The invention discloses a fire situation analysis method. The method includes step 1 of performing smooth processing on the data acquired by a plurality of sensors via Kalman filtering, and then sending to a fuzzy inference system, and improving the system measurement precision to make accurate forecasting and decision; step 2 in the fuzzy inference adaptive noise adjusting processing module, verifying the measured values, and determining and replacing the bad values; step 3 giving adjusted measured noise according to the fuzzy rule and membership functions; and step 4 sending the processed data of each sensor to a global optimal fusion module and obtaining the optimal fusion result based on the combination with Kalman filtering; and sending the feedback of the global optimal fusion result to each sensor data processing unit for the subsequent data fusion. In the method, the fuzzification and defuzzification process, the fuzzy rule, and the fuzzy inference method are designed, and the measured noise can be approximate to the real noise.