Residual life prediction of lithium-ion batteries by non-equidistance grey forecasting model

In order to forecast residual life of lithium-ion batteries, a new non-equidistance grey prediction model is investigated. As for the non-equidistance results of the actual lithium-ion batteries accelerated degradation test, screens out those data with a too large interval, computes the mean value of the residual data, based on the corrected mean interval, segments the data and determines the number of data in each segment, two different methods are adopted respectively to transform each segment to equidistance time series according to the number of data in it, grey GM(1, 1) model of residual life about lithium-ion batteries is established after accumulating procession, the method of precision inspection is introduced. Finally, with the results of the actual lithium-ion batteries accelerated degradation test, the non-equidistance grey forecasting model is verified, the value of posterior error ratio and small error probability is given, and compared with the threshold value, which indicates that the non-equidistance grey forecasting model´s accuracy is much better than the existing model.