基于Stacking多模型融合的储能电站电池预制舱消防预警  

Fire Warning for Prefabricated Battery Compartments in Energy Storage Power Stations Based on Stacking Multi-Model Fusion

作　　者：崔巍 CUI Wei(North China Electric Power Design Institute Co.,Ltd.,China Electric Power Engineering Consulting Group,Beijing 100032,China)

机构地区：[1]中国电力工程顾问集团华北电力设计院有限公司,北京100032

出　　处：《自动化应用》2024年第17期81-83,共3页Automation Application

摘　　要：由于储能电站电池预制舱监测温度由多个传感装置独立组成,导致对高温异常进行消防预警的灵敏度偏低,为此,提出基于Stacking多模型融合的储能电站电池预制舱消防预警方法研究。引入了Stacking集成学习方法,对多元储能电站电池预制舱温度数据进行融合处理,并利用正则系数对融合后数据的偏差进行校正;在消防预警阶段,结合温度参数表现的发展趋势以及消防预警的时间尺度要求,设置了个性化的预警温度,根据融合结果与预警温度之间的关系,判断是否作出预警处理。在测试结果中,设计方法对于不同程度异常温度均实现了有效的预警,具有良好的灵敏度。Due to the independent composition of multiple sensing devices in the temperature monitoring of prefabricated battery compartments in energy storage power stations,the sensitivity of fire warning for high temperature anomalies is low.Therefore,a research on fire warning method for prefabricated battery compartments in energy storage power stations based on Stacking multi-model fusion is proposed.Introduced Stacking ensemble learning method to fuse temperature data of prefabricated battery compartments in multi-dimensional energy storage power stations,and used regularization coefficients to correct the deviation of the fused data.In the fire warning stage,personalized warning temperatures are set based on the development trend of temperature parameters and the time scale requirements of fire warning.Based on the relationship between the fusion result and the warning temperature,it is determined whether to issue a warning.In the test results,the design method achieved effective early warning for different degrees of abnormal temperature and had good sensitivity.

关 键 词：Stacking多模型融合 消防预警 正则系数 偏差校正 预警温度 

分 类 号：TM912[电气工程—电力电子与电力传动]