The drone small cell (DSC) network has become a key technology for air-to-ground wireless communications in a variety of temporary or emergency situations. Based on mobile users, frequently changing DSC topologies have important challenges such as severe co-channel interference and limited battery capacity. However, temporarily dispatched drones cannot obtain labeled and historical data in advance, while they only obtain real-time operational data. The observed data can be analyzed by unsupervised learning methods to find useful information for resource management. In this paper, an interference-aware power control (IPC) framework is designed using affinity propagation clustering (APC). The APC method is one of the unsupervised learning methods. The numerical results show that our proposed IPC framework using the APC method can reduce system interference and significantly improve the energy efficiency of DSC networks.