Battery storage time is constrained, for example, in data center backup application, because a backup-battery is required to provide at least a minimum capacitance, as shown in FIGS. 3 and 4. At different storage temp, the storage time varies before hitting Cmin, a set minimum remaining capacity for end of life. The average capital expense of battery per day can be determined based on the capital expense of a battery relative to the degradation that results over storage time.

An optimizationobjectivecanincludefindingtheoptimaltemperaturesuchthattotal cost of cooling and battery capital expense is minimized. As can be seen from FIG. 3, for different temperatures, T1?T5, at the minimum capacity, Cmin, the storage time varies, but have a positive correlation to temperature. Thus, to achieve high resolution, a regression model can be generated based on limited temperature samples. Thus, a temperature to storage time relation can be found with unlimited resolution for fine tuning. For cycle sensitive applications, one can replace storage time in days (FIG. 3) with cycles (FIG. 4) and apply the same algorithm described above (but with cycles instead of storage time) to achieve the same result. For both storage time and cycle sensitive application, since these two parameters both have same correlation to temp, they can be combined together as two parameters in a function of temperature. In such a case, the average capital expense of battery can be determined based on the capital expense of a battery, the storage time, and the number of battery cycles.