A New Sensorless Drive Control System for Transmissionless EVs Using a Permanent-Magnet Synchronous Motor

One of the most important technologies for electric vehicles (EVs) will be drive control technology for main traction motors. It is desired for the drive control system to have the following characteristics: 1) it is based on vector control methods that allow quick, efficient and/or precise torque production; 2) it can produce 150-250% rated torque at standstill and can have wide driving-range, consequently can allow development of EVs with no variable-transmission; 3) it does not require a position/speed sensor for controlling motor drive, which has been mounted on a rotor shaft. This paper proposes a new sensorless drive control system for such EVs using a permanent-magnet synchronous motor as a main traction motor. Potential usefulness of the proposed drive control system is examined through the development of a prototype EV capable of driving on public road.


INTRODUCTION
One of the most important technologies for electric vehicles (EVs) will be a drive control technology for main traction motors. As traction motors of EVs, ac motors such as induction motor (IM), permanent-magnet synchronous motor (PMSM) are usually employed. It is desired for a drive control system to have the following characteristics: 1) It does not require a position/speed sensor for controlling motor drive, which has been mounted on a rotor shaft. 2) It is based on vector control methods that allow quick, efficient and/or precise torque production. 3) It can produce 150-250% rated torque at standstill and can have wide driving-range, consequently can allow development of EVs with no variable-transmission. Most of conventional vector control methods can allow the drive control system to have characteristics 2) and 3), but they require a position/speed sensor to be mounted on a rotor shaft for them. Realization of characteristic 1) allows to have some benefits such as increasing of drive control system reliability, increasing mechanical design flexibility, decreasing motor size in direction of shaft, elimination of wiring sensor cable, decreasing a variety of cost related to the sensor etc. Compared with the motor itself, the position/speed sensor is less reliable from electric, mechanical and thermal viewpoints. Dependence on the sensor results in decrease of total reliability of driving control systems, which will be the most important factor of human-ride vehicles. In addition to improvement in system reliability and mechanical design flexibility, cost and size reductions are attractive in EVs where the storage space is often very restricted.
Due to the attractive features of sensorless drive, developments of sensorless-vector-controlled EVs with no variable transmission (in the following, refereed to as sensorless and transmissionless EVs or shortly to as ST-EVs) have been tried so far [1]- [5]. For ST-EVs using an IM, a successful development allowed to drive on public roads in 2001 by Shinnaka et al. was reported [2]. For ST-EVs using a PMSM, such a successful development seems not to be reported so far, although developments keep going [1]- [5]. Under such circumstances, the authors have succeeded in development of such a ST-EV allowed to drive on public road in 2004.
This paper proposes a new drive control system for ST-EVs using a PMSM in a concretive manner by referring to the newly developed prototype ST-EV and examines its potential usefulness. This paper is organized as follows. Section 2 presents a mathematical model and characteristics of PMSM, which is a direct object to be controlled by vector control methods. The model and characteristics form foundations for establishing the drive control system. Section 3 proposes a new configuration of the drive control system. Section 4 proposes a phase estimator that produces good estimates of phase and speed of the traction PMSM in wide speed-range from standstill to overrated speed instead of a position/speed sensor mounted on a rotor shaft. Section 5 proposes a command generator and a command converter that allow efficient and wide speed-range drive with no variable transmission. Section 6 presents experimental results by both a test-bed setup and the prototyped ST-EV on a chassis dynamometer. Section 7 concludes this paper.

MATHEMATICAL MODEL AND CHARCTERISTICS OF EV MOTOR
A mathematical model and characteristics of PMSM, which is a direct object to be controlled by vector control methods, form foundations for establishing the drive control system. Consider the general reference frame where orthogonal coordinates rotating at an arbitrary instant angular velocity as shown in Fig. 1. Rotating polarity is defined such that the direction from principal axis ( -axis) to secondary axis ( -axis) is positive. Note that all of the following 2x1 vector signals related to PMSMs are defined in the general reference frame. (SP-PMSMs), which consists of three consistent basic equations such as circuit equation, torque evolution equation, and energy transmission, can be described as follows [8], [9]: . (9) where 2x1 vectors , , v 1 i 1 1 are voltage, current and flux of the stator respectively, 2x1 vectors i , m are components of the stator flux 1 , more precisely, i indicates flux evolved directly by the stator current and i 1 m is flux due to the rotor magnet, I is a 2x2 identity matrix, J is a 2x2 skew symmetric matrix such as

Energy transmission equation (the third basic equation)
is the evolved torque, 2n , m 2 are the electrical and mechanical speeds of the rotor, is the stator resistance, is the "in-phase and mirror-phase inductances" having the following relation to the d, q-axes inductances [9] Table 1. One-hour rated output of the traction SP-PMSM is 16-kw, and the maximum output is 40-kw, 250% of one-hour rating. One-hour rating of evolved torque is 40-Nm, and the maximum torque is 100-Nm, 250% of one-hour rating. Rated speed is 400-rad/s in mechanical sense, and the maximum speed is 1000-rad/s, 250% of the rating. Fig. 2 illustrates a basic configuration of the proposed drive control system for ST-EVs. The traction SP-PMSM is coupled to the rear wheels though a differential gear with a constant gear ratio, but no variable transmission. Note that speed ratio between the rotor and the rear wheels is constant in all speed ranges including backward drive. The drive control system features the block circled by a dashed line. The block produces electrical power to be injected to the SP-PMSM according to directions by a d-n-r hand lever and an acceleration foot pedal. The speed meter is driven by speed estimate generated by the drive control system. The main differences between the ST-EV drive control system and the conventional system are the phase estimator and the high frequency voltage generator (indicated as HFV) for sensorless drive, and the command generator and the command converter for energy-efficient and wide speed-range drive. The following sections explain these components in detail, which are indicated by shaded sub-blocks in Fig. 2.

Frequency-hybrid combination
The phase estimator plays a role estimating rotor phase and speed instead of a position/speed sensor, which will be the most important for sensorless drives. It is a matter of common understanding that no single estimation method can appropriately estimate the rotor phase in a wide speed range from zero to overrated speed, and that frequency-hybrid combination of two estimation methods is practical, either of which can operate appropriately in the zero-to-low-speed, or low-to-high-speed range [6], [7].    frequency-hybrid combination by two phase-estimates produced by different estimation methods. The frequency-dependent weight in (a) varies mildly according to frequency, on the other hand, quickly in (b). The proposed drive control system in Fig. 2 employs type (b) with frequency-dependent weight of unity or zero, and selects weight-changing frequency to be about 600-rad/s in electrical sense for the traction SP-PMSM characterized by Table 1. Fig. 4 shows a configuration of the phase estimator based on type (b) frequency-hybrid combination with frequencydependent weight of unity or zero. The phase estimator produces estimates of the rotor phase and speed from inputs of the stator voltage, and current in a rotational reference frame, and the frequency of super-injected high frequency voltage. At electrical frequency about 600-rad/s, the "switch" turns from H-side to L-side or vice versa, one of two the initial phase estimates is always selected. The phase synchronizer, which plays a role producing the final phase estimate and the rotor speed estimate, is shared by both phase estimation methods.

.2 Phase estimation in low-to-high speed range
In low-to-high speed range, for example electrical speed 600~4000-rad/s for the SP-PMSM in Table 1, the rotor phase is estimated using fundamental components of the stator voltage and current, which contributes torque evolution. In the range, the high frequency voltage generator (indicated as HFV) is in principle turned off, and the initial phase estimate is determined by the "D-state-observer" originally proposed by one of the authors [8]. The D-state-observer has the following attractive characteristics [8].
1 ) It is a flux-state observer requiring no additional steady-state condition for the dynamic mathematical model of the PMSMs. 2) Its order is the minimum second.
3) Observer gain guaranteeing proper estimation in four quadrants over a wide operating range except singular zero-speed is a simple constant, and can be easily designed (default value is a unity). 4) It utilizes motor parameters in a very simple manner. 5) Its structure is very simple and it can be realized at very low computational load. 6) It can be applied to both of salient pole and non-salient pole PMSMs. 7) The maximum error of steady-state phase estimate can remain within 4 / -rad even if the speed estimate with huge constant error is used in the observer. The D-state-observer can be described as follows [8]: Fig. 6. Configuration of phase synchronizer based on the generalized integral-type PLL method.  reference frame converges to the d-q rotor reference frame. The description using d, q suffixes in (13) is due to this convergence. The phase synchronizer in Fig. 4 is realized according to the generalized integral-type PLL method, as in Fig. 6.

Phase estimation in zero-to-low speed range
At very low speed, the level of fundamental motor-driving voltage is very small, and its signal-to-noise ratio seriously decreases, furthermore phase estimation based on the fundamental driving signals is in general very sensitive to stator resistance. Consequently, phase estimation using the fundamental motor-driving signals becomes very difficult in the region. In order to cope with the difficulty, the proposed drive control system employs the "mirror-phase estimation method," which was originally proposed by one of the authors [9]. It is a kind of high frequency voltage injection methods and has the following attractive high performance and features. 1) It is directly based on a principle called as the "mirror-phase characteristics", which were recently found to be inherent in ac motors with pole-saliency (refer to Fig. 7 in the folowing).
2) It can allow 250% rated torque at standstill, under which strong flux saturation usually occurs. G 0 const ; ) sgn( 2 g g n J I . ( 1 6 ) Fig. 5 shows the D-state-observer. It is worthy of noting that if the constant observer gain is selected to be zero, i.e.
, the D-state-observer results in so-called the "voltage model" in (1)-(4). This fact means that the D-state-observer is the most compact flux-state observer. The phase lock can be attained by the "generalized integral-type Phase-Locked-Loop (PLL) method," which can be described as follows [9]: 0 5) It can allow instant injection of the rated load even for zero-speed control. 6) It can allow accommodation of a load with huge moment of inertia. 7) It is insensitive to all of motor parameters such as stator resistance, stator inductance and rotor magnet flux. 8) It is also insensitive to saturation phenomena of stator flux.

Generalized Integral-Type PLL Method
9) It is so simple that it can be implemented with no additional hardware. In the zero-to-low speed range, the high frequency voltage generator (HFV) in Fig. 2 generates a rotational high frequency voltage command such as mirror-phase current . According to the "mirror-phase characteristics," the rotor phase locates right at the center of the in-phase and mirror-phase currents (refer to Fig. 7). Based on the mirror-phase characteristics, the rotor phase can be estimated using the in-phase and mirror-phase currents such as The lower blocks of the phase estimator in Fig. 4 performs phase estimation based on the mirror-phase estimation method. The "in-phase mirror-phase current evolver" extracts the in-phase and mirror-phase currents from the stator current, using information of a constant frequency h of the high frequency voltage command. As shown in Fig. 8, the in-phase mirror-phase current evolver is constructed by means of the "filters in the D-module," which was originally proposed by one of the authors [10]. It is known that the filter in the D-module is robust to distortion of sinusoidal input signal, i.e. it produces almost pure sinusoidal output even for distorted input. This feature provides the phase estimation method using the filter in the D-module with robust to signal distortion caused by flux saturation and so on. The phase-error detector in Fig. 4 produces an initial phase estimate from the in-phase and mirror-phase currents. While the lower blocks of the phase estimator in Fig. 4 performs, the "switch" is in principle turned to L-side.
T he phase estimator, which produces estimates of the rotor phase and speed instead of position sensors, makes sensorless drives possible, but does not energy-efficient and wide speed-range drives. It is made possible by the command generator and the command converter in Fig. 2.

Configuration of command generator
The proposed drive control system produces a signed current-norm command from directions by the d-n-r hand lever and the acceleration foot pedal at the command generator in Fig. 2. The d-n-r hand lever directs drive modes such as d (drive), n (neutral), and r (reverse), and the foot pedal does the magnitude of the current-norm command, i.e. where superscript * indicates commands. For the neutral direction, the signed current command is forced to be zero.

Configuration of command
The command converter produces d-, q-current commands from a signed current norm command so that an energy-efficient and wide speed-range drive can be attained with constraints as where "sgn" means a sign-function whose output is a positive or negative unity [11]. Fig. 9 shows the internal structure of the command converter, which firstly determine analytically a d-current command from a signed current-norm command according to specific strategies, if necessary using rotor speed and dc-link voltage information as well, secondly does a q-current command simply based on (23). Note that d-, q-current commands are determined analytically with no necessity of approximation or recursive computation. The detailed explanations of the command converter are given in the following. One of energy-efficient strategies in the constant torque region where no effective voltage limit exists is the one that maximizes ratio of torque per stator current-norm in order to minimize the copper loss due to the stator resistance. The optimal d-, q-current commands for the strategy commands are analytically obtained as [11]  The most important factor to be considered in drive over the rated speed is the voltage limit. With no consideration of the voltage limit, high-speed drive in the overrating region cannot be attained. The optimal d-, q-current commands is analytically obtained as [11] 2 * 2 * * * Equation (27) yielding the unique analytical solution for maximum voltage strategy cannot be applied to Region A. However, this does not mean that region A is unusable. Region A corresponds to a driving case that rotor speed is too low. In other words, region A is the one where no consideration of voltage limit is necessary and pure energy-efficient strategy in (24) can be applied. Note that allowed total driving region includes both regions A and B, and that right hem of both regions A and B is governed by the allowed current norm.
It is worthy of noting that the state stator current corresponding to (25) is optimal in the sense that it produces the maximum torque with the minimum copper loss under the voltage limit . v c Wide speed-range drives across regions A and B in Fig.  10 can be attained effectively by a combination of two strategies, one of which pursuits pure energy-efficiency under no effective voltage limit; the other does so but under effective voltage limit. Fig. 11 shows a flow chart that shows how two strategies are combined, in other words, is switched. At the first step, the inequality in right hand side of (27) is checked. If it is not satisfied, the optimal solution is located in region A in Fig. 10, and the solution by the maximum torque/current strategy i.e. (24) is selected. If satisfied, the optimal solution is located in region B. Even in region B, especially in neighborhood of the upper curve specified by right hand side of (27), there still exists possibility that the pure energy-efficient strategy can be selected. In order to examine the possibility, go to the second step, and compare two d-current commands by (24) and (25). According to the comparison, select the solution In order that the analytical solution has physical meaning or is practical, the following constraints should be satisfied [11]  Using motor parameters in Table 1, Fig. 10 shows a region as region B, where (27) gives a physically meaningful and/or practical solution. The upper hem curve of region B is governed by the right hand side of (27), and the lower hem straight lines by the left hand side. Region B clearly shows that the achievable maximum speed varies dependently on both voltage limit and the current-norm command.

Experiments on test bed
Figs. 12 and 13 show a prototyped drive control apparatus corresponding to the dashed block in Fig. 2 and a SP-PMSM for a main EV traction motor coupled with a load machine on a test bed, respectively (refer to Table 1 for motor characteristics). The SP-PMSM was driven by the apparatus, which switches frequency-hybrid mode at electrical speed 600-rad/s (mechanical speed 150-rad/s). The high frequency voltage was selected as h -rad/s in experiments on the test-bed. Fig. 14 is a response at the rated mechanical speed 400-rad/s to a signed current-norm command of =233-A capable of producing the rated torque 40-Nm. At the speed, d-and q-current commands are determined based on (24) by the command converter in Fig. 9, the HFV is turned off, no high frequency voltage is superimposed, the "switch" in Fig. 4 is turned to H-side, and an initial phase estimate is determined by the D-state-observer. Fig 14 shows, from the top, q-current, d-current, actual rotor phase, and its final estimate. The average values of q-, d-currents are controlled to be equal to the associated commands, although they have some ripples. Rotor phase is smoothly estimated due to low-pass filtering effects by the D-state-observer. The estimation error to actual phase is about 0.1-rad under the rated load.  Fig. 15 is a response at standstill where signed current-norm, d-current, and q-current commands are the same as in Fig. 14. However, HFV is turned on, and a high frequency voltage command is superimposed on the fundamental voltage command, the "switch" in Fig. 4 is turned to L-side, and an initial phase estimate is determined by the mirror-phase estimation method. Fig 15 shows, from the top, q-current, d-current, d-components of high frequency in-phase and mirror-phase currents , . The ripples on d-and q-currents correspond to the high frequency components, which are filtered out as in-phase and mirror-phase currents. No phase difference between the d-components of the in-phase and mirror-phase currents means that rotor phase is well estimated. The maximum estimation error to actual phase is about 0.2-rad under the rated load, and the average error is nearly zero. i 1hi i 1hm

Experiments by prototype EV
The prototyped drive control apparatus was installed in a ST-EV as well as the traction motor. Fig. 16 shows an appearance of the apparatus mounted on the rear of the ST-EV. Experiments using the ST-EV were carried out on a chassis dynamometer as shown in Fig. 17. Experimental data such as U-phase current, dc-bus voltage, which are accessible from outside of the apparatus, were taken.     18 shows a response at drive speed about 50-km/h (about mechanical rotor speed of 600-rad/s). It indicates, from the top, dc-bus voltage and U-phase current. The stator current is well controlled as observed. At this speed, the HFV is turned off, no high frequency voltage is superimposed, the "switch" in Fig. 4 is turned to H-side, and an initial phase estimate is determined by the D-state-observer. Fig. 19 shows a response at drive speed about 5-km/h (about mechanical rotor speed of 60-rad/s). The meaning of the waveform data is the same as in Fig. 18. High frequency components of 800 h -rad/s due to a high frequency voltage command are clearly observed, i.e., at this speed, HFV is turned on, and a high frequency voltage command is superimposed on the fundamental voltage command, the "switch" in Fig. 4 is turned to L-side, and an initial phase estimate is determined by the mirror-phase estimation method. Note that the amplitude of the high frequency voltage command was reset to a larger value 10 1 h V -V than in the test-bed experiments, in consideration of both robustness to disturbance from road and high frequency acoustic noise.
After the necessary experiments, public-road drive of the prototyped ST-EV was started, which was called as "Shin II" after ST-EV "Shin" using an induction motor. Fig. 20 is a picture of ST-EV Shin II driving in Yokohama Port area on March 25, 2004.

CONCLUSIONS
One of the most important technologies for EVs is drive control one for main traction motors. This paper proposed a new sensorless drive control system that allows realization of ST-EVs using a permanent-magnet synchronous motor as a main traction motor. The proposed drive control system has the following potential characteristics: 1) it is based on vector control methods that allow quick, efficient and/or precise torque production; 2) it does not require a position/speed sensor for controlling motor drive, which has been mounted on a rotor shaft; 3) it can produce high torque even at standstill and can have wide driving-range, consequently can allow development of ST-EVs. Potential usefulness of the proposed drive control system was examined through the development of a prototype ST-EV capable of driving on public road.