Wayside energy recovery systems in DC urban railway grids

In modern electrified and rail-bound mass transit vehicles, a considerable part of the braking energy is still dissipated via resistors. This applies in particular to less connected and low frequented grid sections. For this purpose, wayside energy recovery systems can be used to store excess energy and release it during acceleration of nearby vehicles. They offer further advantages, such as a potential reduction in the power requirement of substations or additional safety and reliability. This paper provides an overview of actual demonstrations of various systems in public transport grids. The focus is on the most important technologies (namely: supercaps, flywheels, batteries and inverters), as well as the respective manufacturers and products. The achievable improvements, based on both real demonstrations and scientific studies, are identified and critically evaluated. Important points for a better assessment of the systems are worked out. A future forecast is given for each of the four technologies. © 2019 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).


Introduction
The transport sector plays a major role in the reduction of CO 2 emissions and, as the demand increases with a growing world population, will have to undergo major changes in terms of electrification. The availability of suitable energy storage technologies makes it nowadays possible to use the electrified systems more efficiently. The focus of this work is therefore on the investigation of braking energy recovery in tram, metro and light rail networks, which are supplied with DC voltage, by using stationary storage systems or bidirectional substations.
Modern vehicles enable regenerative braking. Kinetic energy is converted into electrical and fed back into the grid with help of the motors, operating in generator mode. This is possible if corresponding consumers are nearby, such as other vehicles that are currently accelerating. If these are missing, the feed-in of the regenerated current can lead to an inadmissible increase in the catenary voltage. To avoid this, it is dissipated via brake resistors above a certain voltage threshold and is therefore lost to the environment in form of heat [1].
There are different approaches to reduce these losses, such as timetable optimisation, optimised traffic management or ecodriving techniques [2e4]. These are based on operational adjustments and specifications with regard to driving behaviour. Another option is to use energy storage devices. They can be mounted on the vehicles [1]. This approach has the advantage that the energy can be stored directly on site and used again accordingly, without large transmission paths. Autonomous operation on short catenary-free sections is also possible [5]. A disadvantage is the increased vehicle weight, which generally increases energy consumption, as well as the additional space requirement.
Another option, which we reviewed extensively within this work, is the use of wayside energy recovery systems (WERS), i.e. stationary energy storage systems or reversible substations (inverters). These can be installed at suitable locations in the grid, e.g. in appropriate substations. The space and weight play a subordinate role in comparison to on-board systems, which enables larger dimensioning. In addition, they can absorb the braking energy of several vehicles in the vicinity and thus increase the overall efficiency. Further advantages are voltage stabilisation, a reduction of the substations' peak power demand and emergency supply. In the case of stationary energy storage systems, various technologies are applied, of which we took a detailed look at the three most 2. Use cases for WERS in urban railway grids WERS can be used for different reasons and achieve several improvements. We will introduce the basic working principle and the four most important use cases in this chapter. Fig. 1 shows the basic principle. As mentioned before, braking vehicles can feed energy back into the grid if it is receptive (1). If there are no consumers in the vicinity, e.g. when two adjacent vehicles brake at the same time, the electrical current must be dissipated via braking resistors (2). To prevent this, one or more WERS can be installed along the track. The additional consumer increases the grid receptivity and is able to store braking energy (3) and release it again if required (4), thus relieving the substations. In addition to increasing energy efficiency, WERS also offer other advantages, such as voltage stabilisation [18], peak power reduction [38] or emergency supply [39]. These will be introduced in the following subchapters.

Energy efficiency
The most apparent reason for the application of WERS is to save braking energy and thereby increase the overall system efficiency [1]. This directly saves costs and, depending on the energy mix, CO 2 emissions. These losses occur along the entire route. When applying a stationary storage, instead of on-board, it is of crucial importance to find the most suitable location(s) within the grid, which is part several scientific investigations [10e16].
Diverse studies and demonstrations show that energy, and presumably money, can be saved by WERS applications, assuming favorable conditions. Exact numbers and critically evaluated results are stated in chapters 4 and 5.

Voltage stability
A second reason for the application of WERS is to ensure and/or increase voltage stability [18] within the whole grid or at crucial spots. This is of particular interest when distances between substations are relatively large, which can be in outer areas. The introduction of more powerful vehicles [18] or a smaller headway can influence the grid negatively. To increase operation on certain routes/segments, it can be an alternative to install WERS instead of a new substation. Voltage stability may become even more pronounced in the near future, especially if other consumers are connected to the existing network, such as charging stations for electric buses and cars. This approach has been investigated and demonstrated in the context of the European project Eliptic [41].

Peak power reduction
The (peak) power demand of a substation can be reduced by the application of a nearby storage system. It decreases the grid usage fee or demand charge, which is based both on the installed power and (expected) yearly energy consumption [38]. It also enables the design of smaller substations in the future, which potentially shifts their operating windows to higher efficiencies. This aspect has not yet been researched extensively (see 5.2 and 5.3).

Emergency supply
Stationary storage systems can also be used as emergency power supply in case of a substation failure or power outage. This is, among others, motivated by Japanese company Hitachi and used as additional security system in Tokyo's Metro ( [39], see 4.3.2). After the Tsunami/Fukushima disaster in 2011, the demand for emergency supply systems is especially in Japan a major issue. A similar importance is given in cities with generally rather weak electricity grids.

Technologies
The available WERS (supercaps, flywheels, batteries (Li-Ion, NiMh), inverter) differ significantly regarding their capability to store electrical energy and set it free within a certain amount of time. The basic storage mechanism differ, too. The energy can either be stored in an electric field (capacitor), within electro-chemical compounds (all kinds of batteries) or mechanically, e.g. by means of rotating mass (flywheel). In this sense, an inverter is not an energy store, since it only shifts it from one grid into another. It theoretically has an infinite capacity, which is only limited by the receptivity of the superordinate grid. Fig. 2 shows a basic classification of the general energy and power capabilities. It illustrates the respective peripheral areas. In this simplified representation, however, a reading of actual combinations of both is not possible. Table 1 compares the different technologies regarding their technical and economic parameters. For the electric/electrochemical storages (EDLC, Li-Ion, NiMh), the data refers to the cell level. In the literature many points could not be found or were contradictory indicated. For this reason, the values are primarily to be seen as an orientation and do not define any hard limits.

Electric double layer capacitor (supercap)
The electric double-layer capacitor (EDLC, also named super-or ultracapacitor) is an energy storage technology which is based on the principle of a Helmholtz double-layer [46]. Since the energy is stored directly in the electric field, and not in chemical bonds, EDLCs offer a very high power capability of around 10,000 W/kg, also at low temperatures.
Very high numbers of charging/discharging cycles are achievable, reaching values up to 1 million [41]. This is much more than for electrochemical systems. The round-trip efficiency is high, too [47]. Having these attributes, EDLCs are a well suited candidate for the application as energy storage in railway grids. A significant drawback are the comparatively high investment costs with about 10,000 $/kWh [41] and the low energy density. This results in large and heavy systems to obtain the required capacities [1].

Flywheels
A flywheel (FW) is a mechanical storage device which converts electrical energy into rotational kinetic by storing it within the motion of a rotating mass. In essence, it is comprised of a rotor, electrical machine, power electronics converter and auxiliary components, such as bearings, cooling system and vacuum pump. Recent developments in the fields of material engineering, bearings and power electronics, as well as their rapidly falling prices, have made this technology much more reliable and accessible for various applications. Consequently, attractive use cases started to emerge.
In terms of speed, flywheels can be offered in different variants, with slower systems having 5000 rpm and fast systems up to 100,000 rpm [26]. A central advantage of flywheels is the rapid   charge/discharge process and the, if properly maintained, comparably high cycle durability (more than 100,000 [42]). The calendar lifetime is also quite long (around 20 years), always presuming appropriate maintenance. In comparison to EDLCs, they offer a slightly increased energy density and comparable performance. Flywheels also allow for a straightforward determination of the state of charge (SOC), as it is a function of the rotational speed. They have a wide operating range in terms of temperature. The roundtrip efficiency of 90e95% is also quite high, but auxiliary loads are required and reduce the overall efficiency [26]. The vast power densities at moderate energy levels enable to store, when configured in banks or parallel arrays, up to hundreds of kilowatt-hours (kWh) [1]. The investment costs per kWh are significantly higher than for an electrochemical storage [42]. Unfortunately, we could find no reliable sources about the running costs, but these will presumably also be significantly increased, which makes the systems rather expensive.
For more information on the working principle and applications we refer to the corresponding literature [26,48].

Lithium-ion battery
Lithium-Ion batteries are comprised of two electrodes, which are separated by an ion-conducting but electric isolating separator, soaked in an usually liquid electrolyte [49]. The term lithium-ion battery does not describe a specific cell chemistry. Various active materials can be used in the electrodes [50]. This results in different properties like safety, energy-and power density. For more detailed information, please find the corresponding literature [49,51,52].
Contrary to EDLCs, the energy is stored in chemical bonds. The technical and cost data of lithium ion storages are well documented [44]. They differ greatly with the chemistry and cell format used, but in general offer by far the highest energy density. A drawback are the comparatively low power densities. The total lifetime is strongly limited by the number of achievable full cycles (up to 15,000, depending on the technology). Li-Ion batteries have, compared to EDLCs or flywheels, smaller lifetimes (up to 15 years). This strongly depends on the external conditions, especially temperature and SOC. A description of ageing effects was given by Broussely et al. [53]. The specific cell costs are quite low due to economies of scale and start at around 100 $/kWh. This makes large storage systems economically viable.

Nickel metal hydride
Nickel metal hydride batteries (NiMH) are a further development of nickel cadmium batteries (NiCd), with the aim to replace the poisonous cadmium. The nominal voltage is only 1.2 V, so that a large number of cells must be connected in series to achieve feasible battery voltages. While NiCd batteries still offer good performance at temperatures around À40 C, NiMH batteries are similarly temperature-sensitive as lithium-ion batteries. At temperatures above 50 C, the battery is barely chargeable and below À10 C, the power capability collapses massively [49]. Compared to Li-Ion, NiMh batteries are characterised by lower energy densities and poorer efficiencies. In addition, significantly shorter cycle lifetimes can be achieved at higher cell costs. For these reasons, the application of this type is unlikely to have a future.

Inverter
In the near past, inverters (INV), which are transforming direct current into (three-phase) alternating current (AC), have been applied in order to feed braking energy directly back into the medium voltage AC grid (public or owned by the transport operator) [1,32,45]. Nowadays, insulated-gate bipolar transistors (IGBT) are used as switching elements, which can be directly turned on and off in a controlled manner and enable high performances. They replace (gate turn-off) thyristors, which were commonly used for railway applications. A more detailed explanation can be found in the respective technical literature [32,45].
For inverters, an energy-based specification is not meaningful. The specific power values can be derived from the product data sheets [54] and are around 1 kW/kg. Like Flywheels, they are characterised by a long service life of more than 15 years and high efficiency ( > 95%), but with distinct lower investment costs per kW of installed power [45]. However, it should be mentioned that we could not find reliable sources of precise costs and the data is partly based on estimates [45].

Demonstrations and applications
Recent and past demonstrations of the four investigated WERS technologies are presented in this chapter. Each subsection includes a list of the most important manufacturers and their products, as well as the respective place of use and the technical data. Where available, results are reported. A critical evaluation of these is given in chapter 5.1. Table 2 gives an overview of recent WERS demonstrations in urban rail systems based on EDLCs. They are alphabetically ordered by the respective manufacturers.

ABB Enviline ESS
ABB's Enviline ESS is designed for 600/750/1500/3000 V traction power supply voltage. The maximum system power is 4500 kW with a total efficiency of 94%. A cabinet weighs up to 950 kg and has a useable energy of 1.62 kWh at 750 V. The maximum useable energy of the system is 16.2 kWh [68]. At the moment there are two operations [55]: In the Melbourne Metro with a total power of 2.2 MW and a capacity of 6.66 kWh and at Metro Warsaw, where 10 combined cabinets provide 3.3 MW of power and 11.1 kWh of useable energy. The system claims to be the largest ESS installation based on EDLCs with line voltages below 1000 V. The manufacturer  [42,45] states energy savings of 3 MWh per day [55] (see 5.1).

Adetel NeoGreen/NeoStab
The French company Adetel offers its NeoGreen system for energy saving and NeoStab for voltage stabilisation [56]. A NeoGreen storage system has a rated power of 330 kW and an energy capacity of 1 kWh. A connection of several single units by means of a master control is possible in order to be used in larger applications. The system can work in 750 V and 1500 V grids and has an expected EDLC lifetime of 15 years. Demonstrations took place in Lyon and Tours. NeoStab is directly derived and has the same technical specification, but works in a different operation mode in order to stabilise the grid voltage at crucial spots. Operation commenced in 2015 on Saint-Gervais Vallorcine train line and allowed the service to be doubled [56].

Bombardier EnerGStore
Bombardier's wayside energy storage system has been developed to work with line voltages from 600 V up to 1500 V. A single unit has an energy capacity of 1 kWh and can supply a maximum power of 650 kW. The system was designed to be scalable. The connection of several small units offers an increased capacity of more than 5 kWh and a high flexibility in operation. The manufacturer claims that the system can contribute to reduce overall energy consumption by 20% [57]. A 1 kWh system was installed at a test track in Kingston, Canada [69]. The basic operability of the system should have been proven, but the sources do not provide more detailed information about actual energy savings.

Meiden Capapost
In 2012, Japanese companies Meiden and Sojitz Corporation received an order from Hong Kong rail and metro operator MTR [58]. Part of this order were two units of the Capapost energy recovery system, which is based on proprietary EDLCs [59]. The $318 Million contract over two 2 MW systems is stated to be the largest order for EDLCs ever since [70,71]. The installation is expected to reduce traction power consumption on the South Island Metro line by 10% [70]. Operation has commenced end of December 2016, but yet no published operational results could be found. In 2016, another system was put into operation in Tokyo (Seibu Railway). It was installed at a substation and can store a maximum energy of 19.26 kWh. It can be used in two operating modes: for maximum braking energy recovery (lower average EDLC voltage level) or voltage stabilisation (higher average EDLC voltage level). The system saved a maximum of 3000 kWh per day in test operation [60].

Railway technical research institute
There has been research on stationary energy storage systems based on EDLCs in Japan since more than a decade [61]. Test systems for 600 V and 750 V were built up and proven in the laboratory and on-track, at Osaka and Kamakura. At the latter, the storage was connected to the 600 V grid with the purpose to reduce voltage drops. It had only a small energy content of 0.42 kWh, with a maximum current output of 300 A, which was limited by the power electronics. Contrary to the usual installation at a substation, the storage was located nearly in the middle between two of them. At this point, there are substantial voltage drops when two vehicles are accelerating at the same time. Tests showed a voltage-drop compensation of 30 V [61]. At Osaka, the system was connected to 750 V in order to store braking energy. For this purpose, the storage was considerably larger and had an energy content of 1.39 kWh and a maximum current output of 1 kA, which was again limited by the voltage converter. It was located near to the terminal of a main line, close to a section disconnector. Tests showed nonneglectable losses during charging, since there was a parallel resistor in order to regulate the voltage of the EDLC system. This solution offered potential for further optimisation.

Siemens Sitras SES
Siemens' Sitras Stationary Energy Storage (SES) system has been operated in several cities worldwide. It was comprised of 1152 EDLCs, packed in 20 ESM 125 modules [72], and could be used for several stationary and mobile applications. Each ESM 125 had a useable energy capacity of around 0.125 kWh and was able to withstand/provide peak currents of up to 750 A [72]. Sitras SES had a rated useable energy content of 2.3 kWh and a peak power of 700 kW [73]. It could be used in two different operation modes. For the purpose of grid voltage stabilisation, the SOC is kept at a high level and the storage is discharged when the line voltage falls below a certain threshold. Contrary, when used in energy saving mode, the SOC is on a lower level. This enabled the system to store the braking energy and feed it back at later points in time. The system  [67] was firstly installed in Cologne (Germany) in 2001 and used to stabilise the voltage. It was shown that the event of critical undervoltage can be effectively avoided. The ratio between supplied and absorbed energy was around 30% [74]. The energy consumption of the affected substation in one month was reduced by 15,000 kWh compared to the year before [63]. Between February and June 2002, the system was tested at the Metro in Madrid for the purpose of system voltage stabilisation. Again it was shown, that critical undervoltages could be strongly reduced (U Line < 520 V) or completely eliminated (U Line < 490 V). Field tests also showed, that the average power requirement per train was reduced by 50 kW [7]. An early prototype of the system was tested in Portland (Oregon) in 2002 but failed after 6 months of operation due to a failure in the cooling fan supply and was decommissioned [62]. In 2015 a new system was released in a light rail line between Portland and Milwaukie [65]. In Rotterdam, the system was installed in 2010. The results were below expectations and furthermore there have been severe noise issues [75]. At the moment, Sitras SES is not further distributed by Siemens due to the high invest costs, caused by remaining high prices for EDLCs, and the associated small interest of public transport operators (PTO). As replacement, the company now promotes the inverter based Sitras PCI system (see chapter 4.4.6).

Woojin Industrial Systems Co
Woojin Industrial Systems Co. is a Korean manufacturer for rolling stock and equipment [66]. It offers energy storages for railway systems with different voltage levels (DC 750 V, DC 1500 V, AC 55 kV). Energy contents range from 2.9 kWh to 12.9 kWh. The systems have been applied in metro systems of Seoul, Daegu and Daejeon (1500 V), as well as Incheon (750 V). The manufacturer claimed energy savings of around 20e23%, without any further explanation.

Other
In 2007, two in-house developed EDLC systems were deployed at two substations of the Seibu Railway network (Japan, Tokyo area). Over a period of 15 min, 7.7 kWh were stored in the systems, of which 5.9 kWh were provided again. Without further explanation, an effective contribution to the reduction of primary energy consumption was stated [67]. Table 3 gives an overview of recent flywheel based applications in urban rail and trolleybus (Zurich) systems.

Calnitex Vycon Regen
Calnitex Vycon Regen has been developed with electric and rail authorities in the U.S [90]. The system has a modular structure. One module has an energy capacity of 0.52 kWh and a nominal output power of 125 kW. The flywheel rotational speed ranges from 10,000e20,000 rpm [91]. It is used in Los Angeles' metro network since 2014. As of 2015, the installation had a power of 2 MW and energy capacity of 8.2 kWh and was ready for expansion to 6 MW/ 24.6 kWh. The manufacturer claimed average daily savings of 1.6 MWh measured during operation which translates to predicted yearly savings of almost $100,000 [77,90].

Piller Powerbridge
Piller is a German company that mainly produces flywheel storages for uninterruptible power supply (UPS). It has also undertaken various demonstrations within the public transport sector. This started with the first project in Hanover (Germany) in 1998 [78]. The effective useable energy of the Powerbridge called system is rated at 4.6 kWh. The maximum power ranges up to 1 MW for 16 s. The system is low speed rotating, ranging from 1800e3600 rpm. Several installations followed in the early 2000's in Hamburg, Hanover and Paris. According to manufacturer information, around 450 MWh of yearly savings have been recorded. This corresponded to a payback period of around 7 years and led to a second installation in Hamburg. The system was tested in the Zurich trolley/tram grid in 2009. Due to the already good receptivity of the grid, only 16 MWh per year could be saved under ideal conditions [79]. This was and is not sufficient for an economic operation. A Powerbridge was installed in the light rail network in Bielefeld (Germany) in 2012 [80,81]. The storage was located at the end of a line which was to be extended in the near future [81]. It had a weight of 10 t and was, with 95 dB(A), very noisy. Demonstration results showed that annual savings of around 320 MWh could be achieved. Another system has been installed in Freiburg (Germany) in 2013 [82]. The storage was placed at a line terminal, where power is provided only from one substation in 1.1 km distance. The annual energy savings achieved were roughly 200 MWh, with a storage efficiency of around 80%.

Rosseta
Rosseta was a German company from Dessau-Rosslau (Saxony-Anhalt). It became insolvent in 2013(*). They built high speed rotating flywheels (15,000e25,000 rpm), specially designed as braking energy storage in tram grids or in UPS applications. Depending on the power electronics, the system was able to store and provide up to 800 kW of power. It had a useable energy

Siemens
In cooperation with Siemens AG, a flywheel was tested in Cologne at the beginning of the 2000s [85]. It had a maximum power output of 600 kW at an energy storage volume of 6.6 kWh. The system was assigned an energy saving potential of 5e10% per vehicle and 1e2% per fleet [84]. The high investment costs were listed as a significant disadvantage.

Urenco
Urenco is a British manufacturer of centrifugal plants, which has also developed some high-speed systems for operation in railway grids. Three parallel flywheels have been tested on London Underground's Piccadilly line in October 2000. It was able to rotate with 37,800 rpm. The round-trip efficiency was claimed to be about 90% and the lifetime expectancy approximately 20e25 years [7]. The system was able to significantly reduce voltage drops. With the application of a 1 MW WERS, the energy transfer between vehicles should theoretically be increased up to 30%. This would translate to a payback time of 5 years [7]. A 1 MW installation has been applied in New York city [86]. Further applications were in Lyon and Paris [87].

Others
There have been several other applications of systems based on flywheels, mostly in an academic frame. As part of a Spanish research project in 2003, a 350 kW storage for 3000 V DC grids was developed in Madrid [88]. The flywheel had a nominal speed of about 6500 rpm. It was firstly tested in the laboratory and then installed on track. The most critical limitation of the system was heat due to losses. The basic functionality was demonstrated, but no actual statements on increased energy efficiency could be found. In a follow-up project in 2006, a storage system was developed which should provide 5.6 MW at an energy content of 0.88 MWh [89]. It is not clear from the literature we could find whether an actual use has taken place. Table 4 gives an overview of recent demonstrations in urban rail systems based on batteries.

ABB Enviline ESS/Saft Intensum Max (Li-Ion)
In cooperation with Saft America and Viridity Energy, ABB installed a 420 kWh storage system on the Market-Frankford line in Philadelphia. It is based on Saft's Intensum Max containered lithium-ion battery [92]. The maximum power output was 2.2 MW, but the system should be extended to supply 8.75 MW in the near future (as of 2014). Besides working as wayside energy storage, it is operated as public grid storage, too. Measured energy savings ranged from around 900 kWh on average weekdays to 1284 kWh on average weekend days.

Hitachi BeCHOP (Li-Ion)
The first Hitachi BeCHOP was built of Li-Ion batteries formerly used in hybrid cars [39]. The actual energy storage is connected to the grid by means of a step up/down chopper. The rated storage capacity varies from 19 to 114 kWh and the rated power from 500 to 3000 kW. The system has been deployed several times in Japan and Korea. One of the first units was installed in Kobe in 2007. Operational results stated the energy reduction of the 1 MW/ 37.5 kWh to be more than 310 MWh per year [67]. However, this value was given without any background and other sources, which is why the credibility is not confirmed and no further consideration is made (see 5.1).

Kawasaki Gigacell (NiMh)
Kawasaki  Table 5 gives an overview of recent demonstrations in urban rail systems based on inverters.

ABB Enviline ERS
ABB's Enviline ERS is designed to be used with 600, 750 or 1500 V nominal traction supply voltage. It offers a respective power of up to 1 MW for the smaller voltages and 2 MW for 1500 V. So far, the ERS has only been used in test operation for longer periods of time. The results are "very positive" according to the manufacturer [107], without any further explanation. A confirmation by a real application could not be found in our research.

Alstom Hesop
The french Alstom group sells its IGBT based regenerative substations under the name Hesop. The power classes range from 1 to 4 MW (nominal), at line voltages of 600 V to 1500 V [99]. The energy is fed back directly into the public medium-voltage grid. The system is currently in operation in London and Milan, where it is to recover around 15% of the traction energy on the Yellow line [101]. New assignments are currently being planned and under construction in, among others, Milan (tram), Riyadh and Sydney (both metro).

Dynniq (imtech)
Dynniq, formerly Imtech, is a Dutch supplier of integrated mobility and energy solutions and services. They developed and installed two active regeneration units [102] at existing substations in Rotterdam's metro grid in the frame of the Ticket2Kyoto project (T2K [108]). The invest costs were V478,000. With yearly estimated energy savings of 600 MWh, translating to benefits of around V55,000 per year, the expected payback time (including funding) was stated as 5 years [103].

Ingeteam
Spanish company Ingeteam applied its IGBT based energy recovery system in German city of Bielefeld's tram and Brussels' metro grid. The investment costs in Brussels amounted to V1,800,000. Expected energy savings were around 9%, or 3.4 GWh, annually. This led to a payback time of 5 years [106]. Two 1 MW system have been combined with a flywheel storage (see 4.2.2) in Bielefeld. Results from operation were positive. Each inverter was able to recover more than 300 MWh per year [81].

Meiden
Japanese company Meiden offers, besides storage systems (see 4.1.4), also inverters for energy recovery in railway grids. Applications are 750 V and 1500 V grids. The power ranges from 500 to 1000 kW, with 300% overload capability for 1 min [109]. No recent demonstration could be found, though.

Siemens Sitras PCI
Since 2016 S offers the Sitras PCI system, based on selfcommutated IGBTs. It replaces the former TCI (Thyristor based [110]) and SES (Supercapacitor, see 4.2.2) energy recovery systems. It can be used for nominal voltages of 750 V or 1500 V and offers a maximum power of 2.5 MW [54]. The inverter can either be directly connected to the medium voltage grid with a separate transformer, or in parallel to the a rectifier by means of an additional autotransformer. The system is currently used in Stuttgart (Germany) and Riyadh (Saudi-Arabia).

Summary and comparison of different WERS applications
A detailed overview of the available wayside energy recovery systems and their actual applications has been presented in the previous chapters. Here, the technologies will be briefly compared regarding power, energy and year of installation. Fig. 3 compares applications of the three storage systems based on power and energy capacity. Inverters are excluded at this point because of the theoretically infinite energy absorbency. It turns out that the systems based on EDLCs have the lowest capacities, but reach into high power classes. Flywheel solutions vary greatly in performance and energy. It can be clearly seen that systems with a high energetic capacity are mainly based on batteries. This is explained by the significantly higher energy density that goes along with strongly reduced power density. Fig. 4 shows the power and energy of different projects on a time scale. In the beginning years (from 1998 on) only EDLCs and flywheels with comparatively low performance were applied. In the last ten years, especially from 2010 onward, a rising number of battery systems and inverters with higher power and energy (batteries) commenced operation. For example, Siemens has replaced its EDLC based system, which was the most widely used worldwide, with inverters (see chapter 4.1.6). The overall trend is clearly towards systems with higher power and capacity. This suggests that not all energy-saving potentials have yet been exploited, especially in the case of high-performance light rail vehicles. The use of inverters, but also batteries, is increasing significantly, since the cost of power electronics and Li-Ion batteries has fallen sharply in recent years. EDLCs are also still being used, as, for example, illustrated by the 3.3 MW installation in Warsaw in 2016 [55]. They still offer the great advantage of working well even at low temperatures, which also makes them an interesting candidate to be used on the vehicle. Applications of flywheels have decreased significantly in the past years. The energy content of both systems is showing a slight upward trend, but does not reach the capacity of batteries or inverters by far.

Discussion
As we can see from chapter 4, there were and are numerous systems in operation, and the number is increasing. What can not be found in most cases, is a critical and independent evaluation on the actual impacts that can be achieved by using WERS. This chapter intends to fill this gap, by evaluating the achievable improvements based on demonstrations and theoretical studies. We thereby outline important issues in order to make a better consideration in the future, which especially intends to help researchers and operators. Based on all findings, a future prognosis on each technology is given in the end.

Improvements achievable with WERS based on demonstrations
The improvements achievable with WERS are stated according to the published results. No theoretical studies and considerations based on simulations are listed, but only released results of actual applications and demonstrations. The values are subject to great  [54] uncertainty, as the applied measurement methods are often not or only very inadequately explained. In addition, some publications contain information from manufacturers for which a critical and independent presentation is not necessarily guaranteed. For this reason, the credibility of the results is evaluated. Low indicates poorly presented and incomprehensible sources. Medium states more comprehensibility, but still information is missing or independence is not guaranteed. Good sources represent the measurement methodology and substantiate this by suitable graphics, so that a traceability of the results is possible. However, they can also be from the manufacturers, and thus not be 100% independent. To be excellent, it must be a detailed and comprehensible report from an independent institution. There is a lack of such investigations, though. Table 6 gives an overview of the published results that could be found within our research.

Energy efficiency
Starting with a consideration of the results with regard to energy efficiency, energy savings of 3 MWh per day were stated for the EDLC system at a metro substation in Warsaw ( [55], see 4.1.1).
There is no information about the measurement setup and the value comes directly from the manufacturer. For this reason, the credibility of the high value is rather low. The Siemens EDLC system that was tested in Cologne (see 4.1.6) showed energy savings of 15,000 kWh in one month [63]. The measurement methodology was good because the authors compared the energy consumption of the corresponding substation over one month with (August 2000) and without storage (August 2001). However, the theoretically achievable annual energy savings of 500,000 kWh with one energy storage were certainly too optimistic. Particularly in winter, the higher receptivity of the grid due to increased heating demand in the vehicles leads to poorer results for the storage systems. In Seibu [60], the operation of a 19.26 kWh EDLC storage (see 4.1.4) was evaluated. The operation with and without WERS on two consecutive days was compared. The outdoor temperature was given as an important parameter and the energy consumption was compared. Unfortunately, it is not clear where exactly this was recorded. Probably it was the energy consumption of the substation. In this case, the savings effects were rather small. In some cases, even more energy was consumed with storage than without at comparable outside temperatures.
In Ref. [81] (FW in Bielefeld, see chapter 4.2.2), [103] (Inverter in Rotterdam, see chapter 4.4.3) and [106] (Inverter in Brussels, see chapter 4.4.4), results of the European research project Ticket2Kyoto were presented. The sheets were prepared by the involved public transport companies, which suggests independence. The methodology behind the published results was described by Deveaux and Tackoen [75], which is a good and important report on the topic. However, it has to be considered that such publications, which often describe testing of quasi-prototypes in an emerging market, are not expected to be too critical or negative.
The application of a Piller flywheel system in Freiburg (see 4.2.2) has an extensive and good documentation in German language [82]. The methodology was well presented. The stated savings of around 200,000 kWh annually were based on the amount of energy released again by the WERS within a certain period of time. Unfortunately, only periods in the summer half-year were considered and scaled to a whole year. Since the receptivity of the grid is higher in winter due to the additional heating energy demand, this would lead to lower expected annual savings. This effect may be mitigated by the WERS location at the line terminal, as there are only very few vehicles simultaneously. A presentation of the effects over an entire year would therefore be desirable. The average daily energy savings of the flywheel at Los Angeles metro (see 4.2.1) were stated with 1.5 MWh [76,77]. The exact method of measuring the values and effects on the total energy consumption of the substations are not entirely clear [77]. The annual savings achieved by the WERS were estimated to be at least $99,000. Unfortunately, the calculations were not explained in detail and the investment and operating costs of the storage were not offset and remain unknown.
The Li-Ion battery installation in Philadelphia (see 4.3.1) revealed an interesting concept: the dual use of a WERS to absorb braking energy in railway networks and participate in the frequency regulation market (FRM). The average value of the daily recovered energy was between 900 kWh and 1280 kWh. It is not clear to what exactly the value of recovered energy refers, but probably it is the stored amount. Depending on the time of day, the system offered various powers to the FRM, ranging from 100 kW to 500 kW. This generated revenues of $100,000 in 6 months [111]. Further scientific studies on this approach would be of great interest. A 2 MW/76 kWh Li-Ion WERS was installed in Haijima (Japan, see 4.3.2). Energy savings of 1.4 MWh per day were reported in a comprehensible way [112]. Different voltage thresholds were tested. By optimising them, recovery effects could be significantly increased. The paper shows interesting correlations and is well presented. The dependency of savings on the outside temperature was also considered [113]. It is also clarified that the discharged energy of 750 MWh/year is not equal to the actual (substation) energy savings. For the Li-Ion storage in Okegawa, an exact analysis was carried out, including the outside temperature as important factor. It is shown how different measurement and evaluation methods change the results [113]. Daily recovered energy amounts of around 725 kWh were stated for a 1 MW/160 kWh Li-Ion WERS in the metro of Tokyo [39]. No precise information was given on the origin of this value.
An early system of a regenerative substation (1.5 MW inverter) was tested in Bilbao in 2009 [104,105]. A daily energy recovery of 1800e5000 kWh was stated. Unfortunately, the exact determination of these values was not shown and it is difficult to draw conclusions about actual savings. The aforementioned installations in Rotterdam and Brussels stated considerable energy savings, especially the latter. This is probably also the reason why the Brussels transport companies have pursued the concept in further projects [40,114]  comparability of WERS in terms of efficiency can only be achieved by measuring the energy saved in the affected substations during a comparison period. Similar outdoor conditions play a major role, since, for example, heating in winter has a very large influence on energy consumption and the receptivity of the power grid [114,115]. The pure statement about the energy absorbed and delivered by the WERS, neglecting the ambient conditions, is not meaningful. This should be taken into account in future scientific considerations. Examples of good practice can be found in publications of Suzuki et al. [112] and Iino et al. [113].

Voltage stability
Considerably fewer results could be found with regard to voltage stability. A 180 kW/0.42 kWh EDLC was used to prevent voltage drops on a line between Fujisawa and Kamakura (Japan, see 4.1.5). The achievable voltage drop compensation of 30 V is well presented [61]. Voltage drops below 490 V could be eliminated with an 700 kW/2.3 kWh EDLC system in Madrid (see 4.1.6, [63]). The information is comprehensibly presented, but comes directly from a manufacturer's brochure. A 1.9 MW/400 kWh NiMh battery system was tested in New York City (see 4.3.3). The effects regarding voltage stabilisation were proficiently investigated [94]. Underpinned by measurement results, a line voltage drop reduction by 55 V was displayed.

Peak power reduction
Regarding the reduction of power peaks, we could only find one publication. This refers to the flywheel application in Los Angeles. The explanations regarding power savings are very brief and the results difficult to reconstruct. The stated utility cost savings are divided roughly equally between energy efficiency and power demand reduction, which is about 350 kW. At this point, further reliable investigations of real applications would be desirable.

Emergency supply
The 1 MW/160 kWh Li-Ion WERS in the metro of Tokyo [39] was also used for emergency supply and was dimensioned accordingly. A 10 car train was powered to travel 2.7 km from one station to another within 10 minutes. The SOC decreased by 13% in this experiment. The functionality as an emergency power supply could also be demonstrated for the NYC NiMh battery [94]. A train was moved between two substations using the WERS. The maximum power output was 800 kW. At a distance of about 2.5 km, around 5.5% of the charge in the storage was consumed. Based on these numbers the authors derived that in the event of a power supply emergency, up to 17 trains can be brought to a station 1.2 km away.

Improvements achievable with WERS based on studies/ simulation
Numerous publications on the topic can be found in the scientific literature (see 1). To treat them all at this point goes beyond the scope of this work. For this reason, only a selection of articles published in scientific journals that consider three use cases are presented (see Table 7): energy efficiency [30], voltage stability [18] and peak power/demand charge reduction [38].

Energy efficiency
Ceraolo and Lutzemberger [30] conducted a comparative study on stationary and on-board storage systems based on Li-Ion and EDLC technology. The investigations are based on a simulation model implemented in Modelica language. The physical processes are represented by (simplified) systems of equations which are solved in each time step. For example, only the open circuit voltage (OCV) and the internal resistance are considered in the energy storage model. Such a simplification is necessary in the calculation of complex systems in order to reduce the computing time, without negatively affecting the results too much. The model was applied to a tramline in Bergamo (Italy), which is 12 km long and comprises 10 substations. Not only stationary and on-board installations were compared, but also their number, varying from 1 to 10. Based on different control methods, a cost benefit analysis (CBA) was performed. A grid connection of the stationary Li-Ion ESS both with and without DC/DC converter was considered, which requires different dimensioning. The CBA, however, was based only on the analysis of the system with converter. It turned out that from a certain number of installations (here 3) no significant energy and cost savings can be achieved. The CBA showed clear advantages for the Li-Ion systems with about half the payback time compared to EDLCs. The achievable results depend on the control mode. If energy recovery was completely stopped above a certain voltage threshold (here 900 V, called baseline control), the use of 3 Li-Ion ESS proved to be the best solution. The calculated payback time would be 3 years in that case. The difference between stationary and on-board applications was small, with slight advantages for stationary systems. With an advanced control, where the energy recovery is regulated in such a way that the voltage threshold is not exceeded by switching off directly but only reducing the energy feedback, the application of one stationary Li-Ion ESS was the best solution. This offered a payback time of only two years.

Voltage stability
The use of WERS based on EDLC was considered by Iannuzzi et al. [18]. The authors have published numerous of papers with similar problems and approaches [19,21,24,35,116e118]. The focus was on reducing voltage drops in the overhead line, which can lead to train failure in the worst case. For this purpose, a numerical calculation model was introduced with which differently fed network sections can be simulated (substation only at one side, at both sides, in the middle). Emphasis was placed on the calculation of the grid voltage via a discretised distribution of the currents/ current densities. The EDLC-based ESS, which is connected to the grid via DC/DC-converter, was represented as a limited current source. An example calculation was performed on a suburban 3000 V line in Naples (Italy). The WERS were to be dimensioned to keep voltage drops below 300 V. Different configurations were included in three stations, varying from 63.36 to 173.57 F at a rated voltage of 500 V. This resulted in an overall installed capacity of 9.6 kWh along the track and a maximal voltage drop reduction by 68% compared to the system without WERS.

Peak power reduction
An approach to calculate the effect of peak power/demand charge reduction was given by Roch-Dupr e et al. [38]. It is a subject that still lacks investigations. The authors presented the energy bill elements, which depend on consumption and power averages at different time intervals (5 or 15 min) and days/daytimes. With nearly V40 per kW, power prices are highest in the late afternoon/ evening on working days in winter. The headway dependency was investigated, too. The theory was applied to a 600 V line with 11 substations in Spain. Two ESS operation modes were investigated. Energy is discharged as often as possible in the first to reach highest efficiency. In the second, the threshold is increased to supply only in peak power phases. Storage systems based on EDLCs, varying from 5 to 25 kWh of energy and 500 to 4000 kW of power, were investigated. A combination of 2500 kW and 17.5 kWh showed to be the most profitable solution. A cost reduction in the energy bill of 15% per year was stated. The reduction in energy consumption made up the biggest part, with around 82%. Nevertheless, it was also shown that demand charge reduction can significantly contribute to reduce total costs. A net present value (NPV) calculation was performed. The result was negative only in less favorable conditions (presuming high investments costs of around V700,000), but positive for smaller invest costs.

Evaluating the impact of WERS
WERS are used for various reasons (see 2), of which the most important is the recovery of braking energy and thus the increase in efficiency. Many publications, especially those based on demonstrations (5.1), consider only the technical effects without taking an overall economic view. This requires the execution of an investment calculation. From a competitive point of view, however, it is understandable that such sensitive data in an evolving market can hardly be found in public. The results of the Ticket2Kyoto project, for example, are an exception, where cost numbers of different applications were stated [75]. Unfortunately, the investment costs are not too detailed here either, but the different approaches and the results evaluation are described. It serves as an important publication of the actual use of WERS.
The basic methodology for the economic evaluation of such systems could be found in several studies [30,34]. However, this is based on fictitious systems and cost assumptions. A larger number of (independent) studies on the actual economic effect of WERS, based on real demonstrations and cost figures, are desirable for the future. When quantifying savings, care must be taken to use the correct methodology. The author's thoughts on that, in relation to energy savings, have been explained in the note at the end of chapter 5.1.1.
The exact quantification of the effects with regard to increasing the voltage quality, power peak reduction and emergency supply are an important field of research in the future, as there are little publications to be found. Concerning voltage stability, a large number of technical papers could be found [17,18,24,119] which state the positive effects. However, no economic valuation that monetises those could be found. An approach could be to evaluate the statistical drop in vehicle failures/errors due to a less fluctuating voltage when applying a WERS. Roch-Dupr e et al. [38] stated an approach to quantify the effects of peak power/demand charge reduction. The economic effect of the use as emergency power supply is difficult to quantify. It is highly dependent on local conditions and regulations. The application plays an important role in well developed regions/countries which are threatened by natural disasters, such as Japan, since power outages can occur more frequently than in comparably developed regions such as Europe.

Summary comparison and future forecast
The fundamental principles and properties of the four technologies have been described in chapter 3. They are to be compared EDLC -focus on demand charge reduction with ESS -sample calculation for a 600 V line in Spain -15% annual savings of electricity bill with ESS -demand charge reduction contributes to 18% of savings -positive business case for intermediate scenario [38] again briefly here, including not only the theoretical basics but also results from the demonstrations (see 4,5.1) and publications (see 5.2). A 100% reliable cost estimate is not possible as we could find no, or at least not enough, reliable and published information on the actual costs of the systems (see 3,5.3). For this reason, only the rough trends are included in the evaluation. Table 8 summarises the most important advantages and drawbacks of each technology. In addition, a future prognosis is given. EDLCs have been used for WERS since the turn of the millennium (see 4.5). They are characterised by high cycle stability and performance [1], which made them, especially at that time, an excellent candidate. Siemens in particular marketed the Sitras SES, a system that was frequently used and also achieved good results in the beginning years (see 4.1.6). The large number of publications on the subject proves that EDLCs had the greatest relevance in this application the last 15 years (see 1). Disadvantages are the low energy density and the associated high space requirement, which is rather problematic for onboard applications. The biggest disadvantage that still exists today, however, is the high investment costs, especially since the prices for cells remain high [1,41]. Unquotable conversations with various PTOs emerged, that the high initial costs and unclear benefits are a major deterrent. In the worst case, technical defects and problems occur that stand in the way of the economic efficiency of the system [62,75]. Probably for this reason, Siemens has meanwhile discontinued sales of the Sitras SES and is now marketing WERS based on inverters with IGBT technology [54]. However, the large new installations in Warsaw (see 4.1.1) and Hong Kong (see 4.1.4) show that this technology is not yet at its end. It will continue to be used in the future, which will certainly be strongly linked to further price developments.
Flywheels are characterised by high power densities and long service lives, when properly maintained. These two advantages and the point that it is a proven mechanical technology led to the fact that these were particularly used in the first applications (see 4.5). This was also due to a lack of alternatives at that time. A variety of applications, stating good results in terms of energy recovery [81,82], can be found from the German company Piller (see 4.2.2). Disadvantages of the systems are the high space requirement, the maintenance effort, as well as noise and safety concerns [103]. The trend in recent years seems to deviate from the use of flywheels (see Fig. 4). The most important reason for this is the very high (specific) investment costs, especially in terms of energy [42,43]. Since other technologies, such as batteries and inverters, now offer more attractive solutions regarding cost and performance, no further relevant flywheel WERS applications are expected in the future.
With growing importance of the lithium-ion battery in automotive applications and the ongoing improvements in energy, performance and cost, it is becoming more and more important for WERS applications. Advantages are the comparatively high energy density and the associated low investment costs for systems with large energy capacity. This makes the technology especially interesting for a double use, e.g. the participation in the FRM [111] or as emergency power supply [39]. The maintenance effort is classified as low. The installation of used car batteries as second life storage can significantly reduce life cycle costs [39]. Disadvantages are the low power density and the shorter cyclic life due to the basic principle of electro-chemical energy storages, especially when full charge/discharge cycles with high currents are applied [120]. For compensation, the storage systems must be dimensioned with a large capacity (see Fig. 3). However, new battery chemistries make interesting system designs possible. Cells with an anode made of lithium-titanate-oxide (LTO) are characterised by lower energy density, but significantly higher power density and cycle stability, even at high currents and lower temperatures [121]. This enables to make battery based WERS significantly smaller, which is also relevant for an installation on the vehicle. Further developments, such as the use of niobium-titanate-oxide in the anode, could become interesting candidates in the future [122]. Since the NiMh battery has been replaced by Li-Ion in almost all areas and this trend is also transferable to WERS, it will not be treated further at this point.
The use of inverter-based WERS has increased significantly in recent years (see 4). The advantage of those systems is the theoretically infinite energy capacity. They also offer high efficiency and performance. On the other hand, there are the large space requirements and the rather big investment costs [123,124]. In addition, the feeding of electrical energy into public mediumvoltage networks is subject to country-specific conditions that could stand in the way of an application [75]. Such and many other issues were addressed in the European project Eliptic [40]. This threat is avoided if the inverters feed the energy back into a grid which itself belongs to the respective PTO. Such a structure can be found, for example, in Brussels [123]. Currently, a large number of new installations of the systems can be seen, for example Alstom's Hesop substations (124 as of [101]). Installing new substations bidirectionally from the outset seems to be a sensible trend and a growing number of future installations is expected.

Conclusion
The various projects have shown that the systems can be a good way to increase efficiency in local electric transport. Over the years, different technologies have been tested and valuable experience could be gathered. The average claimed energy savings are about 10e15%. However, the investigations depend heavily on secondary conditions like outside temperature and correspondingly heating demand. Furthermore, it is mostly unclear how the indicated values Table 8 Advantages, drawbacks and future prognosis of each WERS technology.

WERStechnology Advantages
Drawbacks Future prognosis EDLC High power density and cycle stability, high efficiency [120], acceptable lifetime (if high temperatures are avoided) Small energy density, increased space demand, high investment costs [1], technical problems [62,75] Technology will still be used, especially if prices are falling FW Very high lifetime if properly maintained, high power density [42] Small energy density, large space demand, maintenance effort, noise and safety problems [103], high invest costs [42] No further application, as other technologies are more attractive BAT (Li-Ion) High energy density [120], comparably low invest costs for high energies [1], combination with other services [111], low maintenance effort, use of used batteries as 2nd life application [39] Lowest lifetime expectancy (mainly due to cyclic aging in many high performance cycles), small power density [120] leads to big storage designs Increasing amount of applications, new chemistries enable new system designs [121,122]

INV
(Theoretically) no limitation regarding energy, very high power and efficiency Large/Increased space demand [32] and investment costs [123,124], legal issues when feeding energy back to a public MV grid can be a threat (country specific) [75] Utilisation will increase, installations of new substations with bidirectional converters were determined. It has also been shown that the application of wayside energy recovery systems increases voltage stability, reduces the power demand charge and that they can serve as emergency supply. Translating these positive effects into economic parameters is an important area of future research, in order to enable a meaningful economic evaluation of the systems. Whereas flywheels and EDLCs were mainly installed at the beginning (around the millennium), battery storage systems and regenerative substations have become increasingly established in recent years (from 2010 on). This is due to the fact, that the cost of power electronics and Li-Ion batteries has fallen sharply. These systems will displace the application of flywheels in the future. The scientific literature shows, that the positioning and sizing of the ESS, in order to recover as large an amount of energy as possible and to provide additional network services, are crucial questions. The optimal operating strategy is an integral part of many studies. Potential new consumers, such as charging stations for electric vehicles, are also changing the situation in the grid. Manufacturers offer assistance in planning for this purpose, but scientific work should stay concerned with the topic, too. Detailed measurements and correspondingly evaluated simulations are essential to determine the best locations in the respective grid. A more detailed examination of the energy storage, especially for high power Li-Ion systems, is also becoming more important and serves as a material for further scientific work.