How Pantograph Electric Arcs affect Energy Efficiency in DC Railway Vehicles

In DC electrified railways pantograph electric arcs represent not only a disturbance, but the step change of the pantograph voltage affects power losses directly and indirectly. The available line voltage is reduced if the train is in traction condition, the arc itself is characterized by ohmic power losses, and the triggered oscillating transient responses are characterized by a net power loss. In addition, if arc occurs during braking the arc voltage suddenly increases the pantograph voltage and may interfere with the dissipative braking chopper, reducing the recovered energy. This work presents the model and analysis of these phenomena with experimental results for arcs measured on a 3 kV dc line in traction and braking conditions.


I. INTRODUCTION
Electric arcs are still nowadays an unavoidable by-product of the current collection by sliding contact in electrified transportation systems and are particularly intense for dc traction supplies. The power level is significant: up to several MW per sliding contact in railways with overhead contact line (OCL), lower in metros with third rail distribution that share at least one sliding shoe per coach. The power transfer from the OCL to the carbon blade of the pantograph during train movement occurs through a contact section of a few square millimetres. As soon as the current collection mechanism worsens due to imperfection of the geometry, defects on surfaces, temporary detachment, debris, oxidation, arcing occurs with air ionization and plasma conduction. The temperature increases to the melting point of materials and only train movement and air convection avoid the degeneration into an extended irreversible damage of surfaces or complete OCL breakage.
Arc intensity and duration depends on several parameters characterizing the train: its speed, the intensity of the dc current, the traction or braking condition. Although dc arcs are more intense than those in ac railways, the latter have recently received much more attention, as pointed out in [1] [2]. DC railways in general (so encompassing all the metro, light railways, tramways, etc. using a dc supply) undergo more intense arcing due to two factors: the supply is a dc voltage of significant amplitude (between 600 V and 3000 Vdc nominal), and current intensity can be an order of magnitude larger than ac railways. DC arcs in electrified transports have been characterized in the literature in terms of voltage-current relationship [1][3] [4], typical waveforms and their spectra [5] [6], and interaction with the traction line impedance and locomotive filter [2] [7], exploiting them also for network impedance estimate [8].
In general, arcs are well known as a source of highfrequency noise potentially affecting the quality and availability of telecommunication and signaling systems [1] [9]. The vast majority of studies and data are related to ac systems [10] [11], in a high-speed modern railway perspective, although also dc systems have relevance, thinking of the many metros and rapid transits with integration of wireless signaling and services.
The presence of arcs at the pantograph-OCL interface (and similarly at the sliding shoe-third rail one) during traction is responsible for a first direct effect: the reduction of the available pantograph voltage Vp , since the arc voltage Varc necessary for current conduction in plasma is in series with the line voltage Vl. Such reduction may in general cause a temporarily reduced train performance. Oppositely, arcs during braking increase the pantograph voltage due to the reverse flow of pantograph current [12], causing possible overvoltage issues.
Then, the arc conduction of the pantograph current Ip under a non-negligible voltage drop Varc is responsible for a simple ohmic power loss, given by the product of the two quantities. For ac systems conduction losses were estimated by calculating the Lissajous diagram area between voltage and charge [13].
In addition, the electric arc as a broadband transient signal triggers the transient response of the traction network and train, causing the appearance of oscillations at the respective resonant frequencies, determined by the resonant filters at substation and onboard, and by the resonant modes of the line at higher frequency [2]. These components may disturb onboard metering and control [2], as well as signaling circuits connected to the track through the coupling of the return current [6] [14].
A not so obvious consequence can be identified during regenerative braking: when arcing occurs, due to the direction of flow of Ip, the value of Vp is further increased and larger than the pre-arc line voltage. Dissipative braking by means of the braking chopper feeding onboard resistors can in this case start due to momentarily excessive pantograph voltage, increasing conducted emissions and power losses.
This work models and analyses these cases including the explanation of the observed phenomena with the support of experimental results recorded over the Italian DC railway network and the Metro Madrid lines during the EMPIR 16ENG04 MyRailS project [12] [15]. Section II discusses the electrical behavior of the arc, including the interaction with the traction line and the onboard filter. Section III then focuses on the power losses mechanisms for the identified phenomena, so as intrinsic arc dissipation, reduction of pantograph voltage and triggering of dissipative braking. Section IV reports and comments the experimental results.

A. Electric arc
At the detachment of the pantograph from the catenary (or equivalently of the sliding shoe from the third rail) an electric arc may establish through one or more plasma columns that allow the flowing of the current through the air gap. The electric arc is maintained as long as the combination of air gap and available voltage is compatible with current conduction; in other words the electric arc requires a minimum voltage that depends on the pantograph current intensity Ip and gap width.
The typical V-I characteristic of dc arcs can be described as follows: i) at low current (e.g. tens of amperes) a constant arc power law can be assigned; ii) for increasing current above a specific threshold, the arc voltage reaches a plateau and does not depend then any longer on the current intensity. More complete dynamic models elaborated starting from Mayr's model can be found e.g. in [16], but the focus here is on power dissipation and so a quasi-static analysis is sufficient.
Of course V-I characteristics vary depending on electrode materials and for electrified transports also the relative movement of the electrodes plays an important role, for example increasing the effective arc length (the arc assumes an inclined position along the direction of movement). Various V-I curves for dc low-voltage arcs with static electrodes were collected by Paukert for current intensity ranging from 100 A to 100 kA [4].
The electric arc can thus be represented by the following Thevenin equivalent circuit: a resistor Rarc, variable with the intensity of the flowing current (getting smaller for increasing current), and a voltage source Varc with values reaching a plateau for large Ip current intensity.
The arc resistance Rarc is in general negligible as shown in [17], Fig. 14, where values of 10 to 100 mȍ are reported. In case of various amounts of wearing after long use, the electric contact at the sliding surfaces worsens and the resistance increase by up to an order of magnitude [18]. The observed values during preliminary measurements [19] are of the same order (about 0.2 to 0.5 ȍ).

B. Line resonances
Electric arcs as broadband transients can excite line resonances up to several tens of kHz (see Fig. 1), as shown in [2] using a parametric distributed-parameter model of the traction line and analyzing the components of the recorded electrical quantities. So, they are a relevant excitation source to probe the high-frequency portion of the line impedance and its resonance modes [8]. The transient terms of the resonant modes overlap in an initial burst following arc ignition (the portion of the arc waveform with the largest frequency content); the decay is also rapid since equivalent line resistance (sum of skin effect of catenary, additional losses in the rails [20] and distributed loads) is larger in that frequency range. As observed in [22], sec. III.C, line behavior at resonance and anti-resonance is purely resistive, increasing significantly the amplitude of harmonic active power terms. The phenomenon cannot be quantified, however, without a complex model of the entire line section and rolling stock items within it [22] [23]. In [23] it was demonstrated that for dc vehicles the amount of harmonic power terms is limited, with amplitude much smaller than for ac railways and negligible.
The line resonance response remains thus relevant as a source of conducted disturbance appearing in the traction return current along the track and possibly coupling to audiofrequency track circuits, as analyzed in [24] [25] for Italian high-speed lines.

C. Onboard filter transient response
The onboard LC filter is also excited by transients occurring at the pantograph interface, as well as by step loading during travel. Filter transient response occurs usually in the 12-25 Hz interval, determined mainly by the inductance and capacitance values of the onboard LC filter, although coupling with line and substation impedance (the latter featuring a similar LC output filter) can slightly modify the response. The large-value series inductor during the transient response behaves as a high impedance current source imposing an oscillating current in the Ip waveform and flowing along the line section. This behavior was analyzed in [2], identifying the conditions for the flow of current back to the substation through the rectifier diodes when positively biased by other trains pulling traction current within the same supply section. The absence of this condition prevents the flow through the otherwise reverse-biased rectifier diodes and causes anticipated arc quenching; at the successive Ip oscillation the arc can re-ignite if the pantograph detachment condition holds. The combined effect of the non-linear conductance of the diodes and the excitation of the first line resonance slightly above 100 Hz can be seen in Fig. 2.

A. Electric arc dissipated power
The arc resistance is extremely small and does not contribute significantly to circuit damping; the arc conduction voltage instead appears suddenly at arc ignition and represents the main electric phenomenon related to the arc event. Power dissipation within the electric arc, as a non-linear resistance element, is thus calculated straightforwardly as Parc=VarcIp.

B. Reduction of useful voltage
As sketched in Fig. 3, the arc voltage is in series to the line voltage with respect to the pantograph, so that Vl -Varc = Vp, having noted the arc voltage with load convention, so towards the line with a positive incoming Ip current. This reveals two opposite behaviors, during traction with a reduction of available voltage (being Varc > 0) and during braking with an increase of Vp beyond line voltage (being Varc < 0), possibly triggering protection by dissipative braking, as discussed below in sec. III.C.
Considering an arc event occurring during a traction phase, Vp rapidly falls down while the voltage downstream the filter inductor is kept constant by the large filter capacitance. This imbalance across the filter inductor triggers a current oscillation at the filter natural frequency with a negative derivative. Therefore, at the occurrence of an arc event during traction, Vp and Ip decrease with different time behavior, and the power fed to the traction unit dramatically decreases for a short time interval. If the arc occurs in correspondence of a small absorbed current Ip, with the superposed oscillation negative values might be reached; however, when the current reaches zero the arc quenching occurs. In this case, with the extinction of the arc at zero current the locomotive is completely disconnected from the catenary and the energy stored in the filter capacitor feeds the circuits downstream; as long as the capacitor discharges the voltage across it Vf gets smaller. Because of the pantograph detachment and the zero flow of current in the filter inductor, the pantograph voltage Vp tracks Vf. This is demonstrated by the change of ripple in Vp: before quenching Vp is characterized by the typical 100, 300 and 600 Hz ripple caused by the substation [2] [6]; with the arc quenching the Vp ripple is equal to the Vf ripple that is caused by the traction inverters.

C. Triggering dissipative braking
As previously mentioned, during regenerative braking the current Ip is applied into the network and the arc voltage sign reverses and increases the pantograph voltage Vp beyond the line voltage Vl.
It is customary to implement a safety mechanism, so that the local line voltage (sensed by measuring Vf) cannot go beyond a given threshold (usually corresponding to the maximum permanent voltage of the EN 50163). During detachment and triggering of electric arc the Vp quantity does not follow closely the line voltage Vl any longer, and the detection of the threshold voltage is anticipated. A phase of dissipative braking by means of the braking chopper starts to keep the supposed line voltage Vp stable, thus causing increased power losses and conducted disturbance.

A. Arc event during a traction phase
Two arc events are analyzed in terms of dissipated energy. The first event is shown in Fig. 4 and a zoom around the arc event in Fig. 5. The event is characterized by a voltage reduction of about 300 V and the absorbed current, because of the superposed filter oscillation, goes from 200 A to 0 A (arc quenching). The arc duration is about 14.9 ms while the quenching phase lasts for about 70 ms. At the end of the quenching the contact between the pantograph and the OCL is re-established with a very fast increase of Vp, reaching again approximately the same amplitude before the arc event (about 3.6 kV). This voltage step, characterized by a slew rate of about 14.8 V/μs, triggers the filter oscillation at about 15 Hz.
The average power dissipated by the arc, computed over its duration (14.9 ms) is 6.2 kW, corresponding to an energy of 93 J. For what concerns the active power associated with the oscillation phase, Table I summarizes the results for each of the first three oscillation periods: the active power terms are calculated by considering Vp and Ip and separating them into a dc and an ac component, assigning to the arc phenomenon those terms that include at least one ac component. The table also provides the dc power that the locomotive should have absorbed if the arc event had not occurred.   Fig. 4. Fig. 5 shows the electric arc event during a regenerative braking where the energy is subdivided into a part of recovered energy used to supply other loads along the line and into a remaining part of energy dissipated by the braking rheostats. As previously mentioned, during the braking phase the arc voltage Varc is negative; therefore, the voltage measured at the pantograph, Vp, will be larger than the line voltage Vl. In Fig. 6(a) the arc flashing phase, highlighted in green, is characterized by a voltage impulse of about 450 V rising in about 10 ms. In the same time span, the current decreases to zero, arriving at the arc quenching phenomenon, highlighted in grey. For about 25 ms, as for the arc during the traction phase examined in sec. III.A, the current is zero and the voltage upstream (Vp) and downstream (Vf) of the filter are equal. As can be seen in Fig. 6(b), in this interval not only the braking type changes from mixed to purely dissipative, but the train control system increases the duty cycle of the chopped rheostatic current to counteract the increase of Vp. In the phase following arc quenching, the pantograph voltage quickly returns to the line voltage, triggering the low frequency oscillation of the filter. The mean power dissipated by the arc is 25 kW. As can be seen in Fig. 7 the arc event with the consequent voltage swell also increases the power dissipated by the rheostat reaching a maximum value of 1.65 MW. The mean value of the additional power dissipated by the rheostat because of arc occurrence is 152 kW. V. CONCLUSIONS This work has considered the occurrence of pantograph arcs of a dc railway system during traction and braking phases. In both cases the voltage waveforms and the physical phenomena have been analyzed, to understand the dynamics of the arc discharge, its evolution and possible quenching when the pantograph current reaches zero.

B. Arc Event during a regenerative braking phase
In particular, the dissipated power terms associated to the arc itself and to the electric phenomena (oscillation of the onboard filter) and events (activation of the dissipative braking chopper) have been evaluated, showing that the peak power levels are notable. Despite the short duration of the arc, also the energy values are not negligible, especially considering the rate of repetition of electric arc in dc traction systems.