You are on page 1of 10

Energy Conversion and Management 69 (2013) 17–26

Contents lists available at SciVerse ScienceDirect

Energy Conversion and Management


journal homepage: www.elsevier.com/locate/enconman

Three-phase three-level grid interactive inverter with fuzzy logic based


maximum power point tracking controller q
Necmi Altin a,⇑, Saban Ozdemir b
a
Department of Electrical-Electronics Eng., Faculty of Technology, Gazi University, Ankara, Turkey
b
Ataturk Vocational School, Gazi University, Ankara, Turkey

a r t i c l e i n f o a b s t r a c t

Article history: In this study, three-phase single stage grid interactive inverter with maximum power point tracking
Received 1 December 2012 capability is proposed. The proposed system consists of three-level neutral point clamped inverter, LCL
Accepted 19 January 2013 output filter, line frequency transformer, PI current regulator and fuzzy logic based maximum power
Available online 27 February 2013
point tracking algorithm. Rate of change of photovoltaic power and voltage are defined as input variables,
and the change in reference current is defined as output variable for the fuzzy logic controller. The pro-
Keywords: posed maximum power point tracking algorithm is robust with respect to parameter variations of pho-
Three-level NPC inverter
tovoltaic system with adaptive feature of fuzzy logic controller. Maximum power point tracking
Grid interactive inverter
MPPT
algorithm determines the inverter current reference depending on the system conditions such as irradi-
Fuzzy logic controller ation level and temperature, and PI regulator shapes the inverter output current. Two capacitors’ voltages
of neutral point clamped inverter are also balanced. Simulation and experimental results show that the
proposed inverter system has fast transient response and can track the maximum power point of PV sys-
tem even if atmospheric condition changes rapidly. Also, the inverter output current is in sinusoidal
waveform and in phase with line frequency and phase. In addition, total harmonic distortion level of
the inverter output current is in the limits of international standards (<5%) and efficiencies of maximum
power point tracking algorithm and total system are measured as 98.78% and 93.12%, respectively.
Ó 2013 Elsevier Ltd. All rights reserved.

1. Introduction system is commonly used in remote areas applications such as sa-


tellite systems, mobile applications or telecommunication systems.
As a result of renewable energy studies, the solar energy has In grid interactive operation mode battery groups are not neces-
gained importance with its superior advantages such as being sary, so cost and size of the system is reduced [1,2].
inexhaustible energy source, accessible nearly all around the Grid interactive PV systems seem as an efficient solution to the
world, pollution free operation and little maintenance require- environmental problems with their decreasing costs. However en-
ment. Solar energy is converted to DC electrical energy via photo- ergy conversion efficiency of the PV systems is still low and this
voltaic (PV) modules. Since most of the loads and public grid are subject attracts many researchers’ and studies’ attention. Mainly
AC, an inverter is required. This PV system can operate in stand- these studies can be grouped into two categories. First one is about
alone mode and grid interactive mode. In stand-alone mode, huge the PV modules, new production technologies, materials and aims
battery groups should be used to obtain the continuity of the en- in order to increase the energy conversion efficiency of the PV
ergy. This requirement increases the size and cost of the system. modules. The second one is about the converter control. PV module
The stand-alone PV system is economically superior in which other cannot be assumed as a constant DC current source. Its output
energy sources are impossible or difficult to use. Therefore, this power is varied depending on the load level, temperature and irra-
diation, and it generates maximum power at a voltage and current
level for a specific temperature and irradiation. Variation of maxi-
q
The short form of this paper was presented in The European Workshop on
mum power point (MPP) with irradiation level for different tem-
Renewable Energy Systems (EWRES 2012), which took place on 17–28 September perature levels is shown in Fig. 1 [3].
2012, Antalya, Turkey. The PV module should be operated at the MPP which is valid for
⇑ Corresponding author. Address: Department of Electrical-Electronics Eng., the temperature and irradiation level to obtain maximum effi-
Faculty of Technology, Gazi University, 06500 Besevler, Ankara, Turkey. Tel.: +90
ciency. However, maximum efficiency can be obtained for only un-
312 202 85 25; fax: +90 312 212 13 38.
E-mail addresses: naltin@gazi.edu.tr (N. Altin), sabanozdemir@gazi.edu.tr (S.
ique operation point. Because other operation points, except MPP,
Ozdemir). decrease the efficiency, load power level should be adjusted

0196-8904/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.enconman.2013.01.012
18 N. Altin, S. Ozdemir / Energy Conversion and Management 69 (2013) 17–26

these inverters becomes popular because of some advantages such


as decreasing size of the filters, lower dV/dt ratio, lower electro-
magnetic interference (EMI) level and lower switching losses [5–
8]. Furthermore, commercially available three-level power mod-
ules decrease the cost and simplify the design of the multilevel
inverters.
In addition, single-stage or two-stage grid interactive inverters
can be designed. In two-stage converters, a DC–DC converter per-
forms MPPT process and DC–AC inverter provides grid interactive
operation. Hence, the control complexity of the converter is re-
duced. Both MPPT process and grid interactive operation are car-
ried out by a DC–AC inverter in single-stage grid interactive
inverters. Although control structure of single-stage grid interac-
tive inverter is complex, total efficiency is improved with elimina-
tion of DC–DC converter requirement.
Fig. 1. Effect of solar irradiation on PV module P–V curve for constant temprature. Various MPPT methods have been discussed in past literature to
obtain maximum energy conversion efficiency for all operation
conditions. These methods can be grouped as off-line methods
and on-line methods. Off-line methods use characteristic values
considering the PV efficiency curve to increase the module effi- of PV system such as open circuit voltage, short circuit current,
ciency. This can be implemented through maximum power point ambient temperature and irradiation level. These values should
tracking (MPPT). Variation of maximum power point (MPP) with be read from PV system. An additional sampling PV module should
irradiation level (Ir) for different temperature levels (T) is shown be installed to obtain these data. Although disconnecting the PV
in Fig. 1 [3]. system from the load periodically is another way to obtain these
In this study, three-phase three-level inverter with MPPT capa- data, switching off the loads is a problem. When off-line methods
bility for grid interactive PV systems is proposed. MPPT is ensured are simple and low cost, requirement of sampling module, switch-
with an intelligent control scheme which is fuzzy logic control ing off the loads and difficulties of obtaining accurate mathemati-
based designed. The fuzzy logic controller (FLC) has high perfor- cal equations are important disadvantages of these methods. On
mance under rapidly changing atmospheric conditions. The pro- the other hand, external effects such as shadowing, aging and pol-
posed system covers three-phase, three-level inverter neutral lution affect the performance and operation at the MPP. So, the
point clamped (NPC) inverter, line frequency transformer, LCL out- best behavior may never be achieved [3].
put filter, fuzzy logic based MPPT controller and PI current regula- On-line methods read the PV system parameters and manipu-
tor. Proposed MPPT controller defines the reference current value late the system to obtain real MPPT independent from environ-
of the inverter and PI regulator shapes the inverter output current. mental conditions and PV system parameters [9]. On-line
Simulation and experimental studies are carried out for 1000 W in- methods have different response speeds depending on their con-
put power. Both simulation and experimental results show that trol schemes, but generally these methods are slower than the
proposed inverter has fast transient response and tracks the max- off-line methods. The perturbation & observation (P&O) algorithm
imum MPP of the PV system. Also, inverter output current is in (also called hill climbing technique) is the most common on-line
phase with line voltage and frequency, and in sinusoidal waveform. method. It can be easily implemented and has relatively high
In addition, current harmonics level is in the limits of international tracking efficiency. On the other hand, this method is unsuitable
standards (<5%). Energy conversion efficiency of the PV system im- for rapidly changing atmospheric conditions. If the irradiation
proved and efficiency of the whole system is measured as 93.12% changes rapidly, the P&O method cannot be fast enough to track
while the MPPT efficiency is 98.78%. the MPP. Also, output power perturbs around the MPP and it
causes power loss and decreases the efficiency [9–11]. Incremental
2. Discussion of previous works conductance (IC) method is proposed to overcome this drawback.
This method tracks the slope of the PV module power versus volt-
Generally voltage source inverter topology is used in grid inter- age curve to determine the MPP. Tracking accuracy and dynamic
active inverters. On the other hand, voltage and current levels and performance under rapidly changing atmospheric conditions are
switching frequencies of the static switches limit the maximum improved by this method. However, the iteration step size of IC
power levels of the inverter. Also, cost of the switches exponen- method determines the accuracy at steady state and the response
tially increases at high power levels. Several methods and modern speed of the MPPT. Thus, this value has to be determined properly
inverter topologies have been proposed to increase the power level [9–11].
of the inverters and overcome these drawbacks. Using multiple Neural networks, which are well adapted for microcontrollers,
inverters is one of these methods. However, this method is not eco- are also used for MPPT. This method improves the tracking perfor-
nomical. Using serial or parallel connected switches increase the mance. However, neural networks require no knowledge of the PV
voltage and/or current, thus inverter power level increases. But module parameters, and the trained neural network can provide
non-identical components, current paths, small differences be- accurate MPPT. Since most PV arrays have different characteristics,
tween the control signals and differentiation of the switch charac- a neural network has to be specifically trained for the PV array with
teristics with temperature may cause unbalanced voltage and/or which it will be used. Besides, characteristics of a PV array chang-
current sharing between the serial and/or parallel connected ing in time imply that the neural network has to be periodically
switches [4]. trained to guarantee accurate MPPT [8,12].
Recently, multilevel inverter structures are used to increase the Fuzzy logic control theory, which is a mathematical discipline
inverter voltage and power level to achieve higher power levels. based on vagueness and uncertainty, allows one to use non-precise
Although number of switches is greater than two level inverters, or ill-defined concepts. Since it is nonlinear and adaptive in nature,
higher voltage and power level can be achieved by using switches fuzzy logic control has robust performance under parameter vari-
which are lower rated values in multilevel inverters. Also, use of ation and load disturbances [1]. Many control approaches and
N. Altin, S. Ozdemir / Energy Conversion and Management 69 (2013) 17–26 19

applications of fuzzy logic control have appeared in the literature


since Mamdani published his experiences using a fuzzy logic con-
troller on a test-bed plant in a laboratory [13]. The fundamental
advantage of the fuzzy logic controller over the conventional con-
troller is less dependent on the mathematical model and system
parameters as known widely [14–18]. So, it is useful for MPPT pro-
cess rapidly changing climate conditions [19].

3. PV model

PV cells compose of integrated thin layered p–n semi-conduc-


tors. Output I–V characteristics of PV cell are very similar to the Fig. 3. Model of PV circuit.
diode characteristic in the dark. Light affects the photons, and
movement of electrons occurs, thus current flow is generated. This
PV module output current is related to the irradiation level. In addi-
current flows through the external load when the PV cell is short
tion, operation temperature of PV module is related to module out-
circuited and flows through the characteristic p–n diode when
put voltage and current directly. Effect of temperature and
the PV cell is open circuited [20]. Electrical equivalent circuit of
irradiation level on module voltage and current values are given
PV cell can be shown with a current source, parallel connected
in the following equations:
resistor and diode, and a serial resistor to this resultant circuit as
seen in Fig. 2. Mathematical model of this circuit can be also writ- V OC ðTÞ ¼ V 0OC  ð1 þ C V  ðT  T ref ÞÞ; ð6Þ
ten as Eq. (1) [21]:
    Ir
qðV þ I  RS Þ V þ I  RS ISC ðT; Ir Þ ¼ I0SC  ð1 þ C I  ðT  T ref ÞÞ  ; ð7Þ
I ¼ IG  I0  exp 1  ; ð1Þ Irref
nkT RP
where I is cell output current (A), IG is light-generated current (A), I0 V MP ðTÞ ¼ V 0MP  ð1 þ C V  ðT  T ref ÞÞ; ð8Þ
is cell reverse saturation current or dark current (A), q is electronic
charge (1, 6  1019 C), V is cell output voltage (V), n is ideality fac-
Ir
tor, k is Boltzmann’s constant (8.65  105 eV/K) and T is cell tem- IMP ðT; Ir Þ ¼ I0MP  ð1 þ C I  ðT  T ref ÞÞ  ; ð9Þ
Irref
perature (K). This model mainly focuses on physical effects. If Fig. 2
is re-arranged for irradiation and temperature levels, Fig. 3 is ob- where V 0OC ; I0SC ; V 0MP and I0MP are PV module voltage and current val-
tained and PV output voltage value of this circuit can be calculated ues for T = Tref and Ir = Irref. Also T is temperature, Tref is temperature
with the following equation: value which is given in datasheet, Ir is irradiation level and Irref is
  NðT;Ir Þ  irradiation value which is given in datasheet, VOC is module open
V OC ðT;Ir Þln 2 I
ISC ðT;Ir Þ circuit voltage, ISC(T, Ir) is module short circuit current, VMP(T) is
lnð2Þ
 ðRS ðT; Ir Þ  ðI  ISC ðT; Ir ÞÞÞ module output voltage at MPP, IMP(T, Ir) is module current at MPP.
VðI; T; Ir Þ ¼ :
1þ RS ðT;Ir ÞISC ðT;Ir Þ CV is a voltage–temperature coefficient and is given by Eq. (10). In
V OC ðT;Ir Þ
addition, CI is a current–temperature coefficient and given by Eq.
ð2Þ (11). These coefficients show the amount of module output voltage
Eqs. (3)–(5) can be written for RS(T, Ir), N(T, Ir) and a(T, Ir) expressions and current variations versus temperature [22]:
in Eq. (2) respectively:
%DV 0OC
V OC ðT; Ir Þ  V MP ðT; Ir Þ CV ¼ ; ð10Þ
RS ðT; Ir Þ ¼ ; ð3Þ DC 
IMP ðT; Ir Þ
  %DI0SC
CI ¼ : ð11Þ
ln 2  2aðT;Ir Þ DC 
NðT; Ir Þ ¼   ; ð4Þ
ln IIMP ðT;Ir Þ
SC ðT;Ir Þ
4. Three-phase three-level NPC inverter
 
V MP ðT;Ir Þ  1 þ RS ðT;I r ÞISC ðT;Ir Þ
þ RS ðT;Ir Þ  ðIMP ðT; Ir Þ  ISC ðT; Ir ÞÞ
V OC ðT;Ir Þ The multilevel inverters are mentioned by Nabae [23]. Since
aðT; Ir Þ ¼ :
V OC ðT;Ir Þ number of components is higher than the conventional inverters,
ð5Þ control difficulties and more complex structures, initially these
inverters cannot be used commonly. However, multilevel inverters
become common especially in high power applications with pro-
gress in microcontroller technologies and there is reduction in
power switching devices’ cost. The multilevel inverters have some
advantages respect to conventional two-level inverters such as
lower output current/voltage harmonics, achieving high voltage
and high power levels by using low voltage switches, lower filter
requirements for same frequency levels, lower dV/dt ratio and low-
er electromagnetic interference (EMI). Nevertheless, the multilevel
inverters have some disadvantages such as higher numbers of cir-
cuit components (power switches, capacitors, isolated gate drivers,
etc.) and complex circuit structures [24].
There are different types of multilevel inverter topologies. These
Fig. 2. Electrical equivalent circuit of PV cell. topologies are classified according to their specifications such as
20 N. Altin, S. Ozdemir / Energy Conversion and Management 69 (2013) 17–26

Table 1 Table 2
Comparison of multi-level inverter structures according to number of components. Output voltage value according to switch positions for a leg of the inverter.

Topology NPC FC CHB Switch position S1 S2 S3 S4 Output voltage


Number of switch 2(n  1) 2(n  1) 2(n  1) P (Positive) 1 1 0 0 1/2 V
Number of clamped diode (n  1)  (n  2) 0 0 O (Medium) 0 1 1 0 0
Number of clamped 0 (n  1)  (n  2)/ 0 N (Negative) 0 0 1 1 1/2 V
capacitor 2
Number of supply capacitor (n  1) (n  1) (n  1)/
2 2 3 2 3 2 3
urn 2 1 1 Sr
6 7 1 6 7 6 7
bidirectional power flow capability and number of phases [25]. The u
4 sn 5 ¼ V 4 1 2 1 5 4 Ss 5 ;
 ð15Þ
6
NPC inverter, the cascaded H-bridge (CHB) inverter and the flying utn 1 1 2 St
capacitor (FC) inverter are well known multilevel inverter topolo-
gies [6]. In NPC topology, DC bus voltage is divided by two capac- where Sr, Ss and St are switch status as given in Eq. (12). When Eq.
itors, and balancing of these two capacitors’ voltages is required to (15) is substituted into Eq. (14), three-level inverter equivalent cir-
obtain high performance. The FC topology requires bigger capacitor cuit model is obtained:
and this requirement increases the system size. Also, CHB structure 2d 3 2 3 2 3 2V 3
i
dt r  RL 0 0 ir 6L
ð3Sr  S0 Þ  ELr
requires isolated DC voltage sources. Required component num- 6d 7 6 7 6 7 6V 7
4 dt is 5¼4 0  RL 0 5  4 is 5 þ 4 6L ð3Ss  S0 Þ  ELs 5; ð16Þ
bers for a leg of the inverter, in which n is number of level, are gi-
R 0
ven comparatively in Table 1 for different structures of multi-level
d
i
dt t
0 0 L it V
6L
E
ð3St  S Þ  Lt
inverters. Because the NPC inverter does not require isolated DC
where S0 is expressed with Eq. (17) [27]:
sources and power switch modules for a leg of the inverter are
commercially available, design and implementation of the NPC in- S 0 ¼ Sr þ S s þ S t : ð17Þ
verter is simple. The NPC inverter scheme supplied from PV is gi-
ven in Fig. 4.
As seen from the figure, all power switches of the NPC inverter 5. Proposed single stage three-level MPPT grid interactive
operate with only half of the DC bus voltage. This situation is valid inverter
for P, O and N positions of the switches. Consequently, there is no
problem about voltage sharing between the power switches [26]. Since the grid interactive inverters inject current to the grid,
Output voltage value according to switch position is given in Table 2. these inverters must meet the standards defined by the grid
Table 2 can be rewritten in a function form as following authorities. Some national and international standards such as
equation: IEC61727, IEEE1547 and EN61000-3-2 have been published.
8 1
9 According to these standards, grid interactive inverter must oper-
< 1 ) ðui ¼ 2 VÞ
> >
=
ate at unity power factor, and harmonic components of inverter
Si ¼ 0 ) ðui ¼ 0Þ i ¼ r; s; t: ð12Þ
>
: 1
>
; output current and/or total harmonics distortion of this current
1 ) ðui ¼  2 VÞ (THD%) must be limited. Also characteristics of the renewable en-
Phase voltage expressions for the NPC inverter are given in Eq. (13) ergy source must be considered by grid interactive inverter to ob-
[27]: tain acceptable power conversion efficiencies. It means that grid
interactive inverter should consider both supply and line specifica-
d tions [1,28].
urn ¼ Rir þ L ir þ Er
dt In this study, three-phase three-level grid interactive inverter is
d proposed for PV supplied systems. Proposed inverter tracks the
usn ¼ Ris þ L is þ Es ð13Þ
dt MPP of PV system by using designed FLC based MPPT algorithm.
d FLC is chosen to obtain adaptive structure and to avoid the effects
utn ¼ Rit þ L it þ Et :
dt of parameter variations of PV system. Block diagram of proposed
Eq. (14) is obtained by expressing Eq. (13) in matrix form: three-phase grid interactive inverter is given in Fig. 5. As can be
2 3 2 3 2 3 2 3 2d 3 2 3 seen in the figure, the proposed system consists of the three-level
urn R 0 ir0 L 0 0 dt r
i Er NPC inverter, the line frequency transformer, the phase locked loop
6 7 6 7 6 7 6 7 6 7 6 7
4 usn 5 ¼ 4 0 R 0 5  4 is 5 þ 4 0 L 0 5  4 dtd is 5 þ 4 Es 5: (PLL) circuit, the LCL output filter, the fuzzy logic based MPP con-
utn 0 0 R it 0 0 L d
i
dt t
Et troller, the DC voltage balance controller and the PI current con-
troller. The DC–AC conversion, the grid interactive operation and
ð14Þ
the MPPT action are carried out by proposed single stage NPC in-
Phase voltages of the inverter according to switch position are given verter. FLC has two inputs and an output. Rate of change of PV
in the following equation: power (dP/dt) and rate of change of PV voltage (dV/dt) are defined

Fig. 4. Equivalent circuit model of PV sourced grid interactive NPC inverter.


N. Altin, S. Ozdemir / Energy Conversion and Management 69 (2013) 17–26 21

Fig. 5. Block diagram of the proposed system.

as input variables and change in reference current (diref/dt) is the


controller output variable. Output of the FLC is integrated and
the maximum value of the inverter reference current is deter-
mined. PLL circuit generates reference sine wave synchronized
with grid. Instantaneous value of the inverter reference current is
obtained by multiplying the integrated FLC output and grid voltage
sine wave which is obtained via PLL circuit. The PI current regula-
tor shapes the inverter output current to remove current error. Pro-
portional and integral gains of the PI regulator are determined by
using Ziegler–Nichols method. The LCL filter is used to reduce Fig. 6. Structure of FLC.
the output current THD level preferred, because the resonance fre-
quency of the LCL filter is related to only filter components [1]. Line
frequency transformer is employed to step up the inverter output
voltage to grid voltage level and to obtain galvanic isolation. Also,
DC bus voltage (v) and vdc- is read, and the DC voltage balance con-
troller balances the DC input voltages of the three-level NPC inver-
ter by adding offset value to the reference current value. All from
analogue to digital conversion, control and PWM generation pro-
cesses are carried out by eZdsp28335 controller board. This board
includes TMS320F28335, 32 bit floating point digital signal proces-
sor (DSP). Switching frequency is set to 10 kHz. A 3300 lF filter
capacitors which are represented by Cd1 and Cd2 are used to split
DC voltage. In addition, output LCL filter parameters are as follows:
Lf1 = Lf2 = 0.5 mH, Cf = 10 lF.

Fig. 7. Membership functions for input variables.


5.1. Fuzzy logic based MPPT controller design

Fuzzy Logic Controller is introduced by Zadeh in 1965. It is one


of the most successful applications of fuzzy set theory [29]. Its ma-
jor feature is the use of linguistic variables rather than numerical
variables. The general structure of the FLC is shown in Fig. 6. As de-
picted, the FLC is composed by a fuzzifier, a knowledge base, an
inference engine and a defuzzifier.
Defining the input and output variables is one of the important
steps in the fuzzy controller design. In this study, rate of change of
PV power and voltage are defined as input variables and change in
reference current is the controller output variable. The three vari-
ables of the FLC, the change in PV power, the change in PV voltage
and the change in inverter reference current, have seven triangle Fig. 8. Membership functions for change in duty cycle.
22 N. Altin, S. Ozdemir / Energy Conversion and Management 69 (2013) 17–26

Table 3 where um is the mth input variable, v is the output, Anm is the nth
Rule base of FLC. membership set and Bi is the output membership set belonging to
Rate of change of PV voltage (dV/dt) Rate of change of PV power (dP/dt) ith rule. The rules are set based upon the knowledge and working
NL NM NS Z PS PM PL
of the system.
The rule base adjusts the reference current of the NPC inverter
NL NL NL NM Z PM PL PL
NM NL NM NM Z PM PM PL
based upon the changes in the input of the FLC. The number of
NS NL NM NS Z PS PM PL rules can be set as desired. The rule base includes 49 rules, which
Z Z Z Z Z Z Z Z are based upon the seven membership functions of the input vari-
PS PS PS PS Z NS NM NL ables. Table 3 shows the rule base for the FLC.
PM PM PM PS Z NS NM NL
Inference engine simulates the human decision process and in-
PL PL PL PM Z NM NL NL
fers the fuzzy control action from the knowledge of the control
rules and the linguistic variable definitions. Therefore, the knowl-
edge base and the inference engine are in interconnection during
Table 4
the control process. Firstly active rules are detected by substituting
Normalized values of LCL filter.
fuzzified input variables into rule base. Then these rules are com-
Quantity SREF UREF fREF IREF ZREF LREF CREF bined by using one of the fuzzy reasoning methods. Max–Min
Value Sn Ug fg SREF U 2REF
¼ U REF Z REF 1
2pfREF Z REF
and Max-Product are most common fuzzy reasoning methods.
U REF SREF IREF 2pfREF
The commonly used Min–Max inference method is implemented.
The defuzzifier converts the fuzzy control action that infers from
inference engine to a non-fuzzy control action. Defuzzification is
membership functions for each. The basic fuzzy sets of member- done by using center of gravity method which is given Eq. (19),
ship functions for the variables are as shown in Figs. 7 and 8. The to generate non-fuzzy control signal for change in reference cur-
fuzzy variables are expressed by linguistic variables ‘positive large rent of grid interactive inverter.
(PL)’, ‘positive medium (PM)’, ‘positive small (PS)’, ‘zero (Z)’, ‘nega- P
tive small (NS)’, ‘negative medium (NM)’, ‘negative large (NL)’, for lðzÞ  z
z ¼ P : ð19Þ
all three variables. lðzÞ
The knowledge base consists of a data base with the necessary
linguistic definitions and control rule set. The rule set of knowledge
base consists of some fuzzy rules that define the relations between 5.2. LCL filter design
inputs and outputs. Usually, fuzzy rules are expressed in the form
of IF-THEN fuzzy conditional statements; Different type line filters such as L, LC and LCL are used at the
output of the inverters. Since L filter is first order filter and has
Ri : IF um ¼ Anm and um1 ¼ Anm1 THEN v ¼ Bi ; ð18Þ 20 dB/decade attenuation at whole frequency range, it needs very

Fig. 9. Line voltage, inverter output current and solar irradiation level of the proposed system.
N. Altin, S. Ozdemir / Energy Conversion and Management 69 (2013) 17–26 23

Fig. 10. Step response of the proposed system.

high switching frequencies to obtain acceptable performance. LC inverter reference power, voltage, current and impedance values.
filters are commonly used in uninterruptible power supply appli- Resonance frequency of the LCL filter can be calculated by the fol-
cations. But, the resonance frequency of LC filter varies with line lowing equation:
impedance in grid interactive inverter application [30]. Therefore
10f g 6 fr 6 fsw =2; ð21Þ
LCL filter is used in this study. The resonance frequency of the
LCL filter is related with only values of filter components and is gi- where the fg is the utility line frequency and fsw is switching fre-
ven in the following equation: quency. Similar criteria are required for inductor and capacitor val-
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ues. Normalized values will help to explain these criteria that are
1 L1 þ L2 given in Table 4 [1,30,31].
fr ¼ : ð20Þ
2p L1 L2 C In Table 4, Ug is the line voltage, Sn is the nominal output power
of the inverter and subscript REF represents the normalized values.
Resonance frequency and the components values of the filter is Since filter capacitor C is decreased the power factor and unity
selected according to switching frequency, line frequency, and power factor is desired in grid interactive inverter applications,

Fig. 11. FFT analysis of inverter output current.


24 N. Altin, S. Ozdemir / Energy Conversion and Management 69 (2013) 17–26

L1 ¼ 0:5 mH; ð24Þ

L2 ¼ 0:5 mH; ð25Þ

C ¼ 10 lF: ð26Þ

6. Simulation and experimental results

The proposed single stage, three-phase three-level NPC grid


interactive inverter with fuzzy logic based MPPT controller is mod-
eled and simulated with MATLAB/Simulink by using Fuzzy Logic
Toolbox. Proposed system composes of PV panels, three-phase,
three-level NPC inverter, line frequency transformer, LCL output
filter, PI current regulator and fuzzy logic based MPPT controller.
In this study KYOCERA KC200GT PV modules are modeled, and five
serial connected PV modules are used as PV supply with 1000 W
total power. An IGBT equipped three-phase three-level NPC inver-
Fig. 12. Test rig of the proposed system.
ter is used. SKM 200 MLI 066 T model three-level IGBT modules are
used in power stage. Line frequency transformer steps up the in-
the capacitor value of the filter is usually limited with Eq. (22) verter output voltage to grid voltage level. Also this transformer
[1,30]. ensures the galvanic isolation, and prevents DC ripple injection
to the grid. LCL output filter reduces the high frequency harmonic
C ¼ 0:05C REF : ð22Þ components in inverter output current waveform due to PWM
switching, and reduces the output current THD. Maximum power
Low capacitor values for any resonance frequency increase inductor point of the PV system is tracked with FLC based controller. All
values and also higher inductor values increase the system cost and analog to digital conversion, control algorithm and PWM generat-
size. So, the value of the capacitor is selected near the limits of Eq. ing activities are carried out by TMS320F28335 DSP.
(22). In addition, high total inductance value increases the voltage In this study, solar irradiation level is changed in order to test
drop on the inductors and the required DC voltage level. Higher the MPPT controller performance. Grid voltage, inverter output
DC voltage level will result in higher switching losses. So, total value current and solar irradiation level are seen in Fig. 9. As seen from
of the inductance must be lower than the 10% of LREF [1,30]. the figure, solar irradiation level is changed 400–1000 W/m2 step
Although, different L1 and L2 values can be selected for a unique va- by step and a day is emulated. FLC based MPPT controller deter-
lue of the total inductance, L1/L2 ratio is related to the maximum al- mines the inverter reference current with respect to irradiation le-
lowed value of the current ripple at the inverter and the line sides vel and as a result changes the inverter output current, and tracks
[30,31]. In this study this relation is selected as in the following the maximum power point of system. In addition inverter output
equation: current is synchronized with line voltage phase and frequency
L1 ¼ L2 : ð23Þ and inverter output current is in sinusoidal waveform.
Also in Fig. 10, step response of the proposed system is given.
The LCL filter’s components values of the grid interactive inverter Solar irradiation level is increased 400–1000 W/m2 instantly. In-
which are calculated by using Table 4 and Eqs. (20)–(23) are given verter output current, PV power and solar irradiation level are gi-
in Eqs. (24)–(26). The resonance frequency of the filter is tuned as ven in this figure. Fast transient response of the proposed fuzzy
3183 Hz. logic based MPPT controller is seen from the figure. Also it is clear

Fig. 13. Three-phase output currents (Ch. 2, Ch. 3, and Ch. 4) of proposed three-level inverter and line voltage (Ch. 1).
N. Altin, S. Ozdemir / Energy Conversion and Management 69 (2013) 17–26 25

that fuzzy logic based MPPT controller determines the reference Isc, Imppt, Vmppt, Tn, Tw, operation irradiation level, catalogue irradia-
current quickly and PI regulator shapes the current. Also inverter tion level, irradiation coefficients, temperature, etc. Thus, a PV
output current is in sinusoidal waveform. Furthermore, FFT analy- string with 1000 W total output power is emulated. PPPE generates
sis of the inverter output current is given in Fig. 11. As seen from a P–V curve according to these entered values; in other words, it
this figure total harmonic distortion level of this current is in the changes output current and voltage according to defined PV char-
limits of international standards such as IEEE1547 (3.75% < 5%). acteristic. The irradiation and temperature levels can be changed
MAGNA-POWER TDS III 600-8 model PV Power Emulator with computer interface during the operation, and thus, different
(PPPE), which can act like solar batteries, is used in experimental atmosphere conditions can be developed. Test rig of the proposed
studies. This PV power emulator is a DC power supply which can system is seen in Fig. 12.
give output just like PV. Five series connected commercial avail- In Fig. 13, three-phase output currents of proposed three-level
able KYOCERA KC200GT PV module is defined by using PPPE inter- NPC inverter and line voltage are depicted. Inverter output
face program and some parameters given in datasheets such as Voc, currents are measured from primary side of the line frequency

Fig. 14. Power, power factor, and output currents’ total harmonic distortion levels of proposed three-level inverter.

Fig. 15. PPPE interface program and P–V curve of KYOCERA KC200GT.
26 N. Altin, S. Ozdemir / Energy Conversion and Management 69 (2013) 17–26

transformer. As seen from the figure, inverter output currents are [4] Alvarez R, Fink K, Bernet S. Simulation and experimental investigation of
parallel connected IGBTs. IEEE Int Conf Ind Technol ICIT’10 2010:824–31.
in sinusoidal waveform and in phase with line voltage. Inverter
[5] Rodríguez J, Lai JS, Peng FZ. Multilevel inverters: a survey of topologies,
output current, power, power factor and harmonics levels of inver- controls, and applications. IEEE Trans Ind Electron 2002;49(4):724–38.
ter output currents are given in Fig. 14. It is seen that, unity power [6] Seok BS, Sinha G, Manjrekar MD, Lipo TA. Multilevel power conversion – an
factor operation is obtained and inverter output current harmonics overview of topologies and modulation strategies. In: IEEE proceedings of the
6th international conference on optimization of electrical and electronic
level meets the international standards (<5%). The interface pro- equipments; 1998. p. AD-11–AD-24.
gram of the PPPE and P–V characteristic of defined PV system is gi- [7] Dixon J, Morán L. High-level multistep inverter optimization using a
ven in Fig. 15. This program is used to define the PPPE by entering minimum number of power transistors. IEEE Trans Power Electron 2006;
21(2):330–7.
related values and monitoring the output of the PPPE. It shows the [8] Nami A, Zare F, Ledwich G, Ghosh A, Blaabjerg F. Comparison between
P–V curve, operation point and efficiency of the PV system. So, symmetrical and asymmetrical single phase multilevel inverter with diode-
operation point of the system and effectiveness of MPPT method clamped topology. IEEE Power Electron Spec Conf 2008:2921–6.
[9] Esram T, Chapman PL. Comparison of photovoltaic array maximum power
are visualized and monitored. As seen from the figure, proposed point tracking techniques. IEEE Trans Energy Convers 2007;22(2):439–49.
FLC based MPPT algorithm tracks the MPP of the PV system and en- [10] Liu F, Duan S, Liu F, Liu B, Kang Y. A Variable step size INC MPPT method for PV
ergy conversion efficiency of the PV system is improved. The MPPT systems. IEEE Trans Ind Electron 2008;55(7):2622–8.
[11] Mei Q, Shan M, Liu L, Guerrero JM. A novel improved variable step-size
efficiency is obtained from interface program as 98.78%. Also total incremental-resistance MPPT method for PV systems. IEEE Trans Ind Electron
system efficiency is measured as 93.12%. Both grid interactive 2011;58(6):2427–34.
operation and MPPT processes are carried out by three-phase [12] Tsang KM, Chan WL. Three-level grid-connected photovoltaic inverter with
maximum power point tracking. Energy Convers Manage 2013;65:221–7.
three-level inverter and single stage operation is obtained. Since
[13] Mamdani EH. Application of fuzzy algorithms for control of a simple dynamic
an additional converter requirement is removed, total efficiency plant. IEEE Proc D 1974;121:1585–8.
of the PV system is improved. [14] Alata M, Al-Nimr MA, Qaroush Y. Developing a multipurpose sun tracking
system using fuzzy control. Energy Convers Manage 2005;46(7–8):
1229–45.
_ Fuzzy logic based DSP controlled servo
[15] Bal G, Bekiroğlu E, Demirbasß Sß , Çolak I.
7. Conclusions
position control for ultrasonic motor. Energy Convers Manage 2004;45(20):
3139–53.
In this study, three-phase three-level NPC inverter for grid [16] Paramasivam S, Arumugam R. Hybrid fuzzy controller for speed control of
interactive PV system is proposed. The proposed inverter consists switched reluctance motor drives. Energy Convers Manage 2005;46(9–
10):1365–78.
of NPC inverter structure, line frequency transformer, LCL output [17] Messai A, Mellit A, Massi Pavan A, Guessoum A, Mekki H. FPGA-based
filter and fuzzy logic based MPPT controller. The fuzzy logic based implementation of a fuzzy controller (MPPT) for photovoltaic module. Energy
MPPT controller has two input and single output variables with se- Convers Manage 2011;52(7):2695–704.
[18] Premrudeepreechacharn S, Poapornsawan T. Fuzzy logic control of predictive
ven triangular membership functions for each. Both MPPT process current control for grid-connected single phase inverter. In: Twenty-eighth
and grid interactive operation is carried out by three-level NPC in- IEEE photovoltaic specialists conference; 2000. p. 1715–8.
verter. Since addition DC–DC converter for MPPT action is re- [19] Messai A, Mellit A, Massi Pavan A, Guessoumd A, Mekki H. FPGA-based
implementation of a fuzzy controller (MPPT) for photovoltaic module. Energy
moved, efficiency of the system is improved. LCL filter design is Convers Manage 2011;52:2695–704.
given in detail. Line frequency transformer steps up PV voltage to [20] Walker G. Evaluating MPPT converter topologies using a MATLAB PV model. J
grid voltage, provides galvanic isolation and prevents DC current Electr Electron Eng 2001;21(1):49–56.
[21] Mineiro SJE, Daher S, Antunes FLM, Cruz CMT. Photovoltaic system for supply
injection. Simulation and experimental results show that inverter
public illumination in electrical energy demand peak. In: IEEE 19th annual
output current is in phase with line frequency and phase angle, applied power electronics conference and exposition conference, vol. 3; 2004.
and inverter output current is in sinusoidal waveform. THD level p. 1501–6.
[22] Magna-Power Electronics Inc. Photovoltaic power profile emulator operating
of the inverter output current is below the limits of the interna-
guide. Magna-Power Electronics Inc.; 2009. p. 24–5.
tional standards (<5%) and unity power factor operation is ob- [23] Nabae A, Takashi I, Akagi H. A new neutral-point clamped PWM inverter. IEEE
tained. Also proposed fuzzy logic based MPPT method tracks Trans Ind Appl 1981;17(5):518–23.
maximum power point of the PV modules and has fast transient re- [24] Ozdemir S, Altin N, Sefa I. Single stage three-level MPPT inverter for solar
supplied systems. In: IEEE 21th international symposium on power
sponse. This method keeps the PV system operating at maximum electronics, electrical drives, automation and motion, Sorrento; 2012.
power point for all operation conditions and MPPT efficiency is [25] Pandey A, Singh B, Singh BN, Chandra A, Al-Haddad K, Kothari DP. A review of
achieved as 98.78%. In addition, proposed MPPT method has robust multilevel power converters. J Inst Eng [India] 2006;86:220–31.
[26] Wu B. High-power converters and AC drives. The Institute of Electrical and
structure against the parameter variations. Consequently, total Electronics Engineers, Inc.; 2006.
system efficiency including MPPT, NPC inverter and line frequency [27] Zhezhi Y, Lingzhi Y, Hanmei P, Xi F, Dong D, Study of simplified SVPWM
transformer is measured as 93.12%. algorithm based on three-level inverter. In: IEEE 6th international power
electronics and motion control conference, Wuhan; 2009. p. 876–81.
[28] Sefa I, Altın N. Grid interactive photovoltaic inverters – a review. J Fac Eng Arch
References Gazi Univ 2009;24(3):409–24.
[29] Zadeh LA. Outline of a new approach of the analysis of complex system and
decision processes. IEEE Trans Syst Man Cybern 1963;SMC-3(1):28–44.
[1] Altin N, Sefa I. dSPACE based adaptive neuro-fuzzy controller of grid [30] Raoufi M, Lamchich TM. Average current mode control of a voltage source
interactive inverter. Energy Convers Manage 2012;56:130–9. inverter connected to the grid: application to different filter cells. J Electr Eng
[2] Sefa I, Demirtas M, Çolak I. Application of one-axis sun tracking system. Energy 2004;55(3–4):77–82.
Convers Manage 2009;50:2709–18. [31] Liserre M, Blaabjerg F, Hansen S. Design and control of an LCL-filter-based
_ Özdemir Sß. Multifunctional interleaved boost converter for PV systems.
[3] Sefa I, three-phase active rectifier. IEEE Trans Ind Appl 2005;41(5):1281–91.
IEEE Int Sympos Ind Electron (ISIE) 2010:951–6.

You might also like