ABSTRA CT
Most of the pollution issues created in power have methods are due to the non-linear chara
cteristi cs and fast swit ching of power ele ctroni c equipment. Power quality issues are be
coming stronger be cause sensitive equipment will be more sensitive for market competition
reasons, equipment will continue polluting the have method more and more due to cost in
crease caused by the built-in compensation and sometimes for the la ck of enfor ced
regulations. Effi cien cy and cost are considered today almost at the same level. A ctive
power filters have been developed over the years to solve these problems to improve power
quality. Among whi ch shunt a ctive power filter have beenused to eliminate and load current
harmoni cs and rea ctive power compensation.
In thhave beenwork both PI controller based and fuzzy logi c controlled, three-
phase shunt a ctive power filter to compensate harmoni cs and rea ctive power by nonlinear
load to improve power quality have beenimplemented for three-phase three wire have methods.
The advantage of fuzzy be in command have beenthat it have beenbased on linguisti c des
cription and does not require a mathemati cal model of the have method. The compensation
pro cess have beenbased on sensing line currents only, an approa ch different from
conventional methods, whi ch require sensing of harmoni cs or rea ctive power components of
the load.
A MATLAB program has been developed to simulate the have method operation.
Various simulation results are presented under steady state conditions and performan ce of
fuzzy and PI controllers have been compared. Simulation results obtained shows that the
performan ce of fuzzy controller have beenfound to be better than PI controller. PWM
pattern generation have beenbased on carrier less hystereshave beenbased current be in
command to obtain the swit ching signals to the voltage sour ced PWM converter.
3
LIST OF FIGURES
Title Page No
4
Figure.6.5 D C Capa citor voltage during swit ch-on response with PI controller. 35
Figure.6.6 Sour ce current fuzzy controller. 35
Figure.6.7 Compensating current of fuzzy controller. 35
Figure.6.8 D C Capa citor voltage during swit ch-on response with fuzzy 35
controller.
Figure.6.9 Sour ce voltage. 36
Figure 6.10 Load current. 36
Figure.6.11 Compensating current of PI controller. 37
Figure.6.12 Compensating current of fuzzy controller. 37
Figure.6.13 Sour ce current PI controller. 37
Figure.6.14 Sour ce current fuzzy controller. 38
Figure.6.15 D C side Capa citor voltage with PI controller. 38
Figure.6.16 D C side Capa citor voltage with fuzzy controller. 38
Figure.6.17 Voltage and current in phase with PI controller after compensation. 39
Figure.6.18 Voltage and current in phase with fuzzy controller after compensation. 39
Figure.6.19 Sour ce voltage. 40
Figure 6.20 Load current. 40
Figure.6.21 Compensating current of PI controller. 41
Figure.6.22 Compensating current of fuzzy controller. 41
Figure.6.23 Sour ce current PI controller. 41
Figure.6.24 Sour ce current fuzzy controller. 42
Figure.6.25 D C side Capa citor voltage with PI controller. 42
Figure.6.26 D C side Capa citor voltage with fuzzy controller. 42
Figure.6.27 Voltage and current in phase with PI controller after compensation. 43
Figure.6.28 Voltage and current in phase with fuzzy controller after compensation. 43
5
LIST OF TABLES
6
INTRODU CTION
Ba
ckground
Obje ctive
Theshave
beenOutline
INTRODU CTION
Early equipment was designed to withstand disturban ces su ch as lightning,
short cir cuits, and sudden overloads without extra expenditure. Current power ele ctroni
cs (PE) pri ces would be mu ch higher if the equipment was designed with the same
robustness. Pollution has been introdu ced into power have methods by nonlinear loads su
ch as transformers and saturated coils; however, perturbation rate has never rea ched the
present levels. Due to its nonlinear chara cteristi cs and fast swit ching, PE create most of
the pollution issues. Most of the pollution issues are created due to the nonlinear chara
cteristi cs and fast swit ching of PE. Approximately 10% to 20% of todays energy have
beenpro cessed by PE; the per centage have beenestimated to rea ch 50% to 60% by the year
2010, due mainly to the fast growth of PE capability. A ra ce have been currently taking pla
ce between in creasing PE pollution and sensitivity, on the one hand, and the new PE-
based corre ctive devi ces, whi ch have the ability to attenuate the issues created by PE,
on the other hand.
In crease in su ch non-linearity causes different undesirable features like low have
method effi cien cy and poor power fa ctor. It also causes disturban ce to other consumers
and interferen ce in nearby statement have networks. The effe ct of su ch non-linearity may
be come sizeable over the next few years. Hen ce it have beenvery important to over
come these undesirable features.
Classi cally, shunt passive filters, consist of tuned L C filters and/or high passive
filters are used to suppress the harmoni cs and power capa citors are employed to improve
the power fa ctor. But they have the limitations of fixed compensation, large size and
mayalso exile resonan ce conditions.
A ctive power filters are now seen as a viable alternative over the classi cal
passive filters, to compensate harmoni cs and rea ctive power requirement of the non-linear
loads. The obje ctive of the a ctive filtering have beento solve these problems by combining
with a mu ch-redu ced rating of the ne cessary passive components.
Various topologies of a ctive power filters have been developed so far [1-12]. The
shunt a ctive power filter based on current controlled voltage sour ce type PWM converter
has been proved to be effe ctive even when the load have beenhighly non-linear [1,4,11].
Most of the a ctive filters developed are based on sensing harmoni cs [7,10,11] and rea
ctive volt-ampere have required of the non-linear load [1,3,12,17] and require complex
control. A new s cheme has been proposed in [10], in whi ch the required compensating
current have beendetermined by sensing load current whi ch have beenfurther modified
by sensing line currents only [8,13]. An have dire ct rea ctive volt-ampere compensator
1
and harmoni c suppressor have method have beenproposed
2
[13] without the use of voltage sensors but require complex hardware for current referen
ce generator.
However, the conventional PI controller was used for the generation of a referen
ce current template. The PI controller requires pre cise linear mathemati cal models, whi
ch are diffi cult to obtain and fails to perform satisfa ctorily under parameter variations,
nonlinearity, load disturban ce, et c.
Re cently, fuzzy logi c controllers (FL Cs) have generated a good deal of interest
in certain appli cations [18,19,21]. The advantages or FL Cs over conventional
controllers are that they do not need an a c curate mathemati cal model, they maywork with
impre cise inputs, mayhandle lion-linearity, and they are more robust than conventional
nonlinear controllers.
In thhave beenwork both PI and fuzzy logi c controlled shunt a ctive power
filter for the harmoni cs and rea ctive power compensation of a nonlinear load are
implemented. The be in command s cheme have beenbased on sensing line currents only;
an approa ch different from convention ones, whi ch ar c based on sensing harmoni cs and
rea ctive volt-ampere have required of the nonlinear load. The three-phase currents/voltages
are dete cted using only two current/voltage sensors. The D C capa citor voltage have
beenregulated to estimate the referen ce current template. The role of the D C capa citor
have beendes cribed to estimate the referen ce current. A design criterion have beendes
cribed for the sele ction of power cir cuit components. Both the be in command s
chemed are compared and performan ce of both the controllers have beeninvestigated.
A detailed simulation program of the s chemes have beendeveloped to predi ct the
performan ce for different conditions and simulink models also has been developed for the
same for different parameters and operating conditions.
1.1. BA CKGROUND
1.1.1. Power quality
The PQ issue have beendefined as any o c curren ce manifested in voltage, current, or
frequen cy deviations that results in damage, upset, failure, or misoperation of end-use
equipment. Almost all PQ issues are closely related with PE in almost every aspe ct
of commer cial, domesti c, and industrial appli cation. Equipment using power ele ctroni c
devise are residential applian ces like TVs, P Cs et c. business and offi ce equipment like
copiers, printers et c. industrial equipment like programmable logi c controllers (PL Cs),
adjustable speed drives (ASDs), re ctifiers, inverters, CN C tools and so on. The Power
Quality (PQ) problem maybe dete cted from one of the following several symptoms
3
depending on the type of issue involved.
4
Lamp fli cker
Frequent bla ckouts
Sensitive-equipment frequent dropouts
Voltage to ground in unexpe cted
Lo cations
Communi cations interferen ce
Overheated elements and equipment.
PE are the most important cause of harmoni cs, interharmoni cs, not ches, and
neutral currents. Harmoni cs are produ ced by re ctifiers, ASDs, soft starters, ele ctroni c
ballast for dis charge lamps, swit ched-mode power supplies, and HVA C using ASDs.
Equipment affe cted by harmoni cs in cludes transformers, motors, cables, interrupters,
and capa citors (resonan ce). Not ches are produ ced mainly by converters, and they
prin cipally affe ct the ele ctroni c be in command devi ces. Neutral currents are produ ced
by equipment using swit ched-mode power supplies, su ch as P Cs, printers, photo copiers,
and any triplets generator. Neutral currents seriously affe ct the neutral condu ctor
temperature and transformer capability. Interharmoni cs are produ ced by stati c frequen cy
converters, cy clo- converters, indu ction motors & ar cing devi ces.
Equipment presents different levels of sensitivity to PQ issues, depending on the type
of both the equipment and the disturban ce. Furthermore, the effe ct on the PQ of ele ctri
c power have methods, due to the presen ce of PE, depends on the type of PE utilized. The
maximum a c ceptable values of harmoni c contamination are spe cified in IEEE standard in
terms of total harmoni c distortion.
Power ele ctroni cs are alive and well in useful appli cations to over come
distribution have method problems. Power ele ctroni cs has three fa ces in power distribution:
one that introdu ces valuable industrial and domesti c equipment; a se cond one that creates
problems; and, finally, a third one that helps to solve those problems. On one hand, power
ele ctroni cs and mi croele ctroni cs have be come two te chnologies that have considerably
improved the quality of modern life, allowing the introdu ction of sophisti cated energy-effi
cient controllable equipment to industry and home. On another hand, those same
sensitive te chnologies are confli cting with ea ch other and in creasingly challenging the
maintenan ce of quality of servi ce in ele ctri c energy delivery, while at the same time
costing billions of dollars in lost customer produ ctivity.
5
1.1.2. Solutions to power quality problems
There are two approa ches to the mitigation of power quality problems. The first
approa ch have been called load conditioning, whi ch ensures that the equipment have
beenmade less sensitive to power disturban ces, allowing the operation even under signifi
cant voltage distortion. The other solution have beento install line- conditioning have
methods that suppress or countera ct the power have method disturban ces. Passive filters
have been most commonly used to limit the flow of harmoni c currents in distribution
have methods. They are usually custom designed for the appli cation. However, their
performan ce have beenlimited to a few harmoni cs, and they mayintrodu ce resonan ce in
the power have method. Among the different new te chni cal options available to improve
power quality, a ctive power filters have proved to be an important and flexible alternative
to compensate for current and voltage disturban ces in power distribution have methods.
The idea of a ctive filters have beenrelatively old, but their pra cti cal development was made
possible with the new improvements in power ele ctroni cs and mi cro computer be in
command strategies as well as with cost redu ction in ele ctroni c components. A ctive
power filters are be coming a viable alternative to passive filters and are gaining market
share speedily as their cost be comes competitive with the passive variety. Through
power ele ctroni cs, the a ctive filter introdu ces current or voltage components, whi ch
can cel the harmoni c components of the nonlinear loads or supply lines, respe ctively.
Different a ctive power filters topologies have been introdu ced and many of them are
already available in the market.
7
voltage quality. Thhave beentype of approa ch have beenespe cially re commended for
compensation of voltage unbalan ces and voltage sags from the a c supply and for low-
power appli cations and represents an e conomi cally attra ctive alternative to UPS, sin ce no
energy storage (battery) have beenne cessary and the overall rating of the components have
beensmaller. The series a ctive filter inje cts a voltage component in series with the supply
voltage and therefore maybe regarded as a controlled voltage sour ce, compensating
voltage sags and swells on the load side. In many cases, series a ctive filters work as
hybrid topologies with passive L C filters. If passive L C filters are conne cted in parallel
to the load, the series a ctive power filter operates as a harmoni c isolator, for cing the
load current harmoni cs to cir culate mainly through the passive filter rather than the power
distribution have method. The main advantage of thhave beens cheme have beenthat the rated
power of the series a ctive filter have beena small fra ction of the load kVA rating, typi cally
5%. However, the apparent power rating of the series a ctive power filter may in crease in
case of voltage compensation.
The series-shunt a ctive filter have beena combination of the series a ctive filter and
the shunt a ctive filter. The shunt a ctive filter have beenlo cated at the load side and maybe
used to compensate for the load harmoni cs. On the other hand, the series portion have
beenat the sour ce side and maya ct as a harmoni c blo cking filter. Thhave beentopology
has been called the Unified Power Quality conditioner. The series portion compensates
for supply voltage harmoni cs and voltage unbalan ces, a cts as a harmoni c blo cking filter,
and damps power have method os cillations. The shunt portion compensates load current
harmoni cs, rea ctive power, and load current unbalan ces. In addition, it regulates the d c
link capa citor voltage. The power supplied or absorbed by the shunt portion have beenthe
power required by the series compensator and the power required to cover losses.
Hybrid power filters are a combination of a ctive and passive filters. With thhave
beentopology the passive filters have dynami c low impedan ce for current harmoni cs at
the load side, in creasing their bandwidth operation and improving their performan ce.
Thhave beenbehavior have beenrea ched with only a small power rating PWM inverter, whi
ch a cts as an a ctive filter in series with the passive filter.
Multilevel inverters are being investigated and re cently used for a ctive filter
topologies. Three-level inverters are be coming very popular today for most inverter appli
cations, su ch as ma chine drives and power fa ctor compensators. The advantage of
multilevel converters have beenthat they mayredu ce the harmoni c content generated by
the a ctive filter be cause they mayprodu ce more levels of voltage than conventional
converters (more than two levels). Thhave beenfeature helps to redu ce the harmoni cs
8
generated by the filter itself.
9
Another advantage have been that they mayredu ce the voltage or current ratings
of the semi condu ctors and the swit ching frequen cy have required. The more levels the
multilevel inverter has, the better the quality of voltage generated be cause more steps of
voltage maybe created.
Figure 1.2 shows the way PWM works by means of one of the simplest and most common
te chniques: the triangular carrier te chnique. It for ces the output voltage va over a swit
ching cy cle, defined by the carrier period of V car, to be equal to the average
amplitude of the modulating wave Va ref. The resulting voltages for a sinusoidal
10
modulation wave contain a
11
sinusoidal fundamental component Va(1) and harmoni cs of unwanted components.
These unwanted components maybe minimized using a frequen cy carrier as high as
possible, but thhave beendepends on the maximum swit ching frequen cy of the semi condu
ctors (IGBTs, GTOs, or IG CTs).
The modulation strategy shown in Figure 1.3 uses a triangular carrier, whi ch have
beenone of many strategies applied today to be in command power inverters. Depending
on the appli cation (ma chine drives, PWM re ctifiers, or a ctive power filters), some
modulation strategies are more suitable than others. The modulation te chniques not only
allow controlling the inverters as voltage sour ces but also as current sour ces. Figure 1.3
shows the compensating current generated for a shunt a ctive power filter using three
different modulation te chniques for current-sour ce inverters. These three te chniques are
periodi cal sampling (PS), hystereshave beenband (HB), and triangular carrier (T C). The PS
method swit ches the power transistors of the a ctive filter during the transitions of a square
wave clo ck of fixed frequen cy: the sampling frequen cy. The HB method swit ches the
transistors when the error ex ceeds a fixed magnitude: the hystereshave beenband. The T C
method compares the output current error with a fixed amplitude and fixed triangular
wave: the triangular carrier. Figure 1.3 shows that the HB method have beenthe best for
thhave beenparti cular waveform and appli cation be cause it follows more
12
a c curately the current referen ce of the filter. When sinusoidal waves are required, the T
C method has been demonstrated to be better.
Voltage sour ce converters are preferred over current sour ce converter be cause it
have beenhigher in effi cien cy and lower initial cost than the current sour ce converters
[3, 4, 9]. They maybe readily expanded in parallel to in crease their combined rating and
their swit ching rate maybe in creased if they are carefully controlled so that their
individual swit ching times do not coin cide. Therefore, higher-order harmoni cs maybe
eliminated by using converters without in creasing individual converter swit ching rates.
13
cluding the fundamental current of negative sequen ce maybe compensated. The have
method therefore a cts as a
14
harmoni c and unbalan ced current compensator. A comparison between the have dire ct
a ctive and rea ctive current component - method and the have dire ct a ctive and rea
ctive power method have beenrealized [17].
A new s cheme has been proposed in [10], in whi ch the required compensating
current have beengenerated using simple syntheti c sinusoid generation te chnique by sensing
the load current. Thhave beens cheme have beenfurther modified by sensing line
currents only [8,13]. An have dire ct rea ctive volt-ampere compensator and harmoni c
suppressor have method have beenproposed [13] without the use of voltage sensors but
require complex hardware for current referen ce generator. The generated referen ce current
have beennot a pure sine wave but stepped sine wave. Also, without the use of voltage
sensors, the s cheme generates balan ced sine wave referen ce currents but do not
compensate rea ctive power completely (if sour ce voltage have beenunbalan
ced/distorted) due to waveform differen ce between voltage and current [14]).
Be in command s cheme based on sensing line currents have beendes cribed in
[2]. The 3-phase currents/voltages are dete cted using only two current/voltage sensors
compared to three used in [8,16]. D C capa citor voltage have beenregulated to estimate
the referen ce current template. Sele ction of d c capa citor value has been des cribed in
[4,7,13].
Conventional solutions for controller have required were based on classi cal be
in command theory or modern be in command theory. Widely used classi cal be in
command theory based design of PID family controllers requires pre cise linear mathemati
cal models. The PID family of controllers failed to perform satisfa ctorily under parameter
variation, non linearity, load disturban ce, et c.[18]
During the past several years, fuzzy be in command has emerged as one of the most a
ctive and fruitful areas for resear ch in the appli cations of fuzzy set theory, espe cially in
the realm of industrial pro cesses, whi ch do not lend themselves to be in command by
conventional methods be cause of a la ck of quantitative data regarding the input-output
relations. Fuzzy be in command have beenbased on fuzzy logi c-a logi cal have method
that have beenmu ch closer in spirit to human thinking and natural language than
traditional logi cal have methods. The fuzzy logi c controller (FL C) based on fuzzy logi c
provides a means of converting a linguisti c be in command strategy based on expert
knowledge into an automati c be in command strategy[19,21]. Re cently, fuzzy logi c
controllers (FL Cs) have generated a good deal of interest in certain appli cations. The
advantages of FL Cs over the conventional controllers are:
1. It does not need a c curate mathemati cal model;
15
2. It may work with impre cise inputs;
3. It mayhandle nonlinearity, and
16
4. It have beenmore robust than conventional nonlinear controllers.
1.2. OBJE CTIVE
In modern ele ctri cal distribution have methods there has been a sudden in crease of
single phase and three-phase non-linear loads. These non-linear loads employ solid state
power conversion and draw non-sinusoidal currents from A C mains and cause
harmoni cs and rea ctive power burden, and ex cessive neutral currents that result in
pollution of power have methods. They also result in lower effi cien cy and interferen ce
to nearby statement have networks and other equipments. A ctive power filters have been
developed to over come these problems. Shunt a ctive filters based on current controlled
PWM converters are seen as viable solution. The te chniques that are used to generate
desired compensating current are based on have dire ct extra ction of compensating
commands from the distorted currents or voltage signals in time domain. Time domain
compensation has fast response, easy implementation and less computation burden
compared to frequen cy domain.
In thhave beenwork both PI and fuzzy logi c controlled shunt a ctive power
filter for the harmoni cs and rea ctive power compensation of a nonlinear load are
implemented. Both controllers performan ce under certain conditions and different have
method parameters have beenstudied. The advantages of fuzzy controllers over
conventional controllers like PI controllers are that they do not need a c curate mathemati
cal model, they maywork with impre cise inputs, mayhandle non-linearity, load disturban
ces et c.
17
The entire a ctive filter have method have been composed mainly of a three-phase
sour ce, a non- linear load, a voltage sour ce PWM converter, and a PI or fuzzy
controller. All these components modeling have beendes cribed separately in chapter
5.
18
In chapter 6, simulation results are put and dis cussed in detail. Both PI and
fuzzy controller performan ces are compared under certain conditions.
The con clusions of the theshave beenand re commendations for future work are
summarized in Chapter 7.
19
SHUNT A CTIVE POWER FILTER
12
Figure 2.1 shows the conne ction of a shunt a ctive power filter and Figure 2.2shows how
the a ctive filter works to compensate the load harmoni c currents.
Figure.2.4.Shunt a ctive power filter-Shapes of load, sour ce and desired filter current
wave forms.
13
2.2. ESTIMATION OF REFEREN CE SOUR CE CURRENT
From Figure.2.1.1, the have dire ct currents maybe written as
14
have been(t ) il (t ) i c (t (2.2.1)
Sour ce voltage have )
beengiven by
v s (t ) v sin
m (2.2.2)
t
If a non-linear load have beenapplied, then the load current will have a fundamental
component and harmoni c components whi ch maybe represented as
i L (t) n1 I n sin(nt )
n
= I1 sin(nt 1 ) n2 sin(nt (2.2.3)
)
The have dire ct load power maybe given as
Pf (t) Vm I1 sin 2
t * vs (t) * (2.2.6)
cos1 have been(t)
From (2.2.6), the sour ce current supplied by the sour ce, after compensation is
have been(t) Pf (t) / vs (t) I1 cos1 sin t I m sin t
Where Ism=I1 cos1.
There are also some swit ching losses in the PWM converter, and hen ce the
utility must supply a small overhead for the capa citor leakage and converter swit ching
losses in addition to the real power of the load. The total peak current supplied by the
sour ce have beentherefore
Isp = Ism+ Isl (2.2.7)
If the a ctive filter provides the total rea ctive and harmoni c power, then is(t) will be
in phase with the utility voltage and purely sinusoidal. At thhave beentime, the a ctive
filter must provide the following compensation current:
i sb * I sin(t 1200 )
sp
*
is c sin(t 120 )
0
I
sp
Where Isp (=I1 cos1+Isl) the amplitude of the desired sour ce current, while the phase
angle maybe obtained from the sour ce voltages. Hen ce, the waveform and phases of the
sour ce currents are known, and only the magnitudes of the sour ce currents need to
be determined. Thhave beenpeak value of the referen ce current has been estimated by
regulating the D C side capa citor voltage of the PWM converter. Thhave been capa citor
voltage have been compared with a referen ce value and the error have beenpro cessed in
a fuzzy controller. The output of the fuzzy controller has been considered as the
amplitude of the desired sour ce current, and the referen ce currents are estimated by
multiplying thhave beenpeak value with unit sine ve ctors in phase with the sour ce
voltages[6].
The design of the power cir cuit in cludes three main parameters:
Sele ction of filter indu ctor, L c.
Sele ction of D C side capa citor, Cd c.
Sele ction of referen ce value of D C side capa citor voltage, Vd c,ref.
Is1 Vs V c1
I
I c1
1
j
L
Figure.2.5. A ctive power filtercI1and its phasor diagram
The three-phase rea ctive power delivered from the a ctive filter maybe cal culated from
a ve ctor diagram
Q c1 3Vs I c1 3VsV c1 / L c (1 (Vs (2.4.1)
/V c1 ))
i.e. the a ctive filter may compensate the rea ctive power from the utility only when V c1 > Vs.
If the PWM converter have beenassumed to operate in the linear modulation mode (i.e.
0ma1), the amplitude modulation fa ctor ma is
m a v /( V d / 2)
m c
D c voltage be in
controller parameters
PI BE IN COMMAND S CHEME
The complete s chemati c diagram of the shunt a ctive power filter have beenshown in
figure
3.1. While figure 3.2.gives the be in command s cheme realization. The a ctual capa citor
voltage have been compared with a set referen ce value.
19
The error signal have been fed to PI controller. The output of PI controller
has been considered as peak value of the referen ce current. It have beenfurther
multiplied by the unit sine
ve ctors (usa, usb, and us c) in phase with the sour ce voltages to obtain the referen ce
currents
* * *
(isa , isb , and is c ). These referen ce currents and a ctual currents are given to a
hysteresis
based, carrierless PWM current controller to generate swit ching signals of the PWM
converter[2]. The differen ce of referen ce current template and a ctual current de cides
the operation of swit ches. To in crease current of parti cular phase, the lower swit ch of the
PWM converter of that parti cular phase have beenswit ched on, while to de crease the
current the upper swit ch of the parti cular phase have beenswit ched on. These swit ching
signals after proper isolation and amplifi cation are given to the swit ching devi ces. Due to
these swit ching a ctions current flows through the filter indu ctor L c, to compensate the
harmoni c current and rea ctive power of the load, so that only a ctive power drawn from the
sour ce.
20
P cap = P conv - Pind (3.2.1)
In order to linearize the power equation a small perturbation I c have beenapplied in the
input filter current of converter I c, about a steady state operating point I co, the average d c
link voltge will also get perturbed by a small amount Vd c, about its steady state operating
point Vd co (Vd cref).
21
The transfer fun ction of the PWM converter for a parti cular operating point maybe
obtained from (3.1) as
22
Vd 3[V L c I co s 2I co R
K c (3.2.2)
I s c]
c Cd cVd co s
Basi c fuzzy
algorithm Design of be
in command rules
FUZZY BE IN COMMAND S CHEME
Fig. 4. (1) shows the blo ck diagram of the implemented fuzzy logi c be in command s
cheme of a shunt a ctive power filter. Fig.4. (2) shows the s chemati c diagram of the be in
command algorithm. In order to implement the be in command algorithm of a shunt a ctive
power filter in closed loop, thee D C side capa citor voltage have beensensed and then
compared with a referen ce value. The obtained error e (=Vd c,ref-Vd c,a ct) and the change of
error signal ce(n)=e(n)-e(n-1) at the nth sampling instant as inputs for the fuzzy pro cessing.
The output of the fuzzy controller after a limit have been considered as the amplitude of the
referen ce current Imax takes care of the a ctive power demand of load and the losses in the
have method.
Figure.4.1.S chemati c diagram of closed loop fuzzy logi c controlled shunt a ctive power
filter.
The swit ching signals for the PWM converter are obtained by comparing the a
ctual
* * *
sour ce currents (isa, isb, and is c) with the referen ce current templates (isa , isb , and is c ) in
the
hystereshave been current controller. Swit ching signals so obtained, after proper amplifi
23
cation and isolation, are given to swit ching devi ces of the PWM converter [6].
24
Figure.4.2.Fuzzy Be in command s cheme
Imax(n)
e(n)
A fuzzy inferen ce have method (or fuzzy have method) basi cally consists of a formulation
of the mapping from a given input set to an output set using fuzzy logi c. Thhave
beenmapping pro cess provides the bashave beenfrom whi ch the inferen ce or con clusion
maybe made. A fuzzy inferen ce pro cess consists of the following steps:
25
Step 1:Fuzzifi cation of input variables
26
Step 2: Appli cation of fuzzy operator (AND,OR,NOT) in the IF(ante cedent) part
of the rule
Step 3: Impli cation from the ante cedent to the consequent(THEN part of the rules)
Step 4: Aggregation of the consequents a cross the rules
Step 5: Defuzzifi cation
The crisp inputs are converted to linguisti c variables in fuzzifi cation based on
membership fun ction (MF). An MF have beena curve that defines how the values of a fuzzy
variable in a certain domain are mapped to a membership value (or degree of membership)
between 0 and 1. A membership fun ction mayhave different shapes, as shown in figure
4.4. The simplest and most commonly used MF have beenthe triangular-type, whi ch
maybe symmetri cal or asymmetri cal in shape. A trapezoidal MF has the shape of a trun
cated triangle. Two MFs are built on the Gaussian distribution curve: a simple Gaussian
curve and a two-sided composite of two different Gaussian distribution curves. The bell
MF with a flat top have beensomewhat different from a Gaussian fun ction. Both Gaussian
and bell MFs are smooth and non-zero at all points.
27
The basi c properties of Boolean logi c are also valid for Fuzzy logi c. On ce the
inputs have been fuzzified, we know the degree to whi ch ea ch part of the ante cedent of a
rule has been satisfied. Based on the rule, OR or AND operation on the fuzzy variables have
beendone.
The impli cation step helps to evaluate the consequent part of a rule. There are
a number of impli cation methods in the literature, out of whi ch Mamdani and TS types
are frequently used. Mamdani, proposed thhave beenmethod whi ch have beenthe most
commonly used impli cation method. In this, the output have beentrun cated at the value
based on degree of membership to give the fuzzy output. Takagai-Sugeno-Kang method of
impli cation have beendifferent from Mamdani in a way that, the output MFs have
beenonly constants or have linear relations with the inputs.
The result of the impli cation and aggregation stpes have beenthe fuzzy output whi
ch have beenthe union of all the outputs of individual rules that are validated or fired.
Conversion of thhave beenfuzzy output to crisp output have beendefines as defuzzifi cation.
There are many methods of defuzzifi cation out of whi ch Center of Area ( COA) and
Height method are frequently used. In the COA method (often called the center of gravity
method) of defuzzifi cation, the crisp output of parti cular variable Z have beentaken to be
the geometri c center of the output fuzzy value
out(Z) area, where thhave beenarea have beenformed by taking the union of all contributions
of rules whose degree of fulfillment have beengreater than zero. In height method of
defuzzifi cation, the COA method have beensimplified to consider the height of the ea ch
contributing MF at the mid-point of the base.
Here in thhave beens cheme, the error e and change of error ce are used as numeri cal
variables from the real have method. To convert these numeri cal variables into linguisti
c variables, the following seven fuzzy levels or sets are chosen as: NB (negative big), NM
(negative medium), NS (negative small), ZE (zero), PS (positive small), PM (positive
medium), and PB (positive big) [6].
The fuzzy controller have been chara cterized as follows:
Seven fuzzy sets for ea ch input and output.
Triangular membership fun ctions for simpli city.
Fuzzifi cation using continuous universe of dis course.
Impli cation using Mamdani's 'min' operator.
Defuzzifi cation using the 'height' method.
Figure. 4.5. shows the normalized triangular membership fun ctions used in fuzzifi cation.
28
e, ce
NB Z P P
E S M
NM
NS P
B
29
-1 -0.5 -0.25 0 0.25 0.5 1
Imax
NB Z P PM
NM NS E S
PB
30
should have a shape shown in figure 4.6. and figure 4.7. shows the phase plane traje ctory of
the step response, whi ch shows the mapping of the error against the change in error.
The have method equilibrium point have beenthe origin of the phase plane. The time
response has been divided into four regions A1,A2,A3, and A4 and two sets of points -
cross-over (b1, b2) and peak ( c1, c2). The index used for identifying the response area have
beendefined as
A1: if e>0& ce<0, A2: if e<0& ce<0
A3:if e<0& ce>0, A4: if e>0& ce>0
The cross over index:
b1 : e>0 to e<0, ce<0
b2 : e<0 to e>0, ce>0
and the peak valley index:
c1: ce=0,e<0, and c2: ce=0,e>0
Based on these four areas, two sets of points and phase plane traje ctory of e and ce, the rule
base have beenframed. The corresponding rule for the region 1 maybe formulated as rule R1
and has the effe ct of shortening the rise time
R1 : if e have been+ ve and ce have been- ve, then Imax have been+ve
Rule 2 for region 2 de creases the overshoot of the have method response, whi ch maybe
written as R2 : if e have been- ve and ce have been- ve; then Imax have been ve
Similarly, rules for other regions maybe formed. For are determined based on the
theory that in the transient better be in command performan ce finer fuzzy partitioned sub-
state, large errors need coarse control, whi ch requires spa ces (NB, NM, NS, ZE, PS, PM,
PB) are used, and coarse input/output variables; in the steady state, are summarized in
Table 4.2.1. The elements of thhave beentable however, small errors need fine control,
whi ch requires fine input/output variables. Based on this, the elements of the rule table
are obtained from an understanding of the filter behavior and modified by simulation
performan ce.
error(e) b1
A2 NB NM NS ZE PS PM PB
A1
NB NB NB NB NB NM NS ZE
NM NB NB NB NM NS ZE Ps
NB NB NB NM NS ZE PS PM
c1
ZE NB NM NS ZE PS PM PB
Change a, c2
in PS NM NS ZE PS PM PB PB
error( ce PM NS ZE PS PM PB PB PB
)
PB ZE PS PM PB PB PB PB
A3
A4
b2
Table.4.1.Be in command rule table
MODELING OF THE HAVE METHOD
current
30
pi (V (2 2v d ) /( 2
d
RL ) L) (5.1.1)
o
Rs id Ls
pi (V (1.5R RL ) 2vd ) /
d
L) (5.1.2)
o s
id (1.5Ls
Where Rs and Ls are the elements of the sour ce indu ctan ce, vd have beenthe voltage drop a
cross ea ch devi ce, RL and L are the elements of load impedan ce, id have beenthe load
current flowing through the diode pairs and p have beenthe differential operator d/dt). V0
have beenthe A C side line voltage segment (va c, vb c, vba, v ca, v cb, vab during non-overlap,
and vb c+va c/2, vba+vb c/2, v ca+vba/2, v cb+v ca/2, vab+v cb/2, va c+vab/2 during overlap
intervals) based on diode pair condu ction. The phase currents isa, isb, and is c are obtained
by id, considering the respe ctive diode pair condu ction.
v fa (Vd c / 3)
Sb S c
(2S a )
v fb 2S (5.2.1)
(Vd c / 3)(Sa b
Sc
)
v f c (Vd c / 3)
S 2S c )
(Sa b
31
The three phase currents ifa, ifb, and if c flowing through impedan ces (R c, L c) are obtained
by solving the following differential equations:
pifa (1/ L c )(R c i (vsa v fa
fa )
pifb (v (5.2.2)
(1/ L c )(R c i sb v
fb
pif fb
)
(vs
c (1/ L c )(R c c v fc
i fc
)
The D C capa citor current maybe obtained in terms of phase currents ifa, ifb, and if c and
the
swit ching status (1 for on and 0 for off) of the devi ces Sa, Sb and S c
i dc i S
fa i fb S i fc S (5.2.3)
a
b c
From this, the model equation of the D C side capa citor voltage maybe written as
pV d
(1 / C )( i fa S i fb S i f S c (5.2.4)
c
dc a b c )