Abstract: Outdoor electrical cabinets in substations operate in high load, strong electromagnetic interference and complex climate environment for a long time, which is easy to cause local abnormal temperature rise due to poor contact, aging insulation and insufficient heat dissipation, which seriously threatens the stable operation of power grid. The existing monitoring methods have obvious shortcomings in real-time and false alarm suppression, and it is difficult to meet the requirements of accurate operation and maintenance. Therefore, this paper proposes a multi-scale monitoring method for the temperature rise of outdoor electrical cabinets in substations. The temperature signal acquisition system is constructed by using non-contact infrared thermography technology as data acquisition method, and the robust segmentation of abnormal regions is realized by combining adaptive local threshold binarization processing. Furthermore, the thermodynamic steady-state model is introduced, and the mapping relationship between temperature rise state and equipment operation state is established, and a multi-scale monitoring system from pixel level to system level is constructed. The experimental results show that the monitoring time of this method is less than 2 s in complex environment, and the false detection rate is less than 1%, so it has good applicability.