application of learning methodologies in control of power electronics drives
DESCRIPTION
Application of Learning Methodologies in Control of Power Electronics Drives J. L. da Silva Neto, L.G. Rolim , W. I. S uemitsu, L. O. A. P. H enriques , P.J. C osta Branco, M . G. S imões. Presentation. Introduction Artificial Intelligence Fuzzy Control of Synchronous Motors - PowerPoint PPT PresentationTRANSCRIPT
Application of Learning Methodologies in Control of Power Electronics Drives
J. L. da Silva Neto, L.G. Rolim, W. I. Suemitsu,
L. O. A. P. Henriques, P.J. Costa Branco,
M. G. Simões
Presentation
• Introduction
– Artificial Intelligence
• Fuzzy Control of Synchronous Motors
• Learning controller for torque ripple reduction in SR
drives
• Conclusions
Introduction
Learning methodologies:• Expert Systems: knowledge represented at a
symbolic level.• Genetic Algorithms: Computational models based
on the theory of evolution (fitness, mutation and reproduction)
• Fuzzy Logic Systems: linguistic technique• Neural Networks: mathematical model of artificial
neurons
Fuzzy Logic and/or Neural Networks
• Some characteristics:– Capability to map information with degree of
imprecision– Parameter estimation (torque and flux)– In last ten years, industrial drives have found
a profound influence of Fuzzy and Neural systems.
– Replacement of classical controllers by controllers with learning methodologies.
Electrical Drive Control
• Evolution of digital signal processors, and circuit integration, make possible the implementation of complex control
• Several types of motors can use the features of “Intelligent” drives:
– AC machines (Induction Motors, Synchronous Machine)
– Switched Reluctance Motors
Fuzzy Control of Synchronous Motors
• Fuzzy Logic Adaptation Mechanism (FLAM)– Objective is to change the rules definitions in the
fuzzy logic controller (FLC) base table, according the comparison between the reference model and the system output.
– Composed by a fuzzy inverse model and knowledge base modifier
– I was used to prove the effectiveness of the control in a TMS320C30 DSP-based speed fuzzy control of a permanent magnet synchronous motor (PMSM)
Ke
z-1
e
e
+
Ke
PMSM
ReferenceModel W
m
+
r
K
+
em
emKm Km
z-1
FuzzyInverseModel
+
KuFuzzy
Controller
N
1ii
oldumlu
K
1z
z
Fuzzy Control of Synchronous Motors
Fuzzy Control of Synchronous Motors
T (s)
i q (
A)
e m (
rad/
s)
,
m r
ad/s
Fuzzy Control of Synchronous Motors
i q (
A)
T (s)
e m (
rad/
s)
,
m r
ad/s
Neuro Fuzzy Control of SR Drives
Low cost (material and manufacturing)
Good thermal behavior Fault tolerance Reliable Easy to repair
Torque ripple Nonlinear model
Neuro Fuzzy Control of SR Drives
• Input signals– Motor speed– Rotor position– Reference current
• Output: current increment (I)• Training signal: oscillating torque
Neuro Fuzzy Control of SR Drives
PI C ontro ller++
N euro-Fuzzy B lock
C onverter+
M otor
i
ipi
iref re f
+-
1/s
filte r
e lT~
Neuro Fuzzy Control of SR Drives
Neuro Fuzzy Control of SR Drives
0 50 100 150 200 250 300 350 400 450 5000
2
4
6x 10
-4 no load
with trainningwithout trainning
0 50 100 150 200 250 300 350 400 450 5000
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
torq
ue
no load
ms
with trainningwithout trainning
Conclusions
• The adaptive fuzzy strategy presented applied for PMSM drives has proved to be very effective when applied for motion control applications.
• It has been implemented on a speed control of a PM motor, it can be extended for other kinds closed loop motor control.
• One highlighted characteristic of this algorithm is that it can compensate non-linear load variations without the need of a completely modeled load.
Conclusions
• For the SR drive the neuro-fuzzy strategy has shown to be effective to reduce torque oscillations
• The adaptive algorithm automatically learns a current profile without the need of observers and state estimators.