Abstract:
In the event of a fault in the distribution system, the Provincial Electricity Authority (PEA) must transfer the load or switching of the system in order to limit the power outage area before correcting the fault. It is necessary to find the fault location beforehand in order to command the switching system correctly. This process takes time and affects the PEA's reliability index that is concerned with the time value criteria or SAIDI. In addition, the commander who orders the system switching must be experienced and skilled so that the ordering can be done quickly and safely.
This thesis presented a method of planning the switching of 22kV distribution system with artificial neural networks (ANNs) in order to decide the switching procedure. The data of Uttaradit Power Station was used as a case study. Fault currents of 10 feeders are simulated by the DIgSILENT program together with the ordering number of the feeders and the overcurrent protection status that was used as the training set of ANNs inputs. The outputs or targets of the ANNs were the switching order procedure of each fault location which was obtained from experts.
From the experimental results that the fault simulation was performed and compared to the results of the 5 faults that happened in Uttaradit substation, it was found that the proposed technique could quickly provide an answer as to the switching procedure and tended to be the same direction as the skilled operator gave. In addition, the method was also proposed to 10 experts to try and give an opinion. The results matched those selected by experts when other faults occurred. It can be concluded that the proposed method can help command the switching system quickly and accurately.