Professional Documents
Culture Documents
Design Optimization of A High Step-Up DC-DC Converter For Photovoltaic Microinverters
Design Optimization of A High Step-Up DC-DC Converter For Photovoltaic Microinverters
Abstract—A design optimization of a generalized high step-up Commercial PV microinverters commonly use single-phase
dc-dc converter applied as front-stage in a photovoltaic (PV) architecture and adopt single-stage or two-stage system
microinverter is proposed. The design methodology optimizes the configurations. Nevertheless, to avoid the usage of electrolytic
weighted California Energy Commission (CEC) efficiency of the capacitors as power decoupling storage elements, the two-stage
generalized converter considering not only the classical solution has shown to be the best alternative [2]. In this
parameters, as turns ratio and switching frequency, but also the configuration, normally a dc-dc converter with high step-up
topology structure through different connections of voltage gain is used to adapt the low PV module voltage to the required
multiplier cells. Two design optimizations have been carried out. dc bus voltage of the dc-ac converter employed for ac utility
The first maximizes only the CEC efficiency, obtaining results
connection. Such high step-up dc-dc converter can be isolated
around 98%. In the second it was added terms related to cost and
volume in the objective function, which results in a converter
or non-isolated, however, the non-isolated ones are capable of
with lower CEC efficiencies – around 96% - but with a higher achieving higher efficiency levels and lower cost [3].
power density and lower cost. In the last decade several high step-up non-isolated dc-dc
converters have been proposed in the literature [4]-[12]. Many
Keywords— design optimization; high step-up dc-dc converter; of these converters are based on the classical Boost converter
PV microinverter; weighted efficiency. and use the same high-step-up techniques, as coupled inductors
and voltage multiplier cells. Due to such plethora of similar
I. INTRODUCTION converters, in [13] it was developed a generalization, which
resulted in the so-called Boost Converter with Gain Cell
Microinverters (or ac-modules) have become an important (BCGC). This generalized approach opened the possibility to
portion of the photovoltaic (PV) market, reaching an estimated apply the converter structure as a design feature. For example,
market share of about 2% in 2015 [1]. Although still distant a question that may appear in a project involving the necessity
from string inverters (37%) and central inverters (61%) due to of high voltage gain is whether is better to increase the
its current higher energy cost, the microinverters present some transformer ratio of the coupled inductor or to use more voltage
interesting advantages that help its market increase such as, for multiplier cells. This is a classic optimization problem that is
example: not easy to answer; it will depend of several variables and the
Less reduction on global power due shading, because desired characteristics of the project. To solve this issue, this
the reduction of power is restricted to the micro-inverter paper proposes a design optimization algorithm to maximize
whose panel is shaded; the efficiency of the BGCG applied to PV microinverters, in a
way that the gain cell is used as a design parameter (an
Individual maximum power point tracking (MPPT) and independent variable) together with the turns ratio, the
consequent increase in the system efficiency; switching frequency, the magnetizing inductance, as well as,
electrical components and materials.
Individual maintenance of panels without need for
complete disconnection; The remainder of this study is organized as follows.
Initially, section II introduces the BCGC. Afterward, in section
Flexible initial installation cost; III, it is presented the design optimization problem, followed
Facilitated expansion, panel by panel; and by a description of the loss models in section IV. Lastly,
section V presents some experimental results to validate the
Requirement of only an ac network on rooftop. study and the conclusions are drawn in section VI.
1 an a b nD
M CCM , (1)
1 D
D 1 D Vo 2
2
433
where:
C. Constraints
The design variables are subjected to several constraints
defined according to the design specifications, physical
limitations, and component safety operating areas:
The lower boundary for the switching frequency is 50
kHz and the upper boundary is 150 kHz.
The lower and upper boundary of the turns ratio
depends of the gain cell indexes and must guarantee that
0.2 ≤ D ≤ 0.7.
The voltage/current ratings of the semiconductor
devices must be at least 150% of the maximum applied
voltage/current.
The magnetizing inductance should be between 100%
and 200% of the critical magnetizing inductance at 10%
of the output power (ko = 0.1).
The peak value of the flux density in the coupled
inductor core cannot exceed the maximum value
defined by its material.
The current density in the windings wires must stay
between 200 e 600 A/cm2.
The estimated junction temperature must be no more
than 90% of the maximum junction temperature. No
heat sink is considered. Fig. 3. Optimization algorithm overview.
434
The turn-off time toff has been obtained by: so that m is the harmonic number, Rdc is the dc resistance, and
Rac is ac resistance of the windings. Such resistances can be
Qsw calculated by [18]:
toff , (15)
Ig
N ( w ) 20 1 20 Tmax 20 MLT
Rdc ( w ) , (24)
Vgs VP n( w )
Ig . (16)
Rg
4nl ( w ) 2 1 n( w) 2
where Ig is the gate current, Vgs is the gate-source supply Rac ( w ) Rdc ( w ) FR , m GR , m k F , (25)
12d s 2 2 2 d t ( w )
voltage, VP is the Plateau voltage, and Rg is total gate
resistance.
ber0 ( m )bei1 ( m ) ber0 ( m )ber1 ( m )
ber1 ( m ) 2 bei1 ( m ) 2
B. Diodes FR , m m , (26)
The diode loss model will be resumed in conduction losses: 4 2 bei 0 ( m )ber1 ( m ) bei 0 ( m )bei1 ( m )
ber1 ( m ) 2 bei1 ( m ) 2
PD I D , rms 2 rD I D , avVT . (17)
ber2 ( m )ber1 ( m ) ber2 ( m )bei1 ( m )
so that rD is the dynamic resistance and VT is the threshold m 2 d cu 2 ber0 ( m ) 2 bei 0 ( m ) 2 . (27)
GR , m
voltage: 2 2 bei 2 ( m )bei1 ( m ) bei 2 ( m )ber1 ( m )
ber0 ( m ) 2 bei 0 ( m ) 2
rD rDO kr T j , (18)
where:
VT VTO kvT j . (19)
d cu
m , (28)
m 2
C. Coupled Inductor
As any magnetic device, the losses in the coupled inductor
1
can be divided into core losses and copper losses: m , (29)
mf s 0 m
Pm Pcore Pcopper . (20)
MLT is the mean length per turn, N(w) is the number of turns,
n(w) is the number of litz-wire strands, dt is the diameter of the
1) Core Losses litz wire, dcu is the copper diameter of each strand, ds is the
The core losses were calculated by the NSE method [16] (or total diameter of each strand, nl(w) is the number of layers, ρ20,
iGSE [17]): α20, and σ are well-known parameters of the copper wire. The
correction field factor kF proposed by [19] is given by:
Pcore k N 2 f s B D1 1 D 1 Ve ,
(21)
3.44 0.505 0.688
2
kF . (30)
k
kN 2
, (22)
2 t w 3d wl
2
1
0
cos d
3 w
. (31)
where α, β, and k are the Steinmetz parameters, ΔB is the so that dwl is the distance between the windings and the core
peak-to-peak magnetic induction, and Ve is the volume of the leg, tw is the windings thickness, and w is the width of the core
core. window.
D. Capacitors
2) Copper Losses
The copper losses in each winding (w) can be calculated by: The capacitor losses were calculated by (32), where rESR is
the resistance of its equivalent series resistor.
Pcopper ( w) Rdc ( w) I ( w) ,rms 2 Rac ( w), m I ( w) ,m , rms 2 , (23) PC rESR I C , rms 2 (32)
m 1
435
TABLE I. RESULTS OF OPTIMIZATION I TABLE II. RESULTS OF OPTIMIZATION II
Fig. 4. Effiency results of the BCGC from Optimization I. Fig. 5. Effiency results of the BCGC from Optimization II.
V. OPTIMIZATION RESULTS much higher than needed). To overcome these issues, a second
optimization was carried out.
The optimizations were performed by the GlobalSearch
method presented in the Toolbox Global Optimization [20] of B. Optimization II
the software Matlab®. For the viability of a continuous
variable approach the gain cell indexes abcp are fixed and the A second optimization has been carried out considering
same algorithm is run for different index values (with also cost and volume terms in the objective function:
max(a,b) = 3). Two design optimizations results are described
below. It should point out that for both the optimizations the
min f x k1CEC (x) k2Ve k3 c$ , (27)
capacitors were preselected; the other devices are chosen from
a database.
so that k1, k2, and k3 are constants for weighting (in this case
A. Optimization I was considered k1 = k2 = k3 = 1/3), Ve is the volume of the
core, and c$ is the cost sum of all selected components.
The following specifications were considered for the first
optimization design: Vi = 20-30 V (the middle value is taken in The following specifications were considered for the
the optimization algorithm), Vo = 400 V, Po = 200 W, and second optimization design: Vi = 20-40 V (the middle value is
Ta = 40 oC. Table I presents the optimal results. Fig. 4 shows taken in the optimization algorithm), Vo = 400 V, Po = 250 W,
the efficiency curves of the obtained design. The solid line and Ta = 40 oC. Table II presents the optimal results. Fig. 5
represents the theoretical curve calculated by the power loss shows the experimental efficiency curves of the obtained
model depicted in section IV, whereas the dotted line is design. The weighted CEC efficiencies are also illustrated in
obtained from experimental results. The weighted CEC Fig. 5, demonstrating results around 96%.
efficiencies are also illustrated in Fig. 4, demonstrating results
around 98%. One can note that the second optimization also brought
about only one voltage multiplier cell, demonstrating again that
One can see from Table I that a BCGC with only one a lower component count results in a better efficiency
voltage multiplier cell was chosen, probably due to its lower performance in the case of the BCGC. Moreover, due to the
component count. In addition, it is possible to note that, since inclusion of cost and volume terms in the objective function the
no volume or cost restriction was considered, the optimization efficiency results have reduced in comparison with the ones
resulted in a large E55/28/21 core and in some overrated from Optimization I; although the slightly difference in the
devices (e.g., the diode SCS215AJ has a current rating of 15 A, specifications.
436
Fig. 6 shows a photograph of the BCGC from Optimization
II applied as front-stage in a 250-W microinverter.
VI. CONCLUSION
This paper has proposed a design optimization algorithm
for the generalized Boost converter with gain cell applied to
photovoltaic microinverters. Besides the classical ones (turns
ratio, switching frequency, and so on), the proposed design
optimization also considers the gain cell as a design parameter.
Two optimized designs were performed. The first demonstrated
the accuracy of the power loss models and had CEC efficiency
results near to 98%. The second showed the flexibility of the
design optimization in allowing the modification of the
objective function to include volume and cost merits. Finally,
although recent works have proposed converters with
increasing static gain, the proposed optimization design has
shown that for the highest efficiency is better to increase the Fig. 6. Prototype photografy of a 250-W microinverter with the dc-dc
converter designed by the Optimization II.
turns ratio than add more voltage multiplier cells.
437