Due to its fully enclosed design and complex internal heat transfer processes, GIS (Gas Insulated Switchgear) equipment presents challenges in quickly diagnosing and warning about internal defects when anomalies occur. The internal temperature field becomes intricate and difficult to monitor, leading to significant difficulties in promptly diagnosing and alerting about internal defects, which, in turn, severely impacts the safe operation of the equipment. This paper addresses the issue of overheating defects in GIS equipment by designing an online monitoring scheme for such defects using various monitoring methods. Based on multiple monitoring approaches, the paper calculates the weight allocation of each monitoring dataset by analyzing their correlations. A dual-layer LSTM (Long Short-Term Memory) network is then employed for prediction. Through experimental data and case studies, the proposed method demonstrates more accurate, convenient, and stable diagnosis of overheating defects within GIS isolation switchgear equipment. Compared to traditional heat monitoring techniques, it offers higher detection accuracy and timely alerts. This approach is adaptable to various scenarios involving overheating defects, and it effectively enhances the capability of detecting overheating defects in power equipment, ensuring the secure operation of GIS isolation switchgear equipment.