The aim of this study is to address the limitations of traditional single-task learning models that fail to fully consider the mutual influence and coupling relationships among equipment states.This paper reveals the model's per-formance and potential optimization directions across different time spans through analysis and testing.The research findings indicate that as the time span increases,the root mean square error(RMSE)and mean absolute error(MAE)of the model generally decrease,with particularly accurate performance in short time spans.For long time span data,more iterations may be necessary to adapt to the complexity of the data.
关键词
多任务学习/变电站设备/设备状态评估
Key words
multi task learning/substation equipment/equipment status assessment
万方数据