Smart grid applications, such as teleprotection, synchrophasors, remote condition monitoring, and control of assets, require timely and reliable data communication systems. Many communication architectures and solutions have been proposed to provide support for these smart grid applications, including fiber networks, wireless mobile networks, wireless sensor networks (WSNs), and so on. The use of heterogeneous solutions is an attractive architecture option since it combines the advantages of two or more communication systems to meet the delay and reliability requirements of the smart grid. Fiber-wireless sensor networks (Fi-WSNs) are gaining popularity as a reliable communication infrastructure in many other applications. This is due to the low cost, reliability, availability, the distributed nature of the WSNs, and the high bandwidth and reliability of the optical fiber networks. Although the Fi-WSNs systems can provide a good delay performance, they may not meet the requirements of the above smart grid delay-critical applications. In this paper, we propose a novel adaptive and cross-layer service differentiation mechanism for the Fi-WSNs. The proposed mechanism implements an adaptive scheduling mechanism and allows WSNs to cooperate with the optical network unit (ONU) to reduce the delay for high priority traffic. We determine the effects of the proposed mechanism on the quality of service (QoS) of delay critical smart grid monitoring applications in terms of the end-to-end delay and reliability. We show through simulations that our proposed QoS mechanism can reduce the end-to-end delay in the Fi-WSN system and in the long-reach passive optical networks (LR-PONs). We also show that our mechanism can outperform the existing techniques under the same traffic and network conditions.