Professional Documents
Culture Documents
Modelling and Control of A Novel Zero-Current-Switching Inverter With Sinusoidal Current Output
Modelling and Control of A Novel Zero-Current-Switching Inverter With Sinusoidal Current Output
Modelling and Control of A Novel Zero-Current-Switching Inverter With Sinusoidal Current Output
Research Article
ISSN 1755-4535
Modelling and control of a novel zero-current- Received on 28th April 2015
Revised on 2nd May 2016
switching inverter with sinusoidal current Accepted on 17th May 2016
doi: 10.1049/iet-pel.2015.0296
output www.ietdl.org
Abstract: This study presents a novel topology of charge-transferring zero-current-switching (CT-ZCS) inverter to improve
the efficiency. Its operating principle, optimal zero-current-switching (ZCS) frequency and design guidelines are
elucidated in detail. A reduced-order modelling method of time-charge discretisation, which treats a set of two separate
ZCS as one combined switch cell, is put forward and the topology can be simplified to dual-buck half-bridge
inverter (DBHBI) topology. Since the DBHBI-derived CT-ZCS inverter permits higher comparison frequency of hysteresis
control than traditional hard-switching inverters, the three-level relay control with high-frequency pulse modulation
is recommended. Compared with two-level and three-level hysteresis control, the proposed control can
achieve effective output harmonic reduction, decreasing from 2.335 to 1.66%, and minimise the size of inverter
under the same output condition. In addition, a current controller is designed to optimise the system and enhance
its anti-disturbance performance. Finally, the experiment on a 35-V 2-A prototype verifies the validity of the
proposed method.
2 CT-ZCS inverter and its control system 3 CT-ZCS operation and design guidelines
2.1 Novel CT-ZCS inverter 3.1 Operation principle
To eliminate the disadvantages of conventional hard-switching In Fig. 1b, f1 and T1 are the frequency and period of high-frequency
inverters, the topology of CT-ZCS inverter is recommended and pulse, respectively. Correspondingly, f and T are the resonant
improved compared with the conventional topologies. The frequency and period. The operation principle of proposed inverter
series-resonant characteristic of inductors and capacitors is utilised is analysed in detail and its operation modes and equivalent
to ensure that the switching states change at zero-crossing point of circuits during a complete switching cycle are shown in Fig. 2.
resonant currents directly, thus ZCS is achieved and the switching To commence with the analysis, some assumptions are made as
losses of inverter can be reduced effectively by ZCS technology. follows:
The topology and proposed control system of the CT-ZCS inverter
are presented in Fig. 1a. Each of switches S1–S4 is constructed by (i) All the switches and diodes are ideal.
using a series-connected IGBT and diode. Moreover, the structure (ii) All the inductors and capacitors are ideal.
of CT-ZCS inverter is complementary symmetry. In addition to (iii) The filter capacitor CL is large enough, and the output voltage
small resonant inductors, a smoothing inductor LL is added to the uCL is constant in one switching period.
load side. This technique helps the output AC voltage across filter
(iv) At the beginning of T1, the initial voltage uC1 0− is −Ud.
capacitor CL (uCL ) convert to load current io in nature. (v) L1 = L2 = L3 = L4 = L, C1 = C2 = C ≪ CL, iL1 = iL3 and iL2 = iL4 .
(vi) F is a symbol of real-time charge-transfer state. ‘1’ stands for
2.2 Proposed control strategy positive charge transfer of the CT-ZCS inverter, ‘−1’ for negative
charge transfer and ‘0’ for no charge transfer.
Since the CT-ZCS inverter is under three-level relay control with
HFPM, each switch operates strictly at ZCS turn-on and ZCS Under proposed control, the operation principle can fall into three
turn-off in each comparison period Ts, which means the switching modes as follows. The series resonant frequency of equivalent
1 ⎧
f = √ (1) ⎪
⎪ (Ud /s) − (uC1 (0− )/s)
2p LC ⎪
⎨ iL1 (s) = R + L s + (1/C s)
1 1
1 (2)
where 1/(1/C1 + 1/CL) ≃ C1 = C. ⎪
⎪ Ud uC1 (0− ) 1/C1 s
⎪ uC (s) =
⎩ −
1 s s R1 + L1 s + (1/C1 s)
3.1.1 Mode 1 [uref . uCL + H] [refer to Figs. 2a, g, c and e]: Formula (2) can be solved as follows
In this mode, S1, S2 and D1 of the CT-ZCS inverter work alternately
⎧
from stage 1 to stage 4, and S1 and S2 work in ZCS condition. In ⎨ i (t) = Ud − uC1 (0− ) sin vt
Figs. 2a, g, c and e, the equivalent circuits transfer positive L1
L 1 v0 (3)
charges and energy from power supply to load side via ⎩
uC1 (t) = Ud − [Ud − uC1 (0− )] cos vt
intermediate capacitor C1.
Stage1 [0, T/2] [refer to Fig. 2a]: The series resonance of L1 and C1 where v = v0 1 − (a/v0 )2 , v0 = 1/ L1 C1 , α = R1/2L1. R1 ≃ 0,
happens. S1 is turned on and S2 is off. F = 0 and R1 is parasitic then α = 0, ω = ω0.
resistor. According to the equivalent circuit in Fig. 3a, the According to the expressions (3), the current iL1 is sinusoidal as
differential equations are transformed to complex frequency shown in Fig. 3c. At t = 1/4T, the current peak value iL1m can be
T1 1 1+D
t1 = + arccos (11)
2 v D−3
1
Em2 = C1 Ud2 (12)
2
Fig. 3 Equivalent circuits in the complex frequency domain and transient
The transient current iL2 (t1) can be calculated as
analysis of the CT-ZCS inverter
a Stage 1 (t ≤ T/2)
b Stage 3 (T1/2 < t ≤ t1) C
c Transient waveforms iL2 (t1 ) = 2 2Ud (Ud − uCL ) (13)
L2
expressed as Part of energy stored in L2 is transferred to load side and the partial
energy Em3 is derived as
C1
iL1m = [Ud − uC1 (0− )] (4) 4CDUd (Ud − uCL )
L1 Em3 = (14)
1 + sgn(F)D
At t = T/2, the maximum value of uC1 (uC1m1) can be expressed as
where sgn(.) is the signum function. At t = t1, iS2 decreases to zero.
uC1m1 = 2Ud − uC1 (0− ) (5) The average current IS2 through S2 is equal to IS1 based on charge
balance of C1.
Stage 4 [t1, T1] [refer to Fig. 2e]: The series resonance of C1, L2 and
The capacitor C1 can store maximum energy Em1 which is CL has finished. All switches are turned off. First, iL2 freewheels
calculated as through D1, and F = 1. The moment t2 that iL2 decreases to zero
can be calculated as
1
Em1 = C1 [2Ud − uC1 (0− )]2 (6) √
2 2 2(1 − D)
t2 = t1 + (15)
During the whole period T1, the average current through S1 (IS1 ) is v[1 + sgn(F)D]
equal to the average value of iL1 (IL1 ) which can be calculated as
Second, if t2 < T1, it has the same condition with Stage 2 when t2 ≤
2C1 [Ud − uC1 (0− )] t ≤ T1. However, if t2 ≥ T1, there is no final condition.
IL1 = (7) At t = T1, uC1 decreases to –Ud and the total energy (Em1–Em2 +
T1 Em3) is transferred from power supply to load side. Although
The voltage stresses and current stress of diodes D1 and D2 can be 4 Modelling and control strategy
derived directly as
4.1 Time-charge discretisation
⎧
⎨ VD1 ,max = VD2 ,max = 4Ud
⎪
As for switching states, ‘1’ stands for ZCS turn-on and ‘0’ for
2C (17) turn-off. The relationship between switching symbol mapping and
⎪
⎩ iD1 ,max = iD2 ,max = 2Ud
L charge transfer is shown in Fig. 5, and F can be expressed as F =
(S12 + D1)−(S34 + D2). In the discrete time domain, the CT-ZCS
The resonance remains unaffected by the difference of power factor inverter transfers quantified charges Q(k) to CL in each T1, which
in stage 1. In stage 3, the load side can be treated as a constant results in voltage variation of CL. The total charges Q(k) includes
current source during each T1 with inductive load, and the vast inductor-transferring charges QL(k) and capacitor-transferring
majority of resonant current iL2 or iL4 flows into CL instead of the charges QC(k). They can be calculated as follows
equivalent constant current source, thus the resonance remains
unaffected with inductive load. It can also remain unaffected with ⎧
capacitive load because C is far less than the parallel capacitance ⎨ QC (k) = 4CUd
of capacitive load and CL. Hence, the ZCS operation remains 4C(1 − D) (19)
⎩ QL (k) =
uninfluenced by the difference of power factor, even under Ud [1 + sgn(F)D]
non-unity power factor.
3.2 ZCS operation limitation Time and charges can be quantified by taking T1 as the time-scale
unit and Q as the charge-scale unit. As the switching symbol
To ensure ZCS, iL2 must reset before next Stage 3. Thus the mapping shows in Fig. 5, the turn-off of S1 and S2 or S3 and S4 in
constraint is obtained as the CT-ZCS inverter correspond to that of S12 or S34 in the
√ DBHBI, while switching states of S1 and S2 or S3 and S4 changing
2 2(1 − D) 1 1+D f from 1 and 0 to 0 and 1 correspond to the turn-on of S12 or S34.
Dt = + arccos ≤ 2p (18)
[1 + sgn(F)D] v D−3 f1 Switching states of D1 and D2 in the CT-ZCS inverter correspond
to that of D1 and D2 in the DBHBI. When ignoring the
In Fig. 4, when D ranges from 0 to 1, the theoretical maximum of intermediate resonance for model reduction, the output
ZCS frequency f1 should be lower than f1min to achieve ZCS. characteristics of DBHBI-derived CT-ZCS inverter with hysteresis
Hence, there is no any practical limitation of D for actualising control can be analysed analogically by substituting one combined
ZCS as shown in Fig. 4 as long as f1 < f1min under ideal switch cell S12 or S34 for a set of separate soft-switching S1 and S2
conditions. Nonetheless, if resonant inductors mismatch between or S3 and S4. Either S12 and D1 or S34 and D2 operate during each
L1 and L2 or L3 and L4 is more than 25% or so and considering T1 when uCL , uref − H or uCL . uref + H. Hence, an equivalent
some measurement error for inductance, the series resonance DBHBI with HFPM can be obtained based on time-charge
cannot accomplish completely in stages 1 and 3, which will lead discretisation as shown in Fig. 6a. Furthermore, the higher
to the ZCS condition broken. To assure that series resonance and switching frequency of ZCS operation means the smaller scale
ZCS can be completed in stages 1 and 3, the optimal f1 should be interval of time-charge discretisation, and the operation
further lower than the theoretical maximum of ZCS frequency characteristics between the DBHBI-derived CT-ZCS inverter and
appropriately. DBHBI are more approximate.
4.2 Controller comparison Since the switching losses can be eliminated completely by
employing ZCS technology, the ZCS inverters permit higher
In Fig. 6b, GI(s) is the current regulator, GV(s) is the open-loop comparison frequency and m is permitted to approach 1.
voltage transfer function, and kuf and kif are the feedback One intersection point X1 exists on the two curves of GV( jω) and
coefficients of uCL and io, respectively. The outer-current loop −1/N1(A) when A = H under two-level hysteresis control. At the
achieves voltage pre-estimate which is used as the voltage reference. point X1, the curve of −1/N1(A) completes a transition from
Only when the stability of inner-voltage loop is considered but the unstable region to stable region along with the direction of
ripple of io with resistive load R is neglected, is the power supply of increasing A. Therefore, the non-linear system performs a stable
equivalent DBHBI a voltage pulse sequence with ±Ud and zero periodic motion containing an oscillation with constant amplitude
potential, hence the reduced-order small-signal state equations of H. Similarly, there are also one intersection point X2 and constant
equivalent DBHBI can be obtained as oscillation under three-level hysteresis control, but the output
⎛ ⎞ ⎛ ⎞ fluctuation of three-level hysteresis control is smaller than that of
1
dîL (t) 1 two-level hysteresis control with the same H. However, there is no
−
⎜ dt ⎟ ⎜ 0 L ⎟ intersection point on the curves of GV( jω) and −1/N3(A) under
⎜ ⎟ ⎜ ⎟ îL (t) + − L ûd (t)
⎝ dûC (t) ⎠ = ⎝ 1 1 ⎠ ûC (t) (20) three-level relay control, and −1/N3(A) is not surrounded by the
L − L 0 curve of GV( jω). The non-linear system is stable without constant
dt CL RCL
amplitude oscillation. Furthermore, the output voltage fluctuation
of three-level relay control is smaller than that of both hysteresis
Then the small-signal model of voltage loop can be derived as control.
R
GV (s) = (21)
RLCL s2 + sL + R 4.3 Current regulator design
GV(s) is linear transfer function of open-loop system whose poles are Through large-signal analysis under three-level relay control, uCL is
located in the left-half s plane. constrained near uref + H or uref−H as shown in Fig. 8. Since the
To analyse and contrast the control performance, N(A) adopts comparison frequency fs is much higher than the fundamental
three-level relay controller, two-level and three-level hysteresis frequency of uCL , the large-signal model of voltage loop is a stable
controllers with hysteresis band H, respectively. The amplitude non-linear system.
and phase curves are obtained as shown in Fig. 7. The reason why In Fig. 8b, uref (k−1), uref (k) and uref (k + 1) represent the voltage
three-level relay control is a specific hysteresis control is given as references of previous, current and next moments, respectively, and
follows k uCL l for average output voltage which k uCL l = uref − H. Assuming
⎡
⎤ k uCL l has tracked uref (k−1)−H at (k−1)Ts, it will take one
2 comparison period Ts to make k uCL l increase from uref (k−1)−H to
2Ud ⎣ mH 2 H ⎦
lim 1− + 1− uref (k)−H or from uref (k)−H to uref (k + 1)−H. When fs is high
m1 pA A A enough, the voltage closed-loop can be equivalent to a voltage
follower with the output lagging reference one Ts. Regarding inner
2Ud H loop as a one-order inertia link with the time constant Ts, there is
+j (m − 1) = N3 (A) (22)
p A2 Ks/(sTs + 1) where Ks = 1/kuf.
To improve anti-disturbance performance, the system is changed where h is the logarithmic bandwidth of middle-frequency band
into a typical type-II system by designing current regulator with where the slope is −20 dB/dec. Moreover, the closed-loop transfer
amplitude-limited output. The current regulator is given as function is obtained as
k pi (ti s + 1) io (s) ti s + 1
GI∗ (s) = (23) = (26)
ti s iref (s) (LL Ts ti /Ks k pi )s + (LL ti /Ks k pi )s2 + kif ti s + kif
3
The open-loop transfer function of the whole system is shown as Generally, h ranges from 3 to 10. When h decreases, the
anti-disturbance performance will be enhanced. However, the
(Ks kif k pi /LL ti )(ti s + 1) overshoot will increase. Furthermore, the recovery time will
G(s) = (24) decrease first, and then increase in the aforementioned situation.
s2 (sTs + 1)
Considering various indicators of tracking and anti-disturbance
performance, it is better to fix h = 5.
Then the PI parameters can be calculated as follows
Fig. 8 Transient waveforms of uCL and its enlargement under three-level relay control
a Voltage uCL
b Partial enlargement
the relay control has higher tracking accuracy. The output current and less output fluctuation can be acquired during each T1 because
demonstrates that effective harmonic reduction, decreasing from of the smaller resonant parameters. TMS320F28335 can process
2.335 to 1.66%, has been achieved. If the inverter operates at the generated comparison logic more timely and obtain higher
higher average switching frequency, the less transferred energy regulation accuracy of current tracking under three-level relay
control, thus the lower THD will be achieved. The THD of io with (iii) The three-level relay control with HFPM is implemented based
different CL and LL under different control strategies is shown in on the CT-ZCS inverter. Compared with three-level hysteresis
Fig. 12e. The PI parameters have been adjusted appropriately with control, the recommended method achieves effective harmonic
LL and fs to achieve optimal control according to expression (25). reduction, reduces the size and improves power density.
The THD decreases as the CL or LL increases under both control (iv) The anti-disturbance performance is enhanced by designing the
methods. Moreover, the THD is reduced availably by adopting current regulator.
relay control with the same CL and LL, which validates the relay
control can permit smaller parameters and size of filter capacitor Finally, the proposed modelling and control methods can be applied
and inductor than the hysteresis control, and minimise the size of to other ZCS inverters. To achieve optimal steady and transient
inverter further. performance, the regulator parameters should be adjusted
When the three-level relay control is adopted, Figs. 13a and b appropriately with the circuit parameters and comparison frequency.
show the experimental waveforms of uCL and io at load step The proposed inverter can achieve ZCS operation by utilising series
change mutating from 40 to 20 Ω and from 20 to 40 Ω. The slight resonance, but the voltage stress and current stress of power devices
fluctuation of io validates the CT-ZCS inverter system has strong increase to some degree. Further effort is needed to explore optimal
robustness against the load mutation. The efficiency of CT-ZCS ZCS strategy with less voltage stress and current stress.
inverter, flyback inverter [25], half-bridge inverter and DBHBI
[26] are shown in Fig. 13c. Although the efficiency of flyback
inverter is a little higher than that of CT-ZCS inverter at low
7 References
output power, the efficiency of flyback inverter is the lowest with
the increase of output power. It is mainly caused by isolation 1 Ebrahimi, A., Fathi, H.S., Gholamrezaei, H.: ‘Novel switching pattern for
transformer and full load-cycle operating mode in flyback inverter. single-stage current source inverter for grid-connected photovoltaics’, IET Proc.
It can be observed from Fig. 13c that the CT-ZCS inverter Power Electron., 2014, 7, (10), pp. 2447–2454
possesses higher efficiency than DBHBI and half-bridge inverter 2 Geury, T., Pinto, S., Gyselinck, J.: ‘Current source inverter-based photovoltaic
system with enhanced active filtering functionalities’, IET Proc. Power Electron.,
by ZCS strategy. Moreover, it should be noted that the proposed 2015, 8, (12), pp. 2483–2491
inverter system eliminates the contradiction between hysteresis 3 Cancelliere, P., Colli, V.D., Di Stefano, R., et al.: ‘Modeling and control of a
band and switching frequency existing in hard-switching inverters. zero-current-switching DC/AC current-source inverter’, IEEE Trans. Ind.
Electron., 2007, 54, (4), pp. 2106–2119
4 Li, R.T.H., Chung, S.H., Lau, W.H., et al.: ‘Use of hybrid PWM and passive
resonant snubber for a grid-connected CSI’, IEEE Trans. Power Electron., 2010,
6 Conclusions 25, (2), pp. 298–309
5 Sarnago, H., Lucía, O., Mediano, A., et al.: ‘High efficiency parallel quasi-resonant
current source inverter featuring SiC MOSFETs for induction heating systems with
The validity and feasibility of proposed method are verified by coupled inductors’, IET Proc. Power Electron., 2013, 6, (1), pp. 183–191
simulation and experiment. This paper has the following characteristics. 6 Zhang, J.S., Liu, X.C., Hu, X.Y.: ‘A new topology of soft-switching inverter’,
J. China Univ. Pet., 2011, 35, (5), pp. 167–172
7 Mishima, T., Takami, C., Nakaoka, M.: ‘A new dual mode current
(i) A novel topology of ZCS inverter, which improves the phasor-controlled ZVS high-frequency resonant inverter for induction heating’.
efficiency for low-power applications, is presented. Proc. IEEE PEDS, 2013, pp. 1033–1038
(ii) The reduced-order modelling method of time-charge 8 Wang, C.M., Lin, C.H., Lin, H.Y., et al.: ‘Analysis, design and performance of a
discretisation is applied in simplification of the inverter topology. soft-switching single-phase inverter’, IET Proc. Power Electron., 2014, 7, (9),
pp. 2412–2423
The simplification scheme provides some certain reference value 9 Ahmed, N.A.: ‘High-frequency soft-switching AC conversion circuit with
to the analysis of other complex topologies which the switches dual-mode PWM/PDM control strategy for high-power IH applications’, IEEE
work synergistically. Trans. Ind. Electron., 2011, 58, (4), pp. 1440–1448