You are on page 1of 12

Richard T.

Meyer1
Department of Mechanical and
Aerospace Engineering,
Western Michigan University,
Kalamazoo, MI 49008
e-mail: richard.meyer@wmich.edu
Hybrid Electric Vehicle Fault
Scott C. Johnson
School of Electrical and
Computer Engineering,
Tolerant Control
Purdue University, This paper investigates the supervisory-level, fault tolerant control of a 2004 Prius
West Lafayette, IN 47907 powertrain. The fault considered is an interturn short circuit (ITSC) fault in the traction
e-mail: johns924@purdue.edu drive (a surface mount permanent magnet synchronous machine (SPMSM) for which its
rotor is part of the vehicle’s driveline). ITSC faults arise from electrical insulation fail-

Downloaded from http://asmedigitalcollection.asme.org/dynamicsystems/article-pdf/140/2/021002/6023978/ds_140_02_021002.pdf by guest on 04 July 2023


Raymond A. DeCarlo ures in the stator windings where part of a phase winding remains functional while the
School of Electrical and remaining decoupled windings form a self-contained loop. Because the permanent
Computer Engineering, magnets on the rotor (driveline) shaft are able to induce very large eddy currents in this
Purdue University, self-contained loop if its rotational velocity is left unchecked, the maximum allowable
West Lafayette, IN 47907 driveline speed, and consequently, vehicle speed, must be reduced to avoid exceeding the
e-mail: decarlo@ecn.purdue.edu drive’s operational thermal limits. A method for detecting these ITSC faults and the
induced eddy current in an SPMSM using a moving horizon observer (MHO) is reviewed.
Steve Pekarek These parameters then determine which previously computed, fault-level-dependent
School of Electrical and SPMSM input–output power efficiency map and maximum safe operating speed is utilized
Computer Engineering, by the supervisory-level controller. The fault tolerant control is demonstrated by simulat-
Purdue University, ing a Prius over a 40 s drive velocity profile with fault levels of 0.5%, 1%, 2%, and 5%
West Lafayette, IN 47907 detected at the midpoint of the profile. For comparison, the Prius is also simulated with-
e-mail: spekarek@purdue.edu out a traction motor fault. Results show that the control reduces vehicle velocity upon
detection of a fault to an appropriate safe value. [DOI: 10.1115/1.4037270]
Scott D. Sudhoff
School of Electrical and
Computer Engineering,
Purdue University,
West Lafayette, IN 47907
e-mail: sudhoff@purdue.edu

1 Introduction motor fault is considered in Phillips et al. [11] and Xu et al. [12].
In both works, vehicle operation is restricted in some way with
Hybrid electric vehicles, such as the Toyota Prius, use both an
respect to the fault, as for example, by reducing maximum allow-
internal combustion engine (ICE) and electric machine, e.g., per-
able motor power. Neither of these works specify how variations
manent magnet synchronous machine (PMSM), for propulsion.
in the level of motor fault are incorporated into the vehicle power-
When a fault occurs in a PMSM, it commonly is an interturn short
train control.
circuit (ITSC) in the stator windings [1]. Continued electrical
Herein, a moving horizon observer is used to identify the pres-
excitation of the PMSM during a fault results in hot spots that
ence of a fault and the fault level in a 2004 Toyota Prius traction
may lead to fire [2,3]. Electrical excitation can result from the
PMSM. Given this fault information, the next objective is to
PMSM inverter as well as rotor spin due to vehicle movement. In
develop fault mitigation controller strategies that allow PMSM-
the latter case, stopping the vehicle to eliminate excitation may
based hybrid vehicles to continue operation albeit at a substan-
leave the driver far from service or in an unsafe location.
tially reduced level. Fault mitigation is carried out at the so-called
A review of pertinent PMSM fault detection work is in Refs.
supervisory level. A supervisory-level controller coordinates vehi-
[2,3]. With regards to PMSM fault mitigation after fault detection,
cle control by determining optimized power flows to the individ-
actions are dividable into hardware reconfiguration and fault toler-
ual subsystems such as was explored in Refs. [13–15]. For
ant control [4]. Hardware reconfiguration usually requires some
example, for a diverse set of situations, the supervisory-level con-
redundancy or additional component features that can be activated
troller would determine how best to utilize the electric drive sys-
during a fault to reduce/bypass its effect. This approach requires
tem in coordination with the ICE to meet driving objectives.
additional hardware cost [5,6]. Fault tolerant control can be classi-
For efficient and feasible numerical optimization strategies,
fied as either passive or active [5]. In the passive case, the control
supervisory-level subsystem models are power flow based;
is designed to be robust to certain faults. In contrast, active con-
specifically, the subsystem models are low granularity power flow
trol, initiated by a fault detection capability, modifies component
component models that utilize efficiency maps as opposed to
operation to mitigate the effects of the fault. Past work in active
high-granularity models based on the underlying physics. In the
fault tolerant control is primarily focused on using observers to
case of the PMSM, such an efficiency map depends on whether or
replace information from faulty motor sensors [7–10]. However,
not the motor has a fault as well as on the degree of fault. An
modification of hybrid electric vehicle control strategies during a
advantage of the “online” supervisory-level control is that it can
adapt the controls to different fault levels as they occur without
1
Corresponding author. requiring an exhaustive library of precomputed controls.
Contributed by the Dynamic Systems Division of ASME for publication in the
JOURNAL OF DYNAMIC SYSTEMS, MEASUREMENT, AND CONTROL. Manuscript received
This paper investigates: (i) the hybrid, i.e., switched system,
September 26, 2016; final manuscript received June 13, 2017; published online supervisory control problem in the Toyota Prius; (ii) construction
September 8, 2017. Assoc. Editor: Azim Eskandarian. of efficiency maps and constraints for a PMSM with ITSC faults;

Journal of Dynamic Systems, Measurement, and Control FEBRUARY 2018, Vol. 140 / 021002-1
C 2018 by ASME
Copyright V
Downloaded from http://asmedigitalcollection.asme.org/dynamicsystems/article-pdf/140/2/021002/6023978/ds_140_02_021002.pdf by guest on 04 July 2023
Fig. 1 The Toyota Prius hybrid powertrain architecture with power flows: (yellow) fuel power
consumed, (green) electrical power produced, (red) electrical power consumed, (blue)
mechanical power produced, and (orange) mechanical power consumed

(iii) work for PMSM ITSC fault detection using a moving horizon state of each mode. Since only one mode can be physically active
observer; and (iv) a high-level vehicle power flow control optimi- at a time, we impose the constraint a1 þ a2 þ a3 þ a4 ¼ 1. We note
zation strategy during both normal and ITSC fault operation. that values of ai vary over time and can change instantaneously; a
minimum dwell time is assumed to avoid the possibility of infin-
2 Prius Supervisory-Level Powertrain Description itely fast mode switching.
Given the modes of operation, the supervisory-level optimiza-
Figure 1 shows the architecture of the 2004 Toyota Prius’s tion problem has the form
hybrid powertrain with a 57 kW Atkinson-cycle ICE, a 21 kW
nickel metal hydride (NiMH) battery pack, two SPMSMs,2 and a
min Jðx0 ; xref ; tp;0 ; tp;f ; x; u; aÞ (1a)
power split device (PSD) consisting of a planetary gear system u;ai 2f0;1g
that connects the power flow pathways of the ICE and SPMSMs. i¼1;…;4
One SPMSM is for power generation and engine starting, labeled
“SPMSM1,” with a maximum mechanical power of 30 kW and subject to xðtp;0 Þ ¼ x0 (1b)
the other SPMSM is for traction and regenerative braking, labeled
“SPMSM2,” with a maximum mechanical power of 50 kW. 1 ¼ a1 þ a2 þ a3 þ a4 (1c)
Due to differing efficiency maps for SPMSM1 and SPMSM2
propelling and generating, the Prius architecture is modeled as a X
4

switched system with a finite set of “modes of operation.” Each x_ ¼ ai fi ðt; x; ui Þ (1d)
mode represents a fixed configuration of power flow directions and is i¼1

characterized by potentially different efficiency maps and dynamics: X


4
(i) SPMSM1 propelling (ICE start-up)-SPMSM2 propelling-battery 0¼ ai hi ðt; x; ui Þ (1e)
discharging-ICE off; (ii) SPMSM1 generating-SPMSM2 propelling- i¼1
battery discharging-ICE on; (iii) SPMSM1 generating-SPMSM2
propelling-battery charging-ICE on; and (iv) SPMSM1 generating- X
4

SPMSM2 generating-battery charging-ICE on or off. 0 ai gi ðt; x; ui Þ (1f )


i¼1
In the remainder of this paper, we will consider an ITSC fault only
occurring in SPMSM2, i.e., no faults occur in SPMSM1. SPMSM1,
and SPMS2 models are developed in Sec. 4 and SPMSM2 fault
detection is presented in Sec. 5. Section 6 describes the simulation In the problem statement:
results of the control and interturn short circuit fault. Section 7 con-  Ja ðx0 ; xref ; tp;0 ; tp;f ; x; u; aÞ is a convex performance index,
cludes the paper. In the Appendix, the supervisory-level models are wherein x0 is the state at the current time tp,0, tp,f is the final
constructed for the ICE, battery, vehicle, power split device, and prediction horizon time, tp,f  tp,0 is the prediction horizon
electrical bus. The Appendix also describes the interconnection length, xðÞ 2 Rn is the supervisory-level state consisting of
structure and other modeling issues associated with the Prius. all pertinent dynamic subsystem variables, xref(t) is a refer-
ence state trajectory to be tracked over ½tp;0 ; tp;f ; uðÞ 2
3 Supervisory-Level Power Flow Control Rm4 is the composite of the mode-specific continuous con-
trol input vectors, and a ¼ ½a1 ; a2 ; a3 ; a4 > is the aggregate of
As mentioned, a supervisory-level controller specifies power mode values such that at each instant of time the entries of a
flows in and out of the subsystems/components. The low- sum to one and there is at most one nonzero entry.
granularity power flow control oriented models (as opposed to  x ¼ ½Pfuel ; Pice ; Xice ; W bat ; !v > is the state vector, where Pfuel
first principles high-granularity models) allow for feasible and is the fuel power delivered to the engine, Pice is the ICE out-
more efficient numerical optimizations. put power, Xice is the square of the ICE angular speed, W bat
To notationally identify the modes set forth in Sec. 2, let ai 僆 is the normalized battery state of charge (SOC), and !v is an
{0, 1}, i ¼ 1, 2, 3, 4 denote the active (ai ¼ 1) or nonactive (ai ¼ 0) energy function equal to the square of the vehicle velocity.
The speed variables are squared to represent energy, which is
2
The 2004 Toyota Prius uses two interior PMSMs (IPMSM). In this work, we ordinarily proportional to the square of speed. Further details
substitute SPMSMs with comparable power capabilities for the IPMSMs to take
advantage of existing fault models as per Refs. [2,3], and [16], which are briefly
can be found in the Appendix.
reviewed in the next section. Although the focus here is on SPMSMs, this work  ui ¼ ½uifuel ; ui1 ; ui2 ; uibrk > , wherein uifuel is the commanded nor-
serves as a baseline for any future work which uses IPMSMs. malized fuel in the ith mode of operation; ui1 is the

021002-2 / Vol. 140, FEBRUARY 2018 Transactions of the ASME


commanded fraction of maximum speed-dependent power  0  gi(t, x, ui) represents the ith mode inequality constraints
for SPMSM1 in the mode of operation, which is electrical in pertinent to the optimization problem, for example, bounds
mode 1 and mechanical in the remaining modes; ui2 is the on states, inputs, algebraic variables, etc.
commanded fraction of maximum speed-dependent power
for SPMSM2 in the mode of operation, which is electrical in
modes 1, 2, 3, and mechanical in 4; and uibrk is the com- 3.1 Comments on the Numerical Solution of the Optimization
manded fraction of maximum speed-dependent frictional Problem. The minimization problem in Eq. (1) is numerically ill-
braking power in the ith mode. conditioned, since the optimization requires searching over all
 The performance index (PI), J(), is one of two expressions combinations of switching sequences. As discussed in Ref. [18],
depending on whether the engine is on or off (note that the this is an NP-hard problem that exponentially increases with the
engine on or engine off is a fuel dependent control action number of modes. To avoid this complexity, the switched optimal
and does not represent two distinct modes of operation) control problem in Eq. (1) is replaced with the embedded optimal
 control problem (a partial relaxation of only the modes of opera-
Joff ðÞ; ICE off (2a) tion), which has the form [18,19]
JðÞ ¼
Jon ðÞ; ICE on (2b)
min Jðx0 ; xref ; tp;0 ; tp;f ; xe ; u; aÞ (4a)

Downloaded from http://asmedigitalcollection.asme.org/dynamicsystems/article-pdf/140/2/021002/6023978/ds_140_02_021002.pdf by guest on 04 July 2023


ð tp;f X
4
u;ai 2½0;1
i¼1;…;4
Joff ðÞ ¼ Cv ð!v  !v;ref Þ2 þ Cibrk ai ðPibrk Þ2
tp;0 i¼1 subject to xðt0 Þ ¼ x0 (4b)
X
4 X
4
þCu1 ai ðui1 Þ2 þ Cu2 ai ðui2 Þ2 dt 1 ¼ a1 þ a2 þ a3 þ a4 (4c)
i¼1 i¼1
ð tp;f
X
4
Jon ðÞ ¼ Joff ðÞ þ CSOC ðW bat  W bat;ref Þ2 x_ e ¼ ai fi ðt; xe ; ui Þ (4d)
tp;0
i¼1
2
þCfuel ðPfuel Þ dt
0 ¼ hi ðt; xe ; ui Þ; 8i ¼ 1; 2; 3; 4 (4e)
where !v,ref is the squared reference velocity or energy to
track, W bat;ref is the desired battery state of charge, Cv is the 0  gi ðt; xe ; ui Þ; 8i ¼ 1; 2; 3; 4 (4f )
penalty weight on the vehicle energy error (and implicitly
vehicle tracking error), Cibrk is the mode-dependent penalty Observe the embedded problem state is denoted as xe to distin-
weight on frictional braking use to promote battery regenera- guish it from the original problem state and the algebraic con-
tive braking, Cu1 and Cu2 are the penalty weights on straints, Eqs. (4e) and (4f), are now satisfied in every mode. The
SPMSM1 and SPMSM2 usage to encourage bang-bang solu- distinguishing aspect of the embedded optimization problem rela-
tions in the numerical optimization described shortly,3 CSOC tive to the original optimization problem is the latter’s requirement
is the penalty weight on deviation of the battery state of ai 僆 {0, 1} is relaxed to ai 僆 [0, 1]. This relaxation transforms the
charge from a desired value, and Cfuel is the penalty weight original problem with combinatorial complexity into a traditional
on ICE
P engine fuel consumption. nonlinear optimization. When ai ¼ 1 for some i, then the embed-
 x_ ¼ 44 ai fi ðt; x; ui Þ represents a composite of the mode spe- ded problem solution in Eq. (4) reduces to the switched problem
cific subsystem dynamics without interconnections. Specifically solution in Eq. (1). When ai 僆 (0, 1), one must project the modal
2 3 2 i
3 solution onto the nearest physically realizable mode of operation
P_ fuel ffuel ðPfuel ; Xice ; uifuel Þ using techniques described in Refs. [17–19]. However, the basic
6 7 6 7 idea is to project the embedded fractional value of ai onto the near-
6 P_ ice 7 6 fice i
ðPfuel ; Pice ; Xice Þ 7
6 7 X 4 6 7 est integer value, while maintaining the constraint Eq. (4c).
6 7 6 7
6 X_ ice 7 ¼ a 6 f i
ðP ice ; P i
Þ 7 (3) It was proven in Ref. [19] that the switched state trajectories of
6 7 i6 Xice ice;psd 7
6 _ 7 i¼1 6 7 x of Eq. (1) are dense in the set of embedded system trajectories
6 W bat 7 6 i i
fbat ðW bat ; Pbat Þ 7
4 5 4 5 xe in Eq. (4), so a given embedded trajectory can be approximated
!_ v i i i i
fv ðPbrk ; Pwhl;psd ; P2;L ; !v Þ by a switched system trajectory to any degree of desired accuracy.
Thus, Eq. (4) provides a numerically viable nonlinear optimiza-
wherein Piice;psd is the ICE power routed into the power split tion problem whose solution can be used to generate approximate
device in the ith mode, Pibat is the ith mode battery discharge solutions to Eq. (1). For additional details concerning the embed-
or charge power, Pibrk is the ith mode frictional braking ded optimal control problem, we refer the reader to Ref. [19].
power, Piwhl;psd is the ith mode power delivered to the drive The specific numerical solution uses fmincon within MATLAB,
which requires the optimization problem to be discretized. Briefly,
wheels from the power split split device, Pi2;L is the ith mode
the performance metric is discretized using trapezoidal numerical
SPMSM2 power to the drive wheels, and integration and the dynamics are discretized using collocation.
i i
ffuel ðÞ; fice ðÞ; fXi ice ðÞ; fbat
i
ðÞ, and fvi ðÞ are the right-hand The reader is referred to earlier works for details [17].
sides of Eqs. (A1), (A2), and (A6)–(A8), respectively, in the
Appendix. Also, for reference yi ¼ ½Pibat ; Pibrk ; Pi1;e ;
4 Surface Mount Permanent Magnet Synchronous
Pi1;L ; Pi2;e ; Pi2;L ; Piice;psd ; Piwhl;psd > ; i ¼ 1; 2; 3; 4 is the vector
Machine Supervisory-Level Modeling
of mode-specific algebraic variables that depend upon states
and continuous control inputs. Because there has been no supervisory-level, efficiency-based
 0 ¼ hi(t, x, ui) describes the ith mode interconnections of the modeling development of the SPMSM that accounts for ITSC
subsystem models and any other equality constraints perti- faults in the literature, its development is presented here. The
nent to the optimization problem. Specifically, the con- remainder of the models are presented in the Appendix.
straints are described by Eqs. (A28)–(A35) in the Appendix.
4.1 SPMSM1. At the supervisory level, SPMSM1 is assumed
3
This is similar to the penalty weight on electric drive system use to promote to have fault-free operation with input-to-output efficiency-based
bang-bang solutions in Ref. [17]. representation

Journal of Dynamic Systems, Measurement, and Control FEBRUARY 2018, Vol. 140 / 021002-3
P1;L ¼ gSPMSM;prop;1 ðx1;m ; P1;inv ÞP1;inv (5)

for propelling and for generating


P1;inv ¼ gSPMSM;gen;1 ðx1;m ; P1;L ÞP1;L (6)

where P1,L is the mechanical power, P1,inv is the inverter power, g


denotes efficiencies, which depend on the input power, and
the rotational speed of the rotor, x1,m, which depends on the
SPMSM1 parameters and powertrain operation as set forth in the
Appendix.
The efficiency maps are generated from a detailed model of the
SPMSM1 operation given in Refs. [2,3], and [20]. Specifically

gSPMSM1;prop ðx1;m;sc ; P1;inv;sc Þ ffi a1 þ b1 x1;m;sc


þ c1 x21;m;c þ d1 P1;inv;sc þ e1 w1;m;sc P1;inv;sc

Downloaded from http://asmedigitalcollection.asme.org/dynamicsystems/article-pdf/140/2/021002/6023978/ds_140_02_021002.pdf by guest on 04 July 2023


(7)

gSPMSM1;gen ðx1;m;sc ; P1;L;sc Þ ffi a2 þ b2 x1;m;sc


þ c2 x21;m;sc þ d2 P1;L;sc þ e2 w1;m;sc P1;L;sc (8)
Fig. 2 The power envelope for SPMSM2 in propelling mode
where x1;m;sc ¼x1;m =ð1000p=3Þ; P1;inv;sc ¼P1;inv =3104 ;P1;L;sc ¼ with degree of fault r 5 0.005, 0.01, 0.02, 0.05, and parameters in
P1;L =3 104 , and the coefficients a1,…, e2 are set forth in the Table 7
Appendix.
To do this effectively, one needs to explicate the underling mathe-
4.2 SPMSM2. For SPMSM2 the efficiency terms include matical model of SPMSM2 as a function of the fault level. The
potential faulted operation depending on the fault level r as model set forth does not include saturation effects that can be
follows: incorporated as an additional mode of operation; this inclusion is
beyond the scope of this paper.
P2;L ¼ gSPMSM2;prop ðx2;m ; P2;inv ; rÞP2;inv (9)

for propelling and for generating


5.1 SPMSM2 Component Level Model. The SPMSM2 sta-
P2;inv ¼ gSPMSM2;gen ðx2;m ; P2;L ; rÞP2;L (10) tor voltage equations with a single ITSC fault in phase-a (modi-
fied from those in Ref. [16] for physical realizability of the mutual
Here, P2,L is the mechanical power, P2,inv is the inverter power, inductances) are
and x2,m is the rotor rotational speed, which depends upon the 2 3 2 32 3
vehicle speed. The fault-level-dependent efficiencies have a form vas ð1  rÞRs 0 0 0 ias
6 7 6 76 7
similar to Eqs. (7) and (8) 6 vbs 7 6 0 Rs 0 7
0 76 i 7
6 7¼6 6 bs 7
6v 7 6 0 0 Rs 0 54 ics 7
7 6
gSPMSM2;prop ðx2;m;sc ; P2;inv;sc ; rÞ ffi a1 ðrÞ þ b1 ðrÞx2;m;sc 4 cs 5 4 5
þ c1 ðrÞx22;m;c þ d1 ðrÞP2;inv;sc þ e1 ðrÞw2;m;sc P2;inv;sc (11) 0 0 0 0 rRs ifs
2 3 2 3
ias ea
gSPMSM2;gen ðx2;m;sc ; P2;L;sc ; rÞ ffi a2 ðrÞ þ b2 ðrÞx2;m;sc 6 7 6 7
d6 ibs 7 6 eb 7
þ c2 ðrÞx22;m;sc þ d2 ðrÞP2;L;sc þ e2 ðrÞw2;m;sc PL;sc (12) þ Lf ðrÞ 6 7þ6 7 (13)
dt 6 7 6 7
4 ics 5 4 ec 5
where x2;m;sc ¼ x2;m =ð200pÞ; P2;inv;sc ¼ P2;inv =5  104 ; P2;L;sc ¼ if s ef
P2;L =5 104 . SPMSM2 parameters and efficiency fit coefficients
a1,…, e2 are given in the Appendix. and
Any ITSC fault results in an eddy current, ifs, induced in the 2 3
faulted coils by the permanent magnets on the spinning rotor shaft. ð1  rÞ2 L ð1  rÞM ð1  rÞM rð1  rÞL
The magnitude of ifs depends on the fault level, r. The i2f s R losses 6 7
6 ð1  rÞM L M rM 7
can cause overheating that exceeds the maximum thermal limits for Lf ðrÞ ¼ 6 7
4 ð1  rÞM M L rM 5
safe operation. Thus, the specific fault level, r, constrains the maxi- rð1  rÞL rM rM 2
r L
mum safe vehicle speed. This limit then induces a constraint on the
supervisory-level control strategy as evidenced through the safe (14)
operating regions of Fig. 2. The fault level of SPMSM2 as well as
the induced eddy current, ifs, must be estimated from a dynamic where vfs and ifs denote the stator voltage and current in phase
observer, which is implemented in this work as a moving horizon f ¼ a, b, c. The back emf terms are given by
observer (MHO) [21]. The estimated r is used to determine which 2 3 2 3
efficiency map is now applied at the supervisory level. ea ð1  rÞcosðhr Þ
6 eb 7 6 7
6 7 ¼ x2;m km 6 cosðhr  2p=3Þ 7 (15)
4 ec 5 4 cosðhr þ 2p=3Þ 5
5 Interturn Short Circuit Fault Moving Horizon
ef r cosðhr Þ
Observer
As mentioned, an MHO is used to estimate the fault level r and where hr ¼ h2,mnp/2 is the electrical angular position of the rotor
current ifs. We assume for convenience a single fault in phase-a.4 and np is the number of motor poles. Note that the fault loop back
emf, ef, has the same phase angle as the back emf in the faulted
4
Multiple phase faults are considered in Refs. [2,3]. phase-a. Finally, by Kirchoff’s current law

021002-4 / Vol. 140, FEBRUARY 2018 Transactions of the ASME


ias þ ibs þ ics ¼ 0 (16) continue, the horizon is incremented forward in time and the opti-
mization steps are repeated.
It is assumed that ifs are sinusoidal with the form For this application, the MHO observer correctly identified the
fault which begins at the 20 s mark in the simulations detailed in
ias ¼ Iqa cosðhr Þ þ Ida sinðhr Þ (17) Sec. 6. Fault detection was achieved within 50 ms. Subsequent
fault and fault current estimates are then used to switch between
   precomputed efficiency maps and maximum safe operating speed
2p 2p
ibs ¼ Iqb cos hr  þ Idb sin hr  (18) limits, which modifies the supervisory-level power management
3 3 optimization problem. Results follow in Sec. 6.
   
2p 2p
ics ¼ Iqc cos hr þ þ Idc sin hr þ (19) 6 Control and Interturn Short Circuit Fault
3 3
Simulation Results
where Iqf and Idf are the current amplitudes determined by the This section reviews the performance of the Prius over a 40 s
inverter. Controlling currents of the form in Eqs. (17)–(19) has the trapezoidal drive profile subject to an ITSC fault at 20 s. The trap-
advantage that the derivatives of the currents are known in Eq. ezoidal profile consists of: (i) a constant acceleration increase in

Downloaded from http://asmedigitalcollection.asme.org/dynamicsystems/article-pdf/140/2/021002/6023978/ds_140_02_021002.pdf by guest on 04 July 2023


(13), making the computation of vfs algebraic. This observation is velocity from rest to 20.12 m/s (45 mph) over [0,10)s, (ii) constant
exploited in the development of the MHO in Sec. 5.2. velocity over [10, 30)s, and (iii) constant deceleration to rest over
[30, 40]s. Figure 3 shows the commanded trapezoidal and the
simulated unfaulted vehicle response on a level road, i.e., hv ¼ 0.
5.2 SPMSM2 Fault Moving Horizon Observer. The MHO Drive profile control simulations are performed using techni-
is the minimization of the following quadratic performance index5 ques set forth in Ref. [17]. The optimization problem is discre-
over the time interval [tk  T,tk]: tized with a Tc ¼ 0.25 s interval time partition according to the
ð tk methods in Ref. [17]. In the PI, Eq. (2a), the prediction horizon
J¼ ½ðyM  y^ÞT QðyM  y^Þ þ wp k^
v f s k2 dt (20) is 0.5 s (two partitions) and the penalty weights are Cv ¼ 100,
tk T Cibrk ¼ 1000=ð50Þ2 for i ¼ 1, 2, 3, C4brk ¼ 100=ð50Þ2 ; Cu1 ¼
2; Cu2 ¼ 2; Cfuel ¼ 10=ð383:5Þ2 , and CSOC ¼ 1000, when W bat <
where Q is a weighting matrix and v^f s 6 is the estimated fault loop W bat;ref ¼ 0:58 and zero otherwise.
voltage that must be driven to (approximately) zero during the We note that the vehicle cannot change velocity instantane-
estimation process; yM is a vector of measured voltages and cur- ously, and thus, a velocity reference obtained from the operator at
rents and y^ is the estimates of those voltages and currents. kTc becomes the commanded velocity to achieve at (k þ 1)Tc, i.e.,
Specifically vref((k þ 1)Tc). In the optimization, at time kTc, the present, the
desired velocity at (k þ 1)Tc is known. The value at (k þ 2)Tc is
M M M M M T
yM ¼ ½vM
ab ; vbc ; vca ; ias ; ibs ; ics  (21) unknown and is estimated linearly as

v as  v^bs ; v^bs  v^cs ; v^cs  v^as ; i^as ; i^bs ; i^cs T


y^ ¼ ½^ (22) vref;est ððk þ 2ÞTc Þ ¼ vref ððk þ 1ÞTc Þ
 
vref ððk þ 1ÞTc Þ  vref ðkTc Þ
Henceforth, the superscript M indicates a measured variable and þTc (23)
Tc
hatted variables are estimates. All the terms in y^ depend upon the
fault level, sinusoids in Eqs. (17)–(19), the time derivatives of Thus
Eqs. (13) and (17)–(19). Hence, the minimization of Eq. (20) can
be taken over the fault level r^ and the associated piecewise con- !v;ref ðkTc Þ ¼ ½vref ðkTc Þ2 (24)
stant magnitudes of the sinusoidal currents—I^qa ; I^da ; I^qb ; I^db ; I^qc ,
and I^dc —subject to Eqs. (13)–(19), (21), and (22).
!v;ref ððk þ 1ÞTc Þ ¼ ½vref ððk þ 1ÞTc Þ2 (25)

5.3 Observer Implementation. The minimization of the per- !v;ref;est ððk þ 2ÞTc Þ ¼ ½vref;est ððk þ 2ÞTc Þ2 (26)
formance index, Eq. (20), is over a 50 ms horizon with two parti-
tions, each of 25 ms. Successive minimizations have an overlap of
one partition. The forward most partition of the current minimiza-
tion becomes the trailing partition in the next optimization. Opti-
mal values from the forward most partition of the current
minimization are used to initialize both partitions in the next
optimization.
The steps in the optimization over each horizon are: (i) obtain
the measured values at each partition’s boundaries over the hori-
zon; (ii) generate initial estimates of Iqf, Iqf, and the fault level r;
(iii) compute values of the estimated currents using Eqs. (17)–(19)
and the their derivatives evaluated at the appropriate time points;
(iv) evaluate Eqs. (14)–(15); (v) determine the estimates of line-
to-neutral voltages vfs from Eq. (13); (vi) set v^f s equal to the last
row of the right-hand side of Eq. (13); (vii) form y^ using Eq. (22);
and (viii) use a discrete-time approximation of the integral in the
performance index to compute the cost. Steps (iii)–(viii) are per-
formed and repeated until a minimal cost is achieved within a
nonlinear optimization solver, such as fmincon within MATLAB. To
5
We note that the MHO, first set forth in Ref. [21], is dual to a model predictive
control implementation of the LQR problem. Fig. 3 Simulated Prius velocity tracking without fault: (solid
6
v^f s is equal to the last row of the right-hand side of Eq. (13). line) simulated velocity and (dot) commanded velocity

Journal of Dynamic Systems, Measurement, and Control FEBRUARY 2018, Vol. 140 / 021002-5
The reference velocity is extrapolated from known current and
past values that leads to the extrapolated reference, !v,ref,est, in
Eq. (26). We use extrapolated values because we do not assume
that future values of the drive profile are known. This linear
extrapolation assumption is meant to approximate a driver but
does add a small error to the tracking of reference signals that are
nonpiecewise linear or have “corners.”
The Prius model is simulated again over the trapezoidal drive
profile now with ITSC faults of 0.5% (r ¼ 0.005), 1% (r ¼ 0.01),
2% (r ¼ 0.02), and 5% (r ¼ 0.05) from 20 s onward. Figure 4
shows the vehicle velocity tracking achieved for each fault level.
Every fault is considered to have been detected by the observer of
Sec. 5 at t ¼ 20 s. At 20 s, !v,ref is chosen to decelerate the vehicle
at 2 m/s2 until !v,ITSC,max (the square of 95% of the maximum
fault operation velocity) is reached. Then, the reference is the
minimum of !v,ref and !v,ITSC,max until the fault is removed. Dur-

Downloaded from http://asmedigitalcollection.asme.org/dynamicsystems/article-pdf/140/2/021002/6023978/ds_140_02_021002.pdf by guest on 04 July 2023


ing the deceleration phase, SPMSM2 is unpowered to avoid exac-
erbating thermal runaway and hence unsafe operating conditions.
This velocity modification behavior during faults is exhibited in
Fig. 4 for all of the fault cases.
Figure 5 shows the mechanical power (positive if propelling Fig. 5 Simulated Prius SPMSM1 (generator) and SPMSM2 (trac-
and negative if generating) of SPMSM1 and SPMSM2 over the tion) mechanical powers with a 5% ITSC fault: (blue solid line)
power and (gray dashed line) superimposed commanded velocity
drive profile with the 5% ITSC fault. Similar SPMSM1 and
SPMSM2 responses are displayed with the other fault levels.
SPMSM1 is off until it is used to start the engine from 3 to 3.5 s.
Thereafter, SPMSM1 is driven by the ICE and sends power to the
electrical bus to drive SPMSM2 and recharge the battery. After
the fault occurs, SPMSM1 is on for the next 2 s, while the vehicle
speed decreases and then remains off for the remainder of the
drive profile. We note from 20.25 s onward that the ICE is off as
well. SPMSM1 is still able to provide battery charge power from
20.25 to 22 s, during which the ICE is off, by consuming the ICE
inertia energy. SPMSM2 propels the vehicle during the initial
commanded acceleration from 0 to 10 s and both propel and
recharge the battery from 10 to 20 s during the commanded con-
stant velocity. Upon the detection of the fault, SPMSM2 is off,
while the vehicle speed is decreased to the maximum safe fault
operation velocity. Next, the drive is used to propel the vehicle at
constant velocity from 26.5 s until the commanded deceleration to
rest is encountered at 38 s, and then the drive provides battery
regenerative braking to 39.75 s.
For reference, the ICE output power for the unfaulted and 5%
fault-level cases are provided in Fig. 6. In addition, the projected
mode selections are given in Fig. 7.

Fig. 6 Simulated Prius engine power for no fault and 5% fault


level: (blue solid line) power and (gray dashed line) superim-
posed drive profile

Fig. 4 Simulated Prius velocity tracking with ITSC faults of


0.5%, 1%, 2%, and 5%: (blue solid line) 0.5% fault, (red solid Fig. 7 Projected modes for the Prius simulation: (blue solid
line) 1% fault, (green solid line) 2% fault, (cyan solid line) 5% line) mode and (gray dashed line) superimposed commanded
fault, and (red dot) commanded velocity velocity

021002-6 / Vol. 140, FEBRUARY 2018 Transactions of the ASME


7 Conclusions
In this paper, we have reviewed a moving horizon observer to
detect ITSC faults in a SPMSM, set forth the process to develop
efficiency maps for an SPMSM with and without an ITSC fault
and demonstrated how the knowledge of the presence and degree
of ITSC faults can be used by supervisory-level control to imple-
ment fault tolerant control. Results show that upon detection of a
fault in the traction motor and its degree, the control takes appro-
priate steps, i.e., switching to the applicable traction input–output
power efficiency map and gradually reducing the speed of the
vehicle if it exceeds a predefined limit. These steps ensure that
continued operation of the vehicle has a lesser risk of electrical
system damage and fire than if the vehicle is allowed to continue
on without any fault mitigation. The converse to continued, fault
tolerant operation during an ITSC fault is to shut down the trac-

Downloaded from http://asmedigitalcollection.asme.org/dynamicsystems/article-pdf/140/2/021002/6023978/ds_140_02_021002.pdf by guest on 04 July 2023


tion motor and render the vehicle immobile. However, there are
safety issues associated with this as well, e.g., the motorist may be
stranded in dangerous traffic or in a remote area. Further, we note
the moving horizon observer is implicitly robust; however, in the Fig. 8 Prius ICE power output versus speed with fuel effi-
presence of noisy measurements it may be necessary to increase ciency regions
the number of observer partitions and correspondingly decrease
the partition time frame. In this case, because of the sinusoidal
nature of the voltages and currents, the moving horizon observer
will naturally “average” the estimated magnitudes. Future work Equation (A1) describes the fuel power, Pfuel, delivery dynamics,
will explore this aspect and other robustness aspects of the moving where Pmax
fuel ðxice Þ is the maximum available fuel power for a given
horizon observer. Additional future work includes performing engine angular velocity xice, ufuel 僆 [0, 1] modulates Pmax fuel , and
control strategy verification experiments with a physically real- sfuel ¼ 0.04 s is the fuel delivery system lag [15,23]. In Eq. (A2),
ized powertrain and reducing the computational time of the Pice is the engine mechanical output power, sice ¼ 0.26 s is the
supervisory-level control and moving horizon observer. In part, average engine power lag due to combustion delay and crankshaft
the latter requires optimizing the number of horizons, horizon and flywheel inertias [15,23,24], and gice(Pice, xice) is the com-
width, and the search algorithm. This is a dual formulation to the bined efficiency of the combustion and engine mechanical power
problem in model predictive control of determining optimal hori- delivery in which gice(Pice, xice)Pfuel(xice) represents the indi-
zon parameters. rectly commanded engine output power through the operation of
As computational power in vehicles continues to increase and proc- the fuel delivery system. It follows that
essor prices decrease, we expect that using moving horizon observers
for fault detection will become an increasingly attractive solution to Pmax
ice ðxice Þ
Pmax
fuel ðxice Þ ¼ (A3)
improving electric machine safety, reliability, and repair costs. gice ðPmax
ice ; xice Þ

completes the specification of the variables. Values for Pmax


ice and
Acknowledgment
gice are determined by least squares fits of appropriate engine data
The authors would also like to acknowledge the support of in Ref. [22]. Here, the functions
Greg Shaver and the Hoosier Heavy Hybrid Center of Excellence.
Pmax
ice ðxice Þ ¼ a1 x ice þ a0 (A4)

2
Funding Data gice ðPice ; xice Þ ¼ b20 P ice þ b11 P ice x ice þ b10 P ice
þ b01 x ice þ b00 (A5)
 U.S. Department of Energy (Grant No. DE-EE0005568).
are sufficient, where x ice ¼ xice =maxðxice Þ and P ice ¼
Pice =maxðPmax
ice Þ are normalized by their maximum values for
Appendix numerical solution purposes. Table 1 lists the fit coefficients.
In this section, details of the 2004 Toyota Prius ICE, battery Further, the ICE angular velocity is expressed using Xice ¼ x2ice
pack, vehicle dynamics, mechanical power split device, maximum by the conservation of power equation
drive power, and electrical bus component models are described.
1 dXice
Also, the conditions for ICE operation transitions from on and off Jice ¼ Pice;psd þ Pice (A6)
are set forth. 2 dt

A.1 Internal Combustion Engine. The 2004 Toyota Prius Table 1 Toyota Prius engine maximum output power and effi-
powertrain has a 57 kW ICE with operating range between 1000 ciency fit coefficients
and 5000 rpm [22] as shown in Fig. 8. Broadening the power flow
Parameter Value
modeling ideas in Refs. [15,23], the power dynamics are repre-
sented by two first-order lag equations a1 61.609
a0 4.0470
dPfuel 1 1 max b20 0.98686
¼ Pfuel þ P ðxice Þufuel (A1) b11
dt sfuel sfuel fuel 0.25973
b10 0.88661
dPice 1 1 b01 0.27567
¼ Pice þ g ðPice ; xice ÞPfuel (A2) b00 0.26627
dt sice sice ice

Journal of Dynamic Systems, Measurement, and Control FEBRUARY 2018, Vol. 140 / 021002-7
Note Xice ¼ x2ice defines a Lyapunov energy function Pice,psd is power split device defined shortly, Pbrk is the frictional braking
the soon to be developed power routed through the power split power, qair is the ambient air density, mv is the total vehicle mass,
device, and Jice ¼ 0.13 kg/mp2ffiffiffiffiffiffiffiffi
[25] is the estimated rotational iner- Afr is the vehicle frontal area, Cd is the drag coefficient, Crr is the
tia of the engine; xice ¼ þ Xice because the engine turns in only tire rolling resistance coefficient, and hv is the road grade angle.
one direction. Table 3 lists the Prius vehicle parameters.
The frictional braking power, Pbrk, is equal to a maximum
velocity-dependent braking power modulated by ubrk 僆 [0, 1]; the
A.2 Battery. The Prius powertrain has a 21 kW discharge/ maximum braking power depends upon a smooth function that
charge power NiMH battery pack that provides a secondary power increases with velocity until a maximum of 50 kW braking power
source and allows the capture and storage of excess vehicle is achieved. Specifically
kinetic energy via regenerative braking. The battery’s SOC, W bat ,
dynamics are represented with a validated empirical formula [26], pffiffiffiffiffiffi
max !v
which has been modified [17] to include an additional quadratic ð Þu
Pbrk ¼ Pbrk !v brk ¼ 50tanh ubrk (A9)
power term to more accurately represent the 7.2 V 6.5 Ah six-cell 5
NiMH modules [27,28] used herein

Downloaded from http://asmedigitalcollection.asme.org/dynamicsystems/article-pdf/140/2/021002/6023978/ds_140_02_021002.pdf by guest on 04 July 2023


 A.4 Mechanical Power Split Device. The PSD mechani-
dW bat Pbat cally connects the ICE, generator electric drive (SPMSM1), and
¼ gbat W bat ; Pbat ; vb max
dt nb Wbat traction electric drive system (SPMSM2) via a planetary gear sys-
  tem, displayed in Fig. 9, to propel the vehicle, charge the battery,
Pbat Pbat 2
gbat ¼ ln W bat þ cvb;1
b
þ cvb;2
b vb vb
þ cb;3 þ cb;4 and start the ICE. Specifically, the planetary carrier is affixed to
nb nb
the output of the ICE; the SPMSM1 rotor is attached to the sun
(A7) gear; and the ring gear is connected to the SPMSM2 rotor, which
is linked (via additional gears) to the drive wheels. The planetary
max gear system dynamics are analyzed to develop the aforementioned
In Eq. (A7), Pbat is the battery power input, Wbat is the battery’s
maximum rated storage energy; nb is the number of battery mod- Pice,psd and Pwhl,psd in Eqs. (A6) and (A8), respectively,
ules in the pack; vb ¼ d, Pbat  0, for discharge; vb ¼ c, Pbat 0,
for charge; and cdb;i =ccb;i ; i ¼ f1; 2; 3; 4g are discharge/charge coef-  
ficients obtained by the least-squares fit of instantaneous power ðrs þ rr Þ P1;L
Pice;psd ¼ xice (A10)
efficiencies produced using NiMH module battery data in Refs. rs jx1;m j þ s
[27,28] and efficiency relationships in Ref. [29]. Table 2 lists the
vb ;bnd  
battery model parameters, which includes Pbat , the discharge rr c P1;L
and charge power bounds. Further, the Prius battery state of Pwhl;psd ¼  v (A11)
rs jx1;m j þ s
charge is restricted to W bat 2 ½0:4; 0:8 [30].
where rs is the sun gear radius, rr is the radius of the ring gear,
A.3 Vehicle. Vehicle dynamics have been described with a x1,m is the angular velocity of the sun gear and SPMSM1 (which
point-mass, linear motion dynamical model [15,17]. However, are directly coupled), s is a small constant 1 to prevent divi-
when this model has an input of power, a singularity occurs at sion by zero, and c relates the vehicle velocity to the ring gear
zero velocity. To eliminate this singularity and develop a vehicle angular velocity, i.e., xr ¼ cv. To begin developing Eqs. (A10)
dynamics model consistent with our power flow approach, a Lya- and (A11), the dynamics of the PSD with the ring gear connected
punov energy function is defined, !v ¼ v2, where v is velocity. to the drive wheels are set forth in the context of Fig. 9, where the
Thus, the conservation of power to vehicle motion results in the four planet gears are assumed to experience the same forces and
power flow dynamical model thus can be replaced with a single-planet gear with combined
mass and inertia
1 d!v
mv ¼ Pd ðvÞ þ Prr ðv; hr Þ þ Pb ðhr Þ dx1;m
2 dt Js ¼ Ts  Fps rs (A12)
þ P2;L þ Pwhl;psd  Pbrk dt
 (A8)
Pd ðvÞ ¼ v  0:5qair Afr Cd v2 sgnðvÞ dxp
 4Jp ¼ Fpr rp  Fps rp (A13)
Prr ðv; hv Þ ¼ v  Crr mv g cosðhv ÞsgnðvÞ dt
Pb ðhv Þ ¼ v  ðmv g sinðhv ÞÞ dvp
4mp ¼ Fps  Fpr þ Fpc (A14)
pffiffiffiffiffiffi dt
where v ¼ þ !v considering only forward motion, Pd is the drag
xc
force power, Prr is the rolling resistance, Pb is the body force Jc ¼ ðrs þ rp ÞFpc þ Tc (A15)
power due to gravity, Pwhl,psd is the wheel power from/to the dt

Table 2 Toyota Prius NiMH battery pack model parameters


Table 3 Toyota Prius vehicle dynamics model parameters
Parameter Discharge (vb ¼ d) Charge (vb ¼ c)
Parameter Symbol Value
max
Wbat 1.6848  102 kJ 1.6848  102 kJ
vb ;bnd
Pbat 21 kW 21 kW Frontal area Afr 2.33 m2 [30]
nb 28 28 Drag coefficient Cd 0.26 [30]
cvb;1
b
4.5077 49.511 Rolling resistance Crr 0.00475 [31]
cvb;2
b
0.84091 0.29369 Vehicle mass mv 1469 kga [30]
cvb;3
b
2.7023 4.9132 Air density qair 1.225 kg/m3
cvb;4
b
3.6526 0.10087
a
The vehicle mass includes two 79 kg occupants.

021002-8 / Vol. 140, FEBRUARY 2018 Transactions of the ASME


Fig. 9 Mechanical power split device sun-planet-ring gear system

dv Pd þ Prr þ Pb þ P2;L  Pbrk rs x1;m þ rr cv ¼ ðrr þ rs Þxice (A21)


mv ¼ rr cFpr þ (A16)
dt v

Downloaded from http://asmedigitalcollection.asme.org/dynamicsystems/article-pdf/140/2/021002/6023978/ds_140_02_021002.pdf by guest on 04 July 2023


Table 4 displays the PSD parameters used herein. The PSD
The torque from/to SPMSM1 is denoted by Ts; Tc is the torque gear radii are given in terms of number of gear teeth rather than
supplied by the ICE; Js, Jp, and Jc are the rotational inertias of the length, because the gear pitch is unknown and the pitch cancels
sun gear, planet gear, and planet carrier, respectively; xp and vp out of the Pice,psd and Pwhl,psd expressions under the assumption
are the planet gear angular and translational velocities, respec- all the gears have the same pitch.
tively; and Fps, Fpr, and Fpc are the forces at the interface of the
sun and planet gears, ring and planet gears, and carrier and planet
gear, respectively. Equation (A16) is based on knowledge that the
A.5 SPMSM1 and SPMSM2. Section 4 describes the
force transferred to the ring gear is that which propels the vehicle
SPMSM1 and SPMSM2 efficiency maps development for power
(along with SPMSM2) and the assumption that the ring gear iner-
flow modeling. This section provides the SPMSM1 and SPMSM2
tia is negligible compared to the overall vehicle. We note (i)
efficiency map coefficients and maximum power expressions.
rp ¼ (rr  rs)/2, (ii) the planet carrier has the same angular velocity
as the ICE, i.e., xc ¼ xice, and (iii) the planet carrier inertia
includes the carrier itself plus that of the ICE, since they are A.5.1 SPMSM1 Parameters, Efficiency Map Coefficients, and
joined together and we assume that the planet inertia is small com- Maximum Power. Table 5 lists the SPMSM1 parameters. The
pared to that of the engine, and thus, the totality can be described SPMSM1 efficiency fit coefficients are given in Table 6. Addition-
with Jice. Further, we take Jp, mp, Js as negligible compared to the ally, SPMSM1’s maximum mechanical and electrical power are
vehicle and ICE inertias. After applying the previous and alge- needed. The maximum SPMSM1 mechanical power during pro-
braic manipulations, we obtain pelling and generating is modeled from given maximum torque
versus angular velocity information with the maximum power
xice rr þ rs
Jice ¼ Ts þ Tc (A17) modified by mildly extending the zero speed power to 1 kW, so
dt rs engine start up is possible from zero speed and adding a curve
segment centered at the torque region boundary at 60p rad/s to
dv rr c Pd þ Prr þ Pb þ P2;L  Pbrk obtain a continuous first derivative function of x1,m
mv ¼ Ts þ (A18)
dt rs v

Next, we relate the sun gear torque to the SPMSM1 power applied
Table 4 Toyota Prius mechanical power split device model
using Ts ¼ P1;L =ðjx1;m j þ es Þ; P1,L is divided by the absolute
parameters
value of x1,m to obtain the expected response of the ICE and vehi-
cle given that both P1,L and x1,m can take positive and negative Parameter Symbol Value
values. For example, (i) if P1,L is negative, power is being con-
sumed by SPMSM1 and one expects the ICE speed to go down Ring gear radius rr 78 teeth [32]
regardless of the sign of x1,m (assuming Tc is constant), and (ii) if Sun gear radius rs 30 teeth [32]
P1,L is positive, SPMSM1 is starting the ICE and its speed is Conversion factor c 14.2097 rad s/(m s) [32]
expected to increase regardless of the sign of x1,m. Finally, apply-
ing the expression for Ts, recognizing Tcxc ¼ Pice, recasting Eqs.
(A17) and (A18) into power flow equations and employing Lya- Table 5 SPMSM1 parameters
punov energy functions, Xice ¼ x2ice and !v ¼ v2, results in
  Parameter Symbol Value
1 Xice ðrr þ rs Þ P1;L
Jice ¼ xice þ Pice
2 dt rs jx1;m j þ es Self inductance L 0.1899 mH
Mutual inductance M 0.09497 mH
¼ Pice;psd þ Pice (A19) Magnet strength km 113 mV s
  Stator resistance Rs 39.81 mX
1 !v rr c P1;L Poles np 8
mv ¼ v þ Pd þ Prr þ Pb
2 dt rs jx1;m j þ es
þP2;L  Pbrk
¼ Pwhl;psd þ Pd þ Prr þ Pb þ P2;L  Pbrk (A20) Table 6 SPMSM1 efficiency map fit coefficients: i 5 1 for pro-
pelling, i 5 2 for generating, and R2 is R-square error
verifying Pice,psd and Pwhl,psd in Eqs. (A10) and (A11), Parameter ai bi ci di ei R2
respectively.
Additionally, an expression for x1,m is needed to determine Propel 0.9296 0.2191 0.1648 0.1969 0.2227 0.9985
Pice,psd and Pwhl,psd. Gears sharing a point of contact have the Generate 0.941 0.2243 0.175 0.2759 0.2843 0.9991
same tangential velocity at that point, thus x1,m is available from

Journal of Dynamic Systems, Measurement, and Control FEBRUARY 2018, Vol. 140 / 021002-9
Table 7 SPMSM2 parameters Table 9 SPMSM2 maximum mechanical power curve fit
coefficients
Parameter Symbol Value
r a b c  103 d  105
Self inductance L 0.21407 mH
Mutual inductance M 0.10703 mH 0.005 4.663  1010 0.3241 1.048 1.272
Magnet strength km 184 mV s 0.01 1.985  109 0.3057 2.105 2.956
Stator resistance Rs 30.5 mX 0.02 5.445  109 0.2713 3.862 6.438
Poles np 8 0.05 0.7528 0.2301 7.251 16.85

Pmax
1;L ðx1;m Þ where xs1 s2
8 2;m ¼ 1198p=30 rad=s; x2;m ¼ 1202p=30 rad=s, and
> 0:1547x1;m þ 1; 0 x1;m xs1 xu2;m ¼ 200p rad=s. Maximum mechanical power data at each
>
> 1;m
>
> ITSC fault level are obtained from the control simulations in Ref.
< 1:061  107 x3  0:1847x2 [3]. This maximum power data at each fault level are approxi-
1;m 1;m
¼ mated using a cubic equation with coefficients determined via a
>
>
3
þ69:68x1;m  6:544  10 ; xs1
1;m < x1;m xs2

Downloaded from http://asmedigitalcollection.asme.org/dynamicsystems/article-pdf/140/2/021002/6023978/ds_140_02_021002.pdf by guest on 04 July 2023


>
> 1;m
least-squares data fit
>
: 30:16; xs2
1;m < x1;m xu1;m
(A22) Pmax 2 3
2;L ðx2;m ; rÞ ¼ aðrÞ þ bðrÞx2;m þ cðrÞx2;m þ dðrÞx2;m (A25)

where xs1 s2
1;m ¼ 1798p=30 rad=s; x1;m ¼ 1802p=30 rad=s, and where Table 9 lists the fit coefficients found for each fault level.
xu1;m ¼ 1000p=3 rad=s. The maximum electrical power during The maximum electrical power during propelling (generating is
propelling (generating is not needed in the supervisory-level not needed in the supervisory-level model) is obtained from the
model) is obtained from the efficiency and maximum mechanical efficiency and maximum mechanical power
power
Pmax
2;L ðx2;m Þ
Pmax Pmax
2;inv ðx2;m Þ ¼  (A26)
1;L ðx1;m Þ gSPMSM2;prop Pmax
2;L ðx2;m Þ; x2;m
Pmax
1;inv ðx1;m Þ ¼  (A23)
gSPMSM1;prop Pmax
1;L ðx1;m Þ; x1;m

A.6 Electrical Bus. The battery pack, SPMSM1, and


A.5.2 SPMSM2 Parameters, Efficiency Map Coefficients, and
SPMSM2 electrical power flows come together in the DC–AC
Maximum Power. Table 7 lists the SPMSM2 parameters. The
Inverter. Both of the drives’ electrical power values include the
SPMSM2 efficiency fit coefficients are given in Table 8. Further,
inverter efficiency, which is assumed to also include any electrical
expressions for SPMSM2’s maximum mechanical and electrical
bus losses. Thus, the electrical bus is taken as having loss-less power
power are needed for the no-fault and ITSC-faulted cases. When
transfer efficiency, resulting in
there is no fault, the maximum mechanical power during propel-
ling and generating is modeled from given maximum torque ver- Pbat ¼ P1;e þ P2;e (A27)
sus angular velocity information where the maximum power is
modified by mildly extending the zero speed power to 3 kW, so
where P1,e and P2,e are the SPMSM1 and SPMSM2 electrical
vehicle movement is possible from zero speed and adding a curve
power values that include inverter efficiency, respectively.
segment centered at the torque region boundary at 40p rad/s to
obtain a continuous first derivative function of x2,m
A.7 Interconnection Equations. Interconnection equations
are constraints for each mode that relate the states, algebraic varia-
Pmax bles, and continuous control inputs. First, the SPMSM1 electrical
2;L ðx2;m Þ
8 and mechanical connections are
>
> 0:3761x2;m þ 3; 0 x2;m xs1
>
>
2;m (
>
> Pmax i
>
> 8:2869  105 x32;m  0:4783x22;m 1;e ðx1;m Þu1 ; i¼1
< Pi1;e ¼ (A28)
þ124:3x2;m  7:853  103 ; xs1 xs2 g1;inv gSPMSM1;gen ðPi1;L ; x1;m ÞPi1;L ; i ¼ 2; 3; 4
¼ 2;m < x2;m 2;m
>
>
>
> 1:848  105
x 2
>
> 2;m (
>
>
: 2:927  102 x2;m þ 53:65; xs2 xu2;m g1;inv gSPMSM1;prop ðPi1;e ; x1;m ÞPi1;e ; i ¼ 1
2;m < x2;m Pi1;L ¼ (A29)
i
Pmax
1;L ðx1;m Þu1 ; i ¼ 2; 3; 4
(A24)

Table 8 SPMSM2 efficiency map fit coefficients: R2 is R-square error

Parameter r ai bi ci di ei R2

Propel (i ¼ 1) 0 0.8903 0.4123 0.4701 0.0906 0.2086 0.9985


0.005 0.6983 3.411 11.05 0.337 1.476 0.9985
0.01 0.7917 3.624 28.67 1.894 23.13 0.9964
0.02 0.9296 0.2191 0.1648 0.1969 0.2227 0.9985
0.05 0.5619 14.44 158.4 2.412 33.52 0.9976
Generate (i ¼ 2) 0 0.9215 0.3676 0.4195 0.1217 0.2032 0.9995
0.005 0.5709 5.475 18.56 0.7304 3.514 0.9963
0.01 0.7302 4.732 26.14 1.771 13.57 0.9942
0.02 0.3138 18.33 123 3.585 37.28 0.9961
0.05 0.3327 26.34 323.5 9.353 154.9 0.9865

021002-10 / Vol. 140, FEBRUARY 2018 Transactions of the ASME


with x1,m available from Eq. (A21) and inverter efficiency, g1,inv, from Eq. (A8) by making P^whl ¼ P2;L þ Pwhl;psd and then
of 0.98. Note that the maximum electrical power here is the solving for P^whl using the shooting method such that the
inverter power divided by the inverter efficiency, i.e., output of Eq. (A8) tracks the square of the reference veloc-
Pmax max
1;e ðÞ ¼ P1;inv ðÞ=g1;inv . ity within a negligible error.
Next, the SPMSM2 electrical and mechanical connections are EO3) The estimated electrical power needed by SPMSM2 is less
( than or equal to what can be supplied by the battery alone
i Pmax i
2;e ðx2;m Þu2 ; i ¼ 1; 2; 3 over a time interval
P2;e ¼ (A30)
g2;inv gSPMSM2;gen ðPi2;L ; x2;m ÞPi2;L ; i ¼ 4 P2;e ðxr;ref ðDtÞÞ Pd;bnd ¼ 21 kW (A41)
bat
(
g2;inv gSMPSM2;prop ðPi2;e ; x2;m ÞPi2;e ; i ¼ 1; 2; 3 Consequent to the ICE off, Eq. (A37) is enforced; the power
Pi2;L ¼ i (A31) values immediately go to zero because the fuel from which power
Pmax
2;L ðx2;m Þu2 ; i¼4
is derived is no longer input.
The engine is started if the reference velocity is nonzero and
where the inverter efficiency is g2,inv ¼ 0.98. Note that the maxi- nondecreasing and any of the following conditions are met:
mum electrical power here is the inverter power divided by the
ES1) The battery SOC is below a threshold at tp,0

Downloaded from http://asmedigitalcollection.asme.org/dynamicsystems/article-pdf/140/2/021002/6023978/ds_140_02_021002.pdf by guest on 04 July 2023


inverter efficiency, i.e., Pmax max
2;e ðÞ ¼ P2;inv ðÞ=g2;inv .
The electrical bus connections for each mode are
W bat ðtp;0 Þ < W bat;iceon ¼ 0:50 (A42)
Pibat ¼ Pi1;e þ Pi2;e (A32) ES2) The estimated wheel power power needed to meet the ref-
erence velocity is greater than what can be supplied by the
The power split device mechanical power connections are traction motor, SPMSM2, alone from tp,0 to tp,f
!
i ðrs þ rr Þ Pi1;L Pmax ^
2;L ðxr;ref ðDtÞÞ < P whl ðDtÞ (A43)
Pice;psd ¼ xice ; i ¼ 1; 2; 3; 4 (A33)
rs jx1;m j þ es
ES3) The estimated electrical power needed by SPMSM2
! exceeds that can be supplied by the battery alone over Dt
rr c Pi1;L
Piwhl;psd ¼  v; i ¼ 1; 2; 3; 4 (A34)
rs jx1;m j þ es
P2;e ðxr;ref ðDtÞÞ > Pd;bnd
bat ¼ 21 kW (A44)
Finally, the frictional braking power interconnection equations Upon the aforementioned conditions being met to start the
are engine, the ICE is driven up to speed by requiring that it have the
minimum ICE operating speed of 100p/3 rad/s after a certain
Pibrk ¼ Pmax i
brk ð!v Þubrk ; i ¼ 1; 2; 3; 4 (A35) time, Dice,start ¼ 0.25 s, i.e.,
 2
100p
Xice ðtice Þ  ; tice  tp;0 þ Dice;start (A45)
30
A.8 Engine Operation. The Prius ICE may transition
between off and on during vehicle operation. Practically, when and allowing uifuel ; Pfuel ; Pice  0. Once the engine has started and
the engine is off, only modes 1 and 4 are possible; modes 2 and 3 reached minimum operating speed, mode 1 is no longer permitted,
should be unavailable until the engine has finished start up and is a1 ¼ 0, until the engine is off again.
on. Further, when the engine is off, no fuel usage or power output
is expected, thus References
[1] Gandhi, A., Corrigan, T., and Parsa, L., 2011, “Recent Advances in Modeling
a2 ; a3 ¼ 0 (A36) and Online Detection of Stator Interturn Faults in Electrical Motors,” IEEE
Trans. Ind. Electron., 58(5), pp. 1564–1575.
Pfuel ; Pice ; uifuel ¼ 0; i ¼ 1; 2; 3; 4 (A37) [2] Johnson, S., 2016, “Observability and Observer Design for Switched Linear
Systems,” Ph.D. thesis, Purdue University, West Lafayette, IN.
[3] Johnson, S. C., Meyer, R. T., DeCarlo, R. A., and Pekarek, S., 2016, “Fault
The engine is turned off if all of the following conditions are Detection in Surface PMSM with Applications to Heavy Hybrid Vehicles,”
satisfied: Purdue University, West Lafayette, IN, Technical Report No. 472.
[4] Campos-Delgado, D., Espinoza-Trejo, D., and Palacios, E., 2008, “Fault-
EO1) Either the battery SOC has reached the nominal value Tolerantolerant Control in Variable Speed Drives: A Survey,” IET Electr.
Power Appl., 2(2), pp. 121–134.
nom [5] Bolognani, S., Zordan, M., and Zigliotto, M., 2000, “Experimental Fault-
W bat ðtp;0 Þ  W bat ¼ 0:58 (A38) Tolerant Control of a PMSM Drive,” IEEE Trans. Ind. Electron., 47(5), pp.
1134–1141.
or the reference velocity is decreasing at a rate below a [6] Parsa, L., and Toliyat, H. A., 2007, “Fault-Tolerantolerant Interior-Permanent-
threshold over a time interval Magnet Machines for Hybrid Electric Vehicle Applications,” IEEE Trans. Veh.
Technol., 56(4), pp. 1546–1552.
[7] Diallo, D., Benbouzid, M. E. H., and Makouf, A., 2004, “A Fault-Tolerant Con-
dvref ðDtÞ trol Architecture for Induction Motor Drives in Automotive Applications,”
< aiceoff ¼ 0:5 m=s2 (A39)
dt IEEE Trans. Veh. Technol., 53(6), pp. 1847–1855.
[8] Jeong, Y.-S., Sul, S.-K., Schulz, S. E., and Patel, N. R., 2005, “Fault Detection
where Dt 僆 [tp,0, tp,f]. and Fault-Tolerant Control of Interior Permanent-Magnet Motor Drive System
for Electric Vehicle,” IEEE Trans. Ind. Appl., 41(1), pp. 46–51.
EO2) The estimated power needed to meet the reference velocity [9] Akrad, A., Hilairet, M., and Diallo, D., 2011, “Design of a Fault-Tolerant Con-
is less than or equal to what can be supplied by SPMSM2 troller Based on Observers for a PMSM Drive,” IEEE Trans. Ind. Electron.,
over a time interval 58(4), pp. 1416–1427.
[10] Raisemche, A., Boukhnifer, M., Larouci, C., and Diallo, D., 2014, “Two Active
^ Fault-Tolerant Control Schemes of Induction-Motor Drive in EV or HEV,”
Pmax
2;L ðxr;ref ðDtÞÞ  P whl ðDtÞ (A40) IEEE Trans. Veh. Technol., 63(1), pp. 19–29.
[11] Phillips, A. M., Jankovic, M., and Bailey, K. E., 2000, “Vehicle System Con-
where xr,ref(Dt) ¼ cvref(Dt). The estimated power, P^whl is troller Design for a Hybrid Electric Vehicle,” IEEE International Conference on
assumed to be piecewise-constant over Dt. It is obtained Control Applications (CCA), Anchorage, AK, Sept. 25–27, pp. 297–302.

Journal of Dynamic Systems, Measurement, and Control FEBRUARY 2018, Vol. 140 / 021002-11
[12] Xu, L., Li, J., Ouyang, M., Hua, J., and Li, X., 2010, “Active Fault Tolerance [22] Muta, K., Yamazaki, M., and Tokieda, J., 2004, “Development of New-
Control System of Fuel Cell Hybrid City Bus,” Int. J. Hydrogen Energy, Generation Hybrid System THS II-Drastic Improvement of Power Performance
35(22), pp. 12510–12520. and Fuel Economy,” SAE Paper No. 2004-01-0064.
[13] Meyer, R., DeCarlo, R., Meckl, P., Doktorcik, C., and Pekarek, S., 2011, [23] Uthaichana, K., 2006, “Modeling and Control of a Parallel Hybrid Electric
“Hybrid Model Predictive Power Flow Control of a Fuel Cell-Battery Vehicle,” Vehicle,” Ph.D. thesis, Purdue University, Lafayette, IN.
American Control Conference (ACC), San Francisco, CA, June 29–July 1, pp. [24] Guzzella, L., and Amstutz, A., 1998, “Control of Diesel Engines,” IEEE Con-
2725–2731. trol Syst. Mag., 18(5), pp. 53–71.
[14] Meyer, R. T., DeCarlo, R. A., and Pekarek, S., 2016, “Hybrid Model Predictive [25] Mansour, C., and Clodic, D., 2012, “Dynamic Modeling of the Electro-
Power Management of a Battery-Supercapacitor Electric Vehicle,” Asian J. Mechanical Configuration of the Toyota Hybrid System Series/Parallel Power
Control, 18(1), pp. 150–165. Train,” Int. J. Autom. Technol., 13(1), pp. 143–166.
[15] Uthaichana, K., DeCarlo, R., Bengea, S., Pekarek, S., and Zefran, M., 2011, [26] Agarwal, V., Uthaichana, K., Decarlo, R. A., and Tsoukalas, L. H., 2010,
“Hybrid Optimal Theory and Predicitive Control for Power Management in “Development and Validation of a Battery Model Useful for Discharging and
Hybrid Electric Vehicle,” J. Nonlinear Syst. Appl., 2(1–2), pp. 96–110. Charging Power Control and Lifetime Estimation,” IEEE Trans. Energy
[16] Kim, K.-H., Choi, D.-U., Gu, B.-G., and Jung, I.-S., 2010, “Fault Model and Convers., 25(3), pp. 821–835.
Performance Evaluation of an Inverter-Fed Permanent Magnet Synchronous [27] Johnson, V., 2002, “Battery Performance Models in ADVISOR,” J. Power
Motor Under Winding Shorted Turn and Inverter Switch Open,” IET Electr. Sources, 110(2), pp. 321–329.
Power Appl., 4(4), p. 214. [28] Markel, T., Brooker, A., Hendricks, T., Johnson, V., Kelly, K., Kramer, B.,
[17] Meyer, R., DeCarlo, R., and Meckl, P., 2013, “Hybrid Model Predictive Power O’Keefe, A., Sprik, S., and Wipke, K., 2002, “Advisor: A Systems Analysis
Management of a Fuel-Cell Battery Vehicle,” Asian J. Control, 15(2), pp. Tool for Advanced Vehicle Modeling,” J. Power Sources, 110(2), pp. 255–266.

Downloaded from http://asmedigitalcollection.asme.org/dynamicsystems/article-pdf/140/2/021002/6023978/ds_140_02_021002.pdf by guest on 04 July 2023


363–379. [29] Wiegman, H. L. N., and Vandenput, A. J. A., 1998, “Battery State Control
[18] Meyer, R. T., Zefran, M., and DeCarlo, R. A., 2014, “A Comparison of the Techniques for Charge Sustaining Applications,” SAE Paper No. 981129.
Embedding Method With Multiparametric Programming, Mixed-Integer Pro- [30] Nam, E., 2004, “Advanced Technology Vehicle Modeling in PERE,” U.S.
gramming, Gradient-Descent, and Hybrid Minimum Principle-Based Methods,” Environmental Protection Agency, Washington, DC, Report No. EPA420-D-
IEEE Trans. Control Syst. Technol., 22(5), pp. 1784–1800. 04-002.
[19] Bengea, S. C., and DeCarlo, R. A., 2005, “Optimal Control of Switching Sys- [31] Meisel, J., 2006, “An Analytic Foundation for the Toyota Prius THS-II Power-
tems,” Automatica, 41(1), pp. 11–27. train With a Comparison to a Strong Parallel Hybrid-Electric Powertrain,” SAE
[20] Krause, P., Wasynczuk, O., Sudhoff, S., and Pekarek, P., 2013, Analysis of Elec- Paper No. 2006-01-0666.
tric Machinery and Drive Systems, 3rd ed., Wiley-IEEE Press, Piscataway, NJ. [32] Staunton, R., Ayers, C., Marlino, L., Chiasson, J., and Burress, T., 2006,
[21] Michalska, H., and Mayne, D., 1995, “Moving Horizon Observers and “Evaluation of 2004 Toyota Prius Hybrid Electric Drive System,” Oak Ridge
Observer-Based Control,” IEEE Trans. Autom. Control, 40(6), pp. 995–1006. National Laboratory, Oak Ridge, TN, Report No. ORNL/TM-2006/423.

021002-12 / Vol. 140, FEBRUARY 2018 Transactions of the ASME

You might also like