Step-Voltage Testing of Ni-Cd Batteries

Tuesday, 7 October 2014
Expo Center, 1st Floor, Center and Right Foyers (Moon Palace Resort)
M. S. Lipkin, S. M. Lipkin (South Russian State Polytechnical University), A. V. Sedov (Southern Scientific Center of Russian Academy of Sciences), and D. A. Onyshko (South Russian State technical university)
State of charge diagnostics is very important problem for high power batteries, charging by constant voltage during long time. For many types of secondary cells, based on intercalation processes, a relation between capacity and current response on potential step can be obtain. This relation is direct proportional only in ideal case, but for real batteries a lot of different causes transformed this relation to more complex form. In this situation neural network models give some advantages.

Proposed principle of diagnostics is based on applying series of steps of voltage with information about battery’s capacity appears in values of corresponding current steps. Three-step voltage was used therefore three values of current were available for calculating the value of capacity (fig. 1).

In this paper 24 batteries with capacities varying from 0.5 A/h to 1.2 A/h were used, so the training set consisted of 24 three dimensional vectors of current values. Principal components analysis [1] has shown that capacity information (data variance in training set) is approximately evenly distributed over all three current values, therefore all of them can be directly used as an input for calculating value of battery’s capacity.

For such calculation feed forward neural network (ANN) with one hidden layer containing three neurons was used. Considering that speaking of battery’s capacity lower prediction is much less critical than higher one results of ANN work can be shown on the following histogram (fig. 2).

Thus, multiple voltage steps technique is multiparameter diagnostic, which provides determination on-line state of charge of secondary batteries. It is especially important for high power modules in the condition of complex history.


  1. Jolliffe, I. T., Principal Component Analysis, 1986, Springer-Verlag, p 487.