Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 46, NO.

5, SEPTEMBER/OCTOBER 2010

1989

Sensorless Control of Induction Machines at Low and Zero Speed by Using PWM Harmonics for Rotor-Bar Slotting Detection
Reiko Raute, Member, IEEE, Cedric Caruana, Member, IEEE, Cyril Spiteri Staines, Member, IEEE, Joseph Cilia, Mark Sumner, Senior Member, IEEE, and Greg M. Asher, Fellow, IEEE

AbstractThis paper presents the use of the inherent highfrequency pulsewidth modulation (PWM) harmonics for sensorless control of ac machines. The amplitude and position of the PWM voltage harmonics cannot be controlled independently and are determined by the fundamental machine operation. However, they do form a high-frequency excitation and can provide information on saliencies within ac machines. This paper examines the feasibility of sensorless control based on extracting the rotor-bar slot position information for a cage induction machine using PWM harmonics. The position-signal demodulation and compensation schemes used are derived. Experimental results are provided for an off-the-shelf induction motor operating under sensorless current, speed, and position control, including zero excitation and zero speed. Index TermsInduction machine, pulsewidth modulation (PWM), saliency, sensorless control, vector control.

I. I NTRODUCTION OR MANY years, the spatial impedance anisotropy in ac machines has been considered for the determination of the mechanical rotor or ux position. Published methods utilized the machine response to injected test signals, including highfrequency [1][10] and voltage-test-vector [11][13] injection. Even though these methods were successful in implementing true zero-speed sensorless control for ac-machine drives, the use of additional test signals comes with disadvantages, such as additional losses and acoustic noise, due to the test signal creating additional disturbances in the machine currents. Recent approaches [14][16] modied the pulsewidth modulation (PWM) scheme to embed the voltage test vectors into the fundamental PWM cycles and avoid additional
Manuscript received October 2, 2009; revised December 17, 2009; accepted January 26, 2010. Date of publication July 12, 2010; date of current version September 17, 2010. Paper 2009-IDC-327.R1, presented at the 2008 Industry Applications Society Annual Meeting, Edmonton, AB, Canada, October 59, and approved for publication in the IEEE T RANSACTIONS ON I NDUSTRY A PPLICATIONS by the Industrial Drives Committee of the IEEE Industry Applications Society. This work was supported by the European Commission under a Marie Curie Research Training Network for realizing this work under the MEST-CT-2004-504243 Research Project. R. Raute, C. Caruana, C. S. Staines, and J. Cilia are with the Faculty of Electrical Engineering, University of Malta, MSD 06 Msida, Malta (e-mail: cccaru@eng.um.edu.mt). M. Sumner and G. M. Asher are with the School of Electrical Engineering, University of Nottingham, Nottingham, NG7 2RD, U.K. (e-mail: Mark.Sumner@nottingham.ac.uk; greg.asher@nottingham.ac.uk). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TIA.2010.2057495

switching. However, signicant current distortions were still observed [15]. This paper presents a novel technique that exploits the highfrequency signals that are inherently imposed on the machine due to the nature of the inverters modulated output voltage. The high-frequency voltage harmonics, at multiples of the PWM switching frequency, provide sufcient pulsating excitation for the machine to produce a current response with position-dependent modulation [17]. This paper will examine the feasibility of extracting rotor-position information from the resulting high-frequency components for the sensorless control of ac machines. Contrary to normal signal-injection techniques, the highfrequency voltage signal amplitude and position cannot be independently set [18]. The second PWM harmonic was identied as the strongest pulsating HF signal [17] and is therefore utilized as the HF excitation signal for position detection. However, the same technique can be applied to the other harmonics. The nature of the selected PWM harmonic as an excitation signal is described in Section II, and the algorithm for using the test signal to derive a rotor-position signal is presented in Section III. Sections IVVIII discuss the practical implementation of the algorithm, including the various signal processing and control algorithms required to extract the desired signal from the raw signal which contains distorting effects, such as higher order harmonic saturation harmonics. It should be emphasized that this paper describes the application of the new sensorless-control technique to the sensorless vector control of unmodied induction motors (IMs), i.e., motors which have a small degree of rotor slotting saliency. The test motor used in the experimental system has a skewed rotor and has semiclosed slots and therefore presents an extremely demanding control challenge. Experimental results presented in Section IX demonstrate that, despite the complexity of the proposed control approach, good sensorless performance can be achieved in the difcult low- and zero-speed regions. II. B EHAVIOR OF PWM H ARMONICS In standard voltage-source inverters (VSIs), the principle of PWM is used to generate the fundamental excitation voltage. The inverter output is switched between the positive and negative potential of the dc-link voltage at a high carrier frequency. The desired fundamental output voltage is the average of the PWM periods. Fig. 1 shows the inverter output voltage for one

0093-9994/$26.00 2010 IEEE

1990

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 46, NO. 5, SEPTEMBER/OCTOBER 2010

Fig. 2. Pulsating v PWM2 vector rotating with fundamental voltage vector v in frame.

In the space-vector approach, the fundamental three-phase voltages can be combined to a complex phasor in the stator xed frame as follows:
Fig. 1. One PWM cycle of the generated inverter output voltage: (a) phase output voltages and (b) , components of resulting voltage vector.

v =

2 4 2 VDC mA + mB ej 3 + mC ej 3 . 3 2

(5)

PWM period. The PWM inverter voltage generation produces a switched voltage pattern with sharp edges. This generated voltage signal has frequency components that is theoretically up to innity. Since the PWM period width is xed and can be regarded as repeated pattern, the Fourier series can be calculated to investigate the frequency components. As a basic rule of repeated patterns, only frequency components of multiples of the pattern frequency will be present in the Fourier series. That means that the complete inverter output voltage can be fragmented into frequency components of multiples of the PWM switching frequency plus a dc component. As a basic approach, the inverter phase-switching duty cycle dx can be calculated by (1), where the modulation index mx depends directly on the desired phase-voltage amplitude (2). The subscript x refers to phases A, B , and C dx = 0.5 + 0.5mx 2vx mx = . VDC (1) (2)

Thus, the three-phase PWM carrier harmonic components can be also combined to the complex phasors in the reference frame to aPWMn =
2 4 2 aPWMnA + aPWMnB ej 3 + aPWMnC ej 3 . 3 (6)

It was found that, at low speed, the second PWM harmonic has the largest amplitude. Looking back to Fig. 1 helps in understanding this conclusion since it can be seen that, in each PWM period, the active voltage vectors occur twice, centered from the middle of the PWM cycle. Therefore, only the second PWM harmonic is used in the sensorless algorithm. All variables referred to this PWM harmonic are further denoted as PWM2. From (6) and (4), the second PWM carrier harmonic can be described as a pulsating vector, rotating approximately synchronously with the fundamental voltage vector in the stator xed frame, as shown in v PWM2 =
2 2 VDC sin(mA ) + sin(mB )ej 3 3 4 + sin(mC )ej 3 cos(2PWM t).

According to the general form of a Fourier series, the PWM output voltage per phase vx can be described as vx (t) = aPWM0x + (aPWMnx cos(nPWM t) 2 n=1

(7)

Fig. 2 shows the fundamental voltage vector v (5) and the resulting pulsating HF vector v PWM2 (7). III. U SE OF PWM H ARMONICS FOR S ALIENCY D ETECTION The second PWM harmonic can be regarded as an injected HF pulsating vector. The resulting current PWM2 carrier harmonic, together with the injected HF voltage vector, can be used for detecting the impedance inside the machine, similar to HF pulsating-injection drives [5][8]. However, as the HF pulsating-vector amplitude and phase are now determined

+ bPWMnx sin(nPWM t)) . (3)

The desired fundamental inverter output voltage is the PWM period dc component aPWM0x /2 which is (dx 0.5)VDC . Since the signal is regarded as an even function, all bPWMnx coefcients are zero. The amplitude of each nth PWM harmonic can be calculated by the coefcient aPWMnx aPWMnx = 2VDC sin (n (0.5 + 0.5mx )) . n (4)

RAUTE et al.: SENSORLESS CONTROL OF INDUCTION MACHINES BY USING PWM HARMONICS

1991

TABLE I PARAMETERS OF THE E MPLOYED I NDUCTION M ACHINE

Fig. 3.

Block diagram of PWM2 signal demodulation.

iPWM2

1 = T

(iPWM2 + iPWM2
0

by the fundamental operation, a novel position observer is required. Fig. 3 shows a block diagram of the implemented signal processing. The measured three-phase voltages and currents are directly combined in the time domain to the resulting stator voltage and current vector (v , i ). The components of the vectors are bandpass ltered. The lters center frequency is set to twice the PWM switching frequency. The width of the passband needs to be carefully selected. The bandpass lter removes unwanted frequency components and reduces noise, but it also determines the maximum frequency of the detectable saliency signal modulation. The bandpass-ltered signals are further demodulated by a heterodyning technique. The resulting fPWM2 voltage and current vector can be described by (8) and (9), where 2v PWM2 and 2iPWM2 are the amplitudes of the pulsating signals v PWM2 = 2v PWM2 cos(2fPWM2 t + vPWM2 ) iPWM2 = 2iPWM2 cos(2fPWM2 t + iPWM2 ). (8) (9)

cos(2 2fPWM2 t +2iPWM2 )) dt. (13)

An equivalent-impedance vector z PWM2 can be dened based on the demodulated voltage and current PWM carrier harmonic vectors v PWM2 and iPWM2 (14). The ratio between v PWM2 and iPWM2 takes the uncontrolled variation in the injected HF excitation into account. The complex vector division can be calculated in any reference frame. The result will be only affected by the phase and magnitude difference of the two vectors. The nal HF equivalent-impedance vector z PWM2 is calculated by the complex vector division directly in the frame, i.e., z PWM2 = v PWM2 . iPWM2 (14)

IV. P RACTICAL I MPLEMENTATION The parameters of the off-the-shelf induction machine are stated in Table I. The rotor is a skewed squirrel-cage design with semiopened rotor slots. A controlled dc drive is coupled to the induction machine under test as the load. For the control unit, a Texas Instruments DSK 6713C board, in conjunction with a customized eld-programmable gate array (FPGA) board, is used. The FPGA board provides peripheral interfaces that allow using the DSK 6713C board for the control of electric drives. It contains a standard PWM signal-generation unit for the control of a three-phase twolevel VSI and 16-b analog-to-digital converters (ADCs) for the current and voltage measurements. The analog currentand voltage-measurement signals are ltered by a fourth-order 10-kHz low-pass antialiasing lters before the digital sampling. The generated PWM switching signals are directly fed into a commercial inverter. No dead-time compensation strategy is implemented at this stage. The PWM switching frequency is set to 1.818 kHz. This is a rather low value for the 5.5-kW machine used. This value has been chosen to ensure that the investigation of the basic algorithm and its associated signalprocessing and compensation strategies is not restricted by the processing capabilities of the DSP and FPGA platform used. This low PWM frequency allows the voltage to be sampled at a signicantly higher rate such that a reasonably accurate value for the second PWM harmonic can be obtained online. It is recognized that, if a more realistic PWM frequency were considered (e.g., 10 kHz), then the signal-to-noise ratio may be poorer for the current measurements, but, at present, this is considered to be a feasibility study rather that a full commercialization project. The digital bandpass lter used is a fourth-order

The high-frequency voltage vector is directly multiplied by a cosine reference signal cos(2fPWM2 t + vPWM2 ), and the high-frequency current vector signal is multiplied by cos(2fPWM2 t + iPWM2 ). The cosine reference signals are stored in a lookup table (LUT) for the known signal samples. The phases of the cosine multiplication signals were set on experimental data and ne-tuned manually. The resulting voltage and current signals are therefore given by v PWM2 cos(2fPWM2 t + vPWM2 ) = v PWM2 + v PWM2 cos(2 2fPWM2 t + 2vPWM2 ) (10) iPWM2 cos(2fPWM2 t + iPWM2 ) = iPWM2 + iPWM2 cos(2 2fPWM2 t + 2iPWM2 ). (11)

The HF carrier frequency component, which converts to 2fPWM2 , is removed by a discrete average lter. The average lter calculates the average of a time period T that includes exactly an integer number of periods of the HF signal cos(2 2fPWM2 t). As a result, only the amplitude modulation signals v PWM2 and iPWM2 of the fPWM2 voltage and current signals are derived, as shown in the following: v PWM2 1 = T
T

(v PWM2 + v PWM2
0

cos(2 2fPWM2 t +2vPWM2 )) dt (12)

1992

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 46, NO. 5, SEPTEMBER/OCTOBER 2010

Fig. 4. Visualization of rotor-bar equivalent-impedance modulation and resulting v PWM2 and iPWM2 for steady-state operation.

Butterworth design tuned to a center frequency of 3.636 kHz with 400-Hz bandwidth. All currents and voltages are sampled at 30 kHz. The ADC sampling frequency is exactly 16.5 times the PWM switching frequency. The sensorless algorithm is executed every second PWM cycle. Therefore, the HF signal demodulation is performed over 33 signal samples, and the cycle frequency of the sensorless algorithm is 909.09 Hz. V. M EASURED ROTOR -BAR M ODULATION E FFECTS The focus of this paper is to detect the asynchronous modulation due to the conductor bars embedded in the rotor iron package of the machine. It is assumed that the rotor bars cause a circular equivalent-impedance modulation with an amplitude ZRB . The angle RB is the rotor-bar position within one rotorbar period, which is the distance between two adjacent rotor bars. The resulting voltage equation system for the demodulated PWM2 variables is shown by vPWM2 vPWM2 = Z ZRB cos(RB ) ZRB sin(RB ) Z + ZRB cos(RB ) ZRB sin(RB ) iPWM2 . iPWM2 (15)

frame. The injected pulsating HF voltage vector v PWM2 will rotate with the fundamental frequency. In steady state, the amplitude of this vector will be approximately constant, and the locus will be almost circular in the frame (dashed line). The resulting pulsating HF current vector iPWM2 will follow the imposed voltage vector. However, due to impedance variation, the current-vector locus will be slightly different, as shown by the oval shape (dotted line). The difference between the HF voltage and current vector is proportional to the equivalentimpedance modulation ZRB . This modulation is shown in the drawings for different v PWM2 positions in the frame. The oval iPWM2 locus will rotate by 180 ele per rotor-bar period. As can be seen from the gure, a further rotation occurs in the detected equivalent-impedance modulation depending on the PWM2 vector positions in the stator frame. For the explanation of this additional modulation, the xy reference frame, also shown in Fig. 4, is introduced. The equivalent-impedance tensor ZPWM2 can then be transformed to ZPWM2xy by applying (17), including the rotation matrix (16), which are both shown at the bottom of the page. Equation (18) shows the voltage equation for the demodulated PWM2 quantities in the xy reference frame shown in Fig. 4 |i | vPWM2x = ZPWM2xy PWM2 . 0 vPWM2y

(18)

The result of the complex vector division (14) is stated by (19), which shows that an equivalent-impedance vector with an offset Z and a circular modulation with the radius ZRB rotating backwards with RB + 2iPWM2 occurs, i.e., vPWM2 + jvPWM2 iPWM2 + jiPWM2 vPWM2x + jvPWM2y |iPWM2 | + j ZRB sin RB + 2iPWM2 .

z PWM2 = =

= Z ZRB cos RB + 2iPWM2 (19)

Fig. 4 shows the equivalent-impedance modulation. The PWM2 voltage and current vectors (v PWM2 and iPWM2 ) are shown for the condition that v PWM2 is 45 ele in the stator

After compensating for the offset, the additional 2iPWM2 phase modulation can be easily removed since the HF current vector position iPWM2 is directly known.

B xy =

cos iPWM2 sin iPWM2

sin iPWM2 cos iPWM2 ZRB sin RB + 2iPWM2 Z + ZRB cos RB + 2iPWM2

(16)

ZPWM2xy = B xy ZPWM2 BT xy = Z ZRB cos RB + 2iPWM2 ZRB sin RB + 2iPWM2 (17)

RAUTE et al.: SENSORLESS CONTROL OF INDUCTION MACHINES BY USING PWM HARMONICS

1993

Fig. 5.

Modeled saturation HF modulation.

VI. S ATURATION -M ODULATION E FFECTS In standard induction machines, strong magnetic-saturation effects are present. For simplicity, it can be assumed that the saturation effect on the impedance modulation is superimposed to the rotor-bar modulation. The saturation equivalentimpedance modulation is theoretically also modeled as an ellipse, rotating with the fundamental current vector in the frame. As the HF voltage vector is also rotating with the fundamental frequency, it can be assumed that the resulting saturation effect of the z PWM2 equivalent-impedance vector is static. Therefore, in steady state, the magnetic saturation will only affect the dc component of the vector z PWM2 . Fig. 5 shows the effect of the magnetic saturation inuence to the z PWM2 measurement. The axis of the saturation shape is dened by the maximal magnetic saturation. In steady state, the position of the HF equivalent-impedance measurement frame (iPWM2 ) with respect to the saturation axis S is constant. However, at different speeds, the relative positions of S and iPWM2 change. This is mostly due to the speed-dependent back electromotive force. Therefore, the detected z PWM2 saturation inuence will not be constant under dynamic operation. The total magnetic saturation within the machine will have a further effect on the diameter of the saturation-saliency shape and, therefore, the offset Z . VII. D ECOUPLING OF ROTOR -BAR M ODULATION In the theoretical explanation, it is assumed that the average machine equivalent impedance Z and the modulation effects are superimposed as stated in z PWM2 = Z + z PWM2Sat + z PWM2 RB + z PWM2Inv . (20) In the actual machine used, the measured rotor-bar modulation ZRB is very low (1% to 1.5% of the offset value Z ) due to the rotor construction and, particularly, the skewing. The saturation modulation z PWM2Sat is up to 4% of Z . It was observed that further modulation occurs with each current commutation sector. This effect is assumed to be due to inverter nonlinearity, caused by the dead time during the switching

between the two insulated-gate bipolar transistors per-phase leg and current-clamping effects [2]. The amplitude of this inverter modulation z PWM2 Inv is in the range of about 2% to 3% of Z and is actually higher than the slotting effect to be tracked. Therefore, the decoupling of the desired signal for the rotor-position estimation from the other distorting effects is a critical task. In this paper, a basic LUT compensation scheme is implemented to extract only the desired rotor-bar modulation. As explained in Section VI, the saturation modulation depends on the imposed stator currents and the relative position iPWM2 to the saturation axis position S , as shown in Fig. 5. In the regarded operation, id is kept constant, and thus, the total machine current amplitude depends directly on the torqueproducing current component iq . Since the current controller has a fast dynamic response, i q is used as a reference for the LUT for noise reduction. A further compensation dimension is iPWM2 S . The position of the saturation axis S is not known, but it is assumed that it keeps a constant relationship with the stator current vector i for each particular value of iq . Therefore, iPWM2 i is used as a second reference for the LUT. For the compensation of the inverter nonlinearity effect, i is used as the third LUT dimension. The nal implemented LUT contains the values of z PWM2 LU T (21) addressed by the three reference dimensions z PWM2LU T (iq , i , iPWM2 i) = Z + z PWM2Sat + z PWM2Inv . (21)

For this feasibility study, the LUT was created using a special commissioning test. The drive is operated in fully sensored vector control mode over the whole of the required operating range required for the proposed sensorless algorithm i.e., variable speed and iq . The iq current, the stator current angle i , and the angle iPWM2 i are acquired together with the measured equivalent-impedance vector z PWM2 . Therefore, the z PWM2 modulation is measured, including the rotor bar, saturation, and inverter nonlinearity modulation for the entire operating range. The measured data are processed ofine according to the measured iq , i , and iPWM2 i parameters. Data points with the same iq , i , and iPWM2 i references are averaged. Since the rotor-bar modulation is asynchronous to the measured iq , i , and iPWM2 i parameters and the data were collected over a nite time duration, it is assumed that the rotor-bar modulation is removed due to averaging of several samples per iq , i , and iPWM2 i reference. Thus, a z PWM2 LU T modulation prole is generated, which does not include the rotor-bar modulation. Figs. 6 and 7 show the real and imaginary parts of the measured z PWM2 LU T prole dependent on i and iPWM2 i for the rated current. The parameter range is discretized by 2 ele for both reference parameters. The at sections in the gures show conditions that did not occur during operation. The iPWM2 i range is limited due to the fundamental operation of the machine, and therefore, the saliency function is only detected in a small section of the 360 ele range. Also, during sensorless operation, only points in the detected section can occur, and thus, the knowledge of this compensation prole is sufcient.

1994

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 46, NO. 5, SEPTEMBER/OCTOBER 2010

Fig. 6. Real part of z PWM2 LU T (iq = 12.5 A, i , iPWM2 i).

Fig. 9. Imaginary part of z PWM2 LU T (iq , i = 0 ele , iPWM2 i).

Fig. 10.

Signal tracking PLLs in the sensorless algorithm.

Fig. 7. i).

Imaginary part of z PWM2 LU T (iq = 12.5 A, i , iPWM2

Fig. 11.

Block diagram of PLL1 structure.

Figs. 8 and 9 show the equivalent-impedance compensation values as function of iq and iPWM2 i. The 3-D graphs show the data for the fundamental stator current position i = 0 ele . In Fig. 8, it can be seen that the offset in the real part of z PWM2 also changes with iq . Therefore, it can be assumed that the offset Z is also a function of the machinecurrent amplitude. VIII. ROTOR -P OSITION R ECONSTRUCTION
Fig. 8. Real part of z PWM2 LU T (iq , i = 0 ele , iPWM2 i).

In the shown equivalent-impedance prole of z PWM2 LU T , one can notice the parameter dependence of i and iPWM2 i. It can be seen that the real-part offset Z is about 190 V/A for the given iq . Also, the modulation pattern of the current commutation sectors depends on i , as can be seen. The magnetic saturation-saliency modulation z PWM2 Sat has the expected shape. From these gures, it can be seen that z PWM2 Sat is up to 10 V/A, and the saturation axis, where Re(z PWM2 LU T ) is minimum and Im(z PWM2 LU T ) is zero, is approximately at iPWM2 i = 0 ele . This indicates that the magnetic saturation axis is near the stator current vector.

Disturbances were observed in the LUT decoupling. The disturbances are usually only of small duration. However, a corrupted z PWM2 RB decoupling may cause rotor-bar skipping, which produces a position-estimation step of one rotor-bar period (22.5 ele for the machine used). To reduce the likelihood of rotor-bar skipping, phase-locked loops (PLLs) are used in the sensorless rotor-position estimator, as shown in Fig. 10. One PLL (PLL1) is used to track and lter the measured z PWM2 RB modulation, which contains the rotor-bar modulation signal (RB + 2iPWM2 ). A second PLL (PLL2) is used to condition the nal derived rotor-bar position signal. The second PLL block is also used to obtain the mechanical position and a ltered-speed estimate. Figs. 11 and 12 show the two PLL structures in detail. For the rst PLL, the real and imaginary parts of the decoupled

RAUTE et al.: SENSORLESS CONTROL OF INDUCTION MACHINES BY USING PWM HARMONICS

1995

Fig. 12. Block diagram of PLL2 structure.

z PWM2 RB are directly used as input signals. The output is the position of the z PWM2 RB modulation (RB + 2iPWM2 ). It was observed that, under dynamic operation, the LUT compensation does not provide a perfect signal decoupling in the real part of z PWM2 . Even if the offset compensation error is always small (< 3% of Z ), it can reach absolute values larger than the rotor-bar modulation amplitude ZRB . As a result, the quadrature PLL1 error signals derived from the real and imaginary parts of z PWM2 RB are conditioned by the scaling factors kRe and kIm . kIm is set to a larger value than kRe . Therefore, the combined error signal to the proportionalintegral controller is mainly derived from the imaginary part of z PWM2 RB . The tradeoff is that, even under correct lock in of the PLL, an error signal at twice the input frequency can be seen. The second PLL (PLL2) is applied after the 2iPWM2 phase modulation has been removed. This PLL serves only to further lter the estimated rotor-bar position signal. This reduces noise that might be introduced by the 2iPWM2 phase removal and further smoothes the effect of the incurred ripple in PLL1. The tracking signal of PLL2 is directly converted into a mechanical speed estimate and integrated into the nal e . Furthermore, a low-pass-ltered estimated rotor position R e speed signal R F is provided. The PLL system for the signal conditioning provides a signicant improvement of the rotor-bar position-signal tracking. It reduces the occurrences of rotor-bar skipping and smoothes noise. This signal-conditioning structure could be even more enhanced, for instance, by integrating a mechanical observer [19]. IX. E XPERIMENTAL R ESULTS A. Current Control Figs. 13 and 14 show the results when the sensorless IM drive is operating in sensorless current-control mode. The speed is controlled by the coupled dc machine. Fig. 13 shows an experiment to test the dynamic response of the sensorless drive. The ux-producing current component i d is kept constant, and is changed from 0 to 12.5 A (0% to 100%). The i i q q transient is limited to 100% per second. The mechanical speed is set to 52 rev/min by the coupled dc machine. Since at rated iq current, the slip frequency is about 1.74 Hz, the electrical frequency is approximately zero when the maximum load is applied to the machine.

Fig. 13. Sensorless drive operating in current control at constant speed and varying load. (a) Measured mechanical rotor position R and estimated rotor e . (b) Position-estimation error ( e ). (c) ie current composition R R R dq ponents of the sensorless vector-controlled drive. (d) Measured stator current vector i components. (e) and (f) Real and imaginary components of the complex equivalent-impedance vector z PWM2 and the LUT compensation components.

It can be seen that at full applied i q , the stator currents are almost dc. Plots (e) and (f) show that the LUT compensation values (z PWM2 LU T ) follow the oating mean of the equivalent-impedance vector z PWM2 . However, during the transitions of ie q , the compensation values do not follow the z PWM2 average precisely. Therefore, the correct rotor-bar equivalent-impedance modulation z PWM2 RB decoupling is not always ensured. The LUT for the modulation decoupling was generated using several experiments with constant iq . It was observed that, under iq transients, the equivalentimpedance modulation behaves differently than under steadystate conditions. It is assumed that the saturation effect of z PWM2 does not change instantly with idq transitions. As can be seen, the LUT compensation error is only pronounced in the real part of z PWM2 . However, the PLL signal conditioning ensures that the tracking of the correct rotor-bar position can be achieved even under the iq transient conditions. It was found that the error of the z PWM2 compensation is dependent on the step amplitude and rate of change in iq . Fig. 14 shows the sensorless drive operating in currentcontrol mode during speed transients. The current controller references i d and iq are both set to rated values. The speed is changed from zero to 300 rev/min, 300 rev/min, and zero speed again. The speed transients are limited to 600 rev/min/s. It

1996

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 46, NO. 5, SEPTEMBER/OCTOBER 2010

Fig. 15. Sensorless drive in speed control. (a) Measured mechanical rotor e . (b) Position-estimation error position R and estimated rotor position R e ). (c) ie current components of the sensorless vector-controlled (R R dq drive. (d) Measured and estimated mechanical speed nR and ne R.

Fig. 14. Sensorless drive operating at rated current and varying speed. e. (a) Measured mechanical rotor position R and estimated rotor position R e ). (c) ie current components of the (b) Position-estimation error (R R dq sensorless-vector-controlled drive. (d) Measured and estimated rotor speed. (e) and (f) Real and imaginary components of the complex equivalentimpedance vector z PWM2 and the LUT compensation components.

can be seen that the drive operates stably at all times, including at zero mechanical speed. From plots (e) and (f), it can be seen that the LUT compensation follows the correct z PWM2 values even during the speed changes. B. Speed Control Fig. 15 shows the results of the implemented sensorless IM drive operating in speed-control mode. The coupled dc machine is set to provide a constant electrical-load torque. The sensorless IM drive is controlling the speed by an implemented speed loop providing the reference i q . The estimated speed e provided by PLL2 is used for speed feedback. The speed R F reference is changed between 60 rev/min. The transitions of the reference speed are limited to 120 rev/min/s. The speed controller bandwidth is set to a low value due to the limited iq dynamic caused by the transient effect of the z PWM2 RB decoupling. It can be seen that the reference speed is obtained by the sensorless drive. As can be seen, the drive provides up to 100% of rated current of the induction machine (iq = 12.5 A) in this experiment. The visible stiction at zero speed is due to friction in the mechanical coupling between the test and load machine of the used rig. The same effect occurred also in sensored operation and is therefore not associated to the sensorless algorithm.

Fig. 16. Sensorless drive in position control. (a) Measured mechanical rotor e . (b) Position-estimation error position R and estimated rotor position R e ). (c) ie current components of the sensorless vector-controlled (R R dq drive. (d) Measured and estimated mechanical speed nR and ne R.

C. Position Control Fig. 16 shows the implemented sensorless drive operating in position-control mode. The coupled dc machine is set to provide a constant electrical-load torque. The position controller of the sensorless IM drive provides a reference for the speed-control loop. The reference position is changed by one mechanical revolution (two electrical periods) forward and backward. The reference position ramp is limited to 360 ele per second. The position-control loop contains only a proportional controller since the plant transfer function provides an integral

RAUTE et al.: SENSORLESS CONTROL OF INDUCTION MACHINES BY USING PWM HARMONICS

1997

part. It can be seen that in this experiment, the sensorless drive also provides up to rated iq . X. C ONCLUSION This paper has proposed a new technique for accurate sensorless control of a standard IM at zero and low speeds. The technique detects the rotor slotting saliency but does not require any additional test-signal injection or PWM pattern modication. The pulsating high-frequency excitation due to the second harmonic of the PWM voltage waveform is used to detect the equivalent impedance, which is calculated by a complex division using the demodulated high-frequency current and voltage vector variables. A precalibrated LUT is required to decouple the rotor-position saliency from an offset and other saliencies due to iron saturation and other distorting effects such as inverter nonlinearities. A dual PLL structure is used to improve the signal tracking. Experimental results for a skewed off-the-shelf IM with semiclosed slots show that even though the rotor-bar modulation is very small, sufcient position information can be retrieved for position and speed control at zero and low speed. This paper has demonstrated the feasibility of the proposed algorithm, although a low switching frequency has been employed and detailed precommissioning is required. It is anticipated that with the continuing advances in microprocessor technology, the data sampling and processing algorithms may be implemented at higher frequencies to allow the switching frequency to increase to values on the order of 10 kHz. The subject of precommissioning, online self-tuning, and the use of alternative compensation devices such as expert systems, continues to be a major research area for most types of sensorless drive control based on saliency tracking. R EFERENCES
[1] C. Caruana, G. M. Asher, and M. Sumner, Performance of HF signal injection techniques for zero-low-frequency vector control of induction machines under sensorless conditions, IEEE Trans. Ind. Electron., vol. 53, no. 1, pp. 225238, Feb. 2006. [2] N. Teske, G. M. Asher, M. Sumner, and K. J. Bradley, Analysis and suppression of high-frequency inverter modulation in sensorless positioncontrolled induction machine drives, IEEE Trans. Ind. Appl., vol. 39, no. 1, pp. 1018, Jan./Feb. 2003. [3] Q. Gao, G. Asher, and M. Sumner, Sensorless position and speed control of induction motors using high-frequency injection and without ofine precommissioning, IEEE Trans. Ind. Electron., vol. 54, no. 5, pp. 2474 2481, Oct. 2007. [4] M. W. Degner and R. D. Lorenz, Position estimation in induction machines utilizing rotor bar slot harmonics and carrier-frequency signal injection, IEEE Trans. Ind. Appl., vol. 36, no. 3, pp. 736742, May/Jun. 2000. [5] M. J. Corley and R. D. Lorenz, Rotor position and velocity estimation for a salient-pole permanent magnet synchronous machine at standstill and high speeds, IEEE Trans. Ind. Appl., vol. 34, no. 4, pp. 784789, Jul./Aug. 1998. [6] J.-H. Jang, S.-K. Sul, J.-I. Ha, K. Ide, and M. Sawamura, Sensorless drive of surface-mounted permanent-magnet motor by high-frequency signal injection based on magnetic saliency, IEEE Trans. Ind. Appl., vol. 39, no. 4, pp. 10311039, Jul./Aug. 2003. [7] M. Linke, R. Kennel, and J. Holtz, Sensorless speed and position control of synchronous machines using alternating carrier injection, in Proc. IEMDC, Madison, WI, 2003, pp. 12111217. [8] J. Holtz, Acquisition of position error and magnet polarity for sensorless control of PM synchronous machines, IEEE Trans. Ind. Appl., vol. 44, no. 4, pp. 11721180, Jul./Aug. 2008.

[9] A. Consoli, G. Scarcella, and A. Testa, Industry application of zero-speed sensorless control techniques for PM synchronous motors, IEEE Trans. Ind. Appl., vol. 37, no. 2, pp. 513521, Mar./Apr. 2001. [10] F. Briz, W. Degner, P. Garca, and J. M. Guerrero, Rotor position estimation of AC machines using the zero-sequence carrier-signal voltage, IEEE Trans. Ind. Appl., vol. 41, no. 6, pp. 16371646, Nov./Dec. 2005. [11] M. Schroedl, Sensorless control of AC machines at low speed and standstill based on the INFORM method, in Conf. Rec. IEEE IAS Annu. Meeting, San Diego, CA, Oct. 610, 1996, pp. 270277. [12] J. Holtz, Sensorless position control of induction motorsAn emerging technology, IEEE Trans. Ind. Electron., vol. 45, no. 6, pp. 840852, Dec. 1998. [13] C. Caruana, G. M. Asher, and J. C. Clare, Sensorless ux position estimation at low and zero frequency by measuring zero-sequence current in delta-connected cage induction machines, IEEE Trans. Ind. Appl., vol. 41, no. 2, pp. 609617, Mar./Apr. 2005. [14] J. Holtz and J. Juliet, Sensorless acquisition of the rotor position angle of induction motors with arbitrary stator windings, IEEE Trans. Ind. Appl., vol. 41, no. 6, pp. 16751682, Nov./Dec. 2005. [15] Q. Gao, G. M. Asher, M. Sumner, and P. Makys, Position estimation of AC machines over a wide frequency range based on space vector PWM excitation, IEEE Trans. Ind. Appl., vol. 43, no. 4, pp. 10011011, Jul./Aug. 2007. [16] Y. Hua, G. M. Asher, M. Sumner, and Q. Gao, Sensorless control of surface mounted permanent magnetic machine using the standard space vector PWN, in Conf. Rec. IEEE IAS Annu. Meeting, New Orleans, LA, Sep. 2327, 2007, pp. 661667. [17] R. Raute, C. Caruana, J. Cilia, C. S. Staines, M. Sumner, and G. Asher, Impedance saliency detection for sensorless control of AC machines utilizing existing PWM switching harmonics, in Proc. PEMD, York, U.K., Apr. 24, 2008, pp. 557561. [18] R. Raute, C. Caruana, C. S. Staines, J. Cilia, M. Sumner, and G. Asher, Sensorless control of induction machines by using PWM harmonics for rotor bar slotting detection, in Conf. Rec. IEEE IAS Annu. Meeting, Edmonton, AB, Canada, Oct. 59, 2008, [CD-ROM]. [19] P. L. Jansen and R. D. Lorenz, Transducerless position and velocity estimation in induction and salient AC machines, IEEE Trans. Ind. Appl., vol. 31, no. 2, pp. 240247, Mar./Apr. 1995.

Reiko Raute (S04M09) received the Dipl. Ing. (FH) degree in electrical engineering degree from the University of Applied Science Jena, Jena, Germany, in 2005. From 2005 to 2009 he worked on his Ph.D. degree project about sensorless control of ac machines at the University of Malta, Msida, Malta. He received the Ph.D. degree from the University of Nottingham, Nottingham, U.K., in December 2009. During his undergraduate studies, he visited the University of Nottingham and the University of Adelaide, Adelaide, Australia, working on electric drive projects. He is currently with Carlo Gavazzi (Malta) Ltd.

Cedric Caruana (S01M05) received the Ph.D. degree in electrical engineering from the University of Nottingham, Nottingham, U.K., in 2003. He was with Carlo Gavazzi (Malta) Ltd. and ST Microelectronics (Malta) Ltd. as a Process Engineer and Senior Test Engineer, respectively. He has been with the Faculty of Engineering, University of Malta, Msida, Malta, following receipt of the Ph.D. degree, where he is currently a Senior Lecturer. His main research interests are the control of ac drives, renewable energy conversion, electrical power systems, and energy efciency. Dr. Caruana is a member of the Institution of Engineering and Technology, U.K.

1998

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 46, NO. 5, SEPTEMBER/OCTOBER 2010

Cyril Spiteri Staines (S94M98) received the B.Eng.(Hons.) degree from the University of Malta, Msida, Malta, in 1994, and the Ph.D. degree from the University of Nottingham, Nottingham, U.K., in 1999, with his Ph.D. work on sensorless control of ac drives carried out in the School of Electrical and Electronic Engineering. In 1995, he joined the University of Malta as an Assistant Lecturer, where he became a Lecturer in 1999, Senior Lecturer in 2004, and Associate Professor in 2007, and he currently heads the Department of Industrial Electrical Power Conversion. From 2003 to 2004, he was a Postdoctoral Researcher and Visiting Lecturer at the University of Nottingham. His research interests include sensorless ac motor drives and grid connection of renewable energy sources, in particular, for wind-energy systems. Prof. Staines is a member of the Institution of Engineering Technology, U.K.

Mark Sumner (SM05) received the B.Eng. degree in electrical and electronic engineering from Leeds University, Leeds, U.K., in 1986, and the Ph.D. degree in induction-motor drives from the University of Nottingham, Nottingham, U.K., in 1990. He was with Rolls Royce Ltd., Coventry, U.K., before embarking on his Ph.D. studies. After receipt of the Ph.D. degree, he joined the University of Nottingham as a Research Assistant, and where he was appointed a Lecturer in October 1992, and is currently an Associate Professor and Reader in power electronic systems. His research interests cover control of power electronic systems, including sensorless motor drives, diagnostics and prognostics for drive systems, power electronics for enhanced power quality, and novel power system fault-location strategies.

Joseph Cilia received the Ph.D. degree in sensorless control of ac motors from the University of Nottingham, Nottingham, U.K., in 1997. He has been with the University of Malta, Msida, Malta, where he was a Lecturer in power electronics and drives, following receipt of the Ph.D. degree. His research interests cover ac motor control and encoderless techniques, power converters for motor drive systems, battery management in motivepower applications, grid-connected renewable energy sources, and energy efciency.

Greg M. Asher (SM02F07) received the B.S. and Ph.D. degrees in bond-graph structures and general dynamic systems from Bath University, Bath, U.K., in 1976 and 1979, respectively. He has been with the University of Nottingham, Nottingham, U.K., where he was appointed a Lecturer in control in the School of Electrical and Electronic Engineering in 1984. He developed an interest in motor drive systems, and was appointed Professor of electrical drives in 2000, and is currently the Associate Dean of the Engineering Faculty. He has published over 200 research papers and has received over 5M in research contracts. Dr. Asher is an Associate Editor of the IEEE Industrial Electronics Society (IES). He has been a member of the Executive Committee of the European Power Electronics Association and Chair of the Power Electronics Technical Committee of the IES.

You might also like