Abstract: Long-term extraction of underground metal mineral resources has resulted in the formation of numerous and extensive goafs. This poses greater challenges to the safety of personnel and equipment, and the stability of large-scale goafs has become one of the critical bottlenecks restricting the safe and efficient exploitation of such mineral resources. Taking an underground tungsten mine with typical large-scale goafs formed by open stope method mining as the engineering background, this study analyzes the stability and distribution characteristics of the roof and pillars in the large-scale goaf. It investigates the factors influencing goaf stability and their respective degrees of impact, establishes a method for identifying hazardous instability zones in large-scale goafs. From the perspective of microseismicity, the spatiotemporal distribution characteristics of ground pressure manifestations in large-scale goafs are revealed, the accuracy and reliability of the identification method are verified, targeted goaf treatment measures are proposed, and the long-term evolution characteristics of ground pressure are predicted and analyzed using the b-value and maximum microseismic magnitude. The results indicate that over 50% of the roof area is in a relatively unsafe or unsafe state, while the overall safety condition of the pillars is relatively good, with the risk of roof instability being significantly higher than that of the pillars. The hydraulic radius shows positive correlations with roof area, roof span, and roof perimeter. The safety factor of pillars exhibits positive correlations with pillar width-to-height ratio, pillar area, and area of the rock mass supported by pillar, but negative correlations with pillar burial depth. Roof span and pillar burial depth are the primary controlling factors affecting the stability of the goaf roof and pillars, respectively. Most non-Class I pillars are located within the hazardous instability zones, and approximately 50% of these hazardous instability zones are situated within the strip areas, suggesting that strip areas may be zones with a higher probability of goaf instability risk. Microseismic monitoring data show that the core strip area distributed in an "anticline-type" pattern along the orebody strike is the most densely active zone of mining-induced seismicity, and there is a significant temporal correlation between microseismic activity and large-scale roof collapse events in the goaf. The number of small seismic events is relatively low, while the number of large seismic events is relatively high, indicating a high intensity of ground pressure activity. The Gumbel Type I extreme value distribution principle proves effective in predicting the probability of extreme local microseismic magnitudes, with future local microseismic magnitudes mostly expected to fall within the range of -4.75 to -2.75. The research findings provide key theoretical basis and technical support for the stability control and disaster early warning of large-scale goafs.