This study uses complex network analysis and the Leiden community detection algorithm to investigate the dynamic spatiotemporal patterns of shared bicycle usage in Shanghai. We construct time-sliced travel networks using grid centre points as nodes and origin-destination flows as weighted edges to reveal clustering patterns and their temporal evolution. The results show that there is distinct community segmentation across different periods. During the weekday morning and evening rush hours, for example, the network exhibits strong spatial heterogeneity, with compact, commuting-oriented clusters. At midday, however, the communities merge into larger, more integrated structures, driven by leisure and short-distance trips. At weekends, community structures become more scattered and hierarchical, reflecting greater randomness in travel behaviour. These findings demonstrate that travel community boundaries are dynamic rather than fixed, being shaped by demand fluctuations and commuting rhythms. The study has practical implications for shared bicycle operators, highlighting the need for adaptive dispatch strategies that account for temporal variations, such as intra-community reallocation at weekends and cross-regional coordination during peak weekday periods.