A valid model of a rechargeable battery is required in several applications, especially to determine its internal charge. Due to the excessive variability of battery behavior, a large amount of data, experimentally obtained and suitably coded, should accompany the model. Exploiting the power of the fuzzy neural approach is a usual methodology to meet this requirement. However, also a strictly circuital approach is feasible, as proposed in the present paper. We suggest the use of a simple circuit model constituted by the series of a nonlinear capacitor and a memristor. Their different characteristics, experimentally determined, are compacted under the form of two small sets of vectors to be associated with the battery model.