.3167
Notch 4 632.77 566.49 40.8315 0.0834 0.1503 2.4458 129.3824
Notch 5 787.15 694.13 51.6475 0.1136 0.3057 3.2341 163.4197
Notch 6 965.83 844.70 64.2834 0.1593 0.9703 4.0096 202.2778
Notch 7 1161.70 1004.73 79.8195 0.2585 2.2294 4.8944 249.5078
Notch 8 1312.80 1121.67 94.4690 0.3314 4.1517 5.2350 292.7589
value is chosen to be equal for the battery-hybrid counterpart in This additional rule could be used to turn the engine off instead of
order to maintain the same traction characteristics and driving idling in order to further improve fuel savings. However, due to the
performance. However, according to equation (18), the electrical slow dynamics of the large diesel engine, it is more convenient to
traction power of the battery-hybrid locomotive now represents leave the engine idling, then to frequently turn it on/off. The power
the sum of battery and main generator power. Thus, the chosen demand of SoC controller is calculated according to the eSoC ¼ SoCR -
usable energy capacity of the battery (500 kWh) can be used during SoC difference where the reference SoCR can be selected to be
traction intervals, and the main engine and generator can be constant (SoCRC) or variable (SoCRV).
downsized. It is assumed that the locomotive can operate at the A constant SoC reference value is often chosen for hybrid electric
maximum throttle position for two hours, so during that period the vehicles [25]. However, due to mountainous route being charac-
battery can provide electrical power of Pbatt2h ¼ 250 kW. In that terized by significant variations in elevation profile resulting in
case, the required main generator power is Pg ¼ Ptmax - Pbatt2h, so the notable variations of train potential energy, it would be more
generator can be downsized by the ratio Pg/Ptmax z 0.8. By using convenient to use a variable SoC reference [58]. Having this in mind,
this ratio with the G26CW locomotive data, the final scaling coef- the battery state-of-charge (SoC) reference value SoCRV may be
ficient smgh ¼ 0.8 smg ¼ 0.8226 is obtained, and the resulting calculated as originally proposed in Ref. [58]:
approximate data for the downsized main engine and generator are
given in Table 4. The overall weight of the proposed hybrid loco- h hmin
motive becomes 107.85 t (still within the same load category of 18 t SoCRV ðhÞ ¼ SoCbh $ðSoCbh SoCbl Þ; (20)
hmax hmin
per axle), and it comprises the G26CW locomotive base weight mlb
(see Section 4), battery weight mbatt (see Table 3), and the down- where hmin and hmax are the minimum and maximum elevation for
sized engine weight (m*mg ¼ 17.93 t according to equation (6)). the particular route, SoCbl and SoCbh are the lower and upper
bounds of variable SoC reference. The elevation h can be easily
measured by means of GPS or it can be simply reconstructed based
6.3. Battery-hybrid locomotive model and optimized control
on the a-priori known track profile over the predefined route. The
strategy
SoC controller also comprises a dead-zone DSoC in order to avoid
frequent switching between charging and discharging in the vi-
Fig. 11 shows the block diagram representation of the quasi-
cinity of the SoC target value, and its gain KSoC is chosen to provide
static model of battery-hybrid locomotive, wherein specific sub-
stable closed-loop behavior, while simultaneously being able to
models are framed by dotted lines. The longitudinal dynamics
maintain approximately constant battery SoC within the deep
model, train driver model, traction and power maps within the
discharge area (i.e. when SoC decreases towards 20%). In Ref. [58],
power-train model are the same as for the conventional locomotive
the controller parameters have been derived by iterative search for
(Fig. 6). Again, the driver model calculates the throttle Notch posi-
a particular driving scenario characterized by a fully-loaded train
tion and brake Brk commands based on the difference between the
and a new battery. The iterative search has been performed with
target and actual train velocity. The key difference is in the power-
the goal of obtaining stable overall control system behavior, while
train model. Even though it still comprises the engine-generator
also avoiding deep battery discharges, which are associated with
sub-model providing fuel and emission rates (as in the conven-
low values of SoC controller gains. Obviously, such an approach
tional locomotive model in Fig. 6.), herein the difference between
may not yield an optimal SoC controller parameter set. Therefore, a
the electric generator power Pg and the transmission power Pt feeds
DIRECT2 search algorithm from Ref. [59] is implemented within
the battery model, according to equation (18), which estimates the
Matlab™ environment in order to find the optimal controller
battery state-of-charge (SoC) based on the power input Pbatt and
parameter values. For the infinite number of iterations, a set of
known battery parameters (see Fig. 10).
points sampled with the DIRECT algorithm forms a dense subset of
A rule-based controller (inspired by a similar concepts proposed
a unit hypercube, thus eventually converging towards the global
in Refs. [48,57]) combines the transmission power demand Pt and
optimum [60]. While the DIRECT algorithm can find a globally-
the battery state-of-charge controller power demand PbR in order to
optimal solution, the main drawback is that if there are local op-
calculate the required main engine-generator power PgR, which is
timum solutions, the optimization time is rapidly increasing due to
then related to the throttle NotchRB by means of Notch selector map
the need for exhaustive searching throughout the entire domain of
(cf. Fig. 2a), which is adjusted to the maximum power of the
definition.
downsized engine. An additional rule is introduced, which checks
whether the requested power PgR is lower than the power gener-
ated when NotchRB ¼ 4 is selected. In that case the engine is auto- 2
DIviding RECTangles (or, more precisely, dividing hypercubes) approach de-
matically brought to idling in order to avoid low-efficiency engine scribes the manner in which the optimization algorithm partitions the n-dimen-
operation (cf. operating points in Fig. 3c for first two notch values). sional search space as it converges towards the optimal solution [59].
Fig. 11. Block diagram of quasi-static model of proposed battery-hybrid locomotive.
In this study, the optimization problem solved by the DIRECT larger controller gain KSoC, and wider margins of the controller
algorithm is to find SoC controller parameters (KSoC, DSoC, SoCbl, dead-zone DSoC and SoC reference upper and lower margins SoCbl
SoCbh) which minimize the overall locomotive fuel consumption for and SoCbh, when compared to the fully-loaded train scenario and
the particular driving scenario, subject to the minimum battery SoC brand new battery (third and fifth row in Table 5). Since the opti-
constraint of 20%. Note that for the case of constant SoC reference mization result covering a wide range of conceivable scenarios
the SoCR value is optimized instead of SoCbl and SoCbh bounds.. The would be more robust compared to the specific case of a fully-
optimization was conducted for different simulation scenarios, loaded train and brand new battery, this more general set of opti-
corresponding to a fully and partially loaded train. It should be mized controller parameters is used in the further analysis.
noted that each optimization run, corresponding to a particular
train configuration and load (zero wagons to seven fully-loaded
7. Battery-hybrid locomotive simulation results
wagons), results in its own set of locally-optimal SoC controller
parameters, which would mandate controller scheduling depend-
The hybrid locomotive is again simulated for the previously
ing on the freight train configuration and battery state-of-health.
used driving scenario, shown in Fig. 8, which corresponds to a re-
Thus, in order to simplify the controller parameterization, a near-
turn journey over the mountainous route shown in Fig. 7 with
optimal result is sought which would simultaneously minimize
seven fully-loaded 90-ton cargo wagons. Since the same driver
the fuel consumption as well as the battery SoC control perfor-
model and traction characteristics have been used in simulations
mance, while honoring the battery SoC hard constraint (SoC > 20%)
and the weight of the train was negligibly changed due to slightly
for different loads (from 0 to 7 fully-loaded 90-ton cargo wagons)
larger weight of the hybrid locomotive (107.85 t compared to 103 t
and two distinctive cases of brand new and aged battery, charac-
of the conventional one), similar driving characteristics are ob-
terized by energy (charge) capacity reduced by 20% and internal
tained in terms of train acceleration and velocity. This is illustrated
resistance increased by 100% compared to the nominal case of the
by the velocity difference Dv between the conventional diesel-
brand new battery. The comparative optimization results (opti-
electric locomotive (benchmark) and the battery-hybrid locomo-
mized SoC controller parameters) obtained for the case of a full-
tive used for freight transport, shown in Fig. 12.
loaded train and arbitrary train configuration/load are summa-
Within the battery-hybrid locomotive, the available electrical
rized in Table 5. In particular, optimization results obtained for a
transmission power Pt includes the main electric generator power
wide range of train configurations and the cases of brand new and
Pg and battery power Pbatt. In the case of train accelerating and
aged battery (fourth and sixth row in Table 5) generally results in a
climbing, the downsized engine power is not sufficient to provide
Table 5
Hybrid locomotive battery SoC controller parameters.
Reference type Parameters
KSoC DSoC SoCR SoCbl SoCbh
(SoCRC) Full load only 3866 0.13% 63.13% e e

All loads 8500 4.33% 64.79% e e
(SoCRV) Full load only 26796 0.26% e 41.51% 53.63%
All loads 48786 1.34% e 42.08% 65.12%
Fig. 12. Conventional vs. hybrid-based freight velocity difference.
the required traction power, so battery is being discharged and its provide the required power and maintains the battery SoC above
SoC starts to decrease (cf. Fig. 13 and Fig. 8). In the case of regen- 20%. During battery charging, overcharging is prevented by
erative braking during deceleration or downhill driving the battery disabling the charging process when its SoC reaches 95%, wherein
is being charged and its SoC is increased in turn. Fig. 13a shows the the excess electrical power during electrical braking may be
SoC trace when a constant-valued SoC reference is used, while in diverted to the dissipation resistors grid, unless various auxiliaries
the case of variable (elevation-dependent) SoC reference, battery such as air-conditioning, lighting and similar would be able to
SoC is shown in Fig. 13b. absorb the excess power. Fig. 13c shows the fuel consumption plots
As expected, the variable SoC target results in battery SoC for the conventional locomotive and its battery-hybrid counter-
exhibiting slightly lower-magnitude perturbations compared to the parts with and without battery SoC reference adjustment, assuming
case of constant SoC target, because in the former case the SoC a brand new battery.
target is dynamically adjusted in order to reflect the variations in Additional simulation results for different train loads and for
the available potential energy. In this way, forceful battery charging optimal controller parameters from Table 5 are listed in Table 6. The
is effectively avoided (cf. Fig. 13a and b), which should be beneficial results confirm that the battery SoC controller utilizing the pa-
from the standpoint of battery cycle and calendar life. Moreover, rameters optimized for a wide range of scenarios, including
the variable SoC reference also results in slight improvement of fuel different loads and different battery state-of-health, is able to
efficiency, manifested as fuel consumption reduction (cf. Table 6). maintain the SoC above 20% for all scenarios thus providing a robust
Red dashed lines in Fig. 13a and b represent the battery SoC traces in response. The results obtained for the case of controller optimiza-
the case when battery capacity is reduced by 20% and internal tion based on the specific scenario of a fully-loaded train and a
resistance is increased by 100% with respect to their nominal values brand new battery show that it would be possible to further reduce
(thus emulating battery aging effects according to Ref. [61]). For the fuel consumption compared to the aforementioned near-
both SoC reference types the hybrid locomotive still manages to optimal solution based on a wide range of scenarios. However, in
Fig. 13. SoC variable for the case of constant SoC reference (a) and using variable SoC reference (b), and comparative cumulative fuel consumption of hybrid and conventional
locomotive (c).
Table 6
Simulation results for different loads and different controller parametersa.
Wagons Conv. Vf (L) Hybrid (SoCRC) Hybrid (SoCRV)
Vf (L) SoCMIN (%) Vf (L) SoCMIN (%)
New Aged New Aged New Aged New Aged
0 411 (433)434 (435)435 (54.60)55.87 (53.78)55.81 (385)406 (394)413 (41.20)43.46 (40.31)42.69

1 718 (651)640 (655)644 (56.35)56.64 (55.93)56.36 (583)597 (599)605 (40.82)42.84 (40.31)42.47
2 1082 (942)934 (946)942 (54.33)54.26 (52.86)52.82 (882)889 (903)901 (40.49)42.87 (40.04)42.60
3 1447 (1210)1205 (1214)1218 (46.69)49.76 (44.58)47.75 (1168)1166 (1194)1185 (38.09)40.02 (36.55)39.82
4 1793 (1483)1481 (1497)1506 (42.51)45.15 (38.65)41.70 (1453)1444 (1484)1472 (35.82)39.43 (33.23)37.30
5 2124 (1763)1762 (1788)1791 (37.29)39.94 (32.23)35.03 (1738)1733 (1770)1766 (31.67)37.38 (26.65)34.55
6 2442 (2033)2032 (2057)2069 (32.47)34.82 (26.59)28.30 (2004)2013 (2041)2045 (26.57)32.85 (20.07)27.54
7 2750 (2307)2314 (2331)2357 (24.86)28.22 (12.92)20.00 (2279)2297 (2321)2337 (20.00)27.03 (12.54)20.00
a
Values in brackets correspond to controller parameters optimized for a fully-loaded train and a new battery.
that case the controller may be unable to maintain the battery SoC the locomotive battery energy storage system before an overhaul
above the lower limit of 20% in the case of an aged battery. The needs to be performed. In addition, the aforementioned charge/
results also show that for a zero load (zero wagon case) and con- discharge cycle roughly corresponds to a single return journey over
stant SoC reference case, the hybrid locomotive may have slightly the mountainous railway line, because this particular scenario
larger fuel consumption when compared to the conventional one, contains a similar peak-to-peak charging/discharging of the battery
which is due to sub-optimal yet robust SoC controller tuning which (cf. SoC trace in Fig. 13). The resulting cumulative fuel consumptions
has proven to be suitable for all conceivable fright train configu- and fuel costs of the conventional locomotive and the battery-
rations (i.e. numbers of wagons and loads). In order to achieve a hybrid locomotive with variable SoC reference are listed in
better result, as mentioned above, this would require a rather Table 9. According to the presented results and for the current
complex controller parameter scheduling mechanism based on diesel fuel price of 1.35 EUR/L [62], it is expected that the battery-
comprehensive optimization for all conceivable train configura- hybrid locomotive could reduce fuel costs by about 1.2 million
tions, which might not be practical to implement in the field. EUR (corresponding to 16.47% fuel consumption reduction) over
The final fuel consumption data which are used in further the observed period of utilization. A “rough” estimate of hybridi-
analysis are listed in Table 7, which illustrates the main benefits of zation costs would include the main diesel-generator cost and
battery-hybrid locomotive for the case of fully-loaded train in terms battery cost, and it may be regarded as the overall investment cost
of fuel consumption reduction. In particular, the fuel efficiency (Table 10). In that case, the overall savings achieved by the utili-
improvement amounts to approximately 15% both without and zation of battery-hybrid locomotive are estimated to about 1.06
with SoC target adaptation. Table 7 also lists diesel fuel costs based million EUR (which corresponds to 14.25% expense cost reduction
on current retail fuel price of cfuel ¼ 1.35 EUR/L in Croatia [62]. by means of diesel fuel savings). Based on the data listed in Table 10,
Additional advantage of hybrid locomotive utilization is in the it has been estimated that the diesel-electric locomotive hybridi-
reduction of exhaust emissions which are indicated in Table 8. zation by means of battery energy storage would result in a net gain
amounting to three times the projected investment costs over the
8. Comparative cost assessment useful lifetime of such a hybrid system, with return-of-investment
period being 1/4 of the projected battery system lifetime.
According to Ref. [23], Li-Ion batteries are characterized by cycle Due to the fact that proposed battery and fuel costs could be
lives of up to 2000 cycles wherein the battery is periodically fully different and can also change over time, a maximum acceptable
charged and discharged to 80% depth-of-discharge. This cycle life battery cost for hybridization can be roughly estimated as:
value is chosen in this assessment as the representative lifetime of
Table 7
Comparative fuel consumption data from simulation.
Table 9
Locomotive Fuel consumption Fuel cost (EUR) Fuel cost for a single train run and summarized fuel cost over 2000 runs.
(L)
Locomotive type Cumulative Fuel Cost (EUR)
New Aged New Aged
consumption (L)
Conventional 2750 2750 3712.50 3712.50
1 run 2000 runs
Hybrid (SoCRC) 2314 2357 3123.90 3182.95
Hybrid (SoCRV) 2297 2337 3100.95 3154.95 Conventional 2750 5500000 7425000
Hybrid (SoCRV) 2297 4594000 6201900
Fuel savings 453 906000 1223100
Fuel efficiency improvement [%] 16.47%
Table 8
Comparative exhaust gas emissions data from simulation.
Locomotive Emissions (kg) Table 10

HC CO NOX CO2 Comparative costs of conventional and battery-hybrid diesel-electric locomotives.
Conventional 8.21 94.97 130.43 7257 Locomotive Cost (EUR)

Hybrid (SoCRC) New 6.97 82.02 109.94 6103
Main engine-generator Battery Summarized
Hybrid (SoCRC) Aged 7.09 83.84 111.92 6215
Hybrid (SoCRV) New 6.99 83.23 108.72 6055 Conventional 73845 e 73845
Hybrid (SoCRV) Aged 7.11 84.98 110.55 6160 Hybrid 59076 180000 239076
useful life corresponding to 2000 charging/discharging cycles

within the hybrid locomotive.
Vf ;conv Vf ;hev $Nrun $cfuel
cf ;batt ¼ (21)
Ebatt 9. Conclusion
where Vf,conv and Vf,hev are cumulative fuel consumption in liters for The paper has proposed a backward-looking quasi-static model
the conventional locomotive and for the fully-loaded train scenario of a conventional 1.6 MW heavy-haul diesel-electric locomotive
comprising a battery-hybrid locomotive, respectively (see Tables 6 utilized by the national railway company, based on the known
and 7), Nrun is expected number of exploitation driving cycles locomotive power-train parameters, such as diesel engine rated
(chosen to be 2000 to cover the estimated battery life cycle), cfuel is power and output power throttle control, and electric transmission
fuel price in EUR/L and Ebatt is the proposed battery energy capacity type. In order to derive the fuel consumption model and related
in kWh. green-house gases emissions model, and the traction force and
According to equation (21), using data from Table 9 and taking power characteristics of the drive-train, a similar diesel-engine/
into account the previously defined fuel price, the maximum generator set characteristics and their respective fuel consump-
acceptable battery cost is 1359 EUR/kWh. However, in that case the tion data are taken from the available literature, and are scaled in
benefit of hybridization is only related to the greenhouse gases order to match the target locomotive power ratings. Thus-obtained
emissions reduction potential. conventional locomotive model has been subsequently converted
It should be noted that the presented analysis did not include to its battery-hybrid counterpart by adding a sufficiently-sized
the influence of less-demanding diesel generator operating regimes battery energy storage system in parallel to the generator. In
due to battery being able to take on the peak power-train loads, addition, an appropriate optimized energy management strategy
which would result in reduced strain on the diesel generator set. has been designed in order to maintain the battery state-of-charge
This might provide additional benefits in terms of reduced diesel within pre-defined limits, while simultaneously assisting the con-
generator component wear, and related reduction in maintenance ventional power-train during driving, and also facilitating kinetic
and overhaul costs, as indicated in Ref. [63]. Note also that the energy recovery during descending by means of electrical braking.
above assessment has been carried out using data corresponding to The fuel efficiency performance of the conventional and hy-
a relatively old diesel engine (EMD 16-645-E), which can lead to an bridized locomotive models has been performed for the round-trip
overestimate of the savings compared to hybridization of a more simulation scenario corresponding to the mountainous railway
recent locomotive equipped with more advanced diesel engine route operated by the national railway company (Hrvatske
design. Finally, it should be noted that the presented fuel gain and This scenario includes realistic slope and speed
zeljeznice, HZ).
return-of-investment obtained for the considered mountain rail limitations. The obtained simulation results have shown that the
route might not directly relate to flatter train routes. Namely, these battery-hybrid locomotive can facilitate notable fuel savings which
types of train routes are characterized by smaller changes in amount up to 16.5% fuel cost reduction, which is also reflected in
elevation, where the kinetic energy obtained during acceleration proportional greenhouse gases emissions reductions. Moreover, for
would predominantly determine the regenerative braking energy the expected battery cycle life and related anticipated capacity drop
gains during deceleration, while the exploitation of potential en- of 20%, the hybrid locomotive can maintain the same traction
ergy difference due to elevation profile change would be less performance as its conventional counterpart. Finally it has been
emphasized. estimated that the hybridization investment costs would be
In addition, costs of route electrification are estimated and compensated three times within the expected battery life-time
compared with the costs of acquiring 20 brand new battery-hybrid period based on the current Li-Ion battery installation costs.
locomotives, which corresponds to the number of currently oper- Future work is going to be aimed towards collecting and
ational conventional diesel-electric locomotives of the particular analyzing the data of realistic freight train round-trip driving cycles
type studied herein [34]. It is estimated that capital cost of track for different routes. This experimental data may then be used as a
electrification in the USA roughly amounts to 4.8 million USD/mile basis for a more comprehensive assessment of battery energy
(2.63 million EUR/km) [9]. Thus the total track electrification cost storage system requirements for the battery-hybrid diesel-electric
for the route considered in this paper is estimated to 591 million locomotive, and related analyses of fuel efficiency improvement
EUR. On the other hand, if new diesel-electric locomotives would and related greenhouse gases emission reduction potentials.
be purchased, the unit cost would be around 3 million USD for a
4.500-horsepower unit [64], which corresponds to a specific cost of
Acknowledgements
790 EUR/kW. Therefore, the estimated cost of a single diesel-
electric locomotive, equipped with 1.6 MW power-train would
It is gratefully acknowledged that this research has been sup-
amount to 1.26 million EUR. However, in order to estimate the cost
ported by the European Regional Development Fund under the
of the battery-hybrid locomotive, the aforementioned conventional
grant KK.01.1.1.01.0009 (DATACROSS). Authors would also like to
locomotive cost is adjusted with respect to the reduction of the
express their appreciation of the efforts of the Associate Editor and
main engine-generator size and cost when compared to the con-
anonymous reviewers whose comments and suggestions have
ventional locomotive, and also due to cost increase related to the
helped to improve the quality of the presented subject matter.
battery energy storage system installation (see Table 10). Thus, the
total cost of battery-hybrid locomotive purchase is estimated to
1.43 million EUR, which results in the overall costs of retrofitting of Nomenclature
a 20 hybrid locomotive fleet of 28.6 million EUR, which is still
twenty-times smaller when compared to capital investments for
track electrification. Since this assessment is based on available US Abbreviations
pricing data obtained from the Internet, the actual retrofitting costs AC alternating current
may differ from the above estimate, and this assumption does not CO carbon-monoxide
include the maintenance cost and life duration of track electrifi- CO2 carbon-dioxide
cation which is significantly longer in comparison to the battery DC direct current
EMD Electro-motive division Ft traction force [N]

EU European union FtMAX maximum traction force [N]
EUR Euro currency h elevation [m]
GPS global positioning system l traveled distance [m]
HC hydrocarbons ma train mass [kg]
HZ Hrvatske zeljeznice m_ CO carbon-monoxide emission rate [g/s]
Li-Ion lithium-ion m_ CO2 carbon-dioxide emission rate [g/s]
NOX nitrogen oxides m_ exh overall emission rate [g/s]
USD United states currency mf specific fuel consumption [g/kWh]
m_ f engine fuel consumption rate [g/s]
Parameters m_ HC hydrocarbons emission rate [g/s]
a linear fit coefficient [t/kw] m_ NOX nitrogen oxides emission rate [g/s]
b linear fit offset [t] Notch throttle position []
cbatt specific battery cost [EUR/kW] NotchRB controller throttle position []
Cbatt Total battery cost [EUR] Pbatt battery power [W]
Ce specific engine cost [EUR/kW] Pbatt2h two hours battery power [W]
cf,batt acceptable battery cost [EUR] PbR SoC controller power demand [W]
cfuel retail fuel price [EUR/L] Pf engine fuel-related power [W]
Cg specific electric generator cost [EUR/kW] Pg main generator electric power [W]
E energy [kWh] PgR main engine-generator power [W]
Ebatt battery energy capacity [kWh] Pi input power [W]
g gravity acceleration [m/s2] Pmg main engine power [W]
hmin minimum route elevation [m] Po output power [W]
hmax maximum route elevation [m] Pt electric traction power [W]
k train type configuration coefficient [N/t] r curvature radius [m]
KDr proportional gain [] R battery internal resistance [U]
KSoC SoC controller proportional gain [] s Laplace variable []
m mass [kg] smg scaling factor []
ml overall locomotive weight [t] smgh hybrid locomotive scaling factor []
mlb locomotive base weight [t] SoC battery state of charge []
mmg main engine and generator weight [t] SoCMIN minimum state of charge []
m*mg downsized engine and generator weight [t] SoCR state of charge reference []
N Number of battery cells [] SoCRV variable state of charge reference []
Nrun Number of exploitation driving cycles [] t time [s]
PMAX maximum power [W] Uoc battery open circuit voltage [V]
PmgMAX scaled engine maximum power [W] v longitudinal velocity [m/s]
PmgMAX0 original engine maximum power [W] vref train velocity reference value [m/s]
PtMAX maximum electrical traction power [W] Vf fuel consumption volume [L]
Qmax battery capacity [As] Vf,conv conv. locomotive fuel consumption [L]
r0 curvature resistance constant radii [m] Vf,hev hybrid locomotive fuel consumption [L]
R1 minimum route radii [m] wk specific motion resistance [N/t]
R2 maximum route radii [m] wr curvature motion resistance [N/t]
SoCbh upper bound of variable SoC reference [] a track slope [ ]
SoCbl lower bound of variable SoC reference [] ε emission coefficient []
SoCRC constant state of charge reference [] εCO carbon-monoxide emission coefficient []
vd wind velocity addition [km/h] εCO2 carbon-dioxide emission coefficient []
v0 motion coefficient velocity constant [km/h] εHC hydrocarbons emission coefficient []
wk0 specific motion resistance constant [N/t] εNOX nitrogen oxides emission coefficient []
wr0 curvature resistance constant [Nm/t] Dv velocity difference [km/h]
wr average curvature motion resistance [N/t] hel mechanical-to-electrical efficiency []
b adhesion coefficient parameter [km/h] ma adhesion coefficient []
g adhesion coefficient parameter [km/h] te main engine torque [Nm]
DSoC SoC controller dead zone [] ue main engine speed [rad/s]
rf fuel density [g/dm3]
References
Dynamic variables [1] McCollum D, Krey V, Kolp P, Nagai Y, Riahi K. Transport electrification: a key
Brk brake command [] element for energy system transformation and climate stabilization. Clim
c0 approximation coefficient [g/s] Change 2014;123(3e4):651e64. https://doi.org/10.1007/s10584-013-0969-z.
c1 approximation coefficient [g s1∙(Nm)1] [2] Buzzonii L, Pede G. New prospects for public transport electrification, inter-
national conference on electrical systems for aircraft, railway and ship pro-
c2 approximation coefficient [g s1∙(Nm)2] pulsion. Bologna, Italy: ESARS); 2012. October 16-18, https://doi.org/10.1109/
eSoC state of charge difference [] ESARS.2012.6387464.
Et used electric energy [kWh] [3] Shakya SR, Shrestha RM. Transport sector electrification in a hydropower
resource rich developing country: energy security, environmental and climate
Ebp energy recovery potential [kWh] change co-benefits. Energy Sustain Dev 2011;15(No. 2):147e59. https://doi.
Fb braking force [N] org/10.1016/j.esd.2011.04.003.

[4] Deur J, Skugor B, Cipek M. Integration of electric vehicles into energy and 1016/j.energy.2016.04.051.
transport systems. Autom J Contr Meas Electron Comput Commun [31] Tang R, Wu Z, Li X. Optimal operation of photovoltaic/battery/diesel/cold-
2015;56(No. 4):395e410. https://doi.org/10.7305/automatika.2016.01.1106. ironing hybrid energy system for maritime application. Energy 2018;162:
[5] Spiryagin M, Cole C, Sun YQ, McClanachan M, Spiryagin V, McSweeney T. 697e714. https://doi.org/10.1016/j.energy.2018.08.048.
Design and simulation of rail vehicles. Taylor & Francis Group, LLC; 2014, ISBN [32] Terblanche PJ, Kearney MP, Knights PF. Potential of on-board energy recovery
978-11-380-7370-8. systems to reduce the costs of dieseleelectric mine truck haulage. Mining
[6] Frey S. Railway electrification systems & engineering. first ed. White Word Technology; 2018. https://doi.org/10.1080/25726668.2018.1451611.
Publications; 2012, ISBN 978-81-323-4395-0. [33] Schatz RS, Nieto A, Dogruoz C, Lvov SN. Using modern battery systems in light
[7] Mandi c M, Uglesi
c I, Milardi
c V, Filipovi
c-Gr
ci
c B. Application of regenerative duty mining vehicles. Int J Min Reclam Environ 2015;29(No. 4):243e65.
braking on electrified railway lines in AC traction systems 25 kV, 50 Hz. In: https://doi.org/10.1080/17480930.2013.866797.
12th symposium HRO CIGRE;
2015. Sibenik, Croatia, November 8-11.
[34] Skrobonja D, Berakovic T, Bo zunovic A, Ljubaj Z, Ljubaj N. Applicative moni-
[8] Al-Tony FES, Lashine A. Cost-benefit analysis of railway electrification: case toring of locomotive diesel engine oil. In: Professional paper, fuels and lu-
study for Cairo-Alexandria railway line. Impact Assess Proj Apprais bricants (Goriva i maziva); 2004. p. 291e309. Vol. 43, No. 4, https://hrcak.srce.
2012;18(No. 4):323e33. https://doi.org/10.3152/147154600781767312. hr/7174. [Accessed 8 October 2018].
[9] Technical report: Cambridge Systematics, Inc., prepared for Southern Cali- [35] Vuci
c D. Traffic interconnection of Southern Croatia. In: Presentation (in
fornia association of governments, task 8: analysis of freight rail electrification Croatian) held at vital railway infrastructure panel; November 2014.
in the SCAG region. April 2012. http://www.freightworks.org/Document Dubrovnik, Croatia, http://www.korema.hr/attachments/article/84/5_
Library/CRGMSAIS%20-%20Analysis%20of%20Freight%20Rail%20Electrification Uskotracna%20zeljeznicka%20pruga%20Gabela-Uskopolje-Dubrovnik.pdf.
%20in%20the%20SCAG%20Region.pdf. [Accessed 8 October 2018]. [Accessed 8 October 2018].
[10] Electrified railway lines - European commission: https://ec.europa.eu/ [36] HZ Cargo. HRT 153, List of railway stations and lines (in Croatian). 2014. Valid
transport/facts-fundings/scoreboard/compare/energy-union-innovation/ from 1. april 2014, Zagreb, Croatia, http://www.hzcargo.hr/UserDocsImages/
share-electrified-railway_en#2014, accessed 24.09.2018. Dokumenti/tarife/153.pdf. [Accessed 8 October 2018].
[11] Federici M, Ulgiati S, Basosi R. A thermodynamic, environmental and material [37] UIC e International Union of Railways. Loading Guidelines: code of practice
flow analysis of the Italian highway and railway transport systems. Energy for the loading and securing of goods on railway wagons. first ed., vol. 1; 2017
2008;33. 760-755, https://doi.org/10.1016/j.energy.2008.01.010. 01.04 https://uic.org/IMG/pdf/loading_guidelines_-_volume_1.pdf. [Accessed

[12] Valter Z. Diesel-electric locomotives (in Croatian). Zagreb, Croatia: Skolska 8 October 2018].
knjiga; 1985, ISBN 978-953-0-30697-4. [38] Fritz SG. Evaluation of biodiesel fuel in an EMD GP38-2 locomotive. May 2004.
[13] Kalghatgi G. Is it really the end of internal combustion engines and petroleum NREL REPORT/PROJECT NUMBER: SR-510-33436, https://www.nrel.gov/docs/
in transport? Appl Energy 2018;225:965e74. https://doi.org/10.1016/j. fy04osti/33436.pdf. [Accessed 8 October 2018].
apenergy.2018.05.076. [39] Pourabdollah M. On optimization of plug-in hybrid electric vehicles, thesis for
[14] Bompard E, Carpignano A, Erriquez M, Grosso D, Pession M, Profumo F. Na- the degree of licentiate of engineering. Go € teborg, Sweden: Department of
tional energy security assessment in a geopolitical perspective. Energy Signals and Systems, Chalmers University of Technology; 2012. http://
2017;130:144e54. https://doi.org/10.1016/j.energy.2017.04.108. publications.lib.chalmers.se/records/fulltext/170631/170631.pdf. [Accessed 8
[15] Letnik T, Marksel M, Luppino G, Bardi A, Bo zi
cnik S. Review of policies and October 2018].
measures for sustainable and energy efficient urban transport. Energy [40] Pourabdollah M, Murgovski N, Grauers A, Egardt B. Optimal sizing of a parallel
2018;163:245e57. https://doi.org/10.1016/j.energy.2018.08.096. PHEV powertrain. IEEE Trans Veh Technol 2013;62(No. 6):2469e80. https://
[16] Puksec T, Krajaci
c G, Luli
c Z, Vad Mathiesen B, Dui c N. Forecasting long-term doi.org/10.1109/TVT.2013.2240326.
energy demand of croatian transport sector. Energy 2013;57:169e76. https:// [41] Pourabdollah M, Murgovski N, Grauers A, Egardt B. An iterative dynamic
doi.org/10.1016/j.energy.2013.04.071. programming/convex optimization procedure for optimal sizing and energy

[17] García-Alvarez A, Perez-Martínez PJ, Gonz alez-Franco I. Energy consumption management of PHEVs. In: 19th IFAC world congress; August, 2014. Cape
and carbon dioxide emissions in rail and road freight transport in Spain: a Town, South Africa, https://doi.org/10.3182/20140824-6-ZA-1003.02375.
case study of car carriers and bulk petrochemicals. J Intell Transport Syst [42] Sarvi A, Lyyra €nen J, Jokiniemi J, Zevenhoven R. Particulate emissions from
2013;17(No. 3):233e44. https://doi.org/10.1080/15472450.2012.719456. large-scale medium-speed diesel engines: 1. particle size distribution. Fuel
[18] Lebedevas S, Dailydka S, Jastremskas V, Rapalis P. Research of energy effi- Process Technol 2011;92(No. 10):1855e61. https://doi.org/10.1016/j.fuproc.
ciency and reduction of environmental pollution in freight rail transportation. 2011.04.031.
Transport 2017;32(No. 3):291e301. https://doi.org/10.3846/16484142.2016. [43] Prime Mover EMD 645 e data j Project rail: https://projectrail.wordpress.com/
1230888. 2008/11/12/prime-mover-emd-645-data/, accessed 24.09.2018.
[19] Dominkovi c DF, Ba cekovi c I, Pedersen AS, Krajaci
c G. The future of trans- [44] Atlas copco e QES generators, catalogue, no. 2958 1060 02 e v02. 2017.
portation in sustainable energy systems: opportunities and barriers in a clean https://www.atlascopco.com/content/dam/atlas-copco/construction-
energy transition. Renew Sustain Energy Rev 2017;82:1823e38. https://doi. technique/portable-energy/documents/2_generators/mobile-diesel-
org/10.1016/j.rser.2017.06.117. generator/QES%20mobile%20diesel%20generator%20english.pdf. [Accessed 8
[20] Hemmati R, Saboori H. Emergence of hybrid energy storage systems in October 2018].
renewable energy and transport applications e a review. Renew Sustain En- [45] Dimitrova Z, Mare chal F. Environomic design of vehicle integrated energy
ergy Rev 2016;65:11e23. https://doi.org/10.1016/j.rser.2016.06.029. system eapplication on a hybrid electric vehicle energy system. Chem Eng
[21] Fuhs AE. Hybrid vehicles and the future of personal transportation. Taylor & Trans 2014;39:475e80. https://doi.org/10.3303/CET1439080.
Francis Group, LLC; 2009, ISBN 978-14-200-7534-2. [46] Pichlík P, Zde nek J. Overview of slip control methods used in locomotives.
[22] Doucette RT, McCulloch MD. Modeling the prospects of plug-in hybrid electric Trans Electr Eng 2014;3(No. 2):38e43.
vehicles to reduce CO2 emissions. Appl Energy 2011;88(No. 2):2315e23. [47] Li
cka zeljezni
cka pruga j Hrvatska enciklopedija: http://www.enciklopedija.
https://doi.org/10.1016/j.apenergy.2011.01.045. hr/natuknica.aspx?id¼36402, accessed 24.09.2018.
[23] Sun Y, Cole C, Spiryagin M, Godber T, Hames S, Rasul M. Conceptual designs of [48] Cipek M, Pavkovi c D, Petri
c J. A control-oriented simulation model of a power-
hybrid locomotives for application as heavy haul trains on typical track lines. split hybrid electric vehicle. Appl Energy 2013;101:121e33. https://doi.org/
Proc Inst Mech Eng Part F J Rail Rapid Transit 2013;227(No.5):439e52. https:// 10.1016/j.apenergy.2012.07.006.
doi.org/10.1177/0954409713501655. [49] GPS visualizer: freehand drawing utility: draw on a map and save GPX data:
[24] Spiryagin M, Wolfs P, Szanto F, Sun YQ, Cole C, Nielsen D. Application of http://www.gpsvisualizer.com/draw/, accessed 24.09.2018.
flywheel energy storage for heavy haul locomotives. Appl Energy 2015;157: [50] OpenRailwayMap: https://www.openrailwaymap.org/, accessed 24.09.2018.
607e18. https://doi.org/10.1016/j.apenergy.2015.02.082. [51] Liu J, Peng H. Modeling and control of a power-split hybrid vehicle. IEEE Trans
[25] Technical report: hybrid locomotive e SUSTRAIL. 2014. FP7 No. 265740, Control Syst Technol 2008;16(No. 6):1242e51. https://doi.org/10.1109/TCST.
Deliverable 3.2.1, Hybrid Locomotive, http://www.sustrail.eu/IMG/pdf/d3.2- 2008.919447.
v1-hybrid_locomotive-final_version.pdf. [Accessed 8 October 2018]. [52] Bin Y, Li Y, Feng N. Nonlinear dynamic battery model with boundary and
[26] Donnelly FW, Cousineau RL, Horsley RNM. Hybrid technology for the rail in- scanning hysteresis. In: Proceedings of the ASME 2009 dynamic systems and
dustry, Rail conference. In: proceedings of the 2004 ASME/IEEE joint, Balti- control conference. California, USA: Hollywood; 2009. p. 245e51. No.
more, MD, USA, April 6-8, 2004; 2004. https://doi.org/10.1115/RTD2004- DSCC2009-2745, https://doi.org/10.1115/DSCC2009-2745.
66041. [53] Besselink IJM, Van Oorschot PF, Meinders E, Nijmeijer H. Design of an efficient,
[27] Meinert M, Prenleloup P, Schmid S, Palacin R. Energy storage technologies and low weight battery electric vehicle based on a VW Lupo 3L. In: The 25th world
hybrid architectures for specific diesel driven rail duty cycles: design and battery, hybrid and fuel cell electric vehicle symposium & exhibition. China:
system integration aspects. Appl Energy 2015;157:619e29. https://doi.org/10. EVS-25 Shenzhen; 2010. p. 32e41.
1016/j.apenergy.2015.05.015. [54] Mettlach H. Development of the Li-ion battery for the opel ampera extended-
[28] Spiryagin M, Wu Q, Wolfs P, Sun Y, Cole C. Comparison of locomotive energy range electric vehicle, PPT on GM alternative propulsion center - Europe adam
storage systems for heavy-haul operation. Int J Real Ther 2018;6(No. 1):1e15. opel GmbH. In: 3rd international congress on advanced battery technology;
https://doi.org/10.1080/23248378.2017.1325719. 2010. Wiesbaden, June 22, http://www.selidori.com/tech/00000-04999/1072-
[29] Schmid S, Ebrahimi K, Pezouvanis A, Commerell W. Model-based comparison WHAKu.pdf. [Accessed 8 October 2018].
of hybrid propulsion systems for railway diesel multiple units. Int J Real Ther [55] IRENA. Electricity storage and renewables: costs and markets to 2030. Abu
2018;6(No. 1):16e37. https://doi.org/10.1080/23248378.2017.1390790. Dhabi: International Renewable Energy Agency; 2017. 2017, http://www.
[30] Rupp A, Baier H, Mertiny P, Secanell M. Analysis of a flywheel energy storage irena.org/-/media/Files/IRENA/Agency/Publication/2017/Oct/IRENA_
system for light rail transit. Energy 2016;107:625e38. https://doi.org/10. Electricity_Storage_Costs_2017.pdf. [Accessed 8 October 2018].
[56] Steen M, Lebedeva N, Di Persio F, Boon-Brett L. EU competitiveness in €rkman M, Holmstrom K. Global optimization using the DIRECT algorithm
[60] Bjo
advanced Li-ion batteries for E-mobility and stationary storage applications in matlab. AMO - Adv Model Optim 1999;1(No.2):17e37.
eopportunities and actions. In: JCR science for policy report EUR 28837 EN. [61] Ecker M, Gerschler JB, Vogel J, K€abitz S, Hust F, Dechent P, Sauer DU. Devel-
European Commission; 2017. https://doi.org/10.2760/75757. opment of a lifetime prediction model for lithium-ion batteries based on

[57] Skugor B, Deur J, Cipek M, Pavkovi c D. Design of a power-split hybrid electric extended accelerated aging test data. J Power Sources 2012;215:248e57.
vehicle control system utilizing a rule-based controller and an equivalent https://doi.org/10.1016/j.jpowsour.2012.05.012.
consumption minimization strategy. Proc Inst Mech Eng Part D J Automob Eng [62] Croatia e the trend in gasoline prices: https://autotraveler.ru/en/croatia/
2014;228(No. 6):631e48. https://doi.org/10.1177/0954407013517220. trend-price-fuel-croatia.html, accessed 24.09.2018.
[58] Cipek M, Bitanga P, Mlinari c TJ, Pavkovic D, Kljai
c Z. Comparative analysis of [63] Pavkovi c D, Sedi
c A, Guzovi
c Z. Oil drilling rig diesel power-plant fuel effi-
conventional diesel-electric and hybrid-electric heavy haul locomotive based ciency improvement potentials through rule-based generator scheduling and
on mountain rail route driving scenario. In: Digital proceedings of 3rd south utilization of battery energy storage system. Energy Convers Manag
east Europe (SEE) sustainable development of energy water and environment 2016;121:194e211. https://doi.org/10.1016/j.enconman.2016.05.022.
systems (SDEWES) conference; 2018. Novi Sad, Vojvodina, Serbia. [64] CN to buy 200 locomotives from GE as freight volumes surge - The Globe and
[59] Finkel DE. Global optimization with the DIRECT algorithm, dissertation. North Mail: https://www.theglobeandmail.com/report-on-business/cn-to-buy-200-
Carolina State University; February, 2005. http://www.lib.ncsu.edu/resolver/ locomotives-from-ge-as-freight-volumes-surge/article37418466/, accessed
1840.16/4738. [Accessed 8 October 2018]. 24.09.2018.

Assessment of Battery-Hybrid Diesel-Electric Locomotive Fuel Savings

Uploaded by

Copyright:

Available Formats

Assessment of Battery-Hybrid Diesel-Electric Locomotive Fuel Savings

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Assessment of Battery-Hybrid Diesel-Electric Locomotive Fuel Savings

Uploaded by

Copyright:

Available Formats

What is the paper assessing?

What is the paper assessing?

How does the paper model the conventional and hybrid locomotives?

How does the paper model the conventional and hybrid locomotives?

Energy 173 (2019) 1154e1171

Contents lists available at ScienceDirect

Assessment of battery-hybrid diesel-electric locomotive fuel savings

Fig. 1. Simpliﬁed schematic representation of diesel-electric locomotive power-train.

IDLE 7.83 27.02 0 3.1247 0.0364 0.0531 0.1836 9.7154

percentage of power generated at each “Notch” position listed in

3.2. Normalized diesel engine emissions

where PmgMAX is the main engine maximum power in kW (corre-

IDLE 8.05 0 3.2131 0.0374 0.0546 0.1888 9.9810

Ftmax ¼ ma ml gcosðaÞzma ml g; (10)

b 4.2. Longitudinal train dynamics

Fig. 5. Maximum force (a) and power (b) traction maps.

where Fb is the force applied to the brakes (equal 0 during normal

Fig. 6. Block diagram of conventional diesel-electric locomotive quasi static model.

braking periods, followed by the integration of positive traction

N E (kWh) Qmax (Ah) Pmax (kW) m (kg)

IDLE 6.43 0 2.5704 0.0299 0.0436 0.1510 7.9910

Fig. 11. Block diagram of quasi-static model of proposed battery-hybrid locomotive.

Reference type Parameters

KSoC DSoC SoCR SoCbl SoCbh

(SoCRC) Full load only 3866 0.13% 63.13% e e

Fig. 12. Conventional vs. hybrid-based freight velocity difference.

Wagons Conv. Vf (L) Hybrid (SoCRC) Hybrid (SoCRV)

Vf (L) SoCMIN (%) Vf (L) SoCMIN (%)

New Aged New Aged New Aged New Aged

0 411 (433)434 (435)435 (54.60)55.87 (53.78)55.81 (385)406 (394)413 (41.20)43.46 (40.31)42.69

Locomotive Emissions (kg) Table 10

Conventional 8.21 94.97 130.43 7257 Locomotive Cost (EUR)

useful life corresponding to 2000 charging/discharging cycles

EMD Electro-motive division Ft traction force [N]

You might also like