Professional Documents
Culture Documents
net/publication/283330482
CITATIONS READS
53 1,619
7 authors, including:
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Zhongbei Tian on 06 April 2016.
Research Article
ISSN 2042-9738
Energy evaluation of the power network of a Received on 13th May 2015
Revised on 22nd September 2015
DC railway system with regenerating trains Accepted on 30th September 2015
doi: 10.1049/iet-est.2015.0025
www.ietdl.org
Zhongbei Tian 1 ✉, Stuart Hillmansen 1, Clive Roberts 1, Paul Weston 1, Ning Zhao 1, Lei Chen 1,
Mingwu Chen 2
1
School of Electronic, Electrical and Systems Engineering, University of Birmingham, Birmingham B15 2TT, UK
2
School of Electrical Engineering, Southwest Jiaotong University, Chengdu, People’s Republic of China
✉ E-mail: zxt279@bham.ac.uk
Abstract: Regenerating trains are now in common use on many DC fed railway systems, and train operating companies
are able to get a discount on their energy costs if regeneration is active. The electrical energy consumption in a DC system
is significant, and a comprehensive understanding of how regeneration affects the overall system energy consumption
has not been developed. This study presents a simulation method in which a multi-train analysis is used to determine
the system energy consumption with and without regeneration in operation, as well as the impact on the system
energy consumption of different headways. The results are used to determine a full ‘energy audit’ of the system based
on the data of the Beijing Yizhuang subway line. This includes the energy supplied by the substations, the energy
wasted in the power transmission network, the energy used by the train in traction and regenerated by braking trains.
The initial results show that regenerating trains have a significantly lower substation demand, but slightly more energy
is lost within the network. The results also indicate that, the available regenerative energy and total substation demand
vary with different timetables, and there is a 27% difference between the best and worst headways.
Nomenclature 1 Introduction
M mass of the train (kg) With the continuing focus on rising energy prices and environmental
l rotary allowance concerns, reducing energy consumption and ensuring the
s distance of the train (m) environmental sustainability of railway systems is becoming a key
t time (s) topic which is being paid increasing attention. As a solution which
F tractive effort applied at the wheels (N) aims to minimise traffic congestion and reduce air pollution in
g acceleration due to gravity (m/s2) urban environments, urban rail systems have been developing
R resistance of motion (N) rapidly in recent years. Although the railway system is arguably
K curvature resistance coefficient (Nm) the most efficient form of land based transport, there is still room
r radius of curvature of the track (m) for saving more energy.
A Davis equation constant (N) Several comprehensive approaches for energy saving include
B Davis equation linear term constant (N/(m/s)) energy-efficient driving, regenerative braking, reduction of traction
C Davis equation quadratic term constant (N/(m/s)2) losses, and smart power management, which have been
α gradient angle (rad) investigated and assessed [1]. Regenerative braking and
Pmech mechanical power of the train (W) energy-efficient driving are assumed to have the greatest potential
Pelec electrical power of the train (W) and suitability for reducing energy consumption compared with
ηmotor electrical motor efficiency other energy-saving methods. Studies of driving strategy
Pregen regenerative power of the train (W) optimisation commonly focus on saving traction-energy by
Rsub equivalent resistance of the substation (Ω) applying coasting control. A genetic algorithm (GA) was applied
Vopen circuit no-load voltage of 12-pulse rectifier unit (V) to find proper coasting points to achieve the best energy saving
Vrated rated voltage of 12-pulse rectifier unit (V) results [2]. Both classical and heuristic approaches are presented in
Irated rated current of 12-pulse rectifier unit (A) [3], and multi-coasting points control shows a better energy saving
PT instantaneous power of the train (W) performance in a long inter-station section. Rather than searching
UT instantaneous voltage of the train (V) for the coasting point, a combined function of energy consumption
IT instantaneous current of the train (A) and journey time was optimised by a fuzzy logic technique [4].
T journey time of a single trip (s) Dynamical programming was proposed for optimisation, and
N total number of running cycles performed better than both GA and ant colony optimisation [5]. A
t0 headway (s) multi-train simulator was developed to reduce energy cost and
Ttotal whole-day operation time (s) penalty cost caused by delay [6, 7]. With the development of
ri resistance of the contact line segment i (Ω) communications-based train control systems, designing an optimal
ρ resistivity of contact lines (Ω/m) automatic train operation (ATO) speed profile became a preferred
li length of the contact line segment i (m) research [8, 9]. In an attempt to achieve a fast-response online
optimum control system, a mathematical algorithm proved by
2.1 System overview In general, the curvature resistance is small and it is neglected in this
paper [30]. The train speed curve is generated by the simulator using
Considerable work on simulation and optimisation has been done different driving strategies, which include accelerating, cruising,
over a long time. However, conventional research has often coasting, braking and so on. The accelerating mode is generally
separated the study of train operation and power systems. Most of active at the beginning of the journey with maximal tractive
efforts. Cruising mode is invoked when the train reaches the speed Pregen = Pmech × hmotor (5)
limit and partial power is utilised at this time. When the coasting
mode is utilised, no power is needed to draw the train. Braking
mode is applied when the train is approaching a stop or a lower
speed limit. The maximal feasible braking efforts lower than the
rail adhesion are usually utilised to regenerate electrical energy. 2.3 Power network modelling
These energy-efficiency controls have been proved by the
In a DC railway power network, traction rectifier substations are the
Pontryagin maximum principle under the assumption of train
main electricity source for vehicles. Fig. 3 presents a typical DC
traction simplification [10–12, 31], in which the maximum
traction power network with multiple trains modelling system.
acceleration and braking must be the most energy efficient.
There are four DC substations with rectifiers which can prevent
The driving strategies should fulfil the constraints determined by
regenerative current from flowing back. The resistors present the
tractive ability and speed limits. In Fig. 2, the Beijing Yizhuang
overhead line and return running rail resistance, which are split by
line vehicle tractive effort and train resistance of motion with
the running vehicle or substations. The value of resistance depends
standard passenger load is illustrated. According to the generated
on the length and the resistivity of the conductor, which are 15
speed trajectory, the mechanical power of the train can be
and 10 µΩ/m for overhead line and return rail, respectively. The
computed in (3), which is negative when braking
running trains are located on both up and down tracks, which can
be presented by a power source, and power flow analysis can be
ds utilised to solve the whole system, which will be explained in the
Pmech = F × (3)
dt following section.
the working region of the rectifier units. Thus the simplified voltage all the junctions. Analysis method can be applied to solve the
regulation characteristic becomes linear as shown in Fig. 4. This network in the following equation
diagraph illustrates the output voltage and current at normal load
states. Based on the data from Beijing Yizhuang line, the no-load
[I] = [Y ] × [V ] (7)
voltage of each 12-pulse rectifier unit is 850 V, while the rated
voltage and current are 750 V and 2500 A, respectively. As there
are two 12-pulse rectifiers in parallel, the equivalent resistance for Due to the trains acting as an ideal power source, the train power
the rectifier substation can be calculated in the following equation formula (8) can be utilised as constraint of the power analysis,
where PT is the train power demand
DV Vopen circuit − Vrated 850 − 750
Rsub = = = = 0.02 V (6)
DI Irated − 0 2500 PT = UT IT (8)
2.3.2 Train loads: Previous research often utilised constant current According to the nodal analysis and power load requirements, the
source models or constant efficiency of regenerative braking energy non-linear power flow problem can be solved. The approach to
usage to present trains in a traction power network [13, 33, 34], develop the admittance matrix and to solve the power flow
while in this study, trains are considered as dynamical power sources problem of multi-train power network is presented below.
or power loads for a better simulation performance. Most modern
metros use asynchronous motors, such as the Beijing Yizhuang
metro. There are several working states for trains, as follows: 3.1 Admittance matrix development
(i) Motoring: When the train is motoring, the rotors of the vehicle Defining the substation and train positions is the initial step of
turn slower than the synchronous speed, transforming the input development of the admittance matrix. The locations of substations
electrical power into mechanical power. This input power comes are permanently fixed, whereas according to the single-train
from the power network supplied by substations and other trajectory from the train motion simulation, the location of the
regenerating trains. single-train can be confirmed at any time, which can be explained
(ii) Normal regenerative braking: When the train is braking, the by dtrain(t) where 0 ≤ t ≤ T. Adding an input of the timetable and
rotors are turning faster than the synchronous speed controlled by the the total number of running cycles, the total whole-day operation
driver. Thus, the motor will transform mechanical energy available at time and location of trains can be derived in (9) and (10), where t0
the drive shaft into electrical energy, which can be transferred back to is the headway
the network system to power other motoring trains. For normal
regenerative braking, all of the regenerating energy can be transferred
into the transmission network to power other trains. Ttotal = T + (N − 1) × t0 (9)
(iii) Over-voltage regenerative braking: As regenerative braking
can increase the voltage of a train, a high regen voltage will occur i
Dtrains (t) = dtrain (t − n × t0 ), 0 ≤ t ≤ Ttotal , 0 ≤ (t − n × t0 ) ≤ T
when there are not enough motoring trains absorbing the
regenerative energy in the power network. In the case of a high (10)
voltage hazard, some braking energy cannot be transferred to
contact lines, but is wasted in the on-board braking rheostat as After confirming the location of trains, each resistance of contact
heat when the regen voltage exceeds a safe value. Therefore, it is lines segment split by running trains and substations can be
essential to analyse the amount of usable regenerative energy in computed in the following equation
performance evaluation simulation.
ri = r × li (11)
The auxiliary load of the train is assumed as a constant, including
the air conditioning and lighting power. The vehicle power
requirement is computed by summing the auxiliary load power In terms of a complex circuit model, current sources are easier than
and input power of the motor. voltage sources in iteratively computing the power flow. Using
Norton’s theorem the rectifier substation circuit can be replaced by
a current source with a parallel resistance. As the trains in the
3 Power flow analysis power network are moving all the time, the network circuit should
be reconfigured at every simulation time step. To establish the
To obtain the solution of the power network requires solving a set of complex admittance matrix of the DC power network more
equations which describe the internal electrical relationship between effectively, sparse matrix ordering methods can be utilised [16].
Fig. 6 Up and down direction speed trajectory and electrical power requirement
Train no. Location, km Direction Over voltage Power, kW Final power, kW Voltage, V
location, km 0 2.62 4.71 6.97 9.31 10.66 13.48 14.47 16.46 18.82 20.10 22.73
voltage, V 950 950 950 928 906 895 874 867 855 843 832 786
power, kW 0 0 0 0 0 0 0 0 0 299 780 2739
20,033 s is low, 88.57% of which has been dissipated in the on-board braking power of 3637 kW. The power and voltage output from
braking rheostat. This is because that the first three trains are braking substations at this simulation time is described in Table 5.
and regenerating power at this time, but only the last train 13 is The following section shows some energy evaluation results for
motoring which requires a lot of power. Due to the long distance the whole journey.
between them, only a little of regenerated power can be transferred
which also causes the first three trains reach the over-voltage
4.2 Energy audit
limitation. In Table 3, the substation output results are described.
Only the last three substations output power to supply the train 13. According to the single-train power requirement, whole-day energy
Table 4 shows the results at 20,066 s, where the total traction consumption has been calculated for all trains operating in the
power, braking power and usable regenerating power are 9892, system. This includes the energy supplied by the substations, that
5548 and 4839 kW, respectively. Therefore, 87.22% of braking used by the train in traction, that regenerated by the trains with
power has been reused, and substation output power is 6728 kW. regen and that wasted in the transmission network. For simplicity,
Compared with the previous case, a regenerative efficiency with the headway remains constant all day. As the substation energy
87.22% is better. From the results in Table 4, the voltage of train 7 consumption is compared with different headways, the total
is the lowest with 745 V, because the tractive power train 7 is very number of running cycles is set at 121 to make the total traction
large which makes the line voltage drop considerably. The highest energy consumption and braking energy the same with different
voltage is 947 V, at train 4. This is because train 4 can regenerate a headways. Fig. 7 shows the energy results of the one day’s
large amount of energy and transfer it into other motoring trains operation with regeneration turned off versus different headways,
nearby, thus the voltage of train 4 is high. Due to the over-voltage which range from 300 to 400 s. The total traction energy and
limitation, the voltage of train 4 cannot exceed 950 V; as a result, braking energy are 64.31 and 43.56 MWh, respectively, and they
the output regenerative power is 2929 kW which is less than the are fixed values with different headways. The substation energy
consumption ranges from 68.90 to 70.62 MWh, and there is only
2% difference with various headways. Of the substation output
energy, 7–9% has been dissipated by the substation inner resistor
Table 4 Train power detection at simulation time = 20,066 s and contact lines, with an average value of 5.01 MWh.
Fig. 8 illustrates the energy results with regenerating trains. As
Headway = 300 s, Simulation time = 20,066 s there are no changes in the number of operating trains and train
Train no. Location, Up/ Over Power, Final Voltage,
trajectory, the traction energy and braking energy are also the
km down voltage kW power, V same for the results without regeneration. The energy consumption
kW at the substations is reduced by 22–42% by having regeneration.
The figure shows that the effective use of regenerated energy is
1 1.33 up no 51 51 915 highly variable, and is not simply related to headway. This is
2 2.62 down no 54 54 916
3 4.70 up no 652 652 919
because the stations positions are at unequal distances and the
4 6.80 down yes −3637 −2929 947 braking and accelerating trains randomly overlap. The energy
5 8.40 up no 45 45 895 consumption as measured at the substations ranges from a
6 10.35 down no 45 45 841 minimum of 39.77 MWh (307 s headway) to a maximum of
7 11.84 up no 3727 3727 745
8 13.48 down no 51 51 829
55.02 MWh (350 s headway). In principle, more than 27% of the
9 14.48 up no −1820 −1820 841 energy saving from the substation can be achieved by ‘optimising’
10 16.50 down no 3727 3727 795 the headway. However, in practice, even small deviations in
11 18.81 up no 1938 1938 814 timings of a single train could result in a significant reduction
12 20.10 down no 52 52 832
13 22.13 up no 45 45 840
(or even increase) in the effective use of regenerated power.
The network losses are a little higher, with an average value of
location, km 0 2.62 4.71 6.97 9.31 10.66 13.48 14.47 16.46 18.82 20.10 22.73
voltage, V 915 916 919 936 870 833 829 841 800 814 832 843
power, kW 0 0 0 0 0 733 902 389 2116 1525 756 308
5.60 MWh, when regeneration is turned on, but this is not significant consumption is due to the complex interaction between the
compared to the net energy reduction. In the current model, the headway, the inter-station journey time, and line receptivity. Small
usable regenerated energy accounts for 35–68% of the overall changes in the otherwise constant headway vary the effective use
braking energy. The variability of the substation energy of available regenerated energy significantly (Table 6).