Abstract: In the cable tunnel, the cable tunnel fire caused by the running state of the equipment or the aging of the line has the most serious impact on the operation and maintenance of the cable tunnel. In order to ensure the reliable operation of cable tunnel, a monitoring and early warning method of cable tunnel fire hazards based on nonlinear state estimation is proposed. Through the descending order of the total failure probability of the cable tunnel, the tunnel state is determined. The nonlinear state estimation algorithm is used to build the cable tunnel state prediction model. The embedded monitoring equipment is used to collect the cable tunnel state data, which provides the data source for the cable tunnel state prediction model. The model data is introduced into the database, the data clustering distance is calculated, the K-means clustering method is used to distinguish the fire hazard data, and it is transmitted to the main control terminal to complete the fire hazard early warning work. The simulation results show that the early warning false alarm rate of this method is less than 10% in the case of cable hidden danger or environmental hidden danger, and the accuracy of cable operation state estimation is higher, and the data transmission delay of fire hidden danger is shorter.