Abstract:
Abstract - To design the coordination protection for
passive distribution system is not the tough work, while active
or mesh distribution system which consists many distributed
generators is quite more challenge for protection engineers.
Additionally, the short circuit current will also vary if any DG
in the system is offline, which causes to re-coordinate the relay
protection in the system. To reset the relay protection, the
engineers need more time. However In order to reduce the time
of relay setting calculation, the adaptive protection coordination
is proposed in this study by using artificial neural network. The
study bases on the combinations of DGs’ state and the current
short circuit levels as the input data and low setting of the
directional overcurrent relays (DOCR) as the output data
training. This research is conducted on modified IEEE 9-bus
system equipped with distributed generators. After reaching
convergence of Levenberg-Marquardt Back Propagation
(LMBP) learning process, the results of weights and biases are
compiled into the master controller to control all of the relays in
the whole system. It will generate the relay setting automatically
base on the results of ANN training. The results of this research
have been testified in Etap simulation successfully and it is
obvious that LMBP neural network is a robust method to model
adaptive relay coordination system.
