00998095
00998095
00998095
Abstract
This paper proposes a design method of the L-C output filter for three-phase UPS inverter: The harmonic components of the output voltages are dependent on the L-C value and the system time constant defined by the switching frequency, the speed of control processo< the controller gain, and etc., but their characteristics are complicate and entangled each othel: In this paper: the relation between the L-C value and the system time constant are described with the closed form and the filter values can be calculated to meet the limit of voltage THD considering the system time constant. The schematic design procedure is given by a single flow chart and the proposed filter design algorithm is verified through the simulation and experiments.
1 Introduction
There are two kinds of harmonic sources that make the UPS output voltage waveform be distorted: One is caused by the PWM switching of inverter and the other is caused by the nonlinear characteristics of the load. If any linear loads are connected to the inverter output side, then the output voltage harmonics due to the PWM switching can be eliminated by inserting the L-C filter between the inverter and the loads, and it is well known how to design the optimal values of the filter components [l, 21. But the configuration of the power circuit with L-C filter cannot meet the zero output impedance at all. Therefore, the capacitor voltage is usually collapsed in case of the sudden load change or under the nonlinear load. By using the fast feedback controller for the output capacitor voltage, the harmonic distortion due to the nonlinear load characteristics, such as the capacitor input type rectifier load or transient load change, can
be complemented [3]. But the performance of the feedback controller is dependent on the system time constant defined by the switching frequency, the speed of control processor, and the controller gain. And the harmonics suppression characteristics of the L-C filter and the feedback controller are entangled each other and complicate to analyze. Therefore, it is difficult to design the L-C filter considering the system time constant with the conventional methods. In this paper, the harmonics characteristics due to the L-C filter value and the system time constant are described with the closed form and L-C filter values can be calculated to meet the given specification of voltage THD, considering the system time constant. It means that the harmonic components due to the PWM switching of inverter and the nonlinear load characteristics are decoupled separately, and the L-C filter value and the performance of the feedback controller are determined to meet each allotted ratio of THD, for the harmonics suppression [4, 51. The design procedure of the L-C filter for threephase UPS inverter is described. The transfer function of the filter output voltage to the load current is given as the closed form and it is possible to get the L-C values by solving this closed form. The capacitor value can be calculated with the given system time constant defined by the performance of the system controller. This procedure is more practical to design the power circuit and control circuit of inverter for UPS than the conventional one. And also, the effect of the load current to the voltage distortion can be calculated from the closed form, so it is possible to analyzed the system how much the voltage waveform is distorted under the nonlinear load. Finally, the schematic design procedure is given by a single flow chart and the proposed filter design algorithm is verified through the simulation.
- 941 -
PCC-Osaka 2002
PWM INVERTER
LC FILTER
LOAD
minimize the cost function defined by the summation of the reactive powers. This design procedure can be applied to a linear load well. But under the nonlinear load or transient load change, we cannot neglect the harmonics components due to the second term.
- I I
I
Fig. 2 Single-phase equivalent block diagram. the input and output sides of the filters are transferred to,the dq-values under the stationary reference frame as (1) and (2).
If the inductor value is minimized at almost zero in (3), the capacitor voltage can be independent on the load current characteristics and it means that the UPS system works as an ideal voltage source. To minimize the inductor value, the capacitor value should be maximized. But as the increase of the capacitor value, the power rating of an inverter will be increased due to the increase of the reactive power. So it is permitted to increase the inductor value for the decrease of the capacitor value as far as the harmonic components of the capacitor voltage is controlled with the help of fast feedback voltage controller. It is clear from the above that the filter design value should be compromised as the performance of the system controller. But they are coupled each other and the filter values are adjusted as the performance of the controller by the try-and-error method in the conventional methods.
3
A
1 (i ,
- io)
c/s
The three-phase system can be described by the two single-phase systems of 'd' and 'q' as shown in Fig. 2 without the coupling terms. The dq singlephase equivalent blocks are shown as the MIS0 system with two inputs of the inverter output voltage and the load current in Fig. 2 . We will deal with the equivalent system of 'd' axis only in the equations below and it can be applied to 'q' axis equally. From Fig. 2 , the transfer function of the capacitor output voltage to the inverter output voltage and the load current can be described in (3) neglecting the resistance of inductor [ 5 ] .
Figure 3 shows the system block diagram with controller. The control system is designed based on double regulation loops: the main feature is the inner control loop of the current flowing through the filter capacitor can be limited within a given range and a second feedback has been performed using an outer voltage loop which makes it possible to improve the dynamics of the entire control system. The transfer function is driven as (4).
- 1
In conventional method of the filter design, the second term is treated as the disturbance and only the first term is considered to determine the L-C values. Once the product of L,Cj is calculated from the expecting cut-off frequency of the L-C filter, the exact values of Lfand Cf are determined at values that
In (4), let's consider the input-output transfer characteristics responding to the voltage command
- 942 -
and the load current separately. For the analysis of the effect due to the capacitor command voltage, the effect due to the load current can be neglected by putting zero, I,,=O. Then the controller gains, Kp, Ki, and K, can be designed to meet-the ideal controller. For this, the gain of the transfer function satisfies unit to make all the values of the nominator and the denominator are same within the passing bandwidth. From the above, (4) is simplified as (5).
time constant z and the capacitor value C is given as the closed form of (9). In this equation, the stability factors of are used here as y , =2.5 and y 2=2 which satisfy the Manabes recommendation [5,61. And also, from the same methods of comparison, and K, are all the values of controller gains, Kp, Ki, given as like the following equations.
If the inverter output L-C filter is designed well to eliminate the harmonics included in the inverter output voltage, then the command voltage of (5), Vcd* is substituted by the fundamental component of the inverter output voltage, Vadl as in (6).
4
For the analysis of the effect due to the change of load current only, the first term of the right side of (6) can be neglected and the transfer characteristics of the capacitor voltage to the load current is given as (7).
vcd(jnoo) (in%)
- L P ~ W , ~
Before proceeding of the filter design, the UPS specification and the system parameters are given as followings: 1) Total harmonic distortion factor of voltage, THD, is below 5% in case of both 0.8 linear lagging load and nonlinear rectifier load with the crest factor, CF=1.
IOd
K,K,
(7)
2 ) Modulation index, m y is given as 0.7 considering the variation range of DC link voltage and load.
3.2 Design of controller gain
For the design of controller gain, CDM(Coefficient Diagram Method) controller with observer canonical form is used [6]. The characteristic equation of (4) is defined in CDM controller by (8) to meet the robustness.
The flowchart of filter design procedure is shown in Fig. 4. In the design process, if the calculated total harmonic distortion factor of the output voltage THD, doesnt meet the specification, then it repeats the routine after decreasing the rate of voltage harmonics due to the dominant load current harmonics [4,5].
0 s YI
a r2
aoD
no
With the comparison of the coefficients of two equations, the transfer function of the capacitor voltage to the load current given by (7) is described as (9).
The voltage harmonics of SPWM-VSI is defined ideally as (13) and the fundamental component and dominant harmonics of inverter output voltage normalized by DC input voltage is given as Table I.
(9)
If the specification of the capacitor voltage harmonics caused by the nonlinear load current is given arbitrary, then the relation between the system
Under the assumption of ideal voltage waveform, the nonlinear load current harmonics of capacitor input type rectifier with the crest factor, CF=3 is analyzed using the computer simulation. Table I1 shows the Fourier coefficients of load current normalized by the fundamental component.
- 943 -
0.7
0.70
0.6
0.5
0.50
0.92 0.17
1.08 0.09
f 7 & 2
n
Mag.
1 1.00
3 0.73
7 0.07
9 0.04
11 0.02
13 0.01
0.35
xc
4.2 Determine the impedance ratio of inductor and capacitor
The impedance ratio of inductor and capacitor is determined to satisfy the condition that the ratio of output voltage to input voltage at the specific harmonic order shown in (14) can be smaller than 3[%] of the fundamental component of the output voltage as shown in (15). The specific harmonic order will be given as the same value of switching frequency. (14)
0.01 y2y,2 ~-
1
5 '
n2wiIon
In (16), W is the weighting coefficient determined due to the relative cost of inductor to the capacitor.
After the calculation of THD,, if it does not meet the specification, then the above process should be repeated from the top after modifying the condition as followings: 1) In case of linear load, repeat the calculation process after decreasing the impedance ratio of inductance and conductance from 3[%]. 2 ) In case of nonlinear load, repeat it after decreasing the harmonics ratio caused by the load current from 1["?I.
- 944 -
1
Calculating frequency transfer function of filter input to filter output voltage Determine the impedance ratio of filter inductor and capacitor
4
,
Making closed form of filter capacitance and system equivalent time constant using transfer function with contvoller
using closed form for load current harmonics of the amplitude of dominant
1
4
Selecting controller gains satisfying the system response the impedance ratio of
Analyzing output voltage harmonics under the linear and nonlinear load using transfer function considering the system performance of controller
Simulation results
current waveforms and the harmonic spectra of the output voltage waveform under the capacitor input type three-phase rectifier with CF= 6 are shown Table 111. System parameters for simulation
output Power output Voltage Considered Controller 30[kVA] 22O[V] Switching Frequency Controller time constant Considered 9.54[kHz] 2OO[us]
L/ = 165[uH]
Table I11 shows the system parameters and the design values of L-C filter and the controller gains using the filter design flowchart as shown in Fig. 4. In case of the linear R-L load, the THD, of capacitor output voltage is 3.2[%] when the L-C filter is designed by the conventional method based on the cost function and 3.7[%] when it is designed by the proposed method, respectively. From the results, it is verified that both methods are usefd under the linear load. From Fig. 5 to Fig. 8, the output voltage and
Lf = 4o[uH]
c/ =100[uF]
Rf = 0.05
. Cost function
c/ = 25[uF]
R/ = 0.05
- 945 -
300
300
-300
0.00
0 . 0 1
0.02
0.03
I
0 . 0 1
0.02
0.00
0.03
T i m [sec]
Tims [sec]
Fig. 5 Output voltage and current waveforms with LC filter designed by the conventional method.
Fig. 7 Output voltage and current waveforms with LC filter designed by the proposed method.
..
___
120
0
0
0
20 40 60 80
100
h _
160
120
140
160
20
40
60
80
100
140
Order
Order
. When the filter is designed by the conventional method, the voltage waveform is distorted a little and the its THD, is 8.5[%] with the given system time constant. Therefore, to meet the specification of THD,, the inductor value should be reduced and the capacitor value should be larger. And the inverter power rating will be increased due to the increase of the reactive power.
On the other hand, if the filter is designed by the proposed method, then the output voltage waveform is almost sinusoidal and the THD, comes within the specification with 4.6[%].
References
[l] S. Dewan and P. Ziogas, Optimum filter design for a single phase solid-state UPS system, IEEE Trans. Ind. Appl., vol. IA-15, no. 6, pp. 664-669, 1979.
[2] T. G. Habetler and D. M. Divan, Rectified Inverter reactive component minimization, in Conf. Rec. of IEEE PESC87, pp. 648-657, 1987.
[3] A. Kusko, D. Galler, and N. Medora, Output impedance of PWM UPS inverter-feedback vs. filters,
in Con$ Rec. o f IEEE IAS90, p p . 1044-1048, 1990.
6. Conclusion
This paper proposes a filter design procedure for three-phase UPS inverter. By using the proposed method, the filter values can be calculated from the closed form under the given system time Constant. Therefore, the L-C filter value can be determined quickly through the systematic procedure without using any try and error method.
[4]J. Kim, J. Choi, and F. Blaabjerg, Design and analysis of output filter for UPS, in ConJ: Rec. o f EPE2001, DS3.1, 2001.
[51 J. Kim, Design of output filter and controller for UPS inverter, Ph. D thesis in Chungbuk National University,
- 946 -