Professional Documents
Culture Documents
Streszczenie. W artykule zaproponowano metode sterowania wektorowego trjfazowym silnikiem indukcyjnym z wykorzystaniem regula-
tora rozmytego Takagi-Sugeno zastepuj acego
klasyczne regulatory PI. Przeprowadzono badania symulacyjne i eksperymentalne, werykujace
dziaania sterowania. (Sterowanie wektorowe trjfazowa maszyna indukcyjna z wykorzystaniem logiki rozmytej).
skutecznosc
emf d
2
1
Fig. 4. The structure of a fuzzy logic current controller.
2
s
: offset angle Membership Functions
The Membership Functions (MFs) of the TSF controller
s inputs are shown in Figure 5 and in Figure 6.
Fig. 3. Vector diagram showing the emf and ux linkage relationship. The universe of discourse for the input Eids is dened in
the closed interval [-0.5, 0.5]. The extreme MFs have trape-
Assuming that the magnitude of the feedback vector 2
zoidal shapes but the middle MF takes triangular shape as it
is increased to 2 as shown in Figure 3 due to a dc offset is shown in Figure 5. However, the universe of discourse for
or initial value problem, the phase angle will be greater the input Eiqs is dened in the closed interval [-10, 10] as is
that 90. The quadrature detector will generate a negative shown in Figure 6. The shapes of these MFs are similar to
error signal. The output of the PI regulator cmp is reduced the rst input. For both inputs the linguistic labels are N, Ze
and so is the feedback vector. As a result, the ux vector and P that means Negative, Zero and Positive, respectively.
s
moves toward the original position of 90until the orthogonal
relationship between s and emf
is reestablished. If is less 1
N ZE P
0.5 0.4 0.3 0.2 0.1 0 0.1 0.2 0.3 0.4 0.5
Aggregation
The nal output value uds inferred from n = 9 implica-
N ZE P
1
Ri
0.8 tions is aggregated as the average of all Vds with the weights
Ri
Membership degree
0.6
:
0.4
Ri
ni=1 Ri Vds
0.2
(17) uds = Ri
0
ni=1 Vds
and the nal output value uqs inferred from n = 9 im-
10 8 6 4 2 0 2 4 6 8 10
Iq current error (EIq)
Ri
plications is aggregated as the average of all Vqs with the
Fig. 6. Membership function for the quadrature-axis component of
Ri
the stator current error Eiqs . weights :
ni=1 Ri Vqs
Ri
RV1 ds : if Eids is N and Eiqs is N then (18) uqs =
ni=1 Vqs
Ri
R1
Vds = a Eids + b Eiqs
In our TSF controller, that is Sugeno type, it is not nec-
However, the set of rules for the quadrature component of essary the defuzzication interface [23]. This is so because,
the stator voltage uqs are dened by the rules of the following in Takagi-Sugeno fuzzy controllers, each rule is already crisp
form: and the total result is determined by the weighted sum of
each rule, as is shown in equations (17) and (18). The TSF
controller was programmed in C programming language for
RV1 qs : if Eids is N and Eiqs is N then the simulation because it facilitates its implementation in the
digital signal processor TMS320F28335 from the Texas In-
R1
Vqs = b Eids + a Eiqs
strument.
The a and b coefcients of the rst-order polynomial Simulations and Experimental Results
function typically present in the consequent part of the rs- The simulations activities were performed using MAT-
order Takagi-Sugeno fuzzy controllers. Observe that these LAB R2011b simulation package together with Simulink block
R1 R1
coefcients for Vds and Vqs are the same but with inter- sets and fuzzy logic toolbox. The switching frequency of the
changed order, being it not necessary other different coef- three-phase two level inverter was set to be 10kHz, the direct
cients for each one, as they are repeated for all the rules. For component of the stator current (ids ) was set to be 1.0 pu.
instance, when the consequent part of the rule is a real num- The experimental activities are realized with electronic
ber, we have a zero-order controller, but if the consequent is circuits and three-phase IM. The experimental set-up con-
a linear combination we have a rst-order controller [21]. The sists of a DSP (Texas Instruments TMS320F28335) con-
complete rule base to calculate uds and uqs is shown in Ta- nected to a three-phase squirrel cage IM, driven by a 12kVA
ble 1, where (a=5; b=0.1), (c=6.5; d=0.2), (e=8; f=0.1). The Semikron three-phase inverter (SKS 32F B6U+E1CIF+B6CI
shape and the universe of discourse of the MFs and these co- 12 V06). The mechanical load is imposed by a Foucault brak-
efcients were adjusted through the simulation process using ing system. Conditioning signal boards are necessary to ac-
the trial and error method. quire the motor currents and the Dc link voltage from a high
level of voltage and current to an appropriate voltage level to
Inference Method
be sampled and converted by the internal AD converter.
In general, operators on fuzzy sets use triangular norms,
In order to investigate the effectiveness of the proposed
which may be divided into T-norms (AND operators) and S-
control system and in order to check the closed-loop stabil-
norms (OR operators) [22]. T-norms perform an intersection
ity of the complete system, we performed several tests. The
operation on fuzzy sets and have a particular importance in
simulated scenarios shown in this article covers the following
fuzzy logic control. T-norm is usually denoted as T (a,b). The
situations: trapezoidal prole reversion, triangular prole re-
T-norms used in our TSF controller is the product dened as:
version and step change in speed reference, the three tests
were made with no-load. Besides, the step change in the
(16) Ri = T (iEi , iEiqs ) = iEi iEiqs motor load (from 0 to 1.0 pu and from 0 to 0.5 pu) at forty
ds ds percent of rated speed was made. The parameters of the
for i = 1, ..., n; n = 9 three-phase IM under test is shown in the appendix section.
Figure 7(a) shows the response of rotor angular speed
where iEi and iEi are MFs degrees of the rst and when a trapezoidal prole reversion was applied to the motor
ds qs
second TSF controller inputs, respectively, and Ri is the with no-load. In this test the rotor speed tracked the reference
truth value of the preposition. as expected. Similar behavior is observed in experimental
0.6
0.4
r r r
0.2
(pu)
r
0 C2
C3
r
0.2
0.4
0.6
0.8
1 1.5 2 2.5 3 3.5 4 4.5
Fig. 7. Simulation (a) and Experimental (b) results of r when trapezoidal prole reversion is applied with no-load (C2,C3: 0.2 pu/div, 5s/div).
0.8
0.6
0.4
0.2 r
(pu)
C2
r
0
r C3
r
0.2
r
0.4
0.6
0.8
1 1.5 2 2.5 3 3.5 4 4.5 5 5.5
Fig. 8. Simulation (a) and Experimental (b) results of r when triangular prole reversion is applied with no-load (C2,C3: 0.2 pu/div, 5s/div).
0.8
0.6
r
r
0.4
0.2 r
(pu)
0 C2
r C3
r
0.2
0.4
0.6
0.8
1 2 3 4 5 6 7 8
Fig. 9. Simulation (a) and Experimental (b) results of r when step change in the speed reference is applied with no-load (C2,C3: 0.2 pu/div,
5s/div).
3 ids
ids (pu), iqs (pu)
2
ids
1
0
C2 iqs
1
2 iqs
C3
3
4
1 1.5 2 2.5 3 3.5
Fig. 10. Simulation (a) and Experimental (b) results of ids and iqs when the rated load is applied (C2,C3: 0.5 pu/div, 5s/div).
4
r
ids (pu), iqs (pu), r (pu)
2
r
1 C1
ids
0
ids
1
C2
2
iqs iqs
3
C3
4
1 1.5 2 2.5 3 3.5
Fig. 11. Simulation (a) and Experimental (b) results of r and, ids and iqs when the 0.5 pu of rated load is applied (C1,C3: 0.5 pu/div; C2: 1
pu/div; 5s/div).