In view of the frequent occurrence of major fires and explosions in the production, transportation, and storage of aromatic nitro compounds, the reactive thermal hazards of these compounds have attracted widespread attention in the chemical industry. Thermal hazard assessment is of great significance for the standardized storage of reactive chemicals. Based on adiabatic calorimetry and quantitative structureproperty relationship (QSPR) prediction, a thermal hazard assessment and classification method dominated by the molecular structure of nitro compounds were proposed. While obtaining thermodynamic parameters, stable configuration and molecular descriptors, the prediction models were constructed by multiple linear regression (MLR) and artificial neural network (ANN), respectively, including selfaccelerating decomposition temperature (SADT), maximum power density (MPD), and the apparent activation energy (E a ). On this basis, SADT, H 50 , and E a were selected as the possible characterization parameters of thermal risk of nitro compounds, while MPD and ΔT ad were selected as the characterization parameters of consequence severity. When combined with weight and reference TNT, this method comprehensively evaluated the thermal hazards of nitro compounds and divided them into four types of thermal hazard levels. By comparing the method with the previous classification standards, it is found that the evaluation method is expected to provide theoretical support for the safe standardized storage of nitro compounds in chemical enterprises.