|
On-line fault diagnosis of substation operation is inevitable
in the automation of power system operation. This thesis pro-
poses a new connectionist expert diagnostic system with an
auxiliary artificial neural networks (ANN) diagnostic system
for on-line fault diagnosis of power substation. The
connectionist expert diagnostic system has similar profile of
an expert system, but can be constructed much more easily
from elemental samples. These samples associate fault type with
its primary and secondary protective relays and breakers. This
system can be applicable to the power system control center for
single- or multiple-fault type estimation, even in the cases of
failure operation of relay and breaker, or error-existent data
transmission. Besides , the confidence level of the diagnostic
conclusion and explanation of the answer can be provided to the
users. The auxiliary ANN diagnostic system employs a
hierarchical neu- ral networks structure to help estimate
the fault type when definite conclusion or conclusion with
high confidence level can not be provided by the connectionist
expert system.The peak values of the fault voltage and
current waveforms are used as the input information of the
auxiliary ANN diagnostic system. To improve the efficiency of
the traditional back-propagation training process for large-
size analog ANN,a cascade-correlat- ion learning algorithm
is applied to support the structural implementation of the
system. The proposed approach has been practically verified by
testing on a typical Taipower secondary substation. The test
results, although preliminary, suggest our system can be
implemented by various electric utilities with relatively
low customization effort.
|