Fuzzy Hierarchical Architecture with Additional Deffuzufication Leyer and Applications to the Power Quality Diagnosis
Hierarchical fuzzy systems, generic inference, T-norms and overlap functions,
diagnosis, power quality, wavelet-packet, renewable energy.
Among various existing decision-making methods, hierarchical fuzzy methods have
emerged as a suitable tool for dealing with complex applications which have many input
variables and a high degree of subjectivity. In this context, the product of electric
energy stands out. In general, the diagnosis of energy quality is a difficult practice due
to the subjectivities inherent to the analysis process, nuances among different standards
existing in the world, and uncertainties of evaluation parameters. This thesis proposes a
new methodology for the power quality diagnosis based on the hierarchical fuzzy theory
with a cascade-type architecture. The proposed method analyzes the quality parameters
in steady-state electrical systems based on different existing standards in the world and
performs a linguistic/quantitative diagnosis in which the contributions of the analyzed indices
are weighted on the power quality of the evaluated system. Firstly, the diagnosis
method was implemented from two hierarchical fuzzy architectures known (conventional
and defuzzification free). Posteriorly, a new proposed architecture with additional
defuzzification of layers was developed to aggregate the main advantages of conventional
and defuzzification free in order to make the diagnosis method more complete and
robust. This study proposes that the output of each subsystem obtained from primary
decision-making process is transferred directly between the hierarchical layers, without
loss of linguistic information, to obtain a resultant power quality diagnosis. In addition,
a secondary decision-making process is performed together with an additional defuzzification
method in order to obtain a complementary specific diagnosis at the out of each
hierarchical subsystem. The diagnosis method based on the proposed fuzzy architecture
presented satisfactory results when compared with the two existing architectures. After
validation of the diagnosis method and hierarchical fuzzy architecture, both presented in
this thesis, at the end of research, an adaptive wavelet-fuzzy system with generic inference
method based on extended overlap functions is proposed as a new tool able of monitoring
the power quality in renewable energy systems.