development and analysis of a self-tuned neuro-fuzzy controller for induction motor drives
abstract
induction motors (im) have been widely utilized in industry for variable speed
drives due to some of their advantages, such as rugged construction, low cost and
reliable service with easy maintenance, as compared to conventional dc motors. for
variable speed drive applications, the controller plays an important role so that the
motor can follow the reference trajectories without any significant deviation.
furthermore, a controller which can provide fast speed response and handle
uncertainties and disturbances, is absolutely necessary for high performance drive
systems. traditionally, fixed gain proportional-integral (pi) and some adaptive
controllers have been utilized in industry for a long time. however, there are some
disadvantages of these controllers to handle uncertainties which are inherent to a
nonlinear im. as a result, recently researchers paid their attention to apply intelligent
algorithms to control the im for high performance variable speed drive applications.
intelligent algorithms such as fuzzy logic (fl), neural network (nn), neuro-fuzzy
(nf), etc, have inherent advantages as compared to the conventional controllers.
in this thesis, a novel neuro-fuzzy controller (nfc) has been developed for speed
control o f em. for the complete drive, the indirect field orientation control is utilized
in order to decouple the torque and flux controls. thus, the induction motor can be
controlled like a dc motor and hence the high performance can be achieved without
lacking the advantage o f ac over dc motors. the proposed neuro-fuzzy controller
incorporates sugeno model based fuzzy logic laws with a five-layer artificial neural network (ann) scheme. the controller is designed for low computational burden,
which will be suitable for real-time implementation. furthermore, for the proposed
nfc an improved self-tuning method is developed based on the im theory and its
high performance requirements. the main task o f the tuning method is to adjust the
parameters o f the fuzzy logic controller (flc) in order to minimize the square of the
error between actual and reference output. in this thesis, a model reference adaptive
flux (mraf) observer is also developed to estimate the d-axis rotor flux linkage in
both constant flux and flux weakening regions based on motor voltage, current and
reference trajectories for flux linkage. thus, it provides safe operation to control the
motor at high speeds, especially, above the rated speed. the d-axis reference flux
linkage of the indirect field oriented control is provided by flux weakening method.
furthermore, a proportional-integral (pi) based flux controller is used to provide the
compensation for the reference flux model by comparing the flux reference and the
observed flux from gopinath model flux observer. a complete simulation model for
indirect field oriented control of im incorporating the proposed mraf observer
based nfc is developed in matlab/simulink. in order to prove the superiority of the
proposed controller, the performance of the proposed controller is compared with a
conventional pi as well as fuzzy logic controller (flc) based im drives. the
performance of the proposed im drive is investigated extensively at different
operating conditions in simulation. the performance of the proposed mraf
observer based nfc controller is found robust and a potential candidate for high
performance industrial drive applications.
collections
- retrospective theses [1604]