Vous êtes sur la page 1sur 12

Energy Conversion and Management 45 (2004) 2417–2428

www.elsevier.com/locate/enconman

Performance of a grid connected PV system used as active filter


a,* b
Hugo Calleja , Humberto Jimenez
a
Cenidet––Electronics Department, P.O. Box 5-164, Cuernavaca, Morelos 62050, Mexico
b
IIE––Non-Conventional Energy Sources Department, P.O. Box 1-475, Cuernavaca, Morelos 62091, Mexico
Received 12 March 2003; received in revised form 4 September 2003; accepted 23 November 2003
Available online 30 December 2003

Abstract
In this paper, the performance of a grid connected photovoltaic (PV) system used as an active filter is
presented. Its main feature is the capability to compensate the reactive and harmonic currents drawn by
nonlinear loads while simultaneously injecting into the grid the maximum power available from the cells.
The system can also operate as a stand alone active filter. The system was connected to a 1 kW PV array
and tested with the loads typically found in households: small motors, personal computers and electronic
ballasts. The results show that the system can correct the power factor to values close to unity for all the
cases tested, thereby improving the efficiency of the electric energy supply.
Ó 2003 Elsevier Ltd. All rights reserved.

Keywords: Photovoltaic systems; Power factor correction

1. Introduction

In areas with warm weather, the consumption of electric energy increases during the summer
due to HVAC loads connected to the grid. The summer peak demand is much higher than the
winter demand, but the peak occurs during the hours with the highest solar irradiance [1].
Therefore, a grid connected photovoltaic (PV) system can be used to reduce the peak demand that
the utility must satisfy. In this application, the PV system does not require a battery bank, but the
power stage remains idle during the night.
The feasibility of using PV systems for power conditioning and power factor compensation has
already been demonstrated [2]. However, the vast majority of the applications described in the

*
Corresponding author. Fax: +52-777-312-24-34.
E-mail address: hcalleja@cenidet.edu.mx (H. Calleja).

0196-8904/$ - see front matter Ó 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/j.enconman.2003.11.017
2418 H. Calleja, H. Jimenez / Energy Conversion and Management 45 (2004) 2417–2428

technical literature are for three phase systems using techniques developed for active filters, such as
the d–q transformation [3,4]. Power factor compensation is not usually performed in residential
applications, mainly due to the cost involved, but also because standards set limits to the current
displacement factor and harmonic pollution produced by loads such as electronic ballasts. The
fact, however, is that there is still a large number of devices that were installed prior to enforcement
of the standards. It has been shown that a grid connected inverter, acting as an active filter, can
correct the power factor of a group of nonlinear loads [5]. This application suits well the PV system,
since it can correct the power factor of a household without paying a penalty in added hardware.
Single phase configurations have also been developed, although not as many as the three phase
versions. For instance, a system based on an extension of the p–q theory has been reported, but it
relies on the assumption that the mains voltage is harmonic free [6]. Another approach followed
uses a dual converter configuration and a battery bank, which requires maintenance and lowers
the reliability [7,8]. A system in which the inverter acts as a lamp ballast during the night has also
been reported [9]. It should be noted that many papers only include wave forms, without the
numerical values obtained.
In this paper, the performance of a grid iterative PV system used as power factor corrector is
presented. The system can inject the maximum available power to the grid while simultaneously
compensating reactive and harmonic currents. Its main feature is that the reference signal used to
modulate the system is based on the adaptive interference canceling technique, which lends itself to
a very simple hardware implementation [10]. Two points were considered in the evaluation of the
power factor correction capability. The first one is the ability to correct the reactive power required
for linear loads, such as small motors used for water pumping. The second one is the ability to
compensate the harmonic currents drawn by nonlinear loads, such as personal computers and
electronic ballasts. To make the paper self contained, a brief description of the system is included.

2. PV system description

The operating principle is illustrated in Fig. 1, that shows a lagging square current wave form iL
at the load. The PV system can be modeled as three current sources in parallel. The amplitude IP
of the first one is proportional to the maximum power available from the PV cells, and its fre-
quency and phase are equal to those of the voltage of the mains. The amplitude IQ of the second
source depends upon the reactive power drawn by the load. It has the same frequency as the
voltage of the mains, but it is 90° out of phase with it. The third source generates a wave form
equal to the summation of the harmonics drawn by the load. The behavior of the system is such
that the current iAC supplied by the mains is a purely sinusoidal wave, and the PV system acts as
an active filter for the reactive and harmonic currents drawn by the load.
The circuit block diagram is shown in Fig. 2. The MPP tracking block generates a sinusoidal
signal vP , synchronized with the mains voltage wave form. The amplitude of this signal is pro-
portional to the maximum power available from the PV cells. The IQD block detects the reactive
and harmonic currents drawn by the load and generates the voltage vQD , proportional to these
currents. The signals vP and vQD are added to obtain the modulating signal vM for the power stage,
which is a single phase full bridge inverter. H1 , H2 and H3 are Hall effect current sensors.
H. Calleja, H. Jimenez / Energy Conversion and Management 45 (2004) 2417–2428 2419

Fig. 1. PV system operating principle.

Fig. 2. Circuit block diagram.


2420 H. Calleja, H. Jimenez / Energy Conversion and Management 45 (2004) 2417–2428

Table 1
Polarities of current steps
DILðnÞ DPPV DILðnþ1Þ
(+) (+) (+)
(+) ()) ())
()) (+) ())
()) ()) (+)

The MPP tracking is implemented with a microcontroller. The amplitude of vP is adjusted using
the ‘‘Disturb and Observe’’ technique, which is a simple method that does not require additional
cells or a mathematical model. The tracking is achieved by first measuring the power delivered by
the cells. Later on, a step in the set point of the cells is applied by modifying the amplitude of vP ,
and the resulting power is measured again. The set point is adjusted depending upon the sign of
the difference between the two measured powers, and a new step is applied again. Let DILðnÞ be a
step in the output current, which produces a change DPPVðnÞ in the power at the cells. The sign of
the next current step ILðnþ1Þ is related to the signs in DILðnÞ and DPPVðnÞ as described in Table 1.
The instantaneous values obtained in the power measurements are averaged over an integrating
period equal to an integer number of line cycles. The microcontroller does not check equality in
two successive measurements, instead, it verifies if the last value is larger than the previous one.
With this approach, the PV cells are operated around the MPP with small, low frequency vari-
ations in the current. A side effect of these variations is that the islanding condition is avoided
since the PV system can be exactly matched with the load during only one integrating period.
It should be noted that the current at the DC bus of the inverter is essentially a fully rectified
wave plus high frequency components due to the switching in the power stage. If the current were
drawn directly from the cells, its operating point would shift along the V –I curve, and the power
obtained would not correspond to the maximum available power. Therefore, it is necessary to
include a low frequency L–C filter at the output of the cells (LF and CF in Fig. 2) in such a way that
the current drawn from the cells is, as much as possible, ripple free.
An important issue is related to the active filter performance. There are several options
available to obtain the reference used to modulate the filter. A popular approach is the software
implementation in a digital signal processor, but the process required is quite complex and time
consuming. In fact, it has been shown that the delays involved in signal measurement and soft-
ware processing can produce a significant error in the compensation [11]. The hardware imple-
mentation usually requires a complex analogue circuitry, which is prone to inaccuracies due to
aging, thermal drifts etc. On the other hand, a harmonic and reactive current detecting system,
based on the adaptive interference canceling technique, can be maintained in the best operating
state by continuously self adjusting to changes in the operating parameters. The main advantages
of the technique are that the system becomes almost independent of the parameter variations and
that the resulting circuit is very simple [10].

2.1. Reactive and harmonic currents detection

A block diagram of the circuit used for reactive and harmonic currents detection is shown in
Fig. 3 [10]. The signal vR is a sinusoidal reference, and vH3 is the signal to be processed. The
H. Calleja, H. Jimenez / Energy Conversion and Management 45 (2004) 2417–2428 2421

Fig. 3. Active and reactive currents detection circuit.

feedback loop includes one integrating block and two multipliers. Since the transfer function for
most commercial multipliers includes an attenuation term, the gain blocks G1 and G2 are also
included in the feedback path.
Let VR be the peak amplitude of the sinusoidal reference signal, whose frequency is x. Also, let
s be the integrator time constant. The transfer function for the filter is:
s5 þ 2s3 x2 þ sx4
H ðsÞ ¼ ð1Þ
s5 þ 2s3 x2 þ sx4  kx4
where
G1 G2 VR
k¼ ð2Þ
s
The frequency response corresponds to a band reject filter that eliminates the frequency com-
ponents in vH3 that are in phase with vR .

2.2. Power stage and modulator

The power stage and the modulator are shown in Fig. 4. The frequency response of the system,
and hence its performance as an active filter, depends upon these stages. A pulse width modu-
lation (PWM) technique with hysteresis is used. The output current iO is sensed with the Hall effect
sensor labeled H1 , whose output feeds the comparator U1 . The second input to the comparator is
fed with the modulating signal vM . Because of the feedback, the output current is forced to follow
the shape of the signal vM but including a ripple that depends upon the hysteresis built into the
comparator and the size of the inductor LPV . Although this kind of modulation has the disad-
vantage of providing a nonconstant switching frequency, it offers the advantages of a very simple
implementation, and the output current can be made to take any desired shape within the limits
imposed by the size of the inductor and the hysteresis.
The voltage vH1 at the output of the Hall effect sensor is:
vH1 ¼ z1 iO ð3Þ
where z1 is the transimpedance of the sensor. Assuming that the ripple is small and can be neg-
lected, the modulator forces the voltage vH1 to have the same shape as vM . Therefore, the output
current can be written as:
2422 H. Calleja, H. Jimenez / Energy Conversion and Management 45 (2004) 2417–2428

Fig. 4. Power stage and pulse width modulator.

vM
iO ¼ ð4Þ
z1
The voltage vLPV across the filtering inductor LPV is:
diO
vLPV ¼ LPV ð5Þ
dt
The inductor must be calculated taking into account the amplitude Ih and frequency xh of the
highest current harmonic to be eliminated. The maximum rate of rise associated with this har-
monic is:

diO
¼ xh Ih ð6Þ
dt MAX
The minimum instantaneous value of vLPV occurs when vSEC ¼ nVPV , and vCA ¼ VCA (positive
peak). Therefore, using Eqs. (5) and (6), the maximum value for LPV is:
nVPV  VCA
LPV ¼ ð7Þ
xh Ih
Using a larger inductor will prevent the output current from attaining the rate of rise required to
eliminate the hth harmonic. The size of the inductor has a direct impact on the switching fre-
quency of the power stage, whose maximum value can be calculated with:
nVPV
fMAX ¼ ð8Þ
2BLPV
where B is the width of the hysteresis band [12]. If B is reduced, both the switching frequency and
the losses in the power stage increase. This effect can be compensated by increasing LPV , but it will
H. Calleja, H. Jimenez / Energy Conversion and Management 45 (2004) 2417–2428 2423

be at the expense of reduced bandwidth. Therefore, there is a trade off between the maximum
switching frequency, the desired efficiency and the active filtering performance.

3. Tests and results

A prototype was built using insulated gate bipolar transistors (IGBT) in the power stage and a
voltage transformer with a 1:2 ratio. The system was connected to a 1 kW PV cells bank with an
open circuit voltage of 150 V, and the minimum operational voltage was set at 100 V. Considering
a 120 V ± 10% mains voltage, the maximum value for LPV is 75 mH. In the prototype, the actual
value used was 35 mH. Assuming an hysteresis B ¼ 200 mA, a maximum switching frequency of
14.2 kHz is obtained, which is well within the 20 kHz rating of the transistors.

3.1. Harmonic content

Fig. 5 shows the current wave form obtained when the PV system is delivering 600 W of active
power without reactive and harmonic currents. In order to evaluate the harmonic content pro-
duced by the switching of the power stage, the system is tested unconnected to the grid and
modulated with a purely sinusoidal wave form. The corresponding frequency spectrum is shown
in Fig. 6, where both axes are in a logarithmic scale. Since the modulation technique used does not
provide a constant switching frequency, the current ripple appears as a cluster at frequencies
between 5 and 9 kHz. The total harmonic distortion (THD) obtained in this test was 2.31%. Table
2 lists the THD for different currents.

3.2. Efficiency

The efficiency of the system depends upon the active output power and the amount of reactive
and distortion powers that the system is compensating. The overall efficiency when the system is
CURRENT (Amps)
4

-2

-4
0 2 4 6 8 10 12 14 16 18
TIME (ms)

Fig. 5. Output current wave form.


2424 H. Calleja, H. Jimenez / Energy Conversion and Management 45 (2004) 2417–2428

Normalized amplitude
0
10

-1
10

-2
10

-3
10

-4
10

-5
10

-6
10

-7
10
1 2 3 4
10 10 10 10
Frequency (Hz)

Fig. 6. Frequency spectrum normalized about the fundamental.

Table 2
THD as a function of the output current
IAC (A) THD (%)
1.06 6.39
1.27 5.21
2.04 4.67
2.12 4.63
4.24 4.17
6.36 2.65

Table 3
PV system efficiency
Efficiency (%)
Overall 89
Transformer 96
Power stage 92

delivering 500 W is 89%, distributed between the power stage and the output transformer as listed
in Table 3.

3.3. Power factor correction

Fig. 7 shows the line voltage and current when the PV system is correcting the power factor of a
highly inductive load. The grid rms voltage is 125.6 V. Without compensation, the current sup-
plied by the grid is IACRMS ¼ 3:45 A, yielding an apparent power of 428 VA. With compensation,
the current drops to 1.6 A, yielding an active power of 199.4 W, an apparent power of 201.7 VA
H. Calleja, H. Jimenez / Energy Conversion and Management 45 (2004) 2417–2428 2425

Fig. 7. Power factor correction with an inductive load (VAC : 100 V/div; IAC ¼ 2 A/div).

NORMALIZED AMPLITUDE
1

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1
0
10 100 1k 10 k
FREQUENCY (HZ)

Fig. 8. Mains current spectrum after compensation.

and a power factor of 0.988. Since the losses in the distribution network are proportional to I 2 R,
the losses with compensation are about 22% of the amount obtained without compensation.
Without injecting active power into the grid (that is, as an active filter only), the internal losses are
equal to 32 W.
The harmonic spectrum after compensation is shown in Fig. 8, where the effect of the switching
in the power stage is apparent. In this test, the reference is obtained from the mains voltage wave
2426 H. Calleja, H. Jimenez / Energy Conversion and Management 45 (2004) 2417–2428

form, which has relatively large third and fifth harmonics and an overall THD of 2.5%. Therefore,
these harmonics also appear in the current wave form.

3.4. Active filtering capabilities

Fig. 9 shows the behavior of the system with a nonlinear load connected to the grid. The effect
of the harmonic compensation is apparent. Table 4 lists the harmonic content with and without
the PV system correcting the power factor for two different loads NL1 and NL2 . As in the previous
test, the third and fifth harmonics are produced by the harmonic content in the grid voltage.

3.5. Power flow

Fig. 10 shows the power flow during a sixty minutes test period. The power dissipated by the
load PL is equal to 500 W; the average power delivered by the cells PPV is 651 W; and the power at
the mains PAC is )101 W. Therefore, the grid is absorbing power from the PV system.

Fig. 9. Behavior with nonlinear load (VAC : 100 V/div; IAC ¼ 5 A/div).

Table 4
Harmonic content with nonlinear load
PVS Harmonics (%) THD PF
3° 5° 7° 9° 11° 13° 15° (%)

NL1 No 86.02 66.15 43.5 22.77 10.25 11.64 8.7 120.45 0.625
Yes 10.46 6.0 4.68 2.72 1.52 2.08 2.37 13.76 0.980
NL2 No 76.06 55.75 32.21 14.12 6.67 6.78 5.27 101.53 0.6704
Yes 3.47 2.24 0.95 1.1 <1 <1 <1 5.86 0.986
H. Calleja, H. Jimenez / Energy Conversion and Management 45 (2004) 2417–2428 2427

Power (Watts)

800 PPV

600
PL

400

200

-200
PAC
0 10 20 30 40 50 60
Time (minutes)

Fig. 10. Power flow during a 60 min test.

4. Conclusions

This paper presents the performance of a grid iterative PV system that also acts as an active
filter. During daylight, the system injects energy into the grid, operating the cells at its maximum
power point while simultaneously correcting the power factor of a household. During the night,
the system operates as a stand alone filter. The active filtering capability does not require modifi-
cations to the power stage, although the control circuitry becomes somewhat more complex.
The reference used to compensate the reactive and harmonic currents is based on an adaptive
noise canceling technique, which provides a very simple hardware implementation, and is almost
independent of parameters variations. In fact, there are no critical points in the design of the
circuitry, and it works properly with a wide range of time constant values.
Because of the unavoidable losses in the transformer and the transistors in the power stage, the
PV system draws active power from the grid when it is operating as a stand alone filter. The
amount drawn depends upon the magnitude of the current flowing in the system and represents
the penalty to be paid for the power factor correction feature. However, since harmonic currents
have noxious effects, such as overheating and premature ageing of distribution transformers, it is
always advantageous to suppress them. The tests performed show a large improvement in power
factor, therefore contributing to a more efficient operation of the electric energy distribution grid.
A disadvantage of the system is that in order to obtain the maximum power available from the
PV cells, a low pass filter must be connected at its output. There are alternatives to this filter, but
they usually require an additional switching stage. The configuration used employs a single
switching stage, maximizing both the efficiency and the reliability.

References

[1] Ramos G. Modeling residential customers energy demand profile. IIE Bull 1998;1:11–9 (in Spanish).
[2] Borle L, Dymond M, Nayar C. Development and testing of a 20-kW grid interactive photovoltaic power
conditioning system in Western Australia. IEEE Trans Indust Appl 1997;33(2):502–8.
2428 H. Calleja, H. Jimenez / Energy Conversion and Management 45 (2004) 2417–2428

[3] Cheng L, Cheung R, Leung KH. Advanced photovoltaic inverter with additional active power line conditioning
capability. IEEE Power Electronics Specialists Conference Record, 1997. p. 279–83.
[4] Hwang IH, Ahn KS, Lim HC, Kim SS. Design, development and performance of a 50 kW grid connected PV
system with three phase current-controlled inverter. IEEE Photovoltaic Specialists Conference Record, 2000.
p. 1664–7.
[5] Pottker F, Barbi I. Single phase active power filters for distributed power factor correction. IEEE Power
Electronics Specialists Conference Record, 2000. p. 500–5.
[6] Wu TF, Shen CL, Chang CH, Chiu JY. A 1 W /3 grid-connection PV power inverter with partial active power
filter. IEEE Power Electronics Specialists Conference Record, 2002. p. 1512–7.
[7] Dehbonei H, Nayar CL, Borle L. A combined voltage controlled and current controlled dual converter for a weak
grid connected photovoltaic system with battery energy storage. IEEE Power Electronics Specialists Conference
Record, 2002. p. 1495–500.
[8] Kim S, Yoo G, Song J. A bifunctional utility connected photovoltaic system with power factor correction and UPS
facility. IEEE Photovoltaic Specialists Conference Record, 1996. p. 1363–8.
[9] Wu TF, Chang CH, Chen YK. A multi-function photovoltaic power supply system with grid-connection and
power factor correction features. IEEE Power Electronics Specialists Conference Record, 2000. p. 1185–90.
[10] Lou S, Hou Z. An adaptive detecting method for harmonic and reactive currents. IEEE Trans Indust Electron
1995;42(1):85–9.
[11] Le Roux W, Van Wyk J. The effect of signal measurement and processing delay on the compensation of harmonics
by PWM converters. IEEE Trans Indust Electron 2000;47(2):297–304.
[12] Bode G, Holmes D. Implementation of three level hysteresis current control for a single phase voltage source
inverter. IEEE Power Electronics Specialists Conference Record, 2000. p. 33–8.

Vous aimerez peut-être aussi