Improved LSTM-Based Model of Operation Risk Assessment for High-Current Switchgear Cabinets

To enhance the risk identification and early warning capabilities of high-current switchgear cabinets under complex operation conditions, this paper proposes improved long short-term memory (LSTM)-based model of operation risk assessment for high-current switchgear cabinets. The method introduces a dynamic gating parameter adjustment strategy within the LSTM structure and integrates an attention mechanism to improve the model's perception of temperature rise and short-term load fluctuation characteristics. Through a case study using actual monitoring data from a substation, the model effectively captures nonlinear features and short-term fluctuation trends during operation, enabling dynamic prediction of switchgear risk levels. Experimental results show that the proposed method improves short-term fluctuation capture ability by 4.72% and 7.77%, reduces the mean absolute percentage error (MAPE) by 55.17% and 61.76%, and enhances prediction stability by 61.26% and 67.66%, compared to Baseline 1 and Baseline 2, respectively. In addition, the median prediction error is significantly reduced, and the predicted values are closer to the actual risk trends, indicating superior accuracy and stability. This method provides reliable data support for intelligent operation and maintenance as well as risk prevention of switchgear systems.