Research on lithium battery box fire monitoring system based on wireless sensor network and weight self-adapting decision model

WANG Haoyang; TAN Zhigang; YANG Mingming

doi:10.19956/j.cnki.ncist.2024.06.012

Abstract：This paper focuses on the lithium battery fire monitoring scenario and deeply discusses a lithium battery fire monitoring system that combines a wireless sensor network and a weight self-adaptive decision model.This system integrates ZigBee wireless sensor network,convolutional neural network and random forest model.Among them,ZigBee module is responsible for collecting important indicators such as temperature,smoke concentration and CO concentration in the lithium battery environment by virtue of its advantages of low cost and easy deployment.However,due to the limitation of the effective temperature working range of the ter-minal node,the data above 85℃is seriously distorted and missing.Therefore,the key of this algorithm is to realize fire monitoring when some key data is missing.For the collected data,CNN can not only realize feature extraction,but also complete the weight ratio calculation,and feed back the adaptive weight to the random for-est model in real time.Judging from the experimental results,the random forest model can accurately deter-mine the fire situation according to the received weight information.Compared with the traditional use of infra-red thermal imaging fire detectors to achieve similar data collection and fire prediction functions,which re-quires high costs,this paper cleverly uses the ZigBee module,which not only successfully completes the envi-ronmental data collection and model-based prediction tasks,but also realizes effective cost savings.This inno-vative measure is of great significance in the environmental data monitoring and analysis of lithium battery fire monitoring,and provides an economical and efficient feasible solution for related applications.