1. technology and computing
2. electronic components

R. W. Erickson Department of Electrical, Computer, and Energy Engineering University of Colorado, Boulder An Inverter System
vpv(t)
vbus(t)
DC-DC
Inverter
EMI
PV
vac(t)
H1
d
vac(t)
dinv
PWM
PWM
ibus
Gc3(s)
vbus(t)
iac(t)
Phase-locked loop
H3
Gc3(s)
MPPT
–
+
vpv(t)
–+
Vref-pv
H2
iref(t)
Vref-bus
–
+
Gc2(s)
×
Sinusoid
unit amplitude
phase-locked to vac
Standards
IEEE 1547: standard for connecting a renewable energy source to the utility grid
•  Current harmonic limit (THD < 5%)
•  Anti-islanding (detect loss of grid, shut down within 1 sec)
•  Disconnection when grid frequency or grid voltage is out of bounds
National Electric Code
UL 1741
Weighted Efficiency standards: California Energy Commission (CEC)
Power level,
% of rated
Weight
100%
0.05
75%
0.53
50%
0.21
30%
0.12
20%
0.05
10%
0.04
•  Provides a way to compare products
of different companies
•  Weightings reflect typical distribution
of array power experienced in
California
Microinverters
One inverter per panel
•  Mounted on or near the panel—on roof
•  MPPT on per-panel basis
•  Conventional AC wiring reduces
Balance-of-system cost
•  Straightforward expandability
•  Reliability? Efficiency?
Rated temperature?
Ascension Tech. microinverter, 1998
Enphase
microinverter,
2008
Elements of a Microinverter System
Pdc
i
+
PV
Cells
i
v
v(t)
i(t)
Pac
Energy
storage
DC-DC
Converter
+
DC-AC
Inverter
iac
+
Microinverter
power train:
• DC-DC converter
(high boost ratio)
v
vcap
vac
–
–
–
Pdc
Pac
vac(t)
iac(t)
t
Central box
Microinverter
Microinverter
Power stage
MPPT Current
control
Anti-islanding
Power stage
MPPT Current
control
Anti-islanding
Roof
AC
Transient
protection
AC
utility
AC
disconnect
Communications
Computer
Smart grid
• Energy storage
capacitor
• Inverter
Rooftop system
• Microinverters
include most or all
of grid interface
control
• Central box
boosts the low PV input voltage to a higher voltage. The inverter stage generates the AC current that is injected to the
AC line. Despite various new topologies that have been demonstrated in recent literature [5], the typical low-cost
micro-inverter is still designed either as a full-bridge stage, or as a buck stage with an unfolder stage. The unfolder
Microinverter Approaches
stage, if present, switches at the zero-crossings of the line voltage to convert the rectified sinusoid at the buck output
to a full sinusoid on the AC line.
H-bridge inverter
Buck converter plus unfolder
Unfolder: similar to bridge
rectifier, but power flows in
reverse
Fig. 1. Common micro-inverter power stages. (a) Full bridge. (b) Buck stage with an unfolder
stage. direction. Implemented
using transistors that switch at
ac line frequency
An illustration of the Boundary Conduction Mode (BCM) waveform is shown in Fig. 2. Although it is softswitching, and operates with low RMS current, a disadvantage of BCM is its high average switching frequency, which
causes high switching losses. As demonstrated by equation (1), the BCM waveform has the highest switching
Inverter sinewave synthesis approaches
We can employ any of the approaches we have already discussed for PWM
rectifier systems:
•  Average current control
•  Peak current control
•  Boundary conduction mode
•  Hysteretic control
•  Discontinuous conduction mode control
•  Cycle-by-cycle control
(and there are a few we didn’t discuss, most notably harmonic elimination, that
could be employed for either rectifiers or inverters)
lowest possible peak current in DCM, and as a result the switching frequency in BCM is maximal. Equation (1) also
predicts that the switching frequency of BCM increases at low output powers, creating a switching frequency profile
that causes disproportional switching losses at low powers. This is demonstrated by the last expression in equation (1)
Synthesizing a Sinusoidal Current:
results in a higher switching frequency, so the switching frequency and switching losses in BCM substantially increase
Boundary Conduction Mode (BCM)
at low voltages and low powers.
for which the power-factor is unity and the switching frequency is proportional to Rout. A lower output voltage vout(t)
Inductor current waveform, BCM
Fig. 2. Illustration of the inductor current in BCM, showing soft switching transitions (ZVS, ZCS) and the variations in switching
frequency over the line cycle.
The switching frequencies in DCM and BCM are given by:
DCM:
fs t
vout t
2 L iout t
BCM:
fs t
vout t
2 L iout t
v t
1 out
vdc t
1
2iout t
i pk t
vout t
vdc t
2
Loss components at different solar
irradiance levels, BCM
(300 W, 240 Vac example)
(1)
BCM with unity power factor :
fs t
Fig.
v t
Rout
1 out
2L
vdc t
, where Rout
vout t
iout t
Fig. 3. Distribution of losses in BCM. The vertical bars represent average losses over an AC line cycle. The losses are shown in percent
the average
AC outputat
power.
Switching
losses
dominateThe
at low
powers.
3 shows how the totalrelative
loss into BCM
distributes
various
output
powers.
losses
in this figure are averaged
over a line cycle, and are shown in percent relative to the cycle averaged output power. The total loss is composed of
BCM waveform. The DC operating points are selected with constant ratio of voltage and current vout/iout = Rout, and
thus reside on the same output sinusoid. The efficiency is computed according to the calibrated loss model presented
in section IV, with conditions as follows: Rout = 215.1 Ω, average AC power of 225 W, bus voltage of vbus = 425 V,
and an inductor of 300µH built on a PQ 26/20 core. Each curve is label by its output power pout = voutiout, which is
Discontinuous
conduction
mode
(DCM)
given in percent relative to the maximal instantaneous output power of 450 W.
•  Higher conduction loss
•  program
Lower that
switching loss
To find the optimal values of the inductance L and magnetic flux density ΔBmax, we used a computer
•  A net improvement in
numerically scanned the weighted efficiency of every combination of these two parameters. The program runs over
CEC efficiency
each output power level, and over single DC operating points in the output sinusoid, and computes the efficiency at
each operating point, using the calibrated loss model. The resulting efficiency data is averaged according to the CEC
efficiency formula (see section V). Typical results of this simulation are shown in Fig. 10, for a PQ 26/20 magnetic
core, and conditions as detailed in Table I.
CEC efficiency [%]
Fig. 4. By increasing the peak current 99.2
in DCM, the switching frequency is reduced, while the average inductor current (iout) is unchanged.
99
98.8
98.6
98.4
98.2
98
97.8
97.6
97.4
Weighted efficiency vs. inductor
size, DCM vs. BCM
300 W, 240 Vac example
constant peak current
BCM
250
350
450
550
650
750
850
inductance [µH]
Fig. 10. Weighted CEC efficiency with a PQ 26/20 magnetic core, for a BCM controller and the proposed constant peak current
controller.
0
200
400
DC output power pout [W]
600
Fig. 14. Efficiency measurements at DC operating points, for various average AC powers
Fig. 12.
Inverter
waveforms
at efficiency
a DC operating
Ch1 (yellow)
cycle-by-cycle
capacitor over
voltagea vline
int(t) , Ch2 (blue) auxiliary
At each
power
level,
the AC
is point.
computed
by averaging
theintegration
DC efficiencies
cycle. The results
are
Measured Results: 300 W Microinverter Prototype
winding voltage sensor, at comparator output, Ch4 (green) inductor current iL(t). Conditions: Vdc= 426.8 V, Idc= 0.98 A, vout= 330.1 V, iout=
A.
again averaged by the CEC weighted average formula 1.259
to obtain
the overall CEC efficiency. The AC efficiency is
computed by equation (9), and the results are summarized in Table II.
/ ac
AC line voltage
pout t dt
AC efficiency
0
AC
/ ac
0
pout t
DC
pout t
where pout t
Pac sin 2
ac
t
(9)
Current reference
dt
Filtered inverter current
In this equation, Pac is the average AC power, pout(t) is the output power at a DC operating point, and ηDC(pout) is the
inductor current
efficiency at those operating points, as shown in Fig. 14. The weighted CEC efficiency is found toInstantaneous
be 99.15 %.
Fig. 13. Inverter waveforms over a line cycle. Ch1 (yellow) AC line voltage sensor, Ch2 (blue) AC current iac(t), Ch3 (magenta) reference
signal iref(t), Ch4 (green) inductor current
iL(t). Conditions:
Vdc= 425 V, Idc= 0.462 A, Rload=253 Ω
TABLE
II
CEC Efficiency
CEC
Average
average AC
average AC
power
weight
loss over
power P
The efficiency of the inverter
points.efficiency
These tests are done with a power supply
levelis measured at static DCac operating
AC cycle
100resistor
%
0.05
300 W
2.6 W
99.13 %
at the input and a variable load
(R
load) at the output. To increase the accuracy of the measurements, the meters
75 %
0.53
225 W
1.97 W
99.12 %
at the input and output are filtered
The1.26
efficiency
results
50 % by large
0.21 EMI inductors.
150 W
W
99.16
% are shown in Fig. 14. The various
30 %
0.12
90 W
0.7 W
99.22 %
curves in the figure correspond
to
tests
with
different
average
AC
powers
(P
ac). At each such test the load resistor is
20 %
0.05
60 W
0.45 W
99.24
%
10 instantaneous
%
0.04 output
30power
W
0.24 W in the99.2
% 0 … 2Pac.
set to Rload=Vac,rms2/Pac, and the
is scanned
range
Overall weighted CEC efficiency = 99.15 %
Development of Electrical Model
of the Photovoltaic Cell, slide 1
Photogeneration
Semiconductor material absorbs photons and
converts into hole-electron pairs if
Photon energy h > Egap
(*)
• Energy in excess of Egap is converted to
heat
• Photo-generated current I0 is proportional to
number of absorbed photons satisfying (*)
photon
+
–
Charge separation
Electric field created by diode structure separates holes and electrons
Open circuit voltage Voc depends on diode characteristic, Voc < Egap/q
Development of Electrical Model
of the Photovoltaic Cell, slide 2
Current source I0 models photo-generated current
I0 is proportional to the solar irradiance, also called the “insolation”:
I0 = k (solar irradiance)
Solar irradiance is measured in W/m^2
Development of Electrical Model
of the Photovoltaic Cell, slide 3
Diode models p–n junction
Diode i–v characteristic follows classical
exponential diode equation:
Id = Idss (eLVd – 1)
The diode current Id causes the terminal
current Ipv to be less than or equal to the
photo-generated current I0.
Development of Electrical Model
of the Photovoltaic Cell, slide 4
Modeling nonidealities:
R1 : defects and other
leakage current mechanisms
R2 : contact resistance and
other series resistances
Cell characteristic
Cell output power is Ppv = IpvVpv
At the maximum power point
(MPP):
Vpv = Vmp
Ipv = Imp
At the short circuit point:
Ipv = Isc = I0
Ppv = 0
At the open circuit point:
Vpv = Voc
Ppv = 0
Maximum Power Point Tracking
Automatically operate the PV panel at its
maximum power point
Some possible MPPT algorithms:
• Perturb and observe
• Periodic scan
I-V curve
with partial
shading
• Newton s method, or related hillclimbing algorithms
• What is the control variable? Where is
the power measured?
Power vs. voltage
Example MPPT: Perturb and Observe
•
A well-known approach"
•
Works well if properly tuned"
•
When not well tuned, maximum power
point tracker (MPPT) is slow and can
get confused by rapid changes in
operating point"
•
A common choice: control is switch
duty cycle"
Basic algorithm!
!
Measure power"
Loop:"
• Perturb the operating point in
some direction"
• Wait for system to settle"
• Measure power"
• Did the power increase?"
Yes: retain direction for next
perturbation"
N: reverse direction for next
perturbation"
Repeat"
!
Control Issues:
MPPT by Perturb-and-Observe
,
,
Measured power vs. commanded
PV voltage
Key elements of digital controller
•
•
•
•
•
Find PV voltage that maximizes power output
Switching converter is high noise environment
This “noise” is partly correlated to the control,
and hence isn’t entirely random
The highly-filtered dc control characteristic
exhibits many small peaks (“traps”), where
P&O algorithm gets stuck
More noise makes P&O work better!
Magnified view
Typical experimental data
()
α
()
α
→
• Perturb-and-observe step time of 15 msec
• Perturb-and-observe algorithm may take minutes to find max power
• Weather can change in seconds
• Improved algorithm achieves max power in seconds
• Adaptive algorithm finds max power quickly, then reduces jitter size to
improve equilibrium MPPT accuracy