Wildfires present a considerable challenge to both natural ecosystems and constructed environments, as well as to the safety and economic stability of communities located in areas susceptible to such events. Among the built environments, the electric power grid is particularly vulnerable to the impacts of wildfires, both as a victim and a contributing factor. The high costs and inefficiencies associated with manual inspections of transmission lines hinder power grid companies from implementing effective preventive measures. In recent years, advancements in deep learning technology have led to significant improvements in target detection methods, which are now being widely adopted in various industries. However, the deep learning approach is not without its limitations, including issues related to interpretability. This paper proposes a detection method for wildfire threats that integrates object detection with neural-symbolic reasoning. Experimental results indicate that this approach can deliver relatively stable and efficient detection of wildfire threats within electric grid networks.