Machine Learning Enabled Cluster Grouping of Varistors in Parallel-Structured DC Circuit Breakers

Shuyan Zhao, Yao Wang, Reza Kheirollahi, Zilong Zheng, Fei Lu, Hua Zhang

Research output: Contribution to journalArticlepeer-review


This letter presents the first ever trial of machine learning enabled cluster grouping of varistors for DC circuit breakers (DCCBs). It reveals that the manufacturing discrepancy of varistors is a main challenge in their parallel connection. The proposed cluster grouping concept is introduced to classify varistors according to the interruption characteristic, in which the K-means algorithm is adopted to learn the clamping voltage curves. 70 420 V/50 A V420LA20 varistors are measured in a 120 A transient current interruption platform individually to acquire 70 sets of testing data to train the machine learning engine. Then, 28 new varistors are further tested to verify the trained algorithm, which are classified into 7 clusters using the proposed machine learning method. A 500 V/520 A solid-state circuit breaker (SSCB) is implemented with four parallel varistors in the same cluster. Experiments validate that the current is evenly distributed in varistors, and the difference is limited to 3.1%, which improves parallel varistors lifetime significantly.

Original languageEnglish (US)
Pages (from-to)1003-1010
Number of pages8
JournalIEEE Open Journal of Power Electronics
StatePublished - 2023
Externally publishedYes

All Science Journal Classification (ASJC) codes

  • Electrical and Electronic Engineering


Dive into the research topics of 'Machine Learning Enabled Cluster Grouping of Varistors in Parallel-Structured DC Circuit Breakers'. Together they form a unique fingerprint.

Cite this