1340
IEEE TRANSACTIONS ON PLASMA SCIENCE, VOL. 41, NO. 5, MAY 2013
Comparisons of Three Inductive Pulse Power Supplies Xinjie Yu, Member, IEEE, and Xiangxiang Chu
T1
Abstract— The energy density of the inductive energy storage systems is one order of magnitude higher than that of the capacitive ones. Therefore, they have potential applications in the future. In this paper, three inductive pulse power supply (PPS) topologies, i.e., slow transfer of energy through capacitive hybrid meat grinder, XRAM, and nonmutual inductance PPS, are discussed in depth. First, their principles are analyzed. Then, simulations are made to illustrate the principles and prepare the fundamental data for further comparisons based on a 4-kA charging current, 220-µH total inductance system under the environment of Simplorer 8. Five performance indices reflecting their performances from different aspects are proposed in order to ensure objective and comprehensive comparisons. Based on those indices, the advantages and disadvantages of these topologies are pointed out and some potential improvements are also presented. Index Terms— Comparisons, inductive pulse power supplies, performance indices, slow transfer of energy through capacitive hybrid meat grinder, XRAM.
I. I NTRODUCTION
W
ITH the development of semiconductor and superconducting technology, inductive storage has become a hot topic. On the basis of the same output power, the energy density of the inductive storage is one order of magnitude higher than that of capacitive storage. Until now, there have been two main topologies for inductive pulse power supply (PPS). One is called the slow transfer of energy through capacitive hybrid meat grinder (STRETCH meat grinder) [1], [2], which is presented by the Institute of Advanced Technology (IAT). Another is an XRAM modified by the Institute of Saint Louis (ISL) from the original XRAM [3]. Based on the improvement for the original meat grinder topology [2], [4], the IAT proposed a novel topology STRETCH meat grinder amending the drawback of the overvoltage across the main switch that the original meat grinder encountered [5]. In order to turn off the charging current on the primary side, an integrated gate commutated thyristors (IGCT) was used as the main switch. So far, the IAT has implemented a 3.8-kA charging current, a 20-kA discharging current, and a 1.5-kJ PPS prototype. They planned to design Manuscript received October 27, 2012; revised February 10, 2013; accepted February 18, 2013. Date of current version May 6, 2013. This work was supported in part by the National Science Foundation of China (NSFC) under Project 50507011, the NSFC under Project 50877039, and the Tsinghua University Initiative Scientific Research Program. The authors are with the Department of Electrical Engineering, State Key Lab of Power System, Tsinghua University, Beijing 100084, China (e-mail:
[email protected];
[email protected]). Digital Object Identifier 10.1109/TPS.2013.2248392
Sop
D2 L 1 uC
us
iL1
C D1
iLoad Load
L2 iL2
Fig. 1.
Topology of the STRETCH meat grinder.
an electromagnetic system with 2 MJ muzzle energy in the future. On the basis of Inverse Current Commutation with Semiconductor devices (ICCOS) [6], [7], the ISL has developed XRAM with 8 stages and 20 stages [6], [8], [9]. ICCOS makes sure that the main switch of XRAR, i.e., thyristor, can be turned off safely while bearing low voltage. A PPS system with 3-kA charging current, 60-kA discharging current, and 4.7-kJ energy has been implemented. They also intended to make efforts towards a 20-stage system with 0.5 MJ total energy in the future. Besides, in Tsinghua University some research has been performed on the inductive pulse power supplies. Unlike the STRETCH meat grinder, a topology without mutual inductive inductance is presented in [10]. IGCT is selected as the main switch so as to turn off the primary current actively. II. P RINCIPLES OF T HREE T OPOLOGIES A. STRETCH Meat Grinder The topology of the STRETCH meat grinder is demonstrated in Fig. 1. The basic principle is the same as the traditional meat grinder. By the introduction of capacitor C, the STRETCH meat grinder suppresses the overvoltage across the main switch. The working procedure can be divided into four steps. In step 1, the series-connected inductors L1 and L2 are charged by the primary source us . When the main switch Sop (the active switch IGCT) is on, the constant voltage source begins to charge the seriesconnected inductors L1 and L2 , and the current increases linearly. When the current increases to the designated value, Sop is actively switched off. Then the next step begins.
0093-3813/$31.00 © 2013 IEEE
YU AND CHU: COMPARISONS OF THREE INDUCTIVE PULSE POWER SUPPLIES
In step 2, it works like a meat grinder. Considering that there are mutual inductance and selfinductance between L1 and L2 separately, it is convenient to analyze one by one. As for the energy of mutual inductance, the total flux keeps unchanged at the moment Sop is switched off. So the decrease of flux in L1 will cause the flux in L2 to increase. In general, the inductance value of L1 is ten times larger than that of L2 . Therefore, the decrease of the flux in L1 will cause the current in L2 to increase rapidly. In the meantime, diode D1 begins to conduct. The current through L2 comprises the dominant part of the current multiplication factor, which is defined by the ratio of the maximal load current over the maximal charging current in the primary side. As for the energy related to the leakage flux, the corresponding current has only one path to flow through D1 , Load, C, D2 , and L1 . Capacitor C is inversely charged by L1 , which causes uC to decreases from zero. When the current in L1 decreases to zero, diode D2 was switched off naturally. Then, step 3 begins. In addition, the existence of C limits the voltage of the main switch Sop . In step 3, the voltage across the capacitor is kept constant. The voltage across C attains the maximum at the time when iL1 comes across zero. Although thyristor T1 bears forward voltage, there is no pulse exerted on the gate so that the voltage of C stays unchanged. Inductor L2 supplies all the load current. The pulse given to T1 is controllable, which can be used to regulate the load current and add to some design flexibility. In step 4, T1 is triggered so that C begins to discharge. The discharge circuit contains L1 , T1 , C, Load, and D1 . From the reference direction shown in Fig. 1, applying the KCL principle can get the equation iLoad = iL2 –iL1 . Current iL1 is negative so it helps to increase iLoad . A load current peak will occur due to the decrease of iL2 and the increase of absolute value of iL1 . The capacitor will be charged after its voltage comes across zero. Then, C discharges again along C, D2 , L1 , D1 , and Load. Then, C is inversely charged again after its voltage comes across zero. When iL1 decreases to zero, the voltage of C keeps unchanged. Inductor L2 alone drives the load. In addition, thyristor T1 can be triggered later, if needed, to repeat steps 3 and 4. B. XRAM Based on the ICCOS principle, the ISL put forward the XRAM topology. Without loss of generality, an XRAM with eight stages is chosen to illustrate its basic principle. The topology is drawn in Fig. 2. Unlike STRETCH meat grinder, a thyristor is used as the main switch. The working procedure of XRAM can be divided into three steps. In order to help ICCOS commutate successfully, the capacitors from C1 to C8 are charged with the initial voltage. In step 1, the primary source charges the series-connected inductors L1−8 . Thyristors T h1−8 are triggered. Then, L1−8 are charged so that the current increases. When the current increases to the designated value, T h9 is triggered and step 2 starts. In step 2, ICCOS turns off the main switch.
1341
D1
D 11
Th1
iL1
C1 L1 D12
D2
D21 Th9
Th2
us L2 D8
C2 D 22 D81
Load iLoad
Th8 C8 L8 D82
Fig. 2.
Topology of XRAM with eight stages.
The procedure for ICCOS is so short that the current through the inductors stays almost unchanged. Lk , Dk2 , Load, T h9 , and Ck compose eight parallel discharging circuits. C1 , T h9 , Load, D82 , us , and T h1 compose the first commutating circuit, while Ck , T hk , Dk−1,2 , Load, and T h9 (k = 2 − 8) compose the other seven commutating circuits. Current iThk decreases as iCk increases from zero. When its current comes across zero, T hk is switched off naturally. Owing to the remaining voltage of Ck , it still discharges. This causes diode Dk to carry current, which exerts inverse voltage for T hk . As long as the time is longer than the off time tq , thyristor T hk can be safely switched off. In order to illustrate how diodes Dk (k = 1 − 8) are switched off, we can group them into two parts: D1 and D2−8 . The reason for grouping in such a way is their different voltage compositions. For D1 , its voltage can be computed as follows: uD1 = uC1 –uLoad –us . In this procedure, uC1 decreases, us stays unchanged, and uLoad is positive. So, uD1 decreases. When it decreases below zero, D1 is switched off. The load is composed of low impedance whose current increases. So, the voltage across the load is positive. For diodes D2−8 , it holds that uDk = uCk – uLoad . Voltage uCk decreases and uLoad is positive. Therefore, conclusions can be made that diodes will be switched off. After diodes D1−8 are switched off, Lk , Ck , T h9 , Load, and Dk2 compose a second-order underdamping system. Diode Dk1 is switched on after uCk decreases to zero. Then step 3 starts. In step 3, inductors L1−8 discharge in parallel. C. Nonmutual Inductance PPS The nonmutual inductance PPS topology is shown in Fig. 3. The working procedure can be divided into three steps. In step 1, the primary source charges the series-connected inductors. The main switch Sop (IGCT) is actively switched on. The current through inductors increases until the designated value. Then, Sop is actively switched off, while Scl is switched on at the same time. Then, step 2 starts. In step 2, the current in L1 is inverted.
1342
IEEE TRANSACTIONS ON PLASMA SCIENCE, VOL. 41, NO. 5, MAY 2013 Voltage wave(STRETCH) 1.25 Curve Info
Dcb
Sop
uSop
TR TR
L1 i
L1
iL2
iLoad
Sc1
L2
0.00
t0
t1
Load
uC
-0.63
D
t2
t3
-1.25 7.00
7 .5 0
8.00
8.50
9.00 Time [ms]
9.50
10.00
1 0 .5 0
1 1.00
Topology of nonmutual inductance PPS.
iL2 flows through L2 , Load, and Sc1 all along. The current in inductor L1 charges capacitor C along Scl , Load, and C until the current decreases to zero. Then, C charges L1 through Load, Scl , and L1 while voltage uC decreases. Note that the actual current through L1 is inverted, which benefits the increase of load current. Diode Dcb is switched on after uC comes across zero, which ends step 2. In step 3, the inductors discharge in parallel. L2 –Load–Sc1 and L1 –Dcb –Load constitute the parallel discharging circuit.
Fig. 4.
Voltage waves for the STRETCH meat grinder. Simplorer1
Current wave(STRETCH) 30.00 Name
X
Curve Info
m2
Y
m1
7.5468 4.0981
m2
8.5288 29.1389
iLoad
TR TR
r1.i IGBT1.I
25.00
20.00
Cu rren t [k A ]
Fig. 3.
igbt1.v
0.63
V o lta g e [ k V ]
us
C
c1.v
15.00
III. S IMULATIONS BASED ON S IMPLORER ® 8 Simulations are made to illustrate the principles and prepare the fundamental data for the comparisons under the environment Simplorer 8. In order to compare the advantages and disadvantages more objectively, the designated charging current and the total energy in the inductors are kept the same for the three topologies. In detail, the designated current is about 4.1 kA and the total inductance is about 220 µH. For the STRETCH meat grinder, the inductance values for L1 and L2 are 158 and 5.89 µH, respectively, the coupling factor is 0.94, and the capacitance value is 800 µF. For an XRAM with eight stages, the value of the inductance for a single stage is 30 µH. The capacitance value and the initial voltage are 140 µF and 1000 V, respectively, except C1 , whose capacitance and the initial voltage are 280 µF and 1200 V. Such a various capacitance and voltage selection will ensure the successful commutation. For the nonmutual inductance PPS, the values of inductance L1 and L2 are 44.2 and 176.8 µH, respectively, and the capacitance value is 1000 µF. The load parameters are referred to the published papers issued by the IAT and the ISL, where the 1.5 mΩ, 1 µH low impedance is assigned. Other than that, the 123 V constant voltage source is used as the primary source. The current for driving a real system such as the railgun is about up to hundreds of kA. So, the parameters configured previously for the three topologies can act as one unit. The whole system is composed of ten or more units. The comparisons below are based on such a unit. A. Simulation for the STRETCH Meat Grinder Simulations are made based on Simplorer 8, and the results are shown in Figs. 4 and 5.
10.00
iSop 5.00
m1
0.00 7.00
Fig. 5.
7.50
8.00
8.50
9.00 Time [ms]
9 .5 0
1 0 .0 0
10.50
1 1.00
Current waves for the STRETCH meat grinder.
0–7.55 ms (t0 –t1 ) is step 1 when us charges L1 and L2 . 7.55–7.85 ms (t1 –t2 ) matches step 2. Owing to the capacitance, the voltage across the main switch is about 1.2 kV. A peak of load current occurs at the end of this step. 7.85–8.25 ms (t2 –t3 ) matches step 3. 8.25–11 ms (t3 –end of simulation) matches step 4. In this step, the load current attains the maximum 29.1 kA. B. Simulation for XRAM Simulations are made for XRAM with eight stages based on Simplorer 8. The simulation results are shown in Figs. 6 and 7. 0–9.3 ms (t0 –t1 ) matches step 1 of XRAM. Eight inductors are charged in series. Thyristor T h9 is triggered at the instant when the current increases to the designated value. Then, step 2 starts. 9.3–9.5 ms (t1 –t2 ) matches step 2 of ICCOS. During ICCOS, T h1−9 are on or shorted by the diodes inversely in parallel. So they just bear low voltages while ensuring the reliable turning-off of T h1−8 . In this step, the load current attains the maximum 33.61 kA. 9.5 ms (t2 ) later matches step 3.
YU AND CHU: COMPARISONS OF THREE INDUCTIVE PULSE POWER SUPPLIES
1343
Voltage of Thysistors(XRAM)
Voltage wave(Non_Mutual Indutance)
m6 m5
1.00
Name
0.80
X
uTh9
TH2.V
TR
9.5224 0.1876
m2
6.1793 1.1421
m3
9.5224 0.0645
m4
1.4364 1.1350
m5
2.9714 1.1745
m6
2.9714 1.1745
TH4.V
TR
TH9.V
TR
uTh1 uTh2-8
m1
C1.V_1 IGBT1.V_1
0.25
0.00
t2 uC
t1
m2
0.00
2.50
5.00
7.50 Time [ms]
10.00
1 2 .5 0
1 5.00
Voltage waves for eight-stage XRAM.
-1.00 7.00
7.50
Fig. 8.
8.00
X
m1
9.3528 33.6072
m2
9.4131 4.0410
9 .0 0
9.50
1 0.00
Current wave m1
Y
8.50 Time [ms]
Voltage waves of nonmutual inductance PPS.
Current(8-XRAM) 35.00 Name
TR
-0.75
-0.20
Fig. 6.
TR
-0.50
t2
t1
t0
uSop
m1
-0.25
m3 -0.00
7.8904 -0.8752
0.50
TH8.V
TR
uTh1-8
7.8964 1.0021
m2
TH7.V
TR
0.20
m1
0.75
TH6.V
TR
0.40
Curve Info
Y
TH5.V
TR
V o lta g e [ k V ]
1.00
X
TH3.V
TR
0.60
Name
TH1.V
TR
Y
m1
1.25
Curve Info
m2
m4
V o lta g e [ k V ]
1.20
8.75
Curve Info TR TR
30.00
Name
r1.i L1.I
X
Curve Info
Y
m1
7.5500 4.1042
m2
8.2164 8.0298
iLoad
m2
7.50
iLoad
TR TR
IGBT1.I L3.I
25.00
6.25
20.00
Y 1 [k A ]
Cu rren t [k A ]
5.00
m1
15.00
iTh1
3.75
10.00
2.50
iL1 5.00
m2
1.25
0.00 0.00
Fig. 7.
2.50
5.00
7.50 Time [ms]
1 0.00
1 2 .5 0
1 5.00
Current waves for eight-stage XRAM.
C. Simulation for Nonmutual Inductance PPS Simulations are made for nonmutual inductance PPS based on Simplorer 8. The simulation results can be seen in Figs. 8 and 9. 0–7.55 ms (0–t1 ) matches step 1. When the current rises up to the designated value, step 2 starts. 7.55–8.6 ms (t1 –t2 ) matches step 2. In this step, the load current attains the maximum 8.03 kA. 8.6 ms later matches the third step. IV. C OMPARISONS BASED ON F IVE P ERFORMANCE I NDICES As for inductive PPS, there are many aspects that need to be concerned, such as the voltage pressure on the circuit elements, the lifetime of system, the utility factor of the inductive energy density, and the ability to drive load. In order to make more comprehensive comparisons, five performance indices are used to depict the performance of the different topologies. 1) The voltage across the main switch. 2) The current multiplication factor.
0.00 0 .0 0
Fig. 9.
2.50
5.00
7.50 Time [ms]
1 0 .0 0
1 2.50
1 5.00
Current waves of nonmutual inductance PPS.
3) The ratio of the maximum capacitive energy over the total inductive energy. 4) The current variation rate across the inductance. 5) The integration of the square of the load current. The voltage across the main switch reflects the voltage stress requirements, so lower voltage value is preferred. The larger current multiplication factor means better current amplification capacity. The lower capacitive energy percentage means higher energy utility factor. The lower current variation rate implies better working conditions and lower requirements for the elements. The larger integration of the square of the load current means stronger load-driven capacity. Considering that the load may be launched within 5 ms, 3.5 ms is chosen as the integration time. And the integration starts at the time when the load current begins to increase from zero. All results are shown in Table I. As for the voltage across the main switch, the XRAM has comparative advantage due to the application of ICCOS. Except T h9 , others’ maximal voltages are lower than 200 V. The maximal voltage of T h9 in XRAM is about 1175 V. For STRETCH and nonmutual inductance PPS, the active
1344
IEEE TRANSACTIONS ON PLASMA SCIENCE, VOL. 41, NO. 5, MAY 2013
TABLE I C OMPARISONS OF T HREE T OPOLOGIES F ROM D IFFERENT A SPECTS Indices
Maximal voltage across the main switch/V Current multiplication factor EC /EL (%) Current variation rate/ (A/s) Integration of i2
STRETCH Meat Grinder IGCT Sop 1200
XRAM wit 8 Stages SCR T h9 1175 (ripple) T h1 187 T h2−8 64
Non-Mutual Inductance PPS IGCT Sop 1000
7.1
8.17
1.95
24.17%
38.4%
20.6%
Sop 4.2 × 1010 L1 1.20 × 107 L2 6.44 × 107 6.41 × 105
switch is used to turn off the current actively. Although the capacitor is introduced to limit the voltage, the maximal voltages across IGCT are still about 1000 V. As for the current multiplication factor, the XRAM dominates with 8.17. Inheriting the advantage of a meat grinder, the STRETCH meat grinder is just a little inferior to the XRAM with 7.1. Due to the cancellation of mutual inductance, the third topology has obvious disadvantage with 1.95. Besides, for the STRETCH meat grinder, the higher multiplication is gotten at the expense of the highly coupled inductors with coupling factor 0.94. Limited by the manufacturing level, the larger coupling factor is almost unrealistic. So the further scalability for one single unit is difficult, whereas XRAM is very flexible and can be easily extended to get higher load current just by adding more stages. The XRAM shows weakness for the third index with higher percent 38.4%, while the other two topologies have comparative advantage. The high initial capacitive energy in XRAM ensures the success of ICCOS. In order to make full use of the high density of the inductance, the percentage should be as low as possible while ensuring the success of ICCOS. So, further studies on how to cut down the percentage are meaningful. In view of the current variation rate, XRAM is superior to the others. The current variation rates of the main switch and the inductors are one order magnitude lower than those of the others, which means lower requirements and better working conditions for the elements, thus cutting down the costs and prolonging its lifetime. The amplified current is intended to drive the load. The electrical force is strongly related to the integration of the square of the current. From Table I, XRAM ranks the first with 8.57 × 105 , followed by the STRETCH meat grinder with 6.41 × 105 . The third topology is obviously weak in dealing with driving load. For the STRETCH meat grinder, Integrated GateCommutated Thyristor (IGCT) is used as the main switch. The maximal primary charging current can be actively turned off by the-state-of-the-art IGCT which is about 4 kA. But for XRAM, thyristor is used and the ICCOS is applied. If only
T h9 1.12 × 109 L1 4.35 × 106 L2 2.33 × 106 8.57 × 105
Sop 4.10 × 1010 Scl 1.86 × 107 L1 1.89 × 107 L2 5.35 × 105 1.60 × 105
enough initial capacitive energy is guaranteed, the maximal primary charging current, which is turned off by ICCOS, can be up to tens of, even hundreds of, kA, which has potential future applications. V. C ONCLUSION The principles of three inductive pulse power sources were discussed in detail. On the premise that the charging current and the total inductive energy are the same, simulations were made to illustrate the principles clearly and prepare the fundamental data for comparisons based on Simplorer 8. Five performance indices from different aspects were selected to make more objective and more comprehensive comparisons. In view of the requirements for the circuit elements such as voltage stress and current variation rate, XRAM calls for lower voltage and one order magnitude lower current variation rate than others with comparative advantage. However, it needs more initial capacitive energy to make sure the success of ICCOS, so it does not make full use of the advantage of high energy density like the other two topologies. As for the ability to drive the load, XRAM and STRETCH meat grinder have stronger driving ability than the nonmutual inductance PPS. The STRETCH meat grinder attains this at the expense of high coupling factor, which means limited capability of amplifying the current. Unlike STRETCH meat grinder, XRAM multiplies the current by the means of adding more stages. Some factors were taken into account in dealing with a real system. For example, actively turning-off 4-kA primary current has almost reached the extreme of IGCT. In addition, the cost for such IGCT is very high. So the replacement of IGCT with thyristor will contribute to the further application for the STRETCH meat grinder. Although XRAM has many advantages, how to cut down the capacitive energy percentage while ensuring the success of ICCOS needs further study. Optimization on the topology, the capacitance value, as well as the initial voltage will be taken into consideration in the future.
YU AND CHU: COMPARISONS OF THREE INDUCTIVE PULSE POWER SUPPLIES
R EFERENCES [1] A. Sitzman, D. Surls, and J. Mallick, “Design, construction, and testing of an inductive pulsed-power supply for a small railgun,” IEEE Trans. Magn., vol. 43, no. 1, pp. 270–274, Jan. 2007. [2] D. Giorgi, K. Lindner, J. Long, T. Navapanich, and O. Zucker, “Enhancing the transfer of inductive energy to imploding plasma loads with a single-step meatgrinder circuit,” IEEE Trans. Magn., vol. 23, no. 3, pp. 1913–1918, May 1987. [3] R. D. Ford, R. D. Hudson, and R. T. Klug, “Novel hybrid XRAM current multiplier,” IEEE Trans. Magn., vol. 29, no. 1, pp. 949–953, Jan. 1993. [4] K. Lindner, J. Long, D. Girogi, T. Navapanich, and O. Zucker, “A meatgrinder circuit for energizing resistive and varying inductive loads (EM guns),” IEEE Trans. Magn., vol. 22, no. 6, pp. 1591–1596, Nov. 1986. [5] E. Dierks, I. R. McNab, J. A. Mallick, and S. Fish, “Battery-inductor parametric system analysis for electromagnetic guns,” IEEE Trans. Plasma Sci., vol. 39, no. 1, pp. 268–274, Jan. 2011. [6] P. Dedie, V. Brornmer, and S. Scharnholz, “ICCOS countercurrentthyristor high-power opening switch for currents up to 28 kA,” IEEE Trans. Magn., vol. 45, no. 1, pp. 536–539, Jan. 2009. [7] S. Scharnholz, V. Brommer, G. Buderer, and E. Spahn, “High-power MOSFETs and fast-switching thyristors utilized as opening switches for inductive storage systems,” IEEE Trans. Magn., vol. 39, no. 1, pp. 437–441, Jan. 2003. [8] P. Dedie, V. Brommer, and S. Scharnholz, “Experimental realization of an eight-stage XRAM generator based on ICCOS semiconductor opening switches, fed by a magnetodynamic storage system,” IEEE Trans. Magn., vol. 45, no. 1, pp. 266–271, Jan. 2009.
1345
[9] P. Dedie, V. Brommer, and S. Scharnholz, “Twenty-stage toroidal XRAM generator switched by countercurrent thyristors,” IEEE Trans. Plasma Sci., vol. 39, no. 1, pp. 263–267, Jan. 2011. [10] X. Liu, Z. Wang, and J. Li, “Circuit topology of a new inductive storage pulsed-power supply to drive railgun,” Power Syst. Technol., vol. 33, no. 13, pp. 80–84, Jan. 2009.
Xinjie Yu (M’01) received the B.S. and Ph.D. degrees in electrical engineering from Tsinghua University, Beijing, China, in 1996 and 2001, respectively. He is an Associate Professor of electrical engineering with Tsinghua University. His current research interests include all aspects of pulse power supply, power electronics and computational intelligence.
Xiangxiang Chu received the B.S. degree in electrical engineering from Southeast University, Nanjing, China, and the M.S. degree from Tsinghua University, Beijing, China, in 2010 and 2012, respectively. His current research interests include pulse power supply and smart grids.
