Abstract

The paper describes further developments and application of a new, adaptive battery model based on a re- mapping of the well-known Randles’ Lead-acid variant. This re-mapping of the Randles’ equivalent circuit allows improved modeling capabilities and accurate estimates of dynamic circuit parameters of the battery, when used with Subspace parameter estimation techniques and Kalman Estimation schemes. The techniques correctly identify and converge on voltages associated with the battery State-of-Charge, allowing accurate monitoring of a network of batteries, as typically employed on HEVs and EVs, for instance. Although current state-of-the-art HEVs/EVs employ Li-ion or NiMH technologies, the pre-existing recycling infrastructure for lead means that Valve Regulated Lead Acid batteries remain a credible candidate for the future if their lifetime can be extended. As previously described, the convergent system is able to overcome problems attributed to State-of-Charge drift (incurred by coulomb-counting methods due to over-charging or ambient temperature fluctuations) or erroneous initial conditions, whilst observation of the SoC voltages, and online monitoring of the estimated dynamic model parameters allows battery ageing (State-of-Health) to also be assessed, and thereby cell failure to be predicted and charging scenarios to be optimized to extend lifetime.
It is demonstrated, by experimental measurements, that imminent failure of a single battery can often be masked by the characteristics of healthier units, leading to catastrophic failure in safety-critical systems. Here it is shown that a network of cells can be modeled and monitored as a single unit which, when compared with results from individual batteries, can identify risks associated with potential failure. Although VRLA batteries are used to focus this study, ultimately, the underpinning techniques are more generally applicable to alternative battery chemistries.