Energy storage systems (ESSs), such as lithium-ion batteries, are being used today in renewable grid systems to provide the capacity, power, and quick response required for operation in grid applications, including peak shaving, frequency regulation, back-up power, and voltage support. Each application imposes a different duty cycle on the ESS. This represents the charge/discharge profile associated with energy generation and demand. Different duty cycle characteristics can have different effects on the performance, life, and duration of ESSs. Within lithium-ion batteries, various chemistries exist that own different features in terms of specific energy, power, and cycle life, that ultimately determine their usability and performance. Therefore, the characterization of duty cycles is a key to determine how to properly design lithium-ion battery systems for grid applications. Given the usage-dependent degradation trajectories, this research task is a critical step to study the unique aging behaviors of grid batteries. Significant energy and cost savings can be achieved by the optimal application of lithium-ion batteries for grid-energy storage, enabling greater utilization of renewable grid systems. In this paper, we propose an approach, based on unsupervised learning and frequency domain techniques, to characterize duty cycles for the grid-specific peak shaving applications. Finally, we propose synthetic duty cycles to mimic grid-battery dynamic behaviors for use in laboratory testing.