Académique Documents
Professionnel Documents
Culture Documents
PM
inverter
SM
θ
Id
ADC
DSP
Iq
encoder
Iq*
Haptic
interface
variables that extend results to switching systems. Fig. 3 – Overall regulation system
Moreover this model takes into account the problem of
the different frequencies induced in the system, such as III. PMSM MODEL AND CONTROL STRATEGY
switching/sampling frequencies and inner/outer loops time
periods. A. PMSM Model
Section II gives an overview of the system regulation. The PMSM electrical model, in a synchronous reference
Section III describes the PMSM model and its current frame is described in (1) and (2), where Ed and Eq are
regulation. Section IV develops a hybrid dynamic model of coupling terms. Rms is the stator winding resistance, Ld and
the studied system and section V presents simulations Lq the direct and quadrature axis inductances, whereas Vd,
comparing MATLAB-Simulink results, hybrid model Vq, Id and Iq are the direct and quadrature axis voltage and
results and experimental ones. current components. ϕf is the permanent magnet flux and
Cem the electromagnetic torque.
II. OVER-ALL REGULATION SYSTEM
d
Vd = Rms I d + Ld I d + Ed
The overall regulation system is shown in Fig 3. It dt
(1)
contains three major steps: d
- The first one consists in a current regulation (level 3) Vq = Rms I q + Lq I q + Eq
dt
realized in the DSP with a period Tn of 60 µs. This
Ed = −ω Lq I q
regulation is the image of the desired torque in a force (2)
feedback application, and as such, it must be as good Eq = ω Ld I d + ωϕ f
as possible.
- Cem =
2
3
(
p ϕ f − ( Lq − Ld ) I d I q ) (3)
- The second step is a speed regulation (level 2), also
done in the DSP with a period Tk of 240 µs. This
regulation is only a step to obtain a better force The PMSM mechanical equation is described in (4),
feedback, so it is not very important, and a simple where J and f are the inertia and the friction coefficient of
gain K is enough to control the system. It determines the system, Ω the mechanical speed, and Cr the load torque.
the desired current reference, which is thus updated
every 240 µs. dΩ
J + f .Ω = Cem − Cr (4)
dt
- The third step is realized in the haptic interface (level
1) with a period Ti of 1 ms. First, the reversing block
compensates the speed regulation realized in the Id
I*d Vd* 1/R ms
second step, so that the desired current – image of the PI Ld
desired torque in the force feedback application – is _ - 1+
R ms
p
reflected in the current regulation made in the first
Êd Ed
step. Then, a position and speed regulation is realized
with references obtained from the virtual
environment.
I*q Vq* 1/R ms Iq
PI Lq
The regulators gains and K are calculated to obtain the _ - 1+ p
R ms
desired dynamic in every loop, whereas Km and Bm, are
calculated depending on the desired transparency and Êq Eq
rigidity of the system. This paper only describes the current
regulation corresponding to the first step. In summary, the
current references are updated at a Tk period, and the Control system PMSM
current regulation is realized with a Tn period.
Fig. 4 – Current regulation loop
Electrimacs 2005, Hammamet, Tunisie, April 17-20 2005
B. PMSM control strategy - level 2: the speed regulation loop has a 240 µs time
period corresponding to the sampling period (Tk),
The transfer functions between the stator currents and - level 3: the current regulation loop has a 60 µs time
voltages of the PMSM are first order ones. The coupling period corresponding to the sampling period (Tn),
terms Ed and Eq can be estimated. Two PI regulators are - level 4: the SVPWM creates the duty cycles of the
used to control the current loop, as shown in the block inverter switches (Tn1, Tn2, Tn3), which change at
diagram in Fig. 4. each sampling period and define three new time
1 + τ i .s periods in the system.
C (s) = K (5)
τ i .s A. PMSM and inverter model
However, this regulation is realized in the DSP, and the The shortest sampling period is the current regulation
PI regulator must be discretised in order to implement them one (level 3). Using (1) and (2), a continuous dynamic state
in the DSP algorithm. space model, with the initial condition: X(tn) = Xn, can be
written as:
Tn Tn
τ in = K , Kn = K 1 + (6) dX / dt = A4 ( X ) . X + B4 .U n (7)
τi τi
tk = tn tn + 1 tn+ 2 tn +3 t k + 1 = tn + 4
S w itc h in g d u ty c y c les (T n 1 , T n 2 , T n 3 )
Tn3
T n2
T n1
tn = tn 0 tn1 tn 2 tn 3 t k+ 1 = t n + 4
t (s) 3 3
currents (A)
currents (A)
-15
0 0.005 0.01 0.015 0.02 0.025 2.5 2.5
2 2
1.5 1.5
5 1
Id*S (b) 1
0.5 0.5
Iq*S
0 t (s) 0 t (s)
5 5.2 5.4 5.6 5.8 6 5 5.2 5.4 5.6 5.8 6
currents (A)
-3 -3
x 10 x 10
4.7 4.7
Iq*H Iq*H
4.65 Iq*S 4.65 Iq*S
4.6 4.6
t (s)
currents (A)
-5 4.55
currents (A)
4.55
0 0.005 0.01 0.015 0.02 0.025
4.5 4.5
4.45 4.45
0
(c) 4.4 4.4
-20 4.35
4.35
electrical speed (rad/s)
-80
Fig. 8 – Comparing methods in transient and permanent mode
5
-100
Iq*S,0.25us
(a)
Iq*S,0.5us
-120 0
Iq*S,1us
t (s)
currents (A)
-140 -5
0 0.005 0.01 0.015 0.02 0.025
10 -15
V d*H (a)
-20 t (s)
V q*H 0 0.005 0.01 0.015 0.02 0.025
5
voltages (V)
4.8
0 0.25us
(b) 0.25us
(c)
4 0.5us 0.5us
4.7
1us 1us
-5 3 4.6
currents (A)
currents (A)
4.5
2
t (s)
-10 4.4
0 0.005 0.01 0.015 0.02 0.025
1
4.3
-5
4
-10
3.8
t (s)
-5 -15 3.6
0 0.005 0.01 0.015 0.02 0.025
3.4
-20
0 0.005 0.01 0.015 0.02 0.025 8.55 8.6 8.65 8.7 8.75
t (s) t (s) x 10
-3
0
(c)
-20 4.7
electrical speed (rad/s)
1pt
(c) 1pt (d)
-40 4 2pts 4.65 2pts
3pts 3pts
4.6
-60 3
currents (A)
currents (A)
4.55
-80 4.5
2
4.45
-100
1 4.4
-120
4.35
t (s)
-140 0 4.3
0 0.005 0.01 0.015 0.02 0.025 5 5.2 5.4 5.6 5.8 6 8.55 8.6 8.65 8.7 8.75
t (s) x 10
-3 t (s) x 10
-3
Fig. 7 – Hybrid System Modeling Results Fig. 10 – Hybrid modeling results for different number of points per state
Fig. 10 compares the results obtained with 1, 2 and 3 Fig. 11 gives an overview of the differences obtained
samples per state of the inverter. (b) shows that the between a 1µs simulation step with Simulink and a one
precision of the results is better when the number of points point per state of the inverter simulation with the HDS
between each state of the inverter is increased, however, (c) model. It compares in detail the current ripples obtained
and (d) show that results displayed with a linear with both simulations on a larger scale. (b) and (c) show the
interpolation are identical no matter the number of points. dynamic response of the system in the transient mode,
It is thus useless to increase the number of points to more while (d) and (e) compare results in the permanent mode.
than one point per state. It is important to note that in our The plots in the left column are sampled (discrete) results,
Electrimacs 2005, Hammamet, Tunisie, April 17-20 2005
while the right column shows the same results with a linear 5
(a)
interpolation. We can see that the precision of the Simulink IqS,1us
IqH,1pt
currents (A)
appear between the two methods. 0
currents (A)
currents (A)
the simulation is shorter. 2 2
1 1
Table I – Duration of the simulations
Simulation 0
t (s) 0
t (s)
0.25 0.5 1 5 5.2 5.4 5.6 5.8 6 5 5.2 5.4 5.6 5.8 6
step (µs) x 10
-3
x 10
-3
Simulink
Duration 4.7 4.7
21 9 5 IqH (d) IqH (e)
(s) 4.6
IqS
4.6
IqS
Number of
1 2 3
currents (A)
currents (A)
4.5 4.5
points
HDS
Duration 4.4 4.4
5 9 18
(s) 4.3 4.3
t (s) t (s)
4.2 4.2
8.55 8.6 8.65 8.7 8.75 8.55 8.6 8.65 8.7 8.75
In order to validate both simulations, Fig. 12 shows x 10
-3
x 10
-3
experimental results of the system concerning the current Fig. 11 – Comparing methods in transient and permanent mode
loop with a 60µs sampling period. (c) shows that the 3
Id
(a)
0
(b) shows the dynamic response of the system with a -1
2.6
1.5
responses. 0.5
2.2
0
VI. CONCLUSION
0.05 0.052 0.054 0.056 0.058 0.06 0.0855 0.086 0.0865 0.087 0.0875