Professional Documents
Culture Documents
where vip ≥ 0 and vin ≤ 0 , with i={a, b, c}. The signals with i 0 = d a 0 i a + d b 0 ib + d c 0 i c , (6)
the subscript ‘p’ will only cross the upper carrier, in which d i 0 = si 0 for i={a, b, c}. Assuming that the
p
vcarrier ∈ [0, 1] , and the signals with the subscript ‘n’ will only frequency of the carriers is much higher than the frequency of
n
cross the lower one, vcarrier [− 1, 0] . In Section IV, the process the modulation signals, the duty cycles can be expressed as
follows:
that is implemented when comparing the modulation signals
with the carriers is explained. It is anticipated that the d i 0 = vin+1 − vip where vin+1 = vin + 1 . (7)
connection to the NP (‘0’ level) is produced when: Relationship (7) is obtained from (3) and using basic
vip > vcarrier
p n
and vin < vcarrier , or trigonometry in Fig. 2. As a result:
p n
(3) +1
i0 = van +1
− vap ia + vbn +1
− vbp ib + vcn − vcp ic . (8)
vip ≤ vcarrier and vin ≥ vcarrier .
The two inner transistors of a phase leg of the converter If:
(Fig. 1) are in the on state when the corresponding NP control van − vap = vbn − vbp = vcn − vcp = v x , or (9)
variable si0 ∈{0, 1}, for i={a, b, c}, is activated. In other +1
van +1
− vap = vbn +1
− vbp = vcn − vcp = 1 + v x ,
words, when a variable si0 takes the unity value, the
subsequent output phase is connected to the NP; otherwise it then the averaged NP current would be:
takes the value of zero. When one phase leg is clamped to the i0 = 1 + v x (ia + ib + ic ) (10)
NP, its output current is injected to this point. Therefore, the Since the neutral of the load is open (Fig. 1) or is just a
current i0 can be expressed as follows: triangle-connected load, there is not a zero-sequence of
i0 = s a 0ia + sb 0ib + sc 0ic . (4) current. Subsequently, the sum of the output currents is always
In order to preserve voltage balance, the locally-averaged zero ( ia + ib + ic = 0 ); hence, i0 becomes zero.
NP current must be zero. Therefore, it is necessary to operate In conclusion, the problem of maintaining the locally-
with the averaged NP current instead of the instantaneous averaged voltages on the dc-link capacitors constant is reduced
current. The averaged NP current is obtained by using the to find a proper value for vx. An infinite number of solutions
moving average operator: can be found; however, one especially interesting solution can
t achieve minimum switching frequencies in the devices of the
∫
1
x (t ) = x(τ )dτ , (5) converter. This solution is found by forcing the variables vip
Ts t −Ts and vin to be zero for the maximum time possible, since when
where Ts is the sampling or switching period. Applying this these signals are zero some of the transistors do not switch
operator to (4), one obtains the following: (none of the modulation signals cross a carrier signal).
1055
1.5
Regarding this restriction and relationships (1), (2) and (9), the
va vb vc
following solution is obtained:
1
max(va , vb , vc ) − min(va , vb , vc )
vx = − . (11)
2 0.5
vip = and
2 for i = {a, b, c}. (12)
v − max(va , vb , vc ) -0.5
vin = i ,
2
-1
The algorithm can be easily implemented using the scheme
shown in Fig. 3. -1.5
0 π/3 2π/3 π 4π/3 5π/3 2π
Under the assumption of sinusoidal modulation signals,
Angle
such that:
(a)
v a = m SPWM cos ω t ,
vb = m SPWM cos (ω t − 2π / 3), and
1.5
(13)
v = m
c SPWM cos (ω t + 2π / 3), 1
vap
the solution for vap is:
0.5
- for 0 ≤ ω t ≤ 2π 3 ,
v ap = (v a − v c ) 2 = ( 3 2) m SPWM cos(ω t − π 6 ), 0
- for 2π 3 ≤ ω t ≤ 4π 3 , -0.5
1056
constant. However, this does not imply that these voltages These variables can take three values, xi ∈ {− 1, 0, 1} ,
are equal. In fact, in theory, if the initial voltages on the which are the possible output levels for each phase. In order
capacitors were different, this modulation strategy would to obtain the control functions of the transistors, the
tend to preserve imbalance because the locally-averaged NP following relationships are needed:
current is zero. In practice, the situation is even worse since
dead times, different values and behavior of the components, s i 4 = 1 if x i = 1,
s = 1 if x i = 1 or x i = 0,
etc., can make the voltages drift slowly and without control. i3
Thus, some compensation for voltage imbalances must be (19)
provided. s i 2 = 1 if x i = 0 or x i = −1, and
s i1 = 1 if x i = −1.
A control method for voltage compensation can be simply
shifting the modified modulation signals in accordance to the If the respective above conditions are not met, then
sign of the voltage error (∆vC>0). An inconvenience of this variables si1, si2, si3, and si4 are correspondingly zero. These
method is that when the signals are shifted, they do not variables are associated to the on and off states of the
preserve any interval clamped to zero. As a consequence, the transistors (on=‘1’ and off=‘0’), where two of them are
switching frequencies of the devices increase. Furthermore, always on for each of the valid 27 electrical states of this
the sign of the power flux in the system needs to be known in converter.
order to provide a proper shift to the signals. The balancing
strategy proposed in this paper avoids increasing the V. EXPERIMENTAL RESULTS
switching frequencies of the devices. Moreover, it does not
need to detect the direction of the power flux in the system. The proposed modulation technique has been verified by
This technique is explained below. simulation and experiment. The values of the dc-link voltage
Regarding vap and van in Fig. 4(b), none of the signals are and the capacitors are Vdc=96 V and C=1100 µF,
clamped to zero throughout the intervals π 3 ≤ ω t ≤ 2π 3 respectively. The converter operates over an R-L Wye-
and 4π 3 ≤ ω t ≤ 5π 3 . Therefore, the modified modulation connected load with L= 50 mH and R= 10 Ω. The frequency
signals of phase a can be shifted up or down during these of the carriers is 5 kHz and the modulation index mSPWM=0.9
intervals without increasing the switching frequencies of the in all the results.
devices. Similarly, there are other intervals for phases b and Fig. 5(a) and 5(b) show the results obtained from the
c in which none of the associated modulation signals are application of a standard SPWM. In Fig. 5(a), a line-to-line
clamped to zero, and therefore can be shifted. Although this output voltage (vab) is illustrated, and in Fig. 5(b) the
strategy preserves the switching frequency of the devices variables shown are the voltages on the dc-link capacitors
when the compensator is activated, a significant drawback is (vC1 and vC2) together with two output currents (ia and ib).
the extremely slow voltage-balancing dynamics of the Note that there are significant low-frequency voltage
system. This occurs because only the modified modulation oscillations on the dc-link capacitors. The modulation
signals associated to one phase are shifted at any time. In frequency has been selected to be very low in this example
order to improve the balancing dynamics, the sign of the (fm=20 Hz) in order to emphasize this effect. Fig. 5(c) and
output currents should be sensed. Furthermore, relationship Fig. 5(d) show the same results when the proposed
(2) must be preserved in order to avoid distortion in the modulation technique is applied. In this case, the voltages on
output voltages during the compensation. This occurs if the the dc-link capacitors do not contain any low-frequency
offset applied to the modified modulations signals have oscillation, but only high-frequency ripples related to the
opposite signs. The offset applied to vip is: switching frequency. Note, however, that the switching
frequencies of the devices increase compared to the former
vi _ off = − k p ∆vC ⋅ sign(∆vC ii ) ⋅ sign(vip − vin − 1) . (17) case, since there are intervals in which the line-to-line
The absolute value of the voltage imbalance is multiplied voltages commutate among three states of the converter
by variable kp. The term sign(∆vC·ii) defines a sign for the instead of two. This is produced when neither of the two
compensation. Nevertheless, the final sign of the offset corresponding modified modulation signals is clamped to
depends on sign(vip−vin−1) or sign(vip−v’in), in accordance zero.
with the two possible cases shown in Fig. 2. In Fig. 6, the modulation frequency in this example is f=80
Hz. Observe that the voltages on the dc-link capacitors
become equal very quickly.
IV. CONTROL SIGNALS FOR THE TRANSISTORS Fig. 7 shows a simulation example in which a set of
second and fourth current harmonics have been added to the
The output level activated at each phase of the converter is load currents. These harmonics are very harmful for the
defined by a comparison of the modified modulation signals standard SPWM strategy, since they may introduce
with the carriers, as follows: instability to the NP voltage, which is also a problem when
p
If vip > vcarrier then xip = 1, otherwise xip = 0 operating under standard SV-PWM techniques [10].
n However, when the proposed modulation strategy is applied,
If vin < vcarrier then xin = 1, otherwise xin = 0
the voltages on the dc-link capacitors do not show any
x i = x ip − x in , i ∈ {a, b, c}. (18) negative effect and the system is always stable.
1057
vab vab
(a) (c)
v C1
vC1, vC2
v C2
ia ib ia ib
(b) (d)
v C1
v C2
ia ib
1058
(V), (A) This work made use of Engineering Research Center
60
vC1 Shared Facilities supported by the National Science
Foundation under NSF Award Number EEC-9731677 and
the CPES industry Partnership Program. Any opinions,
40 vC2 Standard SPWM findings and conclusions or recommendations expressed in
becomes unstable this material are those of the authors and do not necessarily
reflect those of the National Science Foundation.
1059