This paper proposes a combination of neural networks with fuzzy logic to produce a fuzzy controller with fast response. The fuzzy neural network uses neural nodes and links to connect input variables to fuzzifiers, to define fuzzy rules and combine outputs into a control action. This new method has been applied to the problem of fuzzy control of a shunt active power filter designed for harmonic current elimination. The performance of the proposed method for training and test data is examined by Matlab simulations and is found to optimize the fuzzy control performance.
A large amount of harmonic current is generated in power systems due to low power based
electronic appliances such as TV sets; personal computers; adjustable speed drives,
temperature and light controllers, solid state AC voltage controllers, UPS, SMPS, etc. The
increase in usage of such nonlinear equipment has caused degradation of power and utility
conditions. Harmonic currents in particular, are receiving more attention as a critical power
quality concern, with an estimated 60% of the electricity now passing through nonlinear
loads. Ironically, the equipment used to boost productivity and efficiency is also increasing
nonproductive power consumption, harmonic distortion, and low power factor. Moreover,
the equipment producing harmonic distortion is also highly susceptible to its own damaging
effects. Active Power Filter (APF) is basically designed on the concept of injecting equal, but
opposite distortion to a system in order to compensate for the harmonics in supply voltage
and/or current. Numerous types of active power filters have been proposed over the years [1]-[3].
Shunt active power filters, using different control strategies have been widely investigated to
provide a viable solution to the problems created by non-linear loads. The main objective of
active power filters is the fast and precise detection of the load current’s harmonic components.
The heart of the control strategy is the control algorithm, responsible for the high level transient and steady state performance of the APF. The control algorithms are implemented
through analog controllers, low-cost microcontrollers, fast digital signal processors,
Application Specific Integrated Circuits (ASIC), depending on the rating and requirements
of the filter. Fuzzy control of Active Power Filter (APF) has also been investigated [4] and
found to be a better alternative to the conventional Proportional-Integral (PI) control. It has
been observed that during the design process, the parameters of the PI control system need to
be adjusted from those derived in the original controller model. Automated tuning is better
suited for both quicker design turnaround time, and an ability to adapt to the observed
characteristics of the control system. Fuzzy controllers are more efficient in meeting the
stringent requirements of dynamic response of this multifunction filter. |
