Elements of Fuzzy Logic in Protective Relays - Digital Signal Processing in Power System Protection and Control - page 228

Fig. 11.7 General
architecture of a fuzzy
protection relay or control
unit

Input
signals

Fuzzification
block

Fuzzy
criteria and
decision
signals

Fuzzy reasoning/
fuzzy signal
processing

Fuzzy
settings

Defuzzification
block

Decision

''high-pass'' depending on the coefficients), while the singleton MF is a point-type
function being non-zero at one point only.
Choosing the type of the MFs for given application is a matter of the designer
experience and good understanding of the phenomena under consideration. The
parameters of MFs are usually set on basis of simulation and/or field-recorded
signals and can be further optimized, e.g. with an evolutionary/genetic procedure.
Practical utilisation of the fuzzy set theory as well as the rules of operations on
fuzzy numbers lead to the general scheme of the fuzzy protection/control device
presented in Fig. 11.7 . The following main blocks can be distinguished here [ 6 ]:

• fuzzification, where the real input signals are converted into their fuzzy coun-
terparts (fuzzy numbers),
• fuzzy reasoning, where the fuzzy criteria signals are processed and—after
comparison with fuzzy settings—some fuzzy decision/output signals are

generated,
• defuzzification, which is understood as conversion of the fuzzy outputs into
crisp numbers (real output signal or decision).

11.2 Boolean Versus Fuzzy Logic

It is common that most of the relay decisions to be taken in protection devices are
of discrete type (0-1). The relay output is either blocking/stabilization that is
issued when the protected plant is healthy (value 0) or tripping/alarm, which is a
proper decision for all internal fault cases (value 1). There are no intermediate
cases, i.e. anything halfway between 0 and 1. In many cases, however, taking
proper decision is difficult due to many reasons and sometimes the ''don't know''
expression would be the best if the premise for given hypothesis is not certain or
the measured value is very close to the decision boundary.