978-1-4673-4362-6/12/$31.00 ©2012 IEEE

Ultracapacitor Assisted Regenerative Braking in

Metropolitan Railway Systems Albert Killer, Andreas Armstorfer, Andrés Emiro Díez, Helmuth Biechl

Universidad Pontificia Bolivariana, Kempten University of Applied Sciences albert.killer@gmx.de, andreas.armstorfer@fh-kempten.de, andres.diez@upb.edu.co, biechl@fh-kempten.de

Abstract—This paper investigates the use of ultracapacitors as an

energy storage system for metropolitan railway systems. The

basic concept is to take advantage of the energy generated during

the braking process of trains, which is otherwise wasted and

transformed into heat. A power system simulation was created as

a tool to analyze the behavior of a stationary ultracapacitor bank

connected to the Metro system. Ultracapacitor modules were

dimensioned adequately, power electronic elements were

adjusted and further a control system was established. Amongst

others the simulation demonstrates the impact of varying the

distance between ultracapacitors and traction power substation.

Furthermore the dependency of the ultracapacitor’s technical

specifications on charging and discharging behavior is shown.

The simulation allows a large number of parameters to be

modified through the easy-to-handle user interface. A scientific

method is provided to analyze ultracapacitor assisted

regenerative braking of metropolitan railway systems. As an

example the Metro of Medellín in Colombia is applied and

rounded out by an estimation of economic viability.

I. INTRODUCTION

The Medellín Metro is a public railway system located in the metropolitan area of Medellín in Colombia since 1995. In 2010 the Metro transported 320 million passengers, reaching rates of up to 25,000 pax/h with an energy consumption of nearly 67 GWh/a where 80 % are used for traction [1].

The operating company cooperates with local industry and universities in order to achieve appropriate solutions and being more independent and flexible in terms of maintenance. In order to investigate possible improvements of the metro’s drive system, a comprehensive simulation of the existing motor control and power electronics was created in the course of a previous project. The results revealed the next steps to be worked on and brought up several projects dealing with subjects like new power electronics using IGBTs and the replacement of break series resistors. Another promising aspect is the integration of modern energy storage systems (ESS) like ultracapacitors in order to take advantage of the energy saving potential of regenerative braking. Parallel to a series of scientific projects at Kempten University concerning the general application of ultracapacitors and their use in public transportation, this work is focused on the implementation of ultracapacitors into the system of the Metro of Medellín.

II. USING THE BRAKING ENERGY

A. Metro system & main objectives

The basic idea of this project is to take advantage of the kinetic energy which is available when the train is braking. The nominal voltage of the trains is 1.5 kV DC, therefore the traction power substations provide a DC voltage of 1.65 kV in order to compensate voltage drop across the overhead line. Because the system isn’t capable of feeding back the energy into the supply grid, up to now the metro system of Medellín only consumes the regenerated energy, if other trains nearby are accelerating in the very same moment when the first train brakes. Otherwise the energy has to be transformed into heat by using the train’s braking resistors. Preventing this loss of energy would be a gain when integrating ultracapacitor energy buffers into the system. The second objective is the stabilization of overhead line’s voltage level. Fluctuations like voltage drops are caused by high starting currents of the motors, while voltage surges are the consequence of the feedback of braking energy into the overhead line. Particularly at sections with low train density these fluctuations are significant, reducing lifespan of sensible equipment. It is going to be analyzed how ultracapacitor substations could be a feasible addition to the existing and comparably high-priced traction power substation in terms of voltage stabilization.

B. Integration of ultracapacitors via DC-DC converter

Due to several considerations a stationary solution is preferred to an onboard ultracapacitor ESS. Amongst various advantages there is the easily feasible integration into the existing system independent of the train design as well as the flexibility in size, weight, number and position. From the economical point of view the possibility of individual sizing for each track carries particular importance. Each track between two stations of the Medellín Metro varies in distance, as well as maximum speed and acceleration. Therefore the resulting kinetic energy of the trains differs from track to track. This directly affects the amount of recoverable energy and demands individual sizing of the ultracapacitors depending on their location along the line in order to enable full capacity utilization. To avoid energy transport losses it is advisable to connect the ultracapacitors close to where the trains usually stop: at the passenger stations. Furthermore the voltage level on the overhead line becomes more unstable the larger the distance to the next traction power substation. Due to this fact the weakest point could be located between two traction power substations and therefore would be the ideal position to install the ultracapacitors in order to stabilize the voltage. With the simulation presented below it is possible to analyze how the

Scenario

1. Low train density < 5 trains > 10 min 19.0 % 3129 Wh

2. High train density 17 trains 4.5 min 2.5 % 412 Wh

N° & frequency Recoverable energy

voltage stabilization potential of the ultracapacitors depends on the distance to the next traction substation.

In order to connect the ultracapacitors to the metro system a bidirectional DC-DC converter is required. It allows to control charging and discharging of the ultracapacitors. The bidirectional DC-DC converter operates as buck converter in one direction, and as boost converter in the opposite direction [2]. The converter’s voltage is controlled by the ratio between on time ton and period duration T, described as duty cycle D according to (1).

D = ton / T = ton ∙ fconv (1)

While charging the ultracapacitor, the buck converter is activated by only using the switch SM while SG is off (Fig. 1 and 2), leading to the following output voltage, assuming steady state conditions (2).

Vcap = Dbuck ∙ Vline (2)

While the ultracapacitor is discharging, SM interrupts and SG starts toggling (Fig. 3 and 4). That is to say the boost converter is active and energy transfer direction has changed. The output voltage can be described by (3).

Vline = 1/(1-Dbuck) ∙ Vcap (3)

The following schemes depict the integration of an ultracapacitor stack and converter into the existing system and show different states of operation:

Braking & charging: Train is braking; ultracapacitor is charging with available energy (Fig. 5).

Reusing energy: The stored energy is used for the next acceleration process (Fig. 6).

C. Recoverable energy & its dependence on train density

For the dimensioning of ultracapacitors it is necessary to consider which amount of energy should be stored. It is assumed that in the majority of cases only one train approaches a passenger station at one point in time. Therefore it is feasible to do an estimation of the maximum usable braking energy which is available from one braking train. For the case of low train density this data was acquired by using the traction system simulation of the previous project and by respective measurements on an isolated test track. It was observed that up to 19 % of the consumed energy could be stored in ultracapacitors. In the case of high train density measurements were performed in daily operation in order to get a statistical idea of the amount of the dissipated energy. The evaluation resulted in an average recoverable energy of approx. 2.5 %, which could be stored by ultracapacitors. The difference between the two cases is related to the fact that the measurements in daily operation reflect energy consumption and generation of one train operating in a very dynamic system of up to 17 trains running at the same time. It was concluded that the more trains operating in the same electrical section the more energy can already be reused without ultracapacitors. On the other hand a lower train density entails a bigger amount of dissipated energy which could be saved by ultracapacitors. That is why it was distinguished between two major scenarios, including two extreme values of energy being recovered due to integration of ultracapacitors (Table I).

III. SIMULATION OF AN ULTRACAPACITOR SUBSTATION

A power system simulation was developed to model the integration of a stationary ultracapacitor substation (SEU) into the system of Medellín Metro, consisting of a converter circuit and an ultracapacitor bank.

Figure 1. Buck converter SM closed.

Figure 2. Buck converter SM open.

Figure 3. Boost converter SG closed.

Figure 4. Boost converter SG open.

Figure 6. Acceleration with two energy sources.

TABLE I. DETERMINED SCENARIOS FOR PRESENT APPROACH.

Figure 5. Using braking energy for charging the ultracapacitors.

A. Basic structure & import of train signal

It is divided into various units represented by modules and submodules interacting in simulation software tool PSCAD. Fig. 7 shows the basic structure of the simulation.

The main module as the initial point contains a traction power substation as the main power supply and a current source representing the train, which acts alternatively as a consumer and supplier of energy. A simplified model of traction power substation was integrated by using an ideal DC voltage source. The ultracapacitor substation can be connected in parallel to the existing model of the system. The overhead line’s length between traction power substation and ultracapacitor substation can be determined by an adjustable resistance. The simulation module “Control Panel” contains the user interface and can be found in the main module.

The simulation models the train’s behavior by using signals of the before mentioned traction system simulation and thus currently only supports a low train density line like line B of Medellín Metro. The traction system simulation provides current signals of trains with different weight, speed and an individual timed sequence of driving and braking. For this investigation the sequence of an average loaded train (150 t) was recorded, representing a full process of braking and accelerating taking 70 s.

B. Calculation of converter circuit parameters

The module “SEU” models the ultracapacitor substation, including DC-DC converter circuit as well as submodules for the ultracapacitor bank and the converter’s control. In terms of sizing the converter’s electrical elements the simulation offers the posibility of manually parameter modification at the module “Control Panel” or calculates them automatically using following equations. Their aim is to avoid the operation in discontinuous mode. Table II contains the conditions for calculation of the circuit’s elements.

The equations results have to be interpreted as suggestions for minimum values. In order to calculate the minimum inductance value for buck converter mode a duty cycle of Dbuck,crit = 0.5 has to be assumed, as the inductor current reaches its maximum value [2]. For the boost converter a similar approach has to be applied using a duty cycle of Dboost,crit = 1/3 [3]. The circuit’s concept allows using the same storage coil for both converters. For this reason the higher value of Lmin,buck (6) and Lmin,boost (7) has to be detected, in order to determine the minimum inductance value, defining the boundaries to converter’s discontinuous mode.

Lmin,buck = Dbuck,crit ∙1/fconv∙ [Vline,max - (Dbuck,crit ∙ Vline,max)]

/ (2∙Itotal,max) (6)

Lmin,boost = [(Vline,r∙1/fconv)/(2∙Itotal,max)]∙Dboost,crit (1-Dboost,crit) 2 (7)

Due to the switching operation both converters create a fluctuating voltage, which has to be smoothened by a capacitor in order to get a consistent voltage signal at the output. How much fluctuation is tolerated can be determined by the allowed voltage ripples Vrip,buck and Vrip,boost. These values directly affect the sizing of the used output capacitors, leading to following equations for minimum output capacitance for buck (8) and boost (9) converter.

Cmin,buck = [Dbuck,crit ∙ Vline,max ∙ (1-Dbuck,crit) ∙ (1/fconv)2]

/ (4 ∙ Lmin,buck ∙ Vrip,buck ) (8)

Cmin,boost = (Itotal,max ∙ Dboost,crit∙ 1/fconv) / Vrip,boost (9)

Using the example values of Table II, following results are obtained for the converter parameters listed in Table III.

C. Model and analysis of ultracapacitor bank

For the simulation of ultracapacitors a commonly used simplified model was chosen, mainly consisting of capacitance and equivalent series resistance (ESR), which is supposed to influence the immediate behavior. The simulation offers an option to integrate leakage resistance for self-discharge, but as this affects the storage of energy rather on long term, it wasn’t taken into account, because integrated in the lines of Medellín Metro, the ultracapacitors will be charged and discharged approx. every 5 to 10 min. In consideration of the amount of energy which has to be stored by the ultracapacitor station and the energy density of current commercially used ultracapacitors, it is necessary to combine various cells or modules to an ultracapacitor bank. The simulation calculates the bank’s parameters as a unit from below listed initial values of single cells or modules (Table IV). First estimates and calculations are made using specification of the 125 V Heavy Transportation Module from Maxwell Technologies [4].

TABLE II. INPUT VALUES FOR CALCULATION OF CONVERTER PARAMETERS.

TABLE III. RESULTS FOR CONVERTER PARAMETERS.

signal name calculated value comment

Lmin_buck 0.31 mH cp. (6)

Lmin_boost 0.17 mH cp. (7)

Cmin_buck 7.20 mF cp. (8)

Cmin_boost 4.80 mF cp. (9)

Figure 7. Division of simulation into modules.

signal name example value comment

fconv 1.0 kHz converter frequency

SoC_min 0.25 min. state of charge

Vline_max 1.800 kV max. input voltage while charging

Vline_r 1.650 kV rated voltage provided from traction power st.

Dbuck_crit 0.5 value for sizing Lmin_buck

Dboost_crit 0.33 value for sizing Lmin_boost

Itotal_max 720 A max. current allowed at ultracapacitor bank

Vrip_buck 50 V fluctuation tolerated in buck's output voltage

Vrip_boost 50 V fluctuation tolerated in boost's output voltage

Nevertheless the PSCAD module provides the option to modify the key parameters and therefore run the simulation by using another type of ultracapacitors.

The suggested number of modules is derived from energy estimations made in Table I and refer to scenario 1, being 19% of the total energy consumed by an average loaded train. The manufacturer typically recommends discharging the modules only to half rated voltage as this will remove 75% of the stored energy. This is because discharging the modules lower than 25% of the state of charge (SoC) is usually not feasible as most converters will over heat due the high current draw at lower voltages. According to that and the general equation for capacitor’s stored energy (10) expresses the amount of usable energy per 125 V HT module and (11) the energy density:

Eusable,module = SoC∙0.5∙C∙V 2

= 0.75∙0.5∙63 F∙(125 V)2 = 102.54 Wh (10)

ρel,usable = (102.54 Wh)/(60.5 kg) = 1.69 Wh⁄kg (11)

From this follows the amount of modules needed for one braking process (12):

Nmodules = Erecoverable / Eusable,module = 31 (12)

Considering required voltage and current levels of the ultracapacitor bank it was decided to determine a first estimated amount of total 30 modules, of which Nparall = 3 rows are connected in parallel with each Nseries = 10 in series, leading to the specifications in Table V, considering (13), (14) and (15).

Vtotal,max = Nseries ∙ Vrated (13)

Itotal,max = Nparall ∙ Imax (14)

ESRtotal = Nparall / (Nseries / ESRmodule) (15)

By identifying the dissipated power (16) which occurs due to the loss at the ESR, the ultracapacitor bank’s theoretical energy conversion efficiency can be calculated (17).

Ploss = ESRtotal ∙ Itotal,max2 (16)

ηel = (Pusable-Ploss) / Pusable (17)

The specified ultracapacitor configuration finally leads to the energy and power values shown in Table VI.

As mentioned before, the recoverable amount of energy provided by a braking train also depends on the position of the passenger station, as distance between passenger stations usually differs significantly.

D. Converter control

During operation in the Metro’s network the ultracapacitor substation has to adjust dynamically to power flow and fluctuations in the overhead line in order to charge and discharge the ultracapacitors appropriately. To do so a control system is introduced which sets the converter’s duty cycles by using a multi-variable decentralized control. The actual value of the overhead line’s voltage Vline is continuously compared to the reference value of traction power substation’s voltage level. Thereby the system decides whether to activate buck or boost converter. At the same time the control system limits the ultracapacitor bank’s current Icap to evade intolerable heating.

E. Voltage stabilization & energy analysis

The simulation module “Control Panel” represents the user interface and allows directly to manipulate key parameters and to examine the simulated system’s reaction for different input scenarios. The idea is to offer instant modifications and analyses to an advanced extend, even to users without deep insight into the simulation’s structure. The control elements provide crucial options like connecting the ultracapacitor substation to the Metro’s system, setting the traction power substation’s supply voltage and varying the distance between both substation types, which directly affects line resistance Rline. The energy storage can be designed for particular demands by determination of the number of modules or cells as well as their interconnection and the setting of min. and max. state of charge.

In order to observe the impact of ultracapacitor integration in terms of voltage stabilization it is appropriate to compare overhead line’s voltage level behavior under changing circumstances. The control panel offers several options to vary the input parameters. However, at this point we will only have a look at voltage fluctuation at different distances to the traction power substation, without and with ultracapacitors connected to the system. The grey signal in Fig. 8 depicts the dependence of these fluctuations on the line resistance which is increasing with the distance to traction power substation by approx. 50 mΩ/km. The dark-gray signal (Fig. 8) is the same line voltage measured after connecting the ultracapacitor substation to the system. As an example for placing the ultracapacitor substation between two traction power substations, the left diagram shows results of a simulation run,

signal name HT module comment

Vrated 125 V rated voltage of one module

Cmodule 63 F rated capacitance measured at 25°C

ESRmodule 18 mΩ DC, max. initial ESR measured at 25°C

Imax 240 A RMS, max. continuous current (ΔT = 40°C)

TABLE IV. ULTRACAPACITOR MODULE SPECIFICATIONS.

signal name Ultracap Bank comment

Vtotal_max 1250 V max. rated voltage at Ultracapacitor Bank

Ctotal 18.9 F total rated capacitance

ESRtotal 60 mΩ total ESR of Ultracapacitor Bank

Itotal_max 720 A RMS, max. continuous current (ΔT = 40°C)

TABLE V. PARAMETERS OF EXAMPLE ULTRACAPACITOR BANK.

signal name Ultracap Bank comment

Estorable_total 4.11 kWh energy storable in the Ultracapacior Bank

Eusable_total 3.08 kWh usable for (dis-)charging braking energy

Eunusable_total 1.03 kWh unused in order to maintain min. SoC

Pusable_min 450 kW available power at min. SoC

Pusable_max 900 kW available power at max. SoC

Ploss 40 kW power loss at ESR

Ptotal_min 419 kW output power at min. SoC

Ptotal_max 869 kW output power at max. SoC

93.1 % degree of efficiency at min. SoC

96.5 % degree of efficiency at max. SoC

TABLE VI. ENERGY AND POWER PROVIDED BY ULTRACAPACITOR BANK. BANK.

applying a distance of 1 km (Rline = 50 mΩ). The reduction of voltage deviation hardly exceeds 25 V. In the right diagram it is extended to 4 km (Rline = 200 mΩ), which is approx. the maximum distance after which a new traction power substation is placed in the line. In this case approx. 100 V can be achieved, proving that ultracapacitor integration only starts to affect voltage fluctuation if placed after a certain distance to overhead line’s main power supply.

The next two graphs (Fig. 9) express the energy and power behavior of the ultracapacitor bank. The left curve shows the charging process of ultracapacitors, starting from minimum state of charge (25%) and reaching the maximum after approx. 20 s. The first two power peaks of the right curve (light gray cp. Fig. 9) while charging directly reflect the progression of the train’s current signal during braking. In the left diagram a significant difference between calculated and actually achieved maximum state of charge can be observed. This is caused by the converter control’s limitation to the maximum continuous current Itotal,max tolerated at the ultracapacitor bank, based on manufacturer’s restrictions to avoid damage to the equipment caused by overheating. Therefore the ultracapacitors don’t charge completely during the predetermined time even though the available energy provided by the braking train would be enough.

This difference can be reduced by rearranging the interconnection of the ultracapacitors as more cells in parallel will multiply the bearable maximum current.

IV. ESTIMATION OF COSTS & BENEFITS

To estimate the capital cost of an ultracapacitor substation it is valid to add up prices of the main components. For the US the averaged example values were obtained during internet research and comparison of distribution offers. Prices for Colombia are calculated by assuming an additional charge of 40 % for importation

1, considering local custom charges and

taxes. Some of the results were additionally verified in direct contact with Colombian distribution partners in Medellín. To get the total cost for the ultracapacitor substation the prices of ultracapacitors are multiplied by the number of cells or modules required and added to an approx. converter price plus a 10 % additional charge, which is estimated for remaining power electronics, connectors and the chassis.

Results of Chap. III.E. predict only negligible effects in terms of voltage stabilization, which is why the estimation of profits at this early state of the project exclusively focuses on energy savings. The annual revenue (TR) due to energy savings is a product of four parameters (18).

TR = Total stops × Electricity rate × Efficiency

× Amount of energy (18)

The Total stops refer to the overall average amount of braking processes detected nearby one passenger station in a year, as this is where the ultracapacitor substations are going to be placed according to Chap. II.B. This value is derived from statistical numbers of stop intervals, taking into account two directions, peak-hours, off-peak hours as well as working days, Saturdays and Sundays [1]. The Electricity rate is determined by price per kWh in Colombia. Therefore the actual rate paid by Medellín Metro in February 2012 serves as a benchmark. Furthermore the price advance is considered by an annual increase of approx. 3 %

2 and therefore demands dynamical

calculation of the annual revenue, taking all periods of the Service life into account. The Amount of energy equals the recoverable energy in each scenario (cp. Table I) and has to be multiplied by the efficiency factor of ultracapacitors and converter, which gets to approx. 90 %.

To complete the Cost-benefit analysis, costs and revenues of the investment were compared to express the results by calculating two significant parameters for capital budgeting. In order to evaluate the investment, IRR (using cash flow method) and payback period (applying a static capital budgeting approach) were calculated for both scenarios, which were defined in the beginning (Table I). With rates of return between approx. 20-50 % and payback periods of only 3-4 years, the asset can be considered a profitable investment, even though the subject has to be researched in further detail, due to the fact that several estimated values had to be used for this first approach.

1 Rule of thumb, confirmed appropriate for preliminary cost calculation by

several professionals the authors worked with during their stay in the

country. 2 Information provided by the engineering department of Medellín Metro. Figure 9. Ultracapacitor’s energy and power while (dis-)charging.

Figure 8. Comparing impact of ultracapacitors on voltage stabilization.

V. CONCLUSION

After describing the stationary integration of ultracapacitor assisted energy storage systems into a metropolitan railway system, a fundamental design and its simulation was discussed. Furthermore, energy estimation revealed that the available energy from braking is dependent to a huge extent on the track’s train density. Therefore it was decided to consider two different scenarios. It was demonstrated how the created simulation helps analyzing certain configurations of an ultracapacitor substation. Several input parameters can be manipulated, creating new circumstances and allowing “what if” analysis. Given a new input signal based on measurements for the train’s current, different train density scenarios or even new types of trains can be simulated by adjusting a couple of parameters at the control panel.

Further ideas can be analyzed, such as the concept of expanding a Metro line by using ultracapacitor substations for voltage stabilization to a certain extent instead of high priced additional traction power substations. It could result in a reduction of overall expansion cost for systems like the Medellín Metro, considering that infrastructure cost is one of the most important barriers for the implementation of electric transportation systems. Usually distance constraints are more related to voltage regulation than power availability of traction power substations; this allows conceiving longer systems with the same installed power capacity. Another interesting thing to remark is that ESS allow reduction of peak demand, smoothing the load curve, in this case more units could operate using the same installed capacity.

The integration of ultracapacitors into the system of Medellín Metro has turned out to be, both technically and economically, an interesting and realistic endeavor. These modern energy storages have proven to bare a huge potential for several applications in public metropolitan railway systems.

ACKNOWLEDGMENT

The authors wish to thank the Universidad Pontificia Bolivariana of Medellín and the Kempten University of Applied Sciences for supporting this research project. Furthermore our thanks go to the Metro de Medellín for providing measurement data and detailed information about the Metro system.

REFERENCES

[1] Metro de Medellín Ltda. - Gerencia de Operaciones, „Informe de

Gestión,“ Empresa de Transporte Masivo del Valle de Aburrá - Metro de

Medellín Ltda., Medellín, 2010.

[2] J. Specovius, Grundkurs Leistungselektronik: Bauelemente, Schaltungen und Systeme, 2 Hrsg., Wiesbaden: Friedr. Vieweg & Sohn, 2008.

[3] R. W. Erickson und D. Maksimović, Fundamentals of Power Electronics,

Norwell (Mass.): Kluwer Acad. Publ, 2001.

[4] Maxwell Technologies Inc., „Data Sheet 125V Heavy Transportation

Modules,“ [Online]. Available:

http://www.maxwell.com/products/ultracapacitors/products/125v-tran-modules. [accessed: July 2012].

[5] M. Ehsani, Y. Gao und A. Emadi, Modern electric, hybrid electric, and

fuel cell vehicles, 2 Hrsg., Boca Raton, Florida: CRC Press Taylor & Francis Group, 2010.