- Email: [email protected]
A new generation of power supplies for pulsed loads
Fusion Engineering and Design xxx (xxxx) xxx–xxx
Contents lists available at ScienceDirect
Fusion Engineering and Design journal homepage: www.elsevier.com/locate/fusengdes
A new generation of power supplies for pulsed loads ⁎
Alessandro Lampasia, Sandro Tenconib, Giuseppe Taddiab, , Filippo Gherdovichb, Luigi Rinaldib a b
National Agency for New Technologies, Energy and Sustainable Economic Development (ENEA), Frascati, Italy OCEM Power Electronics, Valsamoggia, Italy
A R T I C LE I N FO
A B S T R A C T
Keywords: Pulsed power supply Pulsed load energy storage Energy recovery Supercapacitor Inverter
Pulsed power supply systems are employed in many fusion projects and in other scientific applications. A novel power supply approach and a first prototype unit were specifically developed to feed pulsed loads (low duty cycle), as the resistive or superconducting coils used to produce high magnetic fields for some seconds or longer. Thanks to the integrated energy storage, the high pulsed power required for the load is not drawn directly from the electrical grid. In fact, a huge amount of energy can be stored at low power (even through a singlephase 10 A plug, as normally used for household appliance). A significant fraction of the energy stored in the load can be recovered for successive operations. Without such energy storage, all the power supply devices and the upstream chain must be oversized and installed only in locations provided with adequate power. The first power supply unit (2 kA) was able to replace a previous system with comparable performances, but the old system occupies an extremely larger volume and needs a 20-kV connection at the input. Moreover, a specific solution was implemented to achieve a very fast (600 kW) energization of the magnets. The energy storage capability is intrinsically scalable: the setup can be rearranged and the storable energy can be updated even after the installation (for example, to increase the pulse duration).
1. Introduction
energy sources and the expected requirements of electric mobility led to a development and spread of new technologies, as supercapacitors (SCs), lithium batteries and lithium ion capacitors (LICs). In particular, SCs can feature exceptional capacitance values with respect to conventional capacitors, resulting in high power density and fast charge and discharge times. Moreover, the expected lifetime presently exceeds 106 charge/discharge cycles, namely more than 2 order of magnitude higher than best lifetime of lithium batteries [2]. SCs are already in use in transportation systems to sustain power peaks and regenerative electrical braking [3]. They appear as the most promising ESS to manage high power pulses for short times, as required in many tokamak applications. These considerations led to a new PS concept [2], summarized in Fig. 1, specifically developed to feed pulsed loads (low duty cycle). A fraction of the power injected into the load can be recovered at the end of the pulse. This fraction may be relevant in case of superconducting coils, as used in magnetic-confinement nuclear fusion. Last but not least, internal power electronics and upstream infrastructures can be designed for the average power. This concept was recently implemented by developing a compact SC-based PS (SCPS), able to generate arbitrary currents up to 2 kA, in spite of a very low input power (≈1 kW) [4].
Many industrial and research facilities (including nuclear fusion and other plasma applications) require high “pulsed” powers, namely having a typical duty cycles lower than 10% [1,2] or even much less. The considered applications include one or more of the following characteristics: high DC current (up to tens of kiloamperes), high DC voltage (some kilovolts), high peak power (up to hundreds of megawatts), low duty cycle, moderate average power, long useful life (both in terms of years of operation and of number of cycles). The usual way to sustain the short-time peak demands consists in oversizing the power supplies (PSs) and the upstream electric distribution systems, implying high system and operational costs. Since the power sources with the suitable characteristics are often available at times and in places different from the needed ones, a power transfer and time shifting are required. The power delivery requires high-voltage (HV) transmission and medium-voltage (MV) distribution lines, and large intermediate substations, often including variable reactive power compensation at MV. Energy storage systems (ESSs) appear as a viable alternative to shift the power over time. While the classical solutions were mostly based on pumped water and rotating flywheels, the increasing use of renewable
⁎
Corresponding author. E-mail address: [email protected] (G. Taddia).
https://doi.org/10.1016/j.fusengdes.2019.03.066 Received 10 October 2018; Received in revised form 26 February 2019; Accepted 11 March 2019 0920-3796/ © 2019 Elsevier B.V. All rights reserved.
Please cite this article as: Alessandro Lampasi, et al., Fusion Engineering and Design, https://doi.org/10.1016/j.fusengdes.2019.03.066
Fusion Engineering and Design xxx (xxxx) xxx–xxx
A. Lampasi, et al.
Fig. 1. Basic concept (adapted from what proposed in [2]) and actual implementation in the first SCPS unit of the input (low power) and output (high power) connections.
2. Principles for the design of the SC bank Fig. 2 shows a simplified scheme of the SCPS, in the already tested version at two quadrants [4]. The load current is regulated by an Hbridge converter (similar to an inverter) because it is probably the simplest and most reliable configuration. However, other converter topologies could be in principle adopted. The SCPS is able to generate arbitrary currents, provided that the instantaneous energies and voltages are available in the SC bank (SCB) when required. The SCB has a function that is only partially similar to a DC link in a standard H-bridge. First, the input charger is not conceived to replicate the H-bridge power flow in order to keep the voltage on the SCB within a prefixed ripple. Moreover, the SCs do not simply behave as high-capacitance capacitors [2]. The equivalent capacitance decreases with the frequency and with the charge level, while the equivalent series resistance (ESR) is variable with the rate of energy extraction (in practice, with the current shape). A specific DC filter is interposed (see Fig. 2) to limit the AC components drawn by the SCB. In this way, the current flowing from the SCB to the H-bridge tend to follow the average load current that is relatively slow, while the short-time peaks and the ramps, are provided by the filter capacitances. In fact, the derivative of the current supplied by the SCs (red curve) in Fig. 3 is lower than the current required by the load. For short-time peaks or fast ramps (> 75 A/ms), most of the current is provided by the capacitors in the DC filter. On the other hand, like in any other H-bridge, the voltage ripple on the SCB is strongly limited and the power flowing from it to the output is rather constant. Since any SCPS configuration requires several SC modules in the SCB, also the dynamic balance of the module currents must be controlled by proper connections and copper bars and by placing the modules according to their experimental characteristics. The layout of the presented in Section 3 was designed and verified to equalize the
Fig. 3. Effect of the DC filter and SC module current sharing during a rising ramp (load current from 0 to 2 kA in 50 ms). The only way to monitor these characteristics on the actual configuration was to insert flexible Rogowski coils, that are sensitive only to the current derivative.
stray inductances seen by each module. Fig. 3 presents the excellent current sharing achieved during a ramp in the most critical configuration (4 modules in parallel). Only the measurements of the currents flowing in the two parallels of two SC modules were accessible in the final cabinet by inserting flexible Rogowski coils. The voltage balancing of SC modules has less influence on the dynamic current. However, the modules were coupled to minimize the unbalance in the series connection. It is also interesting to stress that the manufacturers declares that modules characteristics intrinsically tend to become more and more similar during the SC lifetime. Since the power flow in the H-bridge can be negative the energy stored in the load can be recovered in the SCB reducing the energy needs. However, the SCB available energy must be significantly higher Fig. 2. Simplified scheme of the first SCPS unit (scalable to larger systems). The parallel/series (“P”/”S”) connections can be modified according to the desired configuration. One output terminal can be grounded or both terminals can be floating. The reported load is an example referred to a copper coil connected to the unit described in Section 3.
2
Fusion Engineering and Design xxx (xxxx) xxx–xxx
A. Lampasi, et al.
than the energy stored in the load inductance to compensate the parasitic effects. The total energy ESC to be stored in the SCB can be estimated by the formula
ESC ≥ EL + ER + Epass + EH + Eage
Table 1 Main features of the first SCPS unit.
(1)
based on sum of the five energy contributions described in the following. The energy stored in the load EL, in practice the energy stored in the inductance LIcoil2/2, depends only on the maximum load current Icoil during the scenario and does not depend on the total scenario time Ttot and ramp durations. ER denotes the energy dissipated in the parasitic elements, especially in the load resistance RL and in the SC equivalent series resistance (ESR). The ER effect can be simplified as ReqIrms2Ttot, where Req is the total equivalent resistance and Irms is the usual root mean square value integrated over the scenario. As a rule of thumb, for RL≤10 mΩ, Req≈2RL [4]. The Epass contribution is significant in nuclear fusion applications, as it includes the power lost in the passive elements as the vessel and the walls (including the other coupled coils) and in the plasma. In principle, the effect of this contribution could be incorporated in Req. EH is the energy necessary to operate the H-bridge or any other power electronics used to control the circuit. For example, a standard H-bridge converter is able to regulate the load current only if the instantaneous SCB voltage vSCB(t) exceeds the instantaneous circuit drop Reqicoil(t). A safety assumption is EH=C(ReqIcoil)2/2. Eage is the safety margin assumed to preserve the SCB from premature ageing. Even though the SCs can be totally discharged, the manufacturers define the lifecycle, the capacitance and the ESR for cycles in the interval Vrated/2÷Vrated (or equivalently in VSCB/2VSCB), corresponding to Eage=ESC/4. The residual voltage VSCB/2 is normally adequate also to cover the EH margin. The voltage of the SCB is related to ESCB=CVSCB2/2, but it must be also sufficiently high to sustain the current rise/fall ramps. In typical cases, VSCB (and the H-bridge voltage) must be increased to sustain the ramps rather than to store enough energy. However, VSCB cannot be accurately optimized, especially at low values, due to the discrete voltage levels of the SC modules available on the market, and the ramp performances also depend on the adopted control strategy and parameters. Typical waveform necessary for the experiments consists in a rectangular (trapezoidal) pulse with fast rise/fall times and a long flat-top phase [1,4]. In this case, the flat-top current IFT and duration TFT can be used to estimate the energy components EL=LIFT2/2 (independent of TFT) and ER≈ReqIFT2TFT. Since the load voltage at the flat-top is minimal, VSCB is even more determined by the pulse rise/fall time.
Characteristic
Value
Maximum output current Peak performances (ramps) Average flat-top performances Best duty cycle Maximum input power
2 kA ± 300 V (≈600 kW) ± 95 V (≈200 kW) ≈10 s/300s 1.8 kW (at 230 Vac)
Table 1, was installed to feed the external coils of the novel spherical tokamak PROTO-SPHERA [5], located in the ENEA laboratories in Frascati. Even though the SCPS was conceived for this application, its embedded SCB was designed and implemented for longer operations (even more than 10 s), also because the SCPS topology and control were selected to be as modular as possible. The ESS of the SCPS consists of four commercial SC modules (Maxwell BMOD0165P048C01), each rated 165 F and 48 V, that could be connected all in parallel (VSCB = 48 V, C = 660 F) or two in parallel and two in series (VSCB = 96 V, C = 165 F) according to load requirements. The configuration can be modified changing the internal bars as sketched in Fig. 2. The SCPS can operate, if necessary, with all the four modules in series with minor hardware modifications. In all the three cases, the maximum total stored energy is about 760 kJ (≈210 Wh) and, considering the Eage margin, the usable energy is about 75% of that. These configurations are able to generate 2 kA with a power of 200 kW and for a time depending on the load request, in any case longer than the target time of 2 s. In order to obtain a movable PS that could be used in every site, the input power is very low and delivered at low voltage. In particular, Fig. 1 shows that the SCPS can be connected to any standard 230 V, single-phase, 50 Hz socket, rated for 10 A, available in every European building. A specific input PS (charger), capable to limit the reactive power and the harmonic pollution, charges the SCB up to the required voltage. The charger can be regulated in voltage, current or power. The charging option (constant power or constant current) is chosen according to the SCB configuration and the desired charging time. The charger can charge a completely discharged SCB in maximum 30 min and can recharge it between two operations in less than 10 min (depending also on the energy dissipated during the previous operation). In Fig. 1 it is also visible a selector close to the input cable, that allows to choose to supply the control and auxiliary systems with a separated line with high reliability, such as an uninterruptible PS (UPS). The movability is also facilitated by the absence of water pipes for cooling. As the SC modules have relatively high thermal resistances and capacitances in air cooling (thermal time constants longer than 1 h), the internal layout was designed considering the specified duty cycle and the air temperature distribution. Nevertheless, the temperature of each module is continuously monitored.
3. Experiences with first SCPS unit The first SCPS unit, shown in Fig. 4(a) and briefly specified in
Fig. 4. Comparison between the SCPS (a) and a previous setup (b) for the same current and duty cycle reported in Table 1. 3
Fusion Engineering and Design xxx (xxxx) xxx–xxx
A. Lampasi, et al.
Fig. 5. Arbitrary waveform, defined by numerical values in the touch screen (a), generated by the SCPS on the load in Fig. 2 (500 μH and 5.5 mΩ), with a detailed analysis of a fast-descending ramp (b).
derivative of 200 A/ms (reported in Fig. 5(b)) after the flat-top at 2 kA is compliant with those necessary in tokamak central solenoids and poloidal coils for plasma breakdown. Fig. 4 compares the SCPS with the previous solution used for the PROTO-SPHERA coils [5], producing the same current (2 kA) for a shorter time (≤1.5 s) with a duty cycle of 1–2 s every 600 s (it is fair to report that the available voltage was higher, ≈350 V). Even though the net energy necessary to supply the load is comparable to a household appliance (< 400 kJ, < 0.1 kW h), the old PS was connected to a 20-kV line to draw the peak power (> 3 MVA). The comparison in terms of total dimensions is summarized in Fig. 4. The SCPS volume is 1.4 m3. The total volume in Fig. 4(b) can be only approximated (for example, since the cubicle containing the circuit breakers and safety switches is used also by another converter, only half of the volume is included in the calculation), but a good estimation is ≈10 m3. Therefore, the SCPS reduced the volume by a factor higher than 7, even getting rid of the 20-kV and 324-V lines and without
The SCPS local control unit (LCU) includes a human machine interface (HMI) on Panel PC (the black box in the right of Fig. 4(a)) with touch screen (see Fig. 5(a)) where the data of the desired waveform can be set. Operation can be started from the LCU or in remote mode (triggered by external signals). Fig. 5(b) shows a waveform measured during a test on a copper dummy load in typical settings (Icoil = 2 kA and Ttot = 2 s). Such waveform is useful to stress two interesting SCPS capabilities: the recovery of the load energy and the high current derivatives. As regards the recovery, the SC charge lost during the rising ramps is partially recovered during the falling ramps (increasing of the SCB voltage in Fig. 5(b)). Even at the end of a 10 s scenario, the observed losses do not exceed 60% in voltage and 80% in energy. Therefore, the SC state-of-charge is usually above Vrated/2. As regards the current derivatives, a specific solution was implemented to achieve 300 V (600 kW) for very fast ramps (< 10 ms) and fast energization of the magnets. In particular, the negative current 4
Fusion Engineering and Design xxx (xxxx) xxx–xxx
A. Lampasi, et al.
requiring a fixed installation with many internal and external connections. The SCPS current ripple is kept within ± 10 A at the flat-top, extremely lower than the old case based on a thyristor rectifier, because the SCPS PWM is controlled at 1.2 kHz, with an equivalent modulation frequency of 2.4 kHz on the load. The high switching frequency also reduced the acoustic noise in the experimental hall. Even in the reduced space, the SCPS safety is preserved in case of a severe fault, as a short-circuit at the output or in a bridge diode. The IGBTs can withstand the current in both cases, because the bridge itself can automatically extinguish fault currents up to 4.8 kA in less than 10 μs without the intervention of the control system. Moreover, also the SC modules can survive after a short circuit at their terminals. Another aspect to be presented is the need of two high-energy resistors that can discharge the SCB in few minutes. They are normally connected during stand-by periods (“Sleep” mode in Fig. 5(a)) to partially discharge the SCB to VSCB/2, as long periods at full voltage or deep discharges could reduce the SC lifetime. The resistors must be used also before accessing the SCPS cabinet (for instance, for maintenance). This safety procedure, necessary to manage the high energy stored in the SCB even after the SCPS disconnection, is balanced by the removal of the MV procedures. All these relevant benefits could be extended to similar cases at different currents, voltages, powers and energies. In fact, scalability and possible upgrading are intrinsic in this technology.
reused. Moreover, the external grid is not affected by nonlinear power converters introducing harmonics and reactive power. This result is relevant, especially when compared with the achievable performances and with the characteristics of previous systems used for the same purposes. Since SCs can feature very long lifetimes and the SCPS can be optimized to preserve them, the SCB is expected to operate for more than 1 million of charge/discharge cycles. The SCPS approach is particularly suitable when applied to nuclear fusion experiments, inductive loads and fixed duty cycles. In particular, the energy recovered from superconducting coils may be significant. Moreover, a specific solution was implemented to achieve very fast ramps on the magnets. Finally, the SCPS approach is intrinsically modular. This is useful to reduce the initial and maintenance costs, to rearrange and update the systems even after the installation, but also to apply this approach to larger facilities as large tokamaks. By series and parallel connections of the SC modules or of the entire SCPS, the SCPS can be extended to higher voltages, higher currents, longer scenarios. In any case, the input power demand and the impact on the external grid are limited with respect to the output performances. References [1] A. Lampasi, S. Minucci, Survey of electric power supplies used in nuclear fusion experiments, 17th IEEE Int. Conf. Environ. Elect. Eng. (EEEIC) (2017). [2] A. Lampasi, et al., ETHICAL: a modular supercapacitor-based power amplifier for high-current arbitrary generation, 16th IEEE Int. Conf. Environ. Elect. Eng. (EEEIC) (2016). [3] A.L. Allegre, et al., Energy storage system with supercapacitor for an innovative subway, IEEE Trans. Ind. Electron. 57 (December (12)) (2010) 4001–4012. [4] A. Lampasi, et al., Compact power supply with integrated energy storage and recovery capabilities for arbitrary currents up to 2 kA, IEEE Trans. Plasma Sci. 46 (October 10) (2018) 3393–3400. [5] A. Lampasi, et al., Progress of the plasma centerpost for the PROTO-SPHERA spherical Tokamak, Energies 9 (7) (2016) 508.
4. Conclusions and future prospects This paper discusses the design issues and the potential benefits of a system able to store a large amount of energy at low power and to deliver it at high power when necessary. Moving from this idea, a first compact and movable unit was developed to generate for some seconds arbitrary currents up to 2 kA, requiring a limited and tunable input power, even in total absence of sources with sufficient power. A fraction of the power delivered to the load can be recovered to be
5
Contents lists available at ScienceDirect
Fusion Engineering and Design journal homepage: www.elsevier.com/locate/fusengdes
A new generation of power supplies for pulsed loads ⁎
Alessandro Lampasia, Sandro Tenconib, Giuseppe Taddiab, , Filippo Gherdovichb, Luigi Rinaldib a b
National Agency for New Technologies, Energy and Sustainable Economic Development (ENEA), Frascati, Italy OCEM Power Electronics, Valsamoggia, Italy
A R T I C LE I N FO
A B S T R A C T
Keywords: Pulsed power supply Pulsed load energy storage Energy recovery Supercapacitor Inverter
Pulsed power supply systems are employed in many fusion projects and in other scientific applications. A novel power supply approach and a first prototype unit were specifically developed to feed pulsed loads (low duty cycle), as the resistive or superconducting coils used to produce high magnetic fields for some seconds or longer. Thanks to the integrated energy storage, the high pulsed power required for the load is not drawn directly from the electrical grid. In fact, a huge amount of energy can be stored at low power (even through a singlephase 10 A plug, as normally used for household appliance). A significant fraction of the energy stored in the load can be recovered for successive operations. Without such energy storage, all the power supply devices and the upstream chain must be oversized and installed only in locations provided with adequate power. The first power supply unit (2 kA) was able to replace a previous system with comparable performances, but the old system occupies an extremely larger volume and needs a 20-kV connection at the input. Moreover, a specific solution was implemented to achieve a very fast (600 kW) energization of the magnets. The energy storage capability is intrinsically scalable: the setup can be rearranged and the storable energy can be updated even after the installation (for example, to increase the pulse duration).
1. Introduction
energy sources and the expected requirements of electric mobility led to a development and spread of new technologies, as supercapacitors (SCs), lithium batteries and lithium ion capacitors (LICs). In particular, SCs can feature exceptional capacitance values with respect to conventional capacitors, resulting in high power density and fast charge and discharge times. Moreover, the expected lifetime presently exceeds 106 charge/discharge cycles, namely more than 2 order of magnitude higher than best lifetime of lithium batteries [2]. SCs are already in use in transportation systems to sustain power peaks and regenerative electrical braking [3]. They appear as the most promising ESS to manage high power pulses for short times, as required in many tokamak applications. These considerations led to a new PS concept [2], summarized in Fig. 1, specifically developed to feed pulsed loads (low duty cycle). A fraction of the power injected into the load can be recovered at the end of the pulse. This fraction may be relevant in case of superconducting coils, as used in magnetic-confinement nuclear fusion. Last but not least, internal power electronics and upstream infrastructures can be designed for the average power. This concept was recently implemented by developing a compact SC-based PS (SCPS), able to generate arbitrary currents up to 2 kA, in spite of a very low input power (≈1 kW) [4].
Many industrial and research facilities (including nuclear fusion and other plasma applications) require high “pulsed” powers, namely having a typical duty cycles lower than 10% [1,2] or even much less. The considered applications include one or more of the following characteristics: high DC current (up to tens of kiloamperes), high DC voltage (some kilovolts), high peak power (up to hundreds of megawatts), low duty cycle, moderate average power, long useful life (both in terms of years of operation and of number of cycles). The usual way to sustain the short-time peak demands consists in oversizing the power supplies (PSs) and the upstream electric distribution systems, implying high system and operational costs. Since the power sources with the suitable characteristics are often available at times and in places different from the needed ones, a power transfer and time shifting are required. The power delivery requires high-voltage (HV) transmission and medium-voltage (MV) distribution lines, and large intermediate substations, often including variable reactive power compensation at MV. Energy storage systems (ESSs) appear as a viable alternative to shift the power over time. While the classical solutions were mostly based on pumped water and rotating flywheels, the increasing use of renewable
⁎
Corresponding author. E-mail address: [email protected] (G. Taddia).
https://doi.org/10.1016/j.fusengdes.2019.03.066 Received 10 October 2018; Received in revised form 26 February 2019; Accepted 11 March 2019 0920-3796/ © 2019 Elsevier B.V. All rights reserved.
Please cite this article as: Alessandro Lampasi, et al., Fusion Engineering and Design, https://doi.org/10.1016/j.fusengdes.2019.03.066
Fusion Engineering and Design xxx (xxxx) xxx–xxx
A. Lampasi, et al.
Fig. 1. Basic concept (adapted from what proposed in [2]) and actual implementation in the first SCPS unit of the input (low power) and output (high power) connections.
2. Principles for the design of the SC bank Fig. 2 shows a simplified scheme of the SCPS, in the already tested version at two quadrants [4]. The load current is regulated by an Hbridge converter (similar to an inverter) because it is probably the simplest and most reliable configuration. However, other converter topologies could be in principle adopted. The SCPS is able to generate arbitrary currents, provided that the instantaneous energies and voltages are available in the SC bank (SCB) when required. The SCB has a function that is only partially similar to a DC link in a standard H-bridge. First, the input charger is not conceived to replicate the H-bridge power flow in order to keep the voltage on the SCB within a prefixed ripple. Moreover, the SCs do not simply behave as high-capacitance capacitors [2]. The equivalent capacitance decreases with the frequency and with the charge level, while the equivalent series resistance (ESR) is variable with the rate of energy extraction (in practice, with the current shape). A specific DC filter is interposed (see Fig. 2) to limit the AC components drawn by the SCB. In this way, the current flowing from the SCB to the H-bridge tend to follow the average load current that is relatively slow, while the short-time peaks and the ramps, are provided by the filter capacitances. In fact, the derivative of the current supplied by the SCs (red curve) in Fig. 3 is lower than the current required by the load. For short-time peaks or fast ramps (> 75 A/ms), most of the current is provided by the capacitors in the DC filter. On the other hand, like in any other H-bridge, the voltage ripple on the SCB is strongly limited and the power flowing from it to the output is rather constant. Since any SCPS configuration requires several SC modules in the SCB, also the dynamic balance of the module currents must be controlled by proper connections and copper bars and by placing the modules according to their experimental characteristics. The layout of the presented in Section 3 was designed and verified to equalize the
Fig. 3. Effect of the DC filter and SC module current sharing during a rising ramp (load current from 0 to 2 kA in 50 ms). The only way to monitor these characteristics on the actual configuration was to insert flexible Rogowski coils, that are sensitive only to the current derivative.
stray inductances seen by each module. Fig. 3 presents the excellent current sharing achieved during a ramp in the most critical configuration (4 modules in parallel). Only the measurements of the currents flowing in the two parallels of two SC modules were accessible in the final cabinet by inserting flexible Rogowski coils. The voltage balancing of SC modules has less influence on the dynamic current. However, the modules were coupled to minimize the unbalance in the series connection. It is also interesting to stress that the manufacturers declares that modules characteristics intrinsically tend to become more and more similar during the SC lifetime. Since the power flow in the H-bridge can be negative the energy stored in the load can be recovered in the SCB reducing the energy needs. However, the SCB available energy must be significantly higher Fig. 2. Simplified scheme of the first SCPS unit (scalable to larger systems). The parallel/series (“P”/”S”) connections can be modified according to the desired configuration. One output terminal can be grounded or both terminals can be floating. The reported load is an example referred to a copper coil connected to the unit described in Section 3.
2
Fusion Engineering and Design xxx (xxxx) xxx–xxx
A. Lampasi, et al.
than the energy stored in the load inductance to compensate the parasitic effects. The total energy ESC to be stored in the SCB can be estimated by the formula
ESC ≥ EL + ER + Epass + EH + Eage
Table 1 Main features of the first SCPS unit.
(1)
based on sum of the five energy contributions described in the following. The energy stored in the load EL, in practice the energy stored in the inductance LIcoil2/2, depends only on the maximum load current Icoil during the scenario and does not depend on the total scenario time Ttot and ramp durations. ER denotes the energy dissipated in the parasitic elements, especially in the load resistance RL and in the SC equivalent series resistance (ESR). The ER effect can be simplified as ReqIrms2Ttot, where Req is the total equivalent resistance and Irms is the usual root mean square value integrated over the scenario. As a rule of thumb, for RL≤10 mΩ, Req≈2RL [4]. The Epass contribution is significant in nuclear fusion applications, as it includes the power lost in the passive elements as the vessel and the walls (including the other coupled coils) and in the plasma. In principle, the effect of this contribution could be incorporated in Req. EH is the energy necessary to operate the H-bridge or any other power electronics used to control the circuit. For example, a standard H-bridge converter is able to regulate the load current only if the instantaneous SCB voltage vSCB(t) exceeds the instantaneous circuit drop Reqicoil(t). A safety assumption is EH=C(ReqIcoil)2/2. Eage is the safety margin assumed to preserve the SCB from premature ageing. Even though the SCs can be totally discharged, the manufacturers define the lifecycle, the capacitance and the ESR for cycles in the interval Vrated/2÷Vrated (or equivalently in VSCB/2VSCB), corresponding to Eage=ESC/4. The residual voltage VSCB/2 is normally adequate also to cover the EH margin. The voltage of the SCB is related to ESCB=CVSCB2/2, but it must be also sufficiently high to sustain the current rise/fall ramps. In typical cases, VSCB (and the H-bridge voltage) must be increased to sustain the ramps rather than to store enough energy. However, VSCB cannot be accurately optimized, especially at low values, due to the discrete voltage levels of the SC modules available on the market, and the ramp performances also depend on the adopted control strategy and parameters. Typical waveform necessary for the experiments consists in a rectangular (trapezoidal) pulse with fast rise/fall times and a long flat-top phase [1,4]. In this case, the flat-top current IFT and duration TFT can be used to estimate the energy components EL=LIFT2/2 (independent of TFT) and ER≈ReqIFT2TFT. Since the load voltage at the flat-top is minimal, VSCB is even more determined by the pulse rise/fall time.
Characteristic
Value
Maximum output current Peak performances (ramps) Average flat-top performances Best duty cycle Maximum input power
2 kA ± 300 V (≈600 kW) ± 95 V (≈200 kW) ≈10 s/300s 1.8 kW (at 230 Vac)
Table 1, was installed to feed the external coils of the novel spherical tokamak PROTO-SPHERA [5], located in the ENEA laboratories in Frascati. Even though the SCPS was conceived for this application, its embedded SCB was designed and implemented for longer operations (even more than 10 s), also because the SCPS topology and control were selected to be as modular as possible. The ESS of the SCPS consists of four commercial SC modules (Maxwell BMOD0165P048C01), each rated 165 F and 48 V, that could be connected all in parallel (VSCB = 48 V, C = 660 F) or two in parallel and two in series (VSCB = 96 V, C = 165 F) according to load requirements. The configuration can be modified changing the internal bars as sketched in Fig. 2. The SCPS can operate, if necessary, with all the four modules in series with minor hardware modifications. In all the three cases, the maximum total stored energy is about 760 kJ (≈210 Wh) and, considering the Eage margin, the usable energy is about 75% of that. These configurations are able to generate 2 kA with a power of 200 kW and for a time depending on the load request, in any case longer than the target time of 2 s. In order to obtain a movable PS that could be used in every site, the input power is very low and delivered at low voltage. In particular, Fig. 1 shows that the SCPS can be connected to any standard 230 V, single-phase, 50 Hz socket, rated for 10 A, available in every European building. A specific input PS (charger), capable to limit the reactive power and the harmonic pollution, charges the SCB up to the required voltage. The charger can be regulated in voltage, current or power. The charging option (constant power or constant current) is chosen according to the SCB configuration and the desired charging time. The charger can charge a completely discharged SCB in maximum 30 min and can recharge it between two operations in less than 10 min (depending also on the energy dissipated during the previous operation). In Fig. 1 it is also visible a selector close to the input cable, that allows to choose to supply the control and auxiliary systems with a separated line with high reliability, such as an uninterruptible PS (UPS). The movability is also facilitated by the absence of water pipes for cooling. As the SC modules have relatively high thermal resistances and capacitances in air cooling (thermal time constants longer than 1 h), the internal layout was designed considering the specified duty cycle and the air temperature distribution. Nevertheless, the temperature of each module is continuously monitored.
3. Experiences with first SCPS unit The first SCPS unit, shown in Fig. 4(a) and briefly specified in
Fig. 4. Comparison between the SCPS (a) and a previous setup (b) for the same current and duty cycle reported in Table 1. 3
Fusion Engineering and Design xxx (xxxx) xxx–xxx
A. Lampasi, et al.
Fig. 5. Arbitrary waveform, defined by numerical values in the touch screen (a), generated by the SCPS on the load in Fig. 2 (500 μH and 5.5 mΩ), with a detailed analysis of a fast-descending ramp (b).
derivative of 200 A/ms (reported in Fig. 5(b)) after the flat-top at 2 kA is compliant with those necessary in tokamak central solenoids and poloidal coils for plasma breakdown. Fig. 4 compares the SCPS with the previous solution used for the PROTO-SPHERA coils [5], producing the same current (2 kA) for a shorter time (≤1.5 s) with a duty cycle of 1–2 s every 600 s (it is fair to report that the available voltage was higher, ≈350 V). Even though the net energy necessary to supply the load is comparable to a household appliance (< 400 kJ, < 0.1 kW h), the old PS was connected to a 20-kV line to draw the peak power (> 3 MVA). The comparison in terms of total dimensions is summarized in Fig. 4. The SCPS volume is 1.4 m3. The total volume in Fig. 4(b) can be only approximated (for example, since the cubicle containing the circuit breakers and safety switches is used also by another converter, only half of the volume is included in the calculation), but a good estimation is ≈10 m3. Therefore, the SCPS reduced the volume by a factor higher than 7, even getting rid of the 20-kV and 324-V lines and without
The SCPS local control unit (LCU) includes a human machine interface (HMI) on Panel PC (the black box in the right of Fig. 4(a)) with touch screen (see Fig. 5(a)) where the data of the desired waveform can be set. Operation can be started from the LCU or in remote mode (triggered by external signals). Fig. 5(b) shows a waveform measured during a test on a copper dummy load in typical settings (Icoil = 2 kA and Ttot = 2 s). Such waveform is useful to stress two interesting SCPS capabilities: the recovery of the load energy and the high current derivatives. As regards the recovery, the SC charge lost during the rising ramps is partially recovered during the falling ramps (increasing of the SCB voltage in Fig. 5(b)). Even at the end of a 10 s scenario, the observed losses do not exceed 60% in voltage and 80% in energy. Therefore, the SC state-of-charge is usually above Vrated/2. As regards the current derivatives, a specific solution was implemented to achieve 300 V (600 kW) for very fast ramps (< 10 ms) and fast energization of the magnets. In particular, the negative current 4
Fusion Engineering and Design xxx (xxxx) xxx–xxx
A. Lampasi, et al.
requiring a fixed installation with many internal and external connections. The SCPS current ripple is kept within ± 10 A at the flat-top, extremely lower than the old case based on a thyristor rectifier, because the SCPS PWM is controlled at 1.2 kHz, with an equivalent modulation frequency of 2.4 kHz on the load. The high switching frequency also reduced the acoustic noise in the experimental hall. Even in the reduced space, the SCPS safety is preserved in case of a severe fault, as a short-circuit at the output or in a bridge diode. The IGBTs can withstand the current in both cases, because the bridge itself can automatically extinguish fault currents up to 4.8 kA in less than 10 μs without the intervention of the control system. Moreover, also the SC modules can survive after a short circuit at their terminals. Another aspect to be presented is the need of two high-energy resistors that can discharge the SCB in few minutes. They are normally connected during stand-by periods (“Sleep” mode in Fig. 5(a)) to partially discharge the SCB to VSCB/2, as long periods at full voltage or deep discharges could reduce the SC lifetime. The resistors must be used also before accessing the SCPS cabinet (for instance, for maintenance). This safety procedure, necessary to manage the high energy stored in the SCB even after the SCPS disconnection, is balanced by the removal of the MV procedures. All these relevant benefits could be extended to similar cases at different currents, voltages, powers and energies. In fact, scalability and possible upgrading are intrinsic in this technology.
reused. Moreover, the external grid is not affected by nonlinear power converters introducing harmonics and reactive power. This result is relevant, especially when compared with the achievable performances and with the characteristics of previous systems used for the same purposes. Since SCs can feature very long lifetimes and the SCPS can be optimized to preserve them, the SCB is expected to operate for more than 1 million of charge/discharge cycles. The SCPS approach is particularly suitable when applied to nuclear fusion experiments, inductive loads and fixed duty cycles. In particular, the energy recovered from superconducting coils may be significant. Moreover, a specific solution was implemented to achieve very fast ramps on the magnets. Finally, the SCPS approach is intrinsically modular. This is useful to reduce the initial and maintenance costs, to rearrange and update the systems even after the installation, but also to apply this approach to larger facilities as large tokamaks. By series and parallel connections of the SC modules or of the entire SCPS, the SCPS can be extended to higher voltages, higher currents, longer scenarios. In any case, the input power demand and the impact on the external grid are limited with respect to the output performances. References [1] A. Lampasi, S. Minucci, Survey of electric power supplies used in nuclear fusion experiments, 17th IEEE Int. Conf. Environ. Elect. Eng. (EEEIC) (2017). [2] A. Lampasi, et al., ETHICAL: a modular supercapacitor-based power amplifier for high-current arbitrary generation, 16th IEEE Int. Conf. Environ. Elect. Eng. (EEEIC) (2016). [3] A.L. Allegre, et al., Energy storage system with supercapacitor for an innovative subway, IEEE Trans. Ind. Electron. 57 (December (12)) (2010) 4001–4012. [4] A. Lampasi, et al., Compact power supply with integrated energy storage and recovery capabilities for arbitrary currents up to 2 kA, IEEE Trans. Plasma Sci. 46 (October 10) (2018) 3393–3400. [5] A. Lampasi, et al., Progress of the plasma centerpost for the PROTO-SPHERA spherical Tokamak, Energies 9 (7) (2016) 508.
4. Conclusions and future prospects This paper discusses the design issues and the potential benefits of a system able to store a large amount of energy at low power and to deliver it at high power when necessary. Moving from this idea, a first compact and movable unit was developed to generate for some seconds arbitrary currents up to 2 kA, requiring a limited and tunable input power, even in total absence of sources with sufficient power. A fraction of the power delivered to the load can be recovered to be
5