Research on Rapid Detection of Battery Health Based on Som Neural Network

Main Article Content

Haifeng Jiang, Chang Wan, Shusheng Lin


This article describes a method for quickly separating the same set of unhealthy batteries and predicting the health
of unknown battery sample by machine learning. The IEC60896-22-2004 and GB/T19638.2-2005 standards
specify the capacity standards and testing methods for lead-acid batteries. The standard capacity test method
requires a constant current discharge of for 10 hours. The test method is difficult to apply to a large number of
in-used battery groups in practice. In this paper, an alternative method based on machine learning is studied. The
method is to perform fast 5 minute high current charging and 5 minute high current discharge on the battery group
which have been voltage balanced, extract the characteristics of charge and discharge, and map each battery
characteristic through SOM neural network to 2D plane and then separated by cluster analysis for batteries of
different capacity properties. Furthermore, through multiple machine learning and acquisition of real capacity
according to standard methods, a training set for supervised learning is established, and SOM neural network
cluster center or a time series similarity search algorithm is used to quickly evaluate the capacity of unknown
battery samples.

Article Details

How to Cite
Haifeng Jiang, Chang Wan, Shusheng Lin. (2021). Research on Rapid Detection of Battery Health Based on Som Neural Network. CONVERTER, 2021(6), 609-614. Retrieved from