Li-ion battery systems have been widely used as an essential power source in many applications. To ensure the safety and longevity of the system, a battery pack thermal model is often used in combination with distributed temperature sensors for thermal management and monitoring purposes. Due to the limited number of sensors and sparse measurement, sensor deployment to maximize the observability of the system thermal dynamics has been a critical topic, which has attracted research attention but remains to be resolved. This paper is devoted to exploring the pattern of optimal sensor locations for scalable battery systems under different observability criteria. A battery pack thermal model is first developed based on a single cell thermal model and considering the thermal interconnection between cells in the pack. Sensor location optimization is then performed by maximizing two Gramian-based observability metrics, which quantify different aspects of system observability. Optimal sensor locations obtained under the two metrics are analyzed and compared for battery systems of different sizes. Based on the results, deployment patterns of optimal sensor locations are extracted and analyzed theoretically by correlating to the physics of the battery thermal dynamics. Moreover, the influences of critical parameters of battery packs on system observability are also analyzed and discussed.