Hysteresis Current Control in Three-Phase Voltage Source Inverter

Mirjana Milo sevi c

Abstract The current control methods play an important role in power electronic circuits, particulary in current regulated PWM inverters which are widely applied in ac motor drives and continuous ac power supplies where the objective is to produce a sinusoidal ac output. The main task of the control systems in current regulated inverters is to force the current vector in the three phase load according to a reference trajectory. In this paper, two hysteresis current control methods (hexagon and square hysteresis based controls) of three-phase voltage source inverter (VSI) have been implemented. Both controllers work with current components represented in stationary (, ) coordinate system. Introduction Three major classes of regulators have been developed over last few decades: hysteresis regulators, linear PI regulators and predictive dead-beat regulators [1]. A short review of the available current control techniques for the threephase systems is presented in [2]. Among the various PWM technique, the hysteresis band current control is used very often because of its simplicity of implementation. Also, besides fast response current loop, the method does not need any knowledge of load parameters. However, the current control with a xed hysteresis band has the disadvantage that the PWM frequency varies within a band because peak-topeak current ripple is required to be controlled at all points of the fundamental frequency wave. The method of adaptive hysteresis-band current control PWM technique where the band can be programmed as a function of load to optimize the PWM performance is described in [3]. The basic implementation of hysteresis current control is based on deriving the switching signals from the comparison of the current error with a xed tolerance band. This control is based on the comparison of the actual phase current with the tolerance band around the reference current associated with that phase. On the other hand, this type of band control is negatively affected by the phase current interactions which is typical in three-phase systems. This is mainly due to the interference between the commutations of the three phases, since each phase current not only depends on the corresponding phase voltage but is also affected by the voltage of the other two phases. Depending on load conditions switching frequency may vary during the the fundamental period, resulting in irregular inverter operation. In [4] the authors proposed a new method that 1

minimize the effect of interference between phases while maintaining the advantages of the hysteresis methods by using phase-locked loop (PLL) technique to constrain the inverter switching at a xed predetermined frequency. In this paper, the current control of PWM-VSI has been implemented in the stationary (, ) reference frame. One method is based on space vector control using multilevel hysteresis comparators where the hysteresis band appear as a hysteresis square. The second method is based on predictive current control where the three hysteresis bands form a hysteresis hexagon. Model of the Three-Phase VSI The power circuit of a three-phase VSI is shown in gure 1. The load model is consisting of a sinusoidal inner voltage e and an inductance (L).
idc Sa ia L
+

ea

Udc

+ -

Sb

ib

L
+

eb

Sc

ic

L
+

ec

Figure 1: VSI power topology To describe inverter output voltage and the analysis of the current control methods the concept of a complex space vector is applied. This concept gives the possibility to represent three phase quantities (currents or voltages) with one space vector. Eight conduction modes of inverter are possible, i.e. the inverter can apply six nonzero voltage vectors uk (k = 1 to 6) and two zero voltage vectors (k = 0, 7) to the load. The state of switches in inverter legs a, b, c denoted as Sk (Sa , Sb , Sc ) corresponds to each vector uk , where for Sa,b,c = 1 the upper switch is on and for Sa,b,c = 0 the lower switch is on. The switching rules are as following: due to the DC-link capacitance the DC voltage must never be interrupted and the distribution of the DC-voltage Udc into the three line-to-line voltages must not depend on the load. According to these rules, exact one of the upper and one of the lower switches must be closed all the time. There are eight possible combinations of on and o switching states. The combinations and the corresponding phase and line-to-line voltages for each state are given in table 1 in terms of supplying DC voltage Udc . If we use the transformation from three-phase (a,b,c ) into stationary (, ) coordinate system:

u u

2 3

1 3 1 3

1 3 1 3

ua ub uc

(1)

this results in eight allowed switching states that are given in table 1 and gure 2. State u0 u5 u3 u4 u1 u6 u2 u7 Sa 0 0 0 0 1 1 1 1 Sb 0 0 1 1 0 0 1 1 Sc 0 1 0 1 0 1 0 1 ua /Udc 0 1/3 1/3 2/3 2/3 1/3 1/3 0 ub /Udc 0 1/3 2/3 1/3 1/3 2/3 1/3 0 uc /Udc 0 2/3 1/3 1/3 1/3 1/3 2/3 0 uab 0 0 -1 -1 1 1 0 0 ubc 0 -1 1 0 0 -1 1 0 uca 0 1 0 1 -1 0 -1 0 u /Udc 0 1/3 1/3 2/3 2/3 1/3 1/3 0 u /Udc 0 1/ 3 1/ 3 0 0 1/ 3 1/ 3 0

Table 1: On and O states and corresponding outputs of a three-phase VSI

b

1 U DC 3

u3(010) S2

u2(110) S1

S3 u0(000) u7(111) S4 S5
2 - U DC 3

uref u1(100) S6
a

u4(011)

1 U DC 3

u5(001) 1 - U DC 3

u6(101) 2 1 U DC U DC 3 3

Figure 2: Switching states of the VSI output voltage

Hexagon Hysteresis Based Control Three hysteresis bands of the width are dened around each reference value of the phase currents (ia , ib , ic ) (gure 3).
ia,b,c d

time

Figure 3: Hysteresis bands around the reference currents ia , ib , ic The goal is to keep the actual value of the currents within their hysteresis bands all the time. As the three currents are not independent from each other, the system is transformed into (, ) coordinate system. With the transformation of the three hysteresis bands into this coordinate system, they result in an hysteresis hexagon area. The reference current vector iref points toward the center of the hysteresis what can be seen in gure 4. In steady state, the tip of the reference current moves on circle around the origin of the coordinate system (gure 4). Therefore, the hexagon moves on this circle too.
bc,b

b SIII SII

ie
SIV

ie
SV SVI

SI

a ab

iref i
ca

Figure 4: Hysteresis hexagon in , plane The actual value of the current i has to be kept within the hexagon area. Each time when the tip of the i touches the border of the surface heading out of the hexagon, the inverter has to be switched in order to force the current into 4

the hexagon area. The current error is dened as:

ie = i iref

(2)

The error of each phase current is controlled by a two level hysteresis comparator, which is shown in gure 5. A switching logic is necessary because of the coupling of three phases.
ia ia,ref
SIV
d

SI

ib ib,ref
SVI
d

SIII

Switching logic

Switches states

ic ic,ref
SII
d

SV

Figure 5: Structure of hysteresis control When the current error vector ie touches the edge of the hysteresis hexagon, the switch logic has to choose next, the most optimal switching state with respect to the following: 1) the current dierence ie should be moved back towards the middle of the hysteresis hexagon as slowly as possible to achieve a low switching frequency; 2) if the tip of the current error ie is outside of the hexagon, it should be returned in hexagon as fast as possible (important for dynamic processes). In order to explain the control method the mathematical equations should be introduced (gure 6).

i uk

Figure 6: The load presentation 1 di = (uk e) dt L 5

(3)

According to equation 2, the current error deviation is given by: die di diref = dt dt dt From equations (3) and (4) we have: 1 die = (uk uref ) dt L where the reference voltage uref is dened by: diref dt

(4)

(5)

uref = e + L

(6)

The reference voltage uref is the voltage which would allow that the actual current i is identical with its reference value iref . In [5] the authors explained why the decisive voltage for the current control is the sum of the inner voltage and the voltage across the inductance of the load. The switching logic for the switches has to select the most optimal out of eight switching states according to the mentioned criteria. For the optimal choice of the switching state, only two pieces of information are required: 1) the sector S1 , S2 , ..., S6 (gure 2) of the reference voltage, 2) the sector SI , SII , ..., SV I (gure 4) in which the current error vector touches the border of the hexagon. For the derivation of the stationary switching table one example would be discussed. Let reference voltage vector uref be somewhere in sector S1 (gure 2). According to equation 5 the current error deviation is somewhere in one of the hatched areas in gure 7. These seven areas describe direction and speed with which the current error deviation can move. 1. If ie touches the border of hexagon in sector SI : To get back towards the middle of the hexagon, ie must move in direction of a negative component. It means that vector uk uref must have a negative component. The hatched areas A0 , A3 , A4 and A5 corresponding in full to this

bc, b

u3
u3-uref A3 u2-uref A2

u2

uref
u1-uref A1

u4-uref A4 ca

u4

u0-uref u7-uref A0,7

u0 u7

u1

u5-uref A5

u5

u6-uref A6

u6

ab

Figure 7: Corresponding areas for uk uref criterion are those that suit for states u0 , u3 , u4 and u5 . The second criterion for the choice of the next optimal state is the length of vector uk uref , which is proportional to the speed of ie . The speed should be as small as possible, which implies that the length of vector uk uref must be the shortest. It can be seen from gure 7 that state u0 is the optimal choice because vector u0 uref has the minimum length. 2. If ie touches the border of hexagon in sector SII : To get back ie towards the middle of hexagon, vector uk uref must be below the ab axis. Hatched areas A0 , A4 , A5 and A6 full this condition (gure 7). Vector u0 uref has the shortest length among vectors uk uref (k=0, 4, 5, 6). Therefore, state u0 is the optimal choice. 3. If ie touches the border of hexagon in sector SIII : To get back ie towards the middle of hexagon, vector uk uref must be below ca axis (gure 7). Areas A1 , A5 and A6 satisfy this condition in full and state u1 has the shortest length of vector u1 uref and this is the optimal choice. 4. If ie touches the border of hexagon in sector SIV : To get back towards the middle of hexagon, vector ie must move in direction of a positive component (gure 7). Only state u1 satises this condition fully and therefore, this is the optimal choice.

5. If ie touches the border of hexagon in sector SV : To get back ie towards the middle of hexagon, vector uk uref must be beyond ab axis. Only state u2 fulls this condition and this is the optimal choice (gure 7). 6. If ie touches the border of hexagon in sector SV I : To get back ie towards the middle of hexagon, vector uk uref must be beyond ca axis. Areas A2 , A3 and A4 (gure 7) full this condition, but state u2 has the shortest length of the corresponding vector u2 uref and this is the optimal choice. Similarly, the optimal switching states for all other reference voltage sectors S2 , S3 , ..., S6 can be determined. Table 2 gives the complete logic for all sectors.

Sectors S1 S2 S3 S4 S5 S6

SI u0,7 u3 u4 u4 u5 u0,7

SII u0,7 u0,7 u4 u5 u5 u6

SIII u1 u0,7 u0,7 u5 u6 u6

SIV u1 u2 u0,7 u0,7 u6 u1

SV u2 u2 u3 u0,7 u0,7 u1

SV I u2 u3 u3 u4 u0,7 u0,7

Table 2: Stationary switching table The switching table for stationary behavior is derived for a movement of the current error ie as slowly as possible [6,7]. Due to the fast changes of current reference value iref , ie can be situated far outside of the hexagon region. In this case, it must be returned as fast as possible back into the hexagon. For the detection of dynamic processes, an additional larger hysteresis hexagon is placed around the existing one (gure 8). In dynamic processes the information about the sector of reference voltage uref is not needed. If vector ie touches one of the borders of the dynamic hexagon in any sector (SI , SII , ..., SV I ) by choosing an inverter voltage uk which directs straight opposite of the direction of current error ie , the speed of getting back ie towards the middle of hexagon will be maximum. For the derivation of the dynamic switching table one example would be discussed. Let current error ie hit the dynamic hysteresis band in sector SI . The state u4 directs straight to the opposite of sector SI and it should be applied in order to get back ie as fast as possible in the inner hysteresis area. There are similar explanations for all other sectors (SII , ..., SV I ), which gives a very simple dynamic table (table 3).

b SIII SII

Dynamic hysteresis hexagon Stationary hysteresis hexagon

SIV SI

d d +h

SV

SVI

Figure 8: Stationary and dynamic hysteresis hexagon

Sector Voltage

SI u4

SII u5

SIII u6

SIV u1

SV u2

SV I u3

Table 3: Dynamic switching table Simulation Results for Hexagon Hysteresis Control The VSI is simulated in MATLAB using PLECS. The simulation result for the explained hexagon hysteresis control is given in gure 9 (steady state). From that gure it can be seen that the vector current error stays within the hexagon area. If we apply step change in reference current that we have that the current error goes outside of the hexagon, because the current changing causes the change in the radius of the circle where the reference current moves on (gure 4), but the hexagon tolerance surface remains the same. The simulation result is presented in gure 10 and the step change can be seen in gure 11.
Hexagon Current Control (steady state) 2

1.5

1 beta component of current error

0.5

0.5

1.5

2 2

1.5

0.5 0 0.5 alpha component of current error

1.5

Figure 9: The current error movement in , plane (steady state)

Hexagon Current Control (with Iref step change) 2

1.5

1 beta component of current error

0.5

0.5

1.5

2 2

1.5

0.5 0 0.5 alpha component of current error

1.5

Figure 10: The current error movement in , plane (with step change)
Threephase current (reference value with step change after 0.001 sec) 4

2 current

0.002

0.004

0.006

0.008

0.01

0.012

0.014

0.016

0.018

0.02

Threephase current (measured value) 3 2 1 current 0 1 2 3 0 0.002 0.004 0.006 0.008 0.01 time (sec) 0.012 0.014 0.016 0.018 0.02

Figure 11: Three-phase VSI current with step change in reference current after 0.001 sec (hexagon control) Square Hysteresis Based Control The employed current control method based on square hysteresis band working in (, ) plane is show in gure 12.
ia ref da
Hysteresis comprator Switching table Hysteresis comprator

Sa Sb Sc
VSI-PWM inverter Load

ib ref

db

ia ib

Figure 12: Block diagram of the used method for current vector control From equation (3) it can be seen that the current vector moves in direction of the voltage across the load inductance, which is the dierence between inverter voltage uk and inner voltage of the load e. 10

In this method we have only two tolerance bands (for and current components). Therefore, the hysteresis surface is a tolerance square for the current error which is shown in gure 13.
b

ie

iref

Figure 13: Square hysteresis area Whenever the current vector touches the border of the surface, another voltage state is applied to force it back within the square. Similarly, as in the case of the hexagon hysteresis control method, here the square tolerance band moves together with the reference current such that the current vector points always in the center of the square. For this purpose two hysteresis comparators for the and components are employed. A simple consideration makes it possible to control the current without any information about the load inner voltage. If the current reaches, for example, the right border of the tolerance square, then another voltage state has to be applied which has the smaller component then the actual state. In this case, regardless of the position of the load inner voltage, the component of the voltage across the load inductance and therefore the current deviation in direction of can be reversed. The complex (, ) plane can be divided into dierent sectors as dened by the dotted lines in gure 2.
da 3
h/2

h/2 h/2

db

0 d

iea

0 d

ieb

Figure 14: Multilevel hysteresis comparators for and components In -axis it is possible to apply four dierent voltage levels of uk ( 2 3 UDC , 11

1 -axis there are three voltage levels of uk ( , 3 of the appropriate voltage vector uk is determined by structure of the and hysteresis comparators and a corresponding switching table (table 4). Hysteresis comparators are depicted in gure 14, where because of the simplicity hysteresis levels are denotes as 0, 1, 2 and 3. 2 1 1 For comparator, level 0 corresponds to level 3 UDC , 1 to 3 UDC , 2 to 3 UDC 1 2 and 3 to 3 UDC . For comparator level 0 corresponds to level 3 , level 1 to 0 1 and level 2 to . 3 The control scheme of this method uses one four level hysteresis comparator for the component and three level hysteresis comparator for the component of the current vector error. Digital outputs of the comparators (d , d ) select the state of the inverter switches Sa , Sb , Sc using the switching table: 1 2 1 3 UDC , 3 UDC and 3 UDC ). In 1 ). The exact selection 0 and 3

, levels 0 1 2

0 u5 u4 u3

1 u5 u0,7 u3

2 u6 u0,7 u2

3 u6 u1 u2

Table 4: Switch logic table for square hysteresis control The practical implementation of three-level hysteresis comparator is given in gure 15. The implementation of four-level comparator is very similar.

d
0

h/2
1

d +
0

d
1

Figure 15: Practical implementation of multilevel hysteresis comparator

12

Simulation Results for Square Hysteresis Control The VSI is simulated in MATLAB using PLECS. The simulation result for the explained square hysteresis control is given in gure 16. It can be seen that the current error vector is in the square area (steady state). The simulation result for the step change in reference current is given in gure 17. It can be seen that the current error vector goes outside of the square (similarly as for hexagon control) due to the step change in reference current because the current changing causes the change in the radius of the circle where the reference current moves on (gure 13) but the square tolerance surface remains the same.
Square Current Control (steady state) 2

1.5

1 beta component of current error

0.5

0.5

1.5

2 2

1.5

0.5 0 0.5 alpha component of current error

1.5

Figure 16: The current error movement in , plane

Square Current Control (with Iref step change) 2

1.5

1 beta component of current error

0.5

0.5

1.5

2 2

1.5

0.5 0 0.5 alpha component of current error

1.5

Figure 17: The current error movement in , plane (with step change)

13

Comments Current regulator techniques based on the hysteresis control together with switch logic are presented. The hysteresis hexagon control requires knowledge of the parameters of the load, while the square method does not require that. The simulation are done for the following data: E =150V, L=1mH, Udc =400V, Iref =0A, (after t =5ms, Iref =2A). The simulation time is 10ms, = 0.8A, h =0.4A for both control techniques. Averaged switching frequency is higher for square method than for hexagon method. The switching frequencies are dierent for dierent phase switches (even for one type of control either for square or for hexagon control). With given parameters of the circuit, for square control, the averaged switching frequency (frequency during the simulation time) for Sa is 107kHz, for Sb is 107.5kHz and for Sc is 100.7kHz and for hexagon control, the averaged switching frequencies are, for Sa 94.7kHz, for Sb 94.3kHz and for Sc 81.4kHz. The hexagon method has smaller switching frequencies because this method is based on the rule to get back current error towards the middle of the hexagon area as slowly as possible, which is not the case for square method. Also, the switching frequencies are dierent for dierent phases because when one switch changes the state (either from 0 to 1 or from 1 to 0) that does not mean that other switches are changing their states, too (for example, if state u1 (100) is applied after state u2 (110) then only switch Sb has to change state from 0 to 1).

14

References 1. S. Buso, L. Malesani, P. Mattavelli, Comparison of Current Control Techniques for Active Filter Applications, IEEE Transaction on Industrial Electronics, Vol. 45, No.5, October 1998., pp.722-729 2. M.P. Kazmierkowski, L. Malesani, Current control techniques for threephase voltage-source PWM converters: a survey, Industrial Electronics, IEEE Transactions on Industrial Electronics, Vol. 45, No. 5, Oct.1998., pp. 691 -703 3. B.K. Bose, An Adaptive Hysteresis-Band Current Control Technique of a Voltage-Fed PWM Inverter for Machine Drive System,IEEE Transactions on Industrial Electronics, Vol. 37, No.5, October 1990., pp. 402-408 4. L. Malesani, P. Mattavelli, Novel Hysteresis Control Method for CurrentControlled Voltage-Source PWM Inverters with Constant Modulation Frequency, IEEE Transaction on Industry Applications, Vol. 26, No. 1, January/February 1990., pp. 88-92 5. A. Ackva, H. Reinold, R. Olesinski, A Simple and Self-Adapting HighPerformance Current Scheme for Three Phase Voltage Source Inverter PECS Toledo 1992. 6. F. Jenni, D. Wuest, Steuerverfahren f ur selbstgef uhrte Stromrichter 7. P. Eichenberger, M. Junger, Predictive Vector Control of the Stator Voltages for an Induction Machine Drive with Current Source Inverter, Power Electronics Specialists Conference, 1997. PESC 97 Record., 28th Annual IEEE, Vol. 2, 22-27 June 1997., pp. 1295-1301

15