Computational Intelligence

14 Eberhart and Shi—Computational Intelligence

The solid line in Figure 12.1 represents a constant current discharge over time,

which starts with an SOC of 1 and decreases linearly to 0. A sigmoidal output PE,

however, trains to something similar to the dotted line in the figure because of the

nonlinearities at the extremes of the sigmoid function. It will, in fact, always exhibit

errors at the ends of the discharge cycle, even under other load conditions. The prob-

lem is that the most important regions of the SOC prediction for many applications

are at the discharge cycle extremes. We want to know accurately when the battery is

almost fully charged ...