Wireless backhaul technology is one of the promising solutions for next-generation mobile communication networks due to its merits of significantly reducing operator's cost, providing user equipment (UE) with fundamental flexibility and improving the network overall spectrum efficiency. By leveraging the Lyapunov stochastic optimization framework and convex optimization theory, this paper proposes a queue-aware power allocation algorithm for in-band full-duplex wireless backhaul networks. Specifically, in each discrete resource scheduling time slot, the algorithm dynamically allocates power for each user's downlink backhaul and access links by comprehensively taking channel state information (CSI) and queue state information (QSI) into consideration, so as to maximize the network average total spectrum efficiency (SE) while ensuring network stability and meeting the quality of service (QoS) requirements of users. In addition, the theoretical analysis and simulation results reveal that the proposed algorithm can flexibly strike a balance between SE and delay by simply tuning an introduced control parameter.