Reducing time operation of protection system is desirable, for the life time of equipment and voltage quality will be much improved. During starting up electrical devices such transformers or electrical motors etc., there must be inrush current. Alternatively, the inrush current and the damaging curves of those are not in the same curve which means that to protect those devices, the minimum time setting of relay protection must be set at least higher than the time delay of inrush current and it can be used the conventional curves of IEC standard. However, it will be much better for the devices and also the power system if nonconventional curve is applied, for it can be regulated to any circumstance of the loads' characteristic curves. Plus, power system protection will be more flexible if this model can be applied in adaptive relay coordination. In this study, Bayesian Regularization Backpropagation Neural Network (BRBPNN) is used in order to model nonconventional curves. BRBPNN is a robust method and is used to model the characteristic curves of overcurrent relay (OCR). It was developed for trend analysis of the input data as load pickup current and tripping time as the target data. It is obvious that the errors between actual data and testing result of simulation and real application of prototype are highly acceptable.

Original languageEnglish
Title of host publication2017 International Electrical Engineering Congress, iEECON 2017
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781509046669
Publication statusPublished - 19 Oct 2017
Event2017 International Electrical Engineering Congress, iEECON 2017 - Pattaya, Thailand
Duration: 8 Mar 201710 Mar 2017

Publication series

Name2017 International Electrical Engineering Congress, iEECON 2017


Conference2017 International Electrical Engineering Congress, iEECON 2017


  • Brbpnn
  • Nonconventional curve
  • Overcurrent relay
  • Protection
  • Time delay


Dive into the research topics of 'Overcurrent relay modeling using artificial neural network'. Together they form a unique fingerprint.

Cite this