Dual Voltage Supply Converter for High-Speed Doubly ... - IEEE Xplore

3 downloads 0 Views 1MB Size Report
Abstract—Small variable speed drives intended for mass-market applications like hand-held power tools and household appliances have to be inexpensive and ...
1016

IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 28, NO. 2, FEBRUARY 2013

Dual Voltage Supply Converter for High-Speed Doubly Salient Reluctance Motors Mohammed A. Elgendy, Volker Pickert, Member, IEEE, Bashar Zahawi, Senior Member, IEEE, Christopher Morton, and Afida Ayob

Abstract—Small variable speed drives intended for mass-market applications like hand-held power tools and household appliances have to be inexpensive and efficient. Doubly salient reluctance motors are known to be cheap and robust, but their use in such mass market applications has been hindered by the cost and limitations of their associated power converters. A new converter concept is proposed in this paper. The new converter has the ability to apply higher voltages across the machine windings than could be obtained with standard circuits to achieve high-speed operation. The demagnetization of the machine coils is load independent ensuring fast current decay and consequently a stable high-speed operation at higher loads. The fast rate of current decay allows a longer duration of positive torque pulses resulting in less pulsating torque. A detailed analysis of the proposed converter is presented and design guidelines are laid out. A 1-kW prototype converter is designed and constructed to experimentally validate the operation of the new circuit. Index Terms—Demagnetization, doubly salient reluctance motors (DSRMs), motor drives.

I. INTRODUCTION OUBLY salient reluctance motors (DSRMs) such as switched reluctance machines and stepper motors have a simple mechanical construction with windings in the stator only. They eliminate the need for rotor windings, brushes, commutators, and any type of rotor bars or permanent magnets. This reduces the total cost, losses, and inertia of these machines and increases their robustness, making them the most suitable for high-speed applications. Single-phase DSRMs have the simplest and cheapest structure. However, they produce a discontinuous torque and are therefore only suitable for applications insensitive to torque pulsations. A smoother torque can be produced by multiphase machines at the expense of higher drive costs [1], [2]. Currently, there are many applications in which the merits of these machines warrant their use. Examples of such applications are the

D

Manuscript received March 8, 2012; accepted June 12, 2012. Date of current version September 27, 2012. Recommended for publication by Associate Editor J. Hur. M. A. Elgendy, V. Pickert, B. Zahawi, and C. Morton are with the School of Electrical and Electronic Engineering, Newcastle University, Tyne and Wear, NE1 7RU, U.K. (e-mail: [email protected]; [email protected]; [email protected]; [email protected]). A. Ayob is with the Department of Electric, Electronic and Systems Engineering, Universiti Kebangsaan Malaysia, 43600 Bangi, Malaysia (e-mail: [email protected]). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TPEL.2012.2205164

textile processing industry and the aerospace industry [1], [4]. However, the high cost of the associated power electronic drive, which outweighs the cost of the machine itself, has been a barrier to the widespread use of DSRMs in domestic applications [5]. Torque production in these types of machines depends mainly on the variation of the magnetic reluctance (or inductance) with rotor position. The inductance varies from a minimum to a maximum as the rotor moves from an unaligned position (where the stator poles of one phase are exactly aligned with the interpolar axis of the rotor) to an aligned position (where the poles of that phase are aligned with a pair of rotor poles) and vice versa. The shape of the inductance waveform depends on machine design, e.g., the polar and interpolar arcs of the stator and rotor. Assuming linear magnetization characteristics, the instantaneous torque at rotor position (θ) is given by dL(θ) 1 i(θ)2 . (1) 2 dθ A positive torque can be produced only when the inductance is increasing. Ideally, the stator current has a square wave shape with a value of zero when the inductance is decreasing. With higher inductance values, the stator current rise (magnetization) and fall (demagnetization) takes longer. This may not be a problem at low speeds where the duration of the inductance increase/decrease period is long. However, for high-speed operation, a fast rate of rise of current must be achieved to allow the current to reach its reference value. A fast rate of fall of current must also be ensured to avoid (or at least reduce) any negative torque pulses resulting from the existence of current after the inductance had started to decrease. Fast rates of change of current can be achieved if the motor is designed with lower inductance. However, this reduces the torque production capability of the machine as illustrated by (1). Applying a high positive voltage across the machine winding during magnetization and a high negative voltage during demagnetization will guarantee fast current rates of rise and fall. Consequently, high-speed operation can be achieved without the need for a lower inductance value. Unfortunately, conventional converters (cf., Fig. 1) provide a limited range of output voltage magnitude. In such converters, the dc-link voltage is produced from a full bridge rectifier fed from the mains supply. Therefore, its magnitude is limited by the maximum value of the ac supply voltage. In an effort to boost the voltage applied across the machine winding, the capacitor boost converter has received a lot of attention in the past [6]–[15]. In these topologies, the energy from the outgoing phase is stored in an additional capacitor providing a boost to the voltage across the motor windings. The

0885-8993/$31.00 © 2012 IEEE

T (θ) =

ELGENDY et al.: DUAL VOLTAGE SUPPLY CONVERTER FOR HIGH-SPEED DOUBLY SALIENT RELUCTANCE MOTORS

Fig. 1.

1017

Conventional single phase DSRM drive converter configuration.

additional auxiliary capacitor can be placed in parallel [11] or in series [12], [13] with the dc-link capacitor. These capacitorboost topologies increase the voltage applied to the machine winding by a value less than the maximum supply voltage. For most of these configurations [7], [10]–[12], the maximum voltage is applied across the machine winding during the magnetization period only. The demagnetization process is load dependent, limiting the motor speed at higher loads. This paper proposes a new converter concept for DSRM drives operating at high-machine voltages. The high voltage across the machine windings is achieved by introducing a simple dual voltage supply to generate high voltages across the machine windings. The need for a high dc-link capacitance value is thus eliminated. For multiphase DSRMs with concentrated windings, mutual coupling between electrical phases can be neglected. This allows the use of the proposed converter with multiphase machines by increasing the number of output stages to match the number of phases. In the following sections, the behavior of the converter is analyzed and simulated. A 1-kW prototype is constructed and tested with a stationary inductive load to experimentally verify the operation of the circuit. Practical laboratory results show very good agreement with those obtained from simulation. II. CONVENTIONAL DSRM DRIVE A single phase DSRM with a pair of stator and rotor poles is considered in this discussion. The inductance is assumed to vary in a sinusoidal pattern. Typical minimum and maximum inductance values of 11.7 and 30 mH and a stator resistance value of 2.18 Ω are used in agreement with those of a practical machine design [11]. Since the focus of this paper is on converter design, machine operation is assumed to be linear for ease of modeling. The instantaneous torque as a function of rotor position is then given by (1). The converter must supply current to the stator winding during the inductance increase periods. This current must be forced to be zero during inductance decrease periods. The most common power converter topology to drive a singlephase machine comprises a single-phase diode bridge rectifier, a dc-link capacitor and an asymmetric bridge inverter as shown in Fig. 1. The machine is represented by its internal resistance rint , its inductance L(θ) as a function of the rotor angle, and the back electromotive force. With this configuration, magnetization is achieved by switching S1 and S2 at the turn-on angle θ0 , connecting the dc-link voltage across the machine winding (see Fig. 2). For low-speed operation, a turn-on angle of 90◦ can be used, assuming that the rotor angle is 90◦ at the

Fig. 2.

Winding current waveform.

Fig. 3.

Control structure for DSRM drive.

unaligned position. When the current reaches the upper limit Iupp er of a hysteresis window predefined with a given reference current Iref , S1 and S2 are switched OFF and the current freewheels through D1 and D2 . In this case, a negative dc-link voltage appears across the machine winding forcing the current to decrease. When the current falls to the lower hysteresis limit Ilower , S1 and S2 are switched ON again to maintain the current within the hysteresis window until the rotor reaches the commutation angle θC at the aligned position (θ = 180◦ ). S1 and S2 are then switched OFF applying a negative voltage across the winding until the current decays to zero. The rotor continues to move through its inertia until it reaches the unaligned position again and the sequence is repeated. The reference current of the hysteresis control is derived from a closed-loop speed controller as shown in Fig. 3 [16]. During demagnetization, the energy is supplied back to the dc-link capacitor. For this reason, the value of the dc-link capacitance must be chosen high enough to avoid a substantial voltage increase across the capacitor terminals. A 2-mF capacitance is used in this investigation. The maximum dc-link voltage of the conventional converter is indicated by the maximum voltage of the ac mains supply (about 340 V for a 240-V supply voltage). At this voltage level, the fast magnetization and demagnetization required for highspeed operation may not be achieved. The slow magnetization prevents the current from reaching its reference value during the entire period of inductance increase and consequently results in a high steady-state speed error. For example, at a reference speed of 15 000 r/m with 0.2 N·m load torque, 90◦ turn-on angle, and 180◦ commutation angle, the actual machine speed is limited to about 6900 r/m; less than half the reference value. The slow demagnetization slows down the current decay resulting in negative torque pulses as shown in Fig. 4. This reduces the average torque and causes oscillations and noise. The negative torque pulses can be reduced by advancing the commutation angle so that the demagnetization starts before the

1018

IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 28, NO. 2, FEBRUARY 2013

Fig. 6.

Fig. 4. Drive waveforms for a single phase DSRM driven by conventional converter (θ 0 = 90◦ and θ C = 180◦ ); high-speed operation.

Half-wave voltage doubler circuit.

[18]–[20]. Voltage doubler circuits consist primarily of diodes and capacitors. Fig. 6 shows a half-wave voltage doubler circuit. Assuming no losses, the operation of the circuit at no load can be briefly described as follows. At t = 0, all capacitors are uncharged. During the negative half cycle of the supply voltage VS , capacitor C1 is charged to a voltage VC 1 given by VC 1 = VS,p eak

(2)

where VS ,p eak is the peak supply voltage. Similarly, during the positive half cycle, capacitor C2 is charged to a voltage VC 2 given by VC 2 = VC 1 + VS,p eak . Fig. 5.

Circuit diagram of the proposed dual voltage converter.

aligned position (θC < 180◦ ). The turn-on angle can also be slightly advanced as the change in inductance with respect to rotor position is low in the vicinity of the unaligned position. As the reference speed increases, however, larger advance angles are required and a larger steady-state speed error may be expected. Also, with a larger advance in commutation angle, a smaller dwell angle (θC − θ0 ) is obtained resulting in more torque pulsations. Current rise and decay rates can be accelerated by applying higher voltage levels across the machine winding. This allows high-speed operation at lower advance angles and consequently smoother torque as described below. III. OPERATING PRINCIPLES OF THE PROPOSED CONVERTER Fig. 5 shows a circuit diagram of the proposed dual voltage converter. The input stage of the converter comprises the ac mains feeding four capacitors as temporary storage elements. The output stage is comprised of one switch S and one diode D. Connected in parallel with the output stage is a charge transfer circuit (shown in red) which is used to transfer the excess charge from C4 to C2 keeping the voltage level on C4 nearly constant. A. Input Stage The input stage of the converter is based on the concept of voltage doubler circuits. Voltage doubler circuits were first introduced by Cockroft and Walton in 1932 [17]. They provide one of the most effective means of generating high voltages at relatively low currents and are very popular in applications such as laser systems, particle accelerators, and X-ray machines

(3)

Substituting for VC 1 , it is clear that the output voltage VC 2 is now double the peak supply voltage VC 2 = 2VS,p eak .

(4)

The input stage of the proposed converter (see Fig. 5) employs two half-wave voltage doubler circuits connected back to back. With this arrangement, the single stage voltage doubler circuit connected to the top rail produces +2VS ,p eak across the machine windings and the single stage voltage doubler circuit connected to the bottom rail produces −2VS ,p eak across the machine windings. When considering ideal circuit components, the capacitor voltages will satisfy (3) and (4) after the first half cycle of charging. However, in real circuits with source resistance and capacitor effective series resistance, several cycles are required to charge the capacitors depending on the capacitance values and the summation of the above resistances (i.e., depending on the charging time constant). The four capacitors must be allowed to charge fully before the inverter starts to operate. The voltages VC 2 and VC 4 are therefore monitored and once both VC 2 and VC 4 have reached a desired threshold value, an enable signal is sent to the controller to commence switching of the inverter stage. B. Output Stage The inverter is controlled by a current hysteresis controller as described earlier for the conventional converter. Assuming that C2 can be considered as a stiff dc source, inverter operation can be described as follows. Mode 1: Magnetization Fig. 7 shows the equivalent circuit of the system during magnetization. S is turned ON causing C2 to discharge and magnetizing the windings. Diode D is reverse biased as long as S1 is ON. During this period, the current through the load increases

ELGENDY et al.: DUAL VOLTAGE SUPPLY CONVERTER FOR HIGH-SPEED DOUBLY SALIENT RELUCTANCE MOTORS

Fig. 7.

1019

Effective circuit during magnetization and current rise periods.

from zero toward a maximum value Im ax , but is limited to Iupp er by the hysteresis controller (switching S OFF). The current through the load windings during magnetization can be derived as i(t) = Im ax 1 (1 − e−t/τ ) where

 Im ax 1 = VC 2

and

rint +

 τ = L (θ)

rint +

dL (θ) ω dθ

(5)

Fig. 8.

Magnetization times at different machine speeds.

Fig. 9.

Effective circuit during demagnetization and current decay periods.

 (6)

 dL (θ) ω . dθ

(7)

Magnetization starts when the rotor is at the unaligned position where the machine has minimum inductance. In this case, approximate values for Im ax1 and τ can be calculated by considering the minimum inductance and the average values of dL(θ)/dθ and VC 2 . The magnetization time Tm ag can then be calculated from (4) by substituting i(t) with Iupp er as   Im ax 1 Tm ag = τ ln . (8) (Im ax 1 − Iupp er ) It is clear from the above equation that the magnetization time depends mainly on the machine inductance, speed, and terminal voltage VC 2 in accordance with (6) and (7). The above equations are also applicable to the conventional converter when replacing VC 2 by the dc-link voltage. At high speeds, the required magnetization time increases significantly as shown in Fig. 8. Shorter magnetization time can be obtained with low-inductance values. However, decreasing the machine inductance deteriorates the output torque as illustrated by (1). Alternatively, the magnetization time can be shortened by feeding the machine with a higher voltage. For example, at a speed of 15 000 r/m and 15 A reference current, the magnetization time is expected to decrease to about one-third of its value when using the proposed converter where the applied voltage is double that of the conventional converter (see Fig. 8). Mode 2: Hysteresis Period In all other subsequent periods during the hysteresis mode of operation, the current increases from an initial value of Ilower . The rising current during the hysteresis period can therefore be described by 

i(t ) = Im ax 1 + (Ilower − Im ax 1 )e−t /τ .

(9)

where t represents the new time reference frame and Im ax1 and τ are given by (6) and (7), respectively. The rising current during the hysteresis period is limited by the maximum threshold value of the hysteresis band Iupp er .

Once the current reaches Iupp er , switch S is switched OFF and diode D starts to conduct, acting as freewheeling diode. With diode D conducting a negative voltage across the machine winding is imposed. The current decays exponentially until it reaches the lower hysteresis limit Ilower . The effective circuit during this part of the hysteresis period is shown in Fig. 9. The decaying current in this period is given by 

i(t ) = Im ax 2 + (Iupp er − Im ax 2 )e−t /τ where

 Im ax 2 = VC 4

rint

 dL (θ) ω . + dθ

(10)

(11)

Once the current has reached Ilower , switch S is switched ON again. With switch S conducting, diode D becomes reversed biased and stops conduction. During the hysteresis period, the machine inductance (and consequently the time constant τ ) increases with time. The time constant τ reaches a maximum value when the rotor reaches the aligned position, i.e., at the position of maximum inductance. Mode 3: Demagnetization The windings are demagnetized when S is switched OFF and D starts to conduct giving the effective circuit shown in Fig. 9. During this period, the current decreases from Iupp er to zero and the energy stored in the machine winding is injected into C4 . The current during the demagnetization period can be expressed by (10). The demagnetization starts at the maximum value of machine inductance, i.e., at the maximum value of τ . Consequently, demagnetization takes longer than magnetization. The demagnetization time Tdem ag can be calculated from (10) by substituting i(t) with 0 as   (Im ax 2 − Iupp er ) Tdem ag = τ ln . (12) Im ax 2

1020

Fig. 10.

IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 28, NO. 2, FEBRUARY 2013

Fig. 11.

Charge transfer circuit.

Fig. 12.

Voltage waveforms across capacitors C 1 , C 3 , C 2 , and C 4 .

Demagnetization times at different machine speeds.

Once again, the demagnetization time depends on the machine inductance, speed, and the applied voltage VC 4 . The above equations are applicable to the conventional converter when replacing VC 4 by the negative value of the dc-link voltage. At high speed, the required demagnetization time increases significantly as shown in Fig. 10. The demagnetization time can be shortened by supplying a higher negative voltage. For example, at a speed of 15 000 r/m and 15-A reference current, the demagnetization time is expected to decrease to about one-third of its value when using the proposed converter. C. Charge Transfer Circuit During the magnetization and current rise periods, capacitor C2 discharges and VC 2 decreases by an amount dependent on the drawn current and capacitance value. The voltage across the capacitor C4 is constant for the duration of the magnetization and current rise intervals. During demagnetization and current decay periods, the freewheeling diode conducts and the resulting machine inductance current charges the capacitor C4 , increasing its voltage VC 4 . To solve this problem, a charge transfer circuit consisting of an inductor LCT with internal resistance rCT , a switch SCT and a diode DCT (see Fig. 11) is added in parallel to the output stage of the converter. This circuit operates independently to transfer the excess charge from C4 into C2 to compensate for the effects of machine loading. When the voltage across C4 exceeds an upper limit VC 4m ax , a 50% duty cycle control signal is enabled to switch SCT ON and OFF at 10 kHz. When SCT is ON, the current id rises toward a maximum value equal to VC 4 /rCT . After 50 μs (50% duty cycle), the switch is turned OFF and the inductor current freewheels through DCT charging C2 . The maximum current through the inductor is a function of the time constant of the discharge circuit and the duration of the on-period of the switch. This current can be limited by using a higher inductance value or a higher switching frequency for SCT . The switching continues until the voltage across C4 falls below a lower limit. IV. CONVERTER OPERATING CHARACTERISTICS To investigate the operating characteristics of the new circuit topology and verify circuit design parameters, simulations were carried out using MATLAB Simulink. For these simulations, all

components were assumed to be ideal except where specified. A stationary 11 Ω–17 mH inductive load is used in the investigation, the approximate average of the inductance values used in Section II. Experimental verification of the results using a 1-kW prototype converter is presented in the next section. The rotor position signal was emulated by 50 and 500 Hz clock signals with 50% duty cycles corresponding to 1500 and 15 000 r/m, respectively. In this investigation, 160 μF capacitors were used for C1 , C2 , C3 , and C4 [21]. Fig. 12 shows the voltages across these capacitors. Capacitor C1 is charged during the negative half cycles and C3 is charged during the positive half cycles of the supply. The voltages across C1 and C3 reach maximum values of about 340 V (VS ,p eak ) and the voltages across C2 and C4 reach values of about 680 V (2VS ,p eak ) after about 130 ms. The converter starts to operate after about 200 ms, when the capacitors C1 –C4 are fully charged. To better explain the operation of the circuit let’s assume that SCT is switched OFF, i.e., the charge transfer circuit is not allowed to work. The resulting waveforms of the voltage across C2 and C4 and of the load current when operating at 50 Hz are shown in Fig. 13. The charging rate of the storage capacitor C2 is fixed by the ac supply frequency while the discharge rate is a function of load current. During the magnetization and load current rise periods, C2 has to supply energy to the load, reducing the average value of VC 2 while increasing its ripple content. The load current dwell angle is synchronized with the positive half cycle of the supply voltage. Fig. 14 shows an expanded view of the load current over one electric cycle (equivalent to half a revolution) together with capacitor currents iC 2 and iC 4 . C2 charges (iC 2 is positive) in

ELGENDY et al.: DUAL VOLTAGE SUPPLY CONVERTER FOR HIGH-SPEED DOUBLY SALIENT RELUCTANCE MOTORS

Fig. 13. Voltage across capacitors C 2 and C 4 and load current; 50-Hz current pulses, charge transfer circuit not in operation.

Fig. 14. Load current, iC 2 and iC 4 during one electric cycle; 50-Hz current pulses, charge transfer circuit not in operation.

the first quarter cycle when the supply voltage is rising, and discharges when the supply voltage is falling. Both the drop in VC 2 and its ripple content are proportional to the average load current and inversely proportional to the value of C2 [22]. In order to reduce the voltage drop and the voltage ripple content across C2 , its capacitance must be increased. However, this increases the price of the capacitor and the time required for it to be fully charged. For the voltage waveforms shown in Fig. 13, the voltage across C2 falls to an average value of about 350 V with a peak-to-peak voltage ripple of about 170 V at an average load current of 2.4 A. Although the value of the load current changes with the hysteresis controller requirements, the direction of current through the load is maintained. That means that C2 provides the energy for magnetizing the windings and C4 absorbs the energy during the demagnetizing period when the freewheeling diode D conducts. Fig. 14 shows that during demagnetization and load current fall periods, the inductor current flows into C4 charging the capacitor. When the charge transfer circuit is not working (SCT is OFF), there is no discharge possibility for C4 (iC 4 is always positive). Its terminal voltage continues to increase as shown in Fig. 13, leading to the eventual failure of the circuit. The charge transfer circuit is needed to overcome this difficulty.

1021

Fig. 15. Voltage across capacitors C 2 and C 4 and load current; 50-Hz current pulses with the operation of the charge transfer circuit.

The charge transfer circuit transfers the excess charge from C4 to C2 . This not only solves the problem of the continuously increasing voltage across C4 keeping it constant, but also reduces the voltage drop and voltage ripple content across C2 . Furthermore, a constant voltage across C4 means that the demagnetization of the machine coil is load independent ensuring stable high-speed operation at higher loads. Fig. 15 shows the resulting circuit waveforms for operation at 1500 r/m. The voltage drop across C2 is caused by the resistive losses in the load inductor. The voltages across C4 is constant (with a small ripple magnitude) at about −680 V. This ensures fast demagnetization rates and consequently allows high-speed operation. The switch SCT starts to operate with a switching frequency of 10 kHz and 50% duty cycle when the voltage across C4 exceeds 685 V (see Fig. 16). Due to the high-inductance value of LCT (4 mH), the inductor current takes time to build up allowing VC 4 to marginally exceed its upper limit reaching 686.5 V. The charge transfer circuit continues to operate until VC 4 falls below 675 V when SCT is switch OFF and the sequence is repeated. More precise control of VC 4 can be achieved when using lower inductance value for LCT . However, this increases the current through the inductor and requires higher current rating devices (SCT and DCT ). Lower inductance values are possible with lower current if current hysteresis control is added for SCT but this increases circuit complexity. The high-frequency ripple of the discharge current should be considered in the choice of the capacitors C2 and C4 . This ripple can be minimized by using higher switching frequency for SCT but this will increase switching losses. The design process is inevitably a compromise between the aforementioned factors when choosing appropriate values for circuit components. When emulating high-speed operation by reducing the period of the current pulses (see Fig. 17), smoother voltage waveforms are obtained for VC 2 and VC 4 due to the shorter operating and nonoperating periods of the switch S. The voltage across C4 oscillates between 675 and 686.5 V as shown in Fig. 18. V. EXPERIMENTAL VERIFICATION A 1-kW, 240-V laboratory prototype converter was constructed and tested to verify the aforementioned analysis. The

1022

Fig. 16.

IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 28, NO. 2, FEBRUARY 2013

Operation of the charge transfer circuit with 50-Hz current pulses. Fig. 19. Experimental results showing voltage across capacitors C 2 and C 4 and load current; 50-Hz current pulses with the operation of the charge transfer circuit.

Fig. 17. Voltage across capacitors C 2 and C 4 and load current; 500-Hz current pulses with the operation of the charge transfer circuit.

Fig. 18.

Operation of the charge transfer circuit with 500-Hz current pulses.

340-V peak ac voltage allows peak voltages on C2 and C4 of 680 and −680 V, respectively. The converter employs rectifier diodes for D1 –D4 , soft recovery diodes for the freewheeling diodes D and DCT , an insulated gate bipolar transistor switch for the switch SCT and a power MOSFET for the switch S. The four capacitors C1 –C4 are chosen with a capacitance value of 160 μF. A 17-mH inductive load was used to represent the machine inductance while a 4-mH inductor is used in the charge transfer circuit for LCT . Currents and voltages were measured using a digital oscilloscope with current probes and high-voltage differ-

ential probes. The feedback current and voltage were measured with Hall Effect sensors: LA55-P and LV25-P, respectively. The control procedure was implemented employing a Texas Instruments TMS320F2812 DSP-based eZdsp kit. This DSP-based control hardware was used for experimental flexibility and ease of programming. In a commercial product, a lower cost microcontroller would be more than adequate to implement the control algorithms under investigation. A 4-kHz cutoff frequency firstorder low-pass filters were used for noise rejection from the current and voltage feedback signal. The output stage of the converter starts to operate after the capacitors are fully charged. The switch S is controlled by hysteresis current control. The upper and lower current limits were programmed as 5.5 and 6.5 A, respectively. The charge transfer circuit starts to operate when the magnitude of the negative voltage on the capacitor C4 exceeds a higher limit of 685 V. The switching frequency of the charge transfer switch SCT is fixed at 10 kHz. The operation of this switch is disabled when the negative voltage on the capacitor C4 falls below 675 V. Fig. 19 shows the waveforms of the load current and the measured voltages across capacitors C2 and C4 when operating with 50-Hz current pulses. The steady-state load current does not follow the hysteresis threshold band closely. In fact, the measured current varies within a wider range of values. This could be a result of the use of the low-pass filter for the load current feedback signal. The filter causes a delay so that the controller changes the switching state of the converter switch S after the actual load current had reached the predefined limit. During this time, the load current continues to increase/decrease overshooting the predefined limits. For the measured data in Fig. 19, the switching frequency of the switch S is about 1.8 kHz. Higher switching frequency can be expected if a lower filter delay is ensured (i.e., by using a filter with a lower cutoff frequency) or if a narrower hysteresis band is used. With digital control, the switching signal usually changes at a rising or falling edge of a clock signal regardless of the instant at which the signal has hit the threshold value. This may cause a slight variation in the current level at which the switching

ELGENDY et al.: DUAL VOLTAGE SUPPLY CONVERTER FOR HIGH-SPEED DOUBLY SALIENT RELUCTANCE MOTORS

Fig. 20. Experimental results showing the operation of the charge transfer circuit with 50-Hz current pulses.

1023

Fig. 22. Experimental results showing the operation of the charge transfer circuit with 500-Hz current pulses.

again about 2 kHz as the load current varies between the same levels during the hysteresis period. The charge transfer circuit operates in the same way at this higher frequency keeping VC 4 constant at about −680 V, as demonstrated in Fig. 22.

VI. CONCLUSION

Fig. 21. Experimental results showing voltage across capacitors C 2 and C 4 and load current; 500-Hz current pulses with the operation of the charge transfer circuit.

state changes resulting in the unsmooth edges of the hysteresis window that appear in Fig. 19. The measured capacitor voltage VC 4 does not follow the predefined limits of −675 and −685 V due to the high inductance value of LCT (4 mH) and the delay introduced by the feedback filter (see Fig. 20). During the hysteresis period, the charge fed into the capacitor C4 increases its voltage up to about 692 V while the charge drawn from C2 reduces its voltage to about 550 V. The charge transfer circuit transfers charges from C4 to C2 so that the voltage of C4 drops to about 670 V and the voltage VC 2 increases up to 630 V before the next hysteresis period. However, this increases the current through the inductor necessitating higher current rating devices for SCT and DCT . Due to the constant inductance of the load and the nearly equal voltage magnitudes across C2 and C4 , demagnetization and magnetization take approximately the same time, as shown in Fig. 20. Figs. 21 and 22 show the experimental results when using 500 Hz current pulses. Smoother voltage waveforms are obtained for VC 2 and VC 4 , in agreement with simulation results. The load current waveform overshoots the hysteresis limits due to the delay caused by the low-pass filter used for the current feedback signal. The switching frequency of the switch S is

A new converter has been proposed for doubly salient reluctance motor drives with the ability to supply double the peak supply voltage to the machine winding. Higher voltages across the machine windings will increase the rate of change of current during the winding magnetizing and demagnetizing periods to enable the machine to work at higher speeds. A dual voltage supply has been integrated into the new converter topology so that the machine windings are supplied from two (positive and negative) voltage rails. The positive high voltage is used to provide the magnetizing current for the windings. The negative voltage is applied across the winding terminals to achieve defluxing. The positive and the negative voltage waveforms produced by the proposed converter circuit are constant and their ripple content is small. This has been achieved by optimizing the design of the converter for high-speed operation. The constant negative voltage applied to the machine windings has the advantageous effect of offering better machine control due to load-independent demagnetization. The change between the positive and the negative high-voltage rails is controlled by only one switch allowing the adoption of a simpler, more reliable controller compared to boost inverter topologies. The proposed converter requires only two switches, the same number used in conventional and boost topologies. However, the new circuit is the only converter that is suitable for high-speed operation at all load conditions. The standard topology cannot provide the high voltages needed for high speeds while capacitor boost converters suffer from a drop in demagnetization voltage at high loads. The operating characteristics of the new circuit have been experimentally verified using 1-kW laboratory converter showing good agreement with simulation and analytical studies.

1024

IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 28, NO. 2, FEBRUARY 2013

REFERENCES [1] J. Kim, K. Ha, and R. Krishnan, “Single-controllable-switch-based switched reluctance motor drive for low cost, variable-speed applications,” IEEE Trans. Power Electron., vol. 27, no. 1, pp. 379–387, Jan. 2012. [2] H. J. Brauer, M. D. Hennen, and R. W. De Doncker, “Control for polyphase switched reluctance machines to minimize torque ripple and decrease ohmic machine losses,” IEEE Trans. Power Electron., vol. 27, no. 1, pp. 370–378, Jan. 2012. [3] P. Greg, “The rise of VSR motors,” in Mechanical Engineering, 1998, pp. 86–87. [4] V. P. Vujiˇci´c, “Minimization of torque ripple and copper losses in switched reluctance drive,” IEEE Trans. Power Electron., vol. 27, no. 1, pp. 388– 399, Jan. 2012. [5] N. Mohan, T. M. Undeland, and W. P. Robbins, Power Electronics: Converters, Applications, and Design, 3rd ed. New York: Wiley, 2003. [6] P. T. J. E. Miller, A. B. Plunkett, and R. L. Steigerwald, “Regenerative unipolar converter for switched reluctance motors using one main switching device per phase,” U.S. Patent US4684867, 1978. [7] P. K. Sood, Power Converter for Switched Reluctance Motor. St. Louis, MO: ESCD-Emerson Electric, 1992. [8] S. Vukosavic and V. R. Stefanovic, “SRM inverter topologies: A comparative evaluation,” IEEE Trans. Ind. Appl., vol. 27, no. 6, pp. 1034–1047, Nov./Dec. 1991. [9] P. D. Webster, “Control circuit and system for a switched reluctance machine and method of operating,” U.S. Patent 5 764 019, Jun. 9, 1998. [10] A. M. Hava, V. Blasko, and T. A. Lipo, “A modified C-dump converter for variable-reluctance machines,” IEEE Trans. Ind. Appl., vol. 28, no. 5, pp. 1017–1022, Sep./Oct. 1992. [11] S. Chan and H. R. Bolton, “Performance enhancement of single-phase switched-reluctance motor by DC link voltage boosting,” IEE Proc.Electric Power Appl., vol. 140, no. 5, pp. 316–322, 1993. [12] J. D. Lewis, H. R. Bolton, and N. W. Phillips, “Performance Enhancement of single and two phase SR drives using a capacitor boost circuit,” in Proc. Eur. Power Electron. Appl. Conf. Rec., 1995, pp. 229–232. [13] M. Barnes and C. Pollock, “Power converter for single phase switched reluctance motors,” Electron. Lett., vol. 31, no. 25, pp. 2137–2138, 1995. [14] J. Liang, D. Lee, G. Xu, and J. Ahn, “Analysis of passive boost power converter for three-phase SR drive,” IEEE Trans. Ind. Electron., vol. 57, no. 9, pp. 2961–2971, Sep. 2010. [15] M. C. Ford, Reluctance Motor. Essex, U.K.: Ford Motor Co. Ltd., 1972. [16] T. Miller, Electronic Control of Switched Reluctance Machines. Oxford, U.K.: Newnes, 2001. [17] J. D. Cockroft and E. T. S. Walton, “Experiments with high velocity positive ions. (I) Further developments in the method of obtaining high velocity positive ions,” Proc. Royal Soc. London. Ser. A, vol. 136, no. 830, pp. 619–630, 1932. [18] S. Iqbal, G. K. Singh, and R. Besar, “A dual-mode input voltage modulation control scheme for voltage multiplier based X-ray power supply,” IEEE Trans. Power Electron., vol. 23, no. 2, pp. 1003–1008, Mar. 2008. [19] F. Hwang, Y. Shen, and S. H. Jayaram, “Low-ripple compact high-voltage DC power supply,” IEEE Trans. Ind. Appl., vol. 42, no. 5, pp. 1139–1145, Sep./Oct. 2006. [20] D. W. Shute, “A hybrid cockroft walton multi-dynode photomultiplier supply for space applications,” IEEE Trans. Nucl. Sci., vol. 17, no. 1, pp. 130–137, Feb. 1970. [21] J. M. Beck, “Using rectifiers in voltage multiplier circuits,” vol. 2007, Vishay Semiconductors, Mumbai, India, Appl. Notes No. 88842, 2002. [22] W. Yan and F. P. Dawson, “DC ignition circuits for a high pressure vortexwater-wall argon arc lamp,” in Proc. 31st IEEE Industry Appl. Soc. Annu. Meet., 1996, vol. 4, pp. 2211–2218. Mohammed A. Elgendy received the B.Sc. degree from Menoufia University, Menoufia, Egypt, in 1997, the M.Sc. degree from Ain Shams University, Cairo, Egypt, in 2003, and the Ph.D. degree from Newcastle University, Tyne and Wear, U.K., in 2010, all in electrical engineering. From June 1998 to May 2006, he was a Research Assistant at the New and Renewable Energy Department, Desert Research Centre, Cairo. He is currently a Research Associate in the School of Electrical and Electronic Engineering, Newcastle University. His research focus is on design and control of power electronic converters for drives and renewable generation schemes.

Volker Pickert (M’03) received the Dipl.Ing. degree in electrical and electronic engineering from Rheinisch-Westfaelische Technische Hochschule Aachen, Aachen, Germany, in 1994, and the Ph.D. degree from Newcastle University, Tyne and Wear, U.K., in 1997. Following his Ph.D., he spent six years in industry starting as a Product Manager at Semikron GmbH and later as a Group Leader of the Electric Drives R&D Group, Volkswagen AG. In 2003, he was a Senior Lecturer in the Power Electronics, Drives and Machines Research Group, Newcastle University, where he became a Full Professor in power electronics in 2011. He has authored and coauthored more than 90 papers in leading international conferences and journals. He is the coauthor of one book. His research interests include power electronics for automotive applications, thermal management and energy management. Dr. Pickert is a member of the executive committee of the IET Power Generation, Conversion and Utilization professional network, and in 2010, he was the Chairman of the Biannual IET International Power Electronics, Machines and Drives Conference, Brighton, U.K. He has received the Denny Medal from the Institute of Marine Engineering, Science & Technology. Bashar Zahawi (M’96–SM’04) received the B.Sc. and Ph.D. degrees in electrical and electronic engineering from Newcastle University, Tyne and Wear, U.K., in 1983 and 1988, respectively. From 1988 to 1993, he was a Design Engineer with a U.K. manufacturer of large variable speed drives and other power conversion equipment. In 1994, he was a Lecturer in electrical engineering at the University of Manchester. In 2003, he joined the School of Electrical and Electronic Engineering, Newcastle University, as a Senior Lecturer. His research interests include small-scale generation, power conversion, and the application of nonlinear dynamical methods to electrical circuits and systems. Dr. Zahawi is a Chartered Electrical Engineer and the recipient of the Crompton Premium awarded by the Institution of Electrical Engineers and the Denny Medal awarded by the Institute of Marine Engineering, Science and Technology. Christopher Morton received the B.Eng. degree in electronic systems design engineering from Northumbria University, Tyne and Wear, U.K., in 2002. He joined Newcastle University as a Research Associate in 2008. He has ten years of work experience in industry. In 2002, he started with Scittexx Engineering Design working on electronic control units. At Scittexx, he was extensively involved in modeling automotive components. He then worked for AT Electronics on embedded controllers for power drives applications. In 2004, he was with the Research and Development Department, Corus Developing Electronic Systems. He moved on working for Turbo Power Systems. At Turbo Power Systems, he specialized in the design of aerospace power electronics Boeing 787 Dreamliner. Before leaving Turbo Power Systems, he conducted research in optimization strategies for the energy management of “more electric” aeroplanes. In 2008, he started as a Research Assistant within the Power Electronics, Drives and Machines Research Group, Newcastle University, in order to pursue a Ph.D. in electrical engineering. He is part of the TSB project S0029D, called Second Generation 7.5 t-12 t Diesel/Electric Hybrid Truck. His role is to design and develop an electric steering system and an electric breaking system for a hybrid electric truck. Afida Ayob received the B.Eng. degree in electrical and electronics engineering from the University of Manchester, Manchester, U.K., in 2000, and the Ph.D. degree in power electronic from Newcastle University, Tyne and Wear, U.K. She was awarded the Malaysian Government scholarship for the Ph.D. study. After graduation, she returned to Malaysia and is currently a Senior Lecturer in the Department of Electrical, Electronic and Systems Engineering, National University of Malaysia, Bangi, Malaysia. Her research focus is on the design and control of power electronic systems, electric drives, and electric vehicle systems.