Abstract: To address the key problems that the traditional fire-fighting facilities of unattended substations cannot adapt to the scenario requirements of no on-site personnel disposal and long remote response time, and it is difficult to realize ultra-early fire warning and closed-loop management, in this paper, the simultaneous thermal analysis‒Fourier transform infrared spectroscopy (STA-FTIR) technology is used to carry out pyrolysis characteristic experiments on four types of typical cable materials commonly used in substations, including ZR-YJLV cable outer sheath and insulation layer, PVC sheath, and N-BV sheath. The thermogravimetric law of materials in the critical temperature range of 30 ℃ to 350 ℃, as well as the release characteristics of characteristic gases such as H2O, CO2, HCl and pyrolysis particles are clarified, and the ultra-early characteristic correlation basis of "temperature‒gas‒pyrolysis particles" is established. Based on the experimental results, an ultra-early fire detection system with a four-layer architecture of "perception‒transmission‒processing‒application" is designed. The core of the system is a pyrolysis particle detector based on cloud chamber principle and optical analysis techno⁃ logy, which integrates multi-source perception devices such as AI vision camera and thermal imaging camera, and constructs an intelligent decision-making system based on cloud‒edge‒end collaboration. Through the Internet of Things platform, the system realizes cloud storage and intelligent analysis of data, and can realize intelligent linkage with substation access control, video surveillance and fan control system. Verified by long-term operation at 5 unattended 110 kV/220 kV substations in Hebei Province, the false alarm rate of the system is less than 0.5%, and more than 30 overheating hazards of electrical equipment have been successfully warned, which effectively ensures the safe and stable operation of unattended substations.

Key words: unattended substation, cloud chamber, pyrolysis particle detector, intelligent linkage, intelligent fire management