Abstract: This study proposes the use of independent or combined spatial arrangements for pyrolysis particle multi-element fire early warning detectors. By applying weighted fusion algorithms or neural network algorithms, it effectively integrates environmental condition sensors, detectors inside electrical cabinets, and thermal imaging temperature measurement sensors to achieve extremely early fire warning in high and low voltage distribution rooms. Experimental data indicate that the early warning detectors can trigger alarms at a very early stage; even during high-power overload tests, the alarm is triggered more than 5 minutes prior to smoke alarm response times. Furthermore, during the combined analysis of detectors within the cabinet and those in the space, the response time for the cable trench overload tests improved by approximately 5 seconds or more, suggesting that spatial parameters play a significant role in very early detection. When utilizing the thermal imaging temperature measurement module in conjunction with the pyrolysis particle multi-element early warning detector, the response time is significantly accelerated for direct sight scenarios of heat-generating components, thereby limiting most thermal disasters to their initial stage.