Review and Improvement of Several Optimal Intelligent Pitch Controllers and Estimator of WECS via Artificial Intelligent Approaches - Complex System Modelling and Control Through Intelligent Soft Computations - page 509

wind

WIND

Wind input
For
Fuzzy if-then
r ule detection

Tm

WIND TURBINE

Control signal
pitch angle

Fuzzy if-then
rule box

Pcap

Fuzzy if-then rule
detector

Fuzzy if-then
rule

Fig. 6 Block diagram of controllers GFS controller diagram

values. The number of removed fuzzy if-then rules is usually the same as that of
added rules in classic genetic algorithm. Figure 6 shows block diagram of the
proposed controller.
According to block diagrams of GFS controller in Fig. 6 controller has two
inputs and one output. This controller provides a suitable pitch angle upon catch
wind speed.

6.5 Simulation Results of GFS Controller

Figure 7 a depicts the established generator torque for GFS controller. The output
mechanical power of GFS is shown in Fig. 7 b.
Figure 7 d shows the blade pitch angle control signal for GFS controller. In
below rated wind speed, optimal power is attained by regulating (optimal perfor-
mance coef

cient); thus, the pitch angle is kept at a mechanical minimum and rotor
speed is controlled in such a way that (optimal tip speed ratio) is always acquired,
akin to MPP tracking.

7 Rule Extraction

The idea of rule extraction from a neural network involves de nite activity. For

example, parameters have read from a network, which is not permitted by a tra-
ditional connectionist structure that these neural networks are established on Gallant
( 1998 ). Gallant presented a routine for extracting propositional rules from a simple
network, in order to explain the inference process in the system at the end of 1980s
Gallant ( 1998 ). This could be surveyed as the origin of the examination on rule
extraction from neural networks. In addition some algorithms have been developed
to extract deterministic

finite-state automata (DFA) from recurrent neural networks
(Giles and Omlin 1993 ; Giles et al. 1992 ).
Rules from each unit in a neural network have been extracted by the decom-
position algorithms and these methods aggregate them. The pedagogical algorithms