Professional Documents
Culture Documents
High Eff Single PH Grid Connected Inverter
High Eff Single PH Grid Connected Inverter
9, SEPTEMBER 2010
TABLE I
PV A RRAY L EVELS FOR A G IVEN G RID VOLTAGE AND T OPOLOGY R ATIO
maximum grid voltage, i.e., nominal value plus 10% [26] and
the topology specific ratio between input and output voltages.
As an example for this factor, a half-bridge or single-phase
multilevel inverter requires a two times higher input voltage
Fig. 3. Voltage ratios for silicon modules available in the market [25]. in relation to the traditional full-bridge configuration for the
same output voltage. The necessary array voltage for European
The leakage currents to ground constitute another important single-phase grid levels (230 V ± 10%) and modulation reserve
issue if one of the terminals of the array cannot be grounded, of 5% values are given in the Table I.
particularly regarding transformerless concepts. Due to their What is important to observe here is that in order to operate
structure, PV modules naturally form a parasitic capacitance with a single stage, multilevel inverters would require a rather
between the cells and the grounded frame [20], [21]. High- large voltage in the input that, for the maximum open-circuit
frequency variations of the cell potential in relation to ground value, is out of the scope of most manufacturers’ specifications
shall therefore be avoided in transformerless circuits since and current standards in Europe [27], [28]. Initial discussions
this leads to large charge/discharge currents partially flowing for the expansion of such limits are nevertheless already taking
through the circuit to the ground, resulting in an increase of the place in Germany [29]. If the American voltage levels were
harmonic content, higher losses, and also, safety and electro- considered, a maximum array voltage superior to 600 V would
magnetic interference problems [22], [23]. Special single-phase be necessary for multilevel converters, falling out of the low-
transformerless topologies with reduced oscillations have been voltage category and, consequently, requiring more complex
developed for such purpose and will be later discussed. In addi- installation equipment [13].
tion, the introduction of frameless panels further contributed to
the reduction of such problem.
E. Electric Conversion Efficiency
This characteristic is sometimes regarded as the most signif-
D. Input Voltage Levels
icant in PV converter circuits, as it directly affects the payback
Unlike other applications employing inverters, PV systems time of the facility.
have as unique characteristic a certain variation on the input Under the necessity of directly comparing different system
voltage, also known as the maximum power point (MPP) concepts and converter topologies with a single value rather
voltage range. The lower limit can be defined under worst than using efficiency curves, the weighted efficiency concept
case conditions with a minimal radiation level and high cell with factors directly related to local radiation levels was de-
temperature (100 W/m2 and 50 ◦ C), while the inverse would be fined. An example is the so-called “European efficiency” [30],
valid for the higher limit (1000 W/m2 and 20 ◦ C). Considering suitable for countries with low radiation levels like Germany
the silicon crystalline modules available in the market [25] and (1000 kW · h/m2 · year), while for regions with higher radi-
the aforementioned conditions, a factor of approximately 0.7 ation levels, like the southeast of the U.S. (over 1800 kW ·
for the division of such limits is obtained. h/m2 · year), other factors were defined under the normally
During system start-up, a higher voltage level, namely, the called “U.S. efficiency” [31].
open-circuit voltage, is impressed in the input of the inverter. The project of converters for PV applications are, as a
From the performed analysis and for a radiation level of consequence, strongly influenced by such factors, as the peak
1000 W/m2 and −10 ◦ C, cell temperature can reach up to efficiency in a good design is specified to reach a maximum
two times the minimal MPP voltage. The calculated ratios are value around 50% and 75% of the nominal input power for,
shown in Fig. 3, as a function of the module rated power. respectively, the European and American values.
For the design of PV inverters, it is therefore necessary to Such characteristic is achieved by a good balance between
consider the factors aforementioned not only for operation but semiconductor switching and conduction losses along with the
also when choosing the voltage rating of the semiconductors. optimization of current freewheeling paths in such a way that
As stated before, when transformers are employed, they pro- the efficiency is not prejudiced by the operation at different
vide some sort of voltage step-up allowing, at the same time, input voltage levels. Of great importance is also the design
flexibility in choosing an optimal input voltage. of the magnetic elements in order to attain reduced core
For highly efficient transformerless topologies composed of losses and, consequently, good performance in the lower power
a single stage, the minimum input-voltage level is tied to the region.
ARAÚJO et al.: SINGLE-PHASE TRANSFORMERLESS INVERTERS FOR GRID-CONNECTED PHOTOVOLTAIC SYSTEMS 3121
F. Semiconductor Technologies
From the previous items was observed that transformerless
inverters operate with higher input voltages when compared
with their counterparts with transformers, which infers the use Fig. 5. Reference output voltage and resultant voltages across the parasitic
capacitors for different modulation strategies in a full-bridge inverter.
of semiconductors with higher blocking voltages. The design
of such circuits with reduced losses has been made possi-
ble mainly due to recent developments toward high-efficiency additional losses due to the internal reactive-power flow inside
semiconductor technologies at higher voltage classes. the inverter. The occurrence of leakage current due to PV array
Regarding silicon switches, a first example are the insulated- oscillations to ground is not a critic issue here, given the use
gate bipolar transistors with Trench-Gate and Field-Stop tech- of a divided output inductor filter, as shown in Fig. 4. The
nologies, characterized by a good balance between conduction resultant voltage across each parasitic capacitor follows the
and switching losses [32], [33]. The superjunction technology output voltage with half the amplitude and offset that is
(also known as CoolMOS) made MOSFETs with blocking volt- equal to half of the input voltage, as shown in Fig. 5. An
ages of up to 900 V a very interesting option due to the reduced example of a commercial device using such approach is the
on-resistance values and high switching speeds, surpassing the SolarMax 4000 [35].
theoretical limits of vertical Si MOSFETs. The use of unipolar modulation [36], also known as three
The recent development of silicon carbide (SiC) switches like level [37], employs two sinusoidal references dephased by
MOSFETs and JFETs has also served as an indication of further 180◦ to modulate each phase leg. Such strategy allows an
advances toward high blocking voltages together with very low additional voltage level with ripple reduction, along with no
losses [34]. internal reactive-power flow due to the freewheeling through
only two antiparallel diodes. In addition, the output-voltage
levels change effectively with the doubling of the switching fre-
III. S INGLE -P HASE T RANSFORMERLESS T OPOLOGIES
quency that leads to a further reduction of the filter size. Despite
Taking into consideration the presented framework, highly such advantages, it cannot be employed in transformerless PV
efficient single-phase inverter topologies that will most likely inverters since the PV array voltage to ground will oscillate at
reach a high level of efficiency at low cost are the ones consti- the switching frequency, as shown in Fig. 5, generating high
tuted by a single-stage without transformer. level of leakage currents.
The classic full-bridge circuit as shown in Fig. 4 would One final alternative to both classical strategies is the use of
constitute a first example, although the performance and ap- the single-phase chopping scheme, where one leg is switched
plicability are strongly influenced by the employed modulation at high frequency and the other at grid frequency for each half-
scheme and the resultant voltage across the parasitic capacitors wave. Of advantage here is not only the additional voltage level
Cp1 and Cp2 between the panels and the ground. but also the reduced switching losses and no internal reactive-
The bipolar modulation [36], also known as two level, [37] power flow. Only one single-output filter inductor may be used
uses a single sinusoidal reference for one group of diagonal in this case. The PV array voltage oscillations to ground take
switches, with the other group using the complementary signal. the form of a square wave with grid frequency, as shown in
As a result, two voltage levels are modulated in the output, Fig. 5, which may require complex filtering solutions to reduce
leading to a high current ripple across the inductors. A further the occurrence of high temporary peaks of leakage current.
drawback is that the current freewheeling happens through Several new single-phase topologies found in the market deal
the antiparallel diodes to the dc-link capacitors, resulting in with the issues discussed earlier, such as the following.
3122 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 57, NO. 9, SEPTEMBER 2010
Fig. 6. Analyzed single-phase high-efficiency PV inverters—from left to right: Heric, H5, Karschny, and Extended input-voltage circuits.
A. HERIC [38] [Fig. 6(a)] such high levels are not yet covered by the actual standards, a
boost stage has been so far applied in the input for commercial
This topology is constituted by a normal full-bridge cir-
products, like, for example, the Triple Lynx Series from the
cuit with each group of diagonal switches being operated at
manufacturer Danfoss [35]. Consequently, the total efficiency
high frequency during one half-wave of the output voltage.
of the topology is significantly reduced.
An additional branch placed in parallel with the filters and
Another critical point regarding the application of such cir-
load has two switches in opposite directions, each one active
cuit in grid-connected appliances is the higher transient voltage
during one whole half-period of the grid waveform. The current
across the middle switches, already observed in commercial
freewheeling now happens through such switches, eliminating
applications. This is mainly justified by the fact that these
the internal flow of reactive power, allowing an additional
switches are not clamped to the dc-link capacitors like the
voltage level, and finally reducing the occurrence of leakage
external ones and are also under direct influence of parasitic
currents, since the PV array is decoupled from the grid during
components in the system layout [41], [42].
the freewheeling phase. The drawbacks here are the increased
As a conclusion, a single-stage transformerless solution of
amount of semiconductors and the incapability of processing
the NPC inverter could also be an interesting option in the near
reactive power with the default switching strategy. Examples of
future only if such problems were addressed, together with a
commercial products with such topology are the NT-series of
revision of the standards toward higher system voltage levels.
inverters from the manufacturer Sunways [35].
Fig. 7. Proposed inverter circuit. Fig. 8. Variant of the proposed inverter circuit.
TABLE II
IV. N OVEL T OPOLOGY C OMPARISON R EGARDING S EMICONDUCTOR C OMPLEXITY
Fig. 10. Operation of the variant circuit: the (a) negative half-wave and
(c) positive half-wave with (b and d) the respective freewheeling stages.
V. T HEORETICAL A NALYSIS
1 1 M +1
= ·D+ · (1 − D) + 1 · 1 = . (3) A. Semiconductors
M M M
The switches operating at high frequency (T1 and T2 ) are
At the top of Fig. 9 are traced the curves representing the under a maximum voltage stress equal to the input voltage. The
factor calculated with (3) as a function of the normalized output effective current through them is calculated in (5), where Tg
voltage for the proposed and for the other previously analyzed is the grid period, Iout (t) the output current, and D(t) is the
topologies. It becomes clear that the new topology achieves duty cycle of the referred switch, which ranges from zero to
the minimum conduction losses among all circuits due to the one following the amplitude of a sinusoidal reference
reduced amount of switches in the current path.
The switching losses can be estimated with the weighted
Tg
The freewheeling diodes need to block the input voltage and TABLE III
P ROTOTYPE S PECIFICATIONS
have an average current given by
Tg
2
1
Iavg_D1 = [Iout (t) · (1 − D(t))] · dt. (6)
Tg
0
Since the dc link also buffers the energy during the free-
wheeling stage at high frequency, it is also advised to employ VI. E XPERIMENTAL R ESULTS
some low-inductance film capacitors with reduced value of A. Specifications and Project
equivalent series resistance (ESR) for reduced high-frequency
voltage ripple. In order to evaluate the performance of the variant of the pro-
posed circuit, a prototype was built following the specifications
given in Table III.
C. Output Filter 1) Semiconductors: Given the maximum MPP voltage and
Knowing that this topology operates similarly to VSIs, it is assuming a derating factor of 60% in order to achieve a high
only necessary to employ inductors in the output to provide level of reliability [48], the switches operating at high frequency
filtering of the output waveform. To reduce the inductor size, need to have a blocking voltage of at least 900 V. The maximum
usually, an additional capacitor is added in parallel with the rms current is equal to 11.9 A. The freewheeling diodes have
load. The complete solution could be considered equivalent to the same voltage rating and an average current of 4.6 A.
the use of an LC filter. The problem regarding the capacitor By considering the same derating factor, the switches oper-
charging current at start-up can be addressed by either using ating at low frequency shall have a blocking voltage of at least
a limiting resistor that is afterward shunted with a relay or by 550 V. Since these switches are clamped to the maximum input
synchronizing the grid connection at the zero-crossing instant. voltage by the diodes D3 and D4 in case of faults or transients
The required inductance value is calculated by considering in the grid side, a slightly higher voltage class of 600 V was
the moment when the ripple in the output current reaches the chosen. The maximum rms value of the current through them is
maximum value. The factor representing such instant depends 13.84 A.
on the modulation factor and is determined by the maximum 2) DC Link: Applying in (11) the values from the specifi-
value of (9), where M represents the voltage gain of the circuit cation results in a dc-link capacitance of at least 1.7 mF. The
(division of the output voltage by the input value) and wg the voltage rating is specified to be 1000 Vdc . In parallel to the
grid angular reference electrolytic capacitors is applied a metallized polypropylene
film capacitor with 20μF, rated for 1000 V that has a reduced
ΔIfactor (M, t) = sin(wg · t) − M · sin2 (wg · t). (9) value of ESR and equivalent series inductance.
3) Filter: In Fig. 11 are shown the curves of ΔIfactor from
The required inductance is then calculated in (10) by choos- (9) for both minimum and maximum modulation indexes based
ing a limit for the ripple in the output current (ΔIL ). Vg on the specified input-voltage range. The highlighted maximum
3126 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 57, NO. 9, SEPTEMBER 2010
B. Control Implementation
The output current is controlled by following a sinusoidal
reference with the angle being extracted from the grid by a
simple phase-locked loop algorithm. The result of the error
between the reference and measured current values is processed
by a proportional–integral plant that afterward goes through
normalization considering the input-voltage value. The whole
algorithm was implemented on a DSpace system.
A small deadband between the signals of the low frequency
switches (100 μs) shall be added to avoid grid short circuit.
Unlike other inverters where both filter inductors are always
active, it is necessary to observe here the requirement of reach-
ing a current equal to zero on one inductor before turning off the Fig. 14. Gate voltage of the switch T1 along with the blocking voltage
waveform and current through the respective inductor.
respective low-frequency switch (so that no overvoltages across
the switch will happen). For such purpose, a dead time between For the initial investigation, the DSpace system was em-
the positive and negative polarities of the reference current ployed to generate all the necessary PWM signals. In Fig. 12 are
waveform was added. This was done by taking the reference shown the gate signals measured in the output of the respective
signal and, respectively, adding and subtracting a certain offset drivers.
value from the negative and positive polarities. Fig. 13 shows the output current along with the current
through one of the inductor filters and the gate signal for the
respective high- and low-frequency switches. For the shown
C. Experimental Waveforms and Curves
waveform, a deadband of 500 μs was employed at the zero-
All the switches were chosen from the CoolMOS Technology crossing instant. On later investigations, a value of 100 μs
(IPW90R120C3 for T1 –T2 and IPW60R045CP for T3 –T4 ), as was employed. The THD of the current waveform for such
by this power level, they are characterized by the lowest pos- conditions was calculated using simulation results and reached
sible conduction and switching losses among all technologies approximately 2%.
employing silicon. Silicon carbide diodes were then employed In Fig. 14 is shown the switching characteristic of one of the
for the D1 –D4 so that the switching losses from the converter high-frequency switches along with the drain-to-source voltage
could be further reduced due to the absence of the reverse and current through the respective inductor. The different rise
recovery effect. and fall times of the gate voltage are explained by different
ARAÚJO et al.: SINGLE-PHASE TRANSFORMERLESS INVERTERS FOR GRID-CONNECTED PHOTOVOLTAIC SYSTEMS 3127
[34] S. Araújo and P. Zacharias, “Analysis on the potential of silicon car- Samuel Vasconcelos Araújo (S’05) received the
bide MOSFETs and other innovative semiconductor technologies in the B.S. degree in electrical engineering from the Fed-
photovoltaic branch,” in Proc. 13th Eur. Conf. Power Electron. Appl., eral University of Ceará, Fortaleza, Brazil, in 2006,
Sep. 2009. and the M.S. degree in renewable energies and en-
[35] B. Burger, “Power electronics for grid connected photovoltaic,” in Proc. ergy efficiency (RE2) from the University of Kassel,
Otti Workshop, Jun. 2008, pp. 163–216. Kassel, Germany, in 2007. He is currently working
[36] N. Mohan, T. M. Undeland, and W. P. Robbins, Power Electronics: toward the Ph.D. degree at the University of Kassel.
Converters, Applications and Design, 2nd ed. New York: Wiley, He is currently working in the Center of Com-
1995. petence for Distributed Electric Power Technology
[37] D. G. Holmes and T. L. Lip, Pulse Width Modulation for Power (KDEE), University of Kassel, and is responsible for
Converters, 1st ed. New York: Wiley, 2003. the development of power electronic circuits in part-
[38] H. Schmidt, S. Christoph, and J. Ketterer, “Current inverter for direct/ nership with industry. His main fields of interest are the design and optimization
alternating currents, has direct and alternating connections with an in- of power circuits and the investigation of innovative semiconductors based on
termediate power store, a bridge circuit, rectifier diodes and a inductive silicon carbide.
choke,” German Patent D E10 221 592 A1, Dec. 4, 2003.
[39] M. Victor, F. Greizer, S. Bremicker, and U. Huebler, “Verfahren
zum Umwandeln einer elektrischen Gleichspannung einer Gleichspan- Peter Zacharias (M’06) received the Dipl.-Ing.
nungsquelle, insbesondere einer Photovoltaik-Gleichspannungsquelle and Dr.-Ing. degrees in electrical engineering from
in eine Wechselspannung,” German Patent DE102004 030 912 B3, the “Otto-von-Guericke” University Magdeburg,
Jan. 19, 2006. Magdeburg, Germany, in 1979 and 1981, respectively.
[40] R. González, E. Gubía, J. López, and L. Marroyo, “Transformerless During 1983–1984, he was a Postdoctoral Re-
single-phase multilevel-based photovoltaic inverter,” IEEE Trans. Ind. searcher at the Polytechnical Institute of Kiev, Kiev,
Electron., vol. 55, no. 7, pp. 2694–2702, Jul. 2008. Ukraine, in power laser engineering. He was with the
[41] Z. Zheng-Yi, Z. Chang-Jiang, H. Yu, X. Ting, and Z. Liang-Bing, “Analy- University of Magdeburg until 1990 as an Associate
sis on voltage unbalance between the inner and outer devices in three level Professor of power electronics in the research fields
IGBT converters,” in Proc. IEEE Int. Conf. Power Electron. Drive Syst., of power electronics for high-voltage applications
Jul. 1999, vol. 1, pp. 218–224. and pulsed power. From 1990 to 1995, he was with
[42] T. Meynard and H. Foch, “Multi-level conversion: High voltage choppers Lambda Physik GmbH, Goettingen, Germany, responsible for power systems
and voltage source inverters,” in Proc. 23rd Annu. IEEE Power Electron. development for excimer lasers. In 1995, he was with the Institute for Solar En-
Spec. Conf., Jun.–Jul. 1992, pp. 397–403. ergy Technology (ISET), Kassel, Germany, and headed the Power Electronics
[43] J. Hantschel, “Direct current–voltage converting method for use in Department until 2001. In 2001, he was with eupec GmbH, Warstein, Germany,
inverter, involves clocking switch units such that high potential and input as Manager of power assemblies. Since 2005, he has been with the University
direct current voltage lie at inputs of storage reactor in magnetized and of Kassel, Kassel, Germany, as a Professor of electric power supply systems.
free-wheel phases, respectively,” German Patent DE102006 010 694 A1, Additionally, he is a member of the Board of ISET, e. V., Kassel.
Sep. 20, 2007.
[44] L. Huber, Y. Jang, and M. M. Jovanovic, “Performance evaluation of
bridgeless PFC boost rectifiers,” IEEE Trans. Power Electron., vol. 23, Regine Mallwitz (M’08) was born in Guestrow,
no. 3, pp. 1381–1390, May 2008. Germany, in 1970. She received the Dipl.-Ing.
[45] J. Yungtaek, M. M. Jovanovic, and D. L. Dillman, “Bridgeless PFC degree from the “Otto-von-Guericke” University
boost rectifier with optimized magnetic utilization,” in Proc. IEEE APEC, Magdeburg, Magdeburg, Germany, in 1994, af-
Feb. 2008, pp. 1017–1021. ter working on pulse power problems in excimer
[46] P. Zacharias, Use of Electronic-Based Power Conversion for Distributed lasers. After developing gas and solid-state lasers at
and Renewable Energy Sources—20 Years of Research on Power Conver- Lambda Physik GmbH, Goettingen, Germany, and
sion Systems. Kassel, Germany: ISET e. V, 2008, pp. 29–35. LISA Laser Products OHG, Lindau, Germany, she
[47] S. B. Kjaer, “Design and control of an inverter for photovoltaic received the Ph.D. degree from the University of
applications,” Ph.D. dissertation, Aalborg Univ., Aalborg, Denmark, Kassel, Kassel, Germany, in 1999.
2005. In 1999, she was with eupec GmbH, Warstein,
[48] J. Dodge, “Reduce circuit zapping from cosmic radiation,” Power Germany, as an R&D Engineer and was responsible for the development of
Electron. Technol., vol. 33, no. 9, pp. 20–27, Sep. 2007. the 1700-V IGBT module family, until 2004. Since 2005, she has been with
[49] D. C. Martins and I. Barbi, Introducao ao Estudo dos Conversores SMA Solar Technology AG, Nietestal, Germany. Her special scientific interest
CC-CA. ‘Florianópolis, Brazil, May 2005, pp. 375–398. is integration of power electronic devices.