改进模糊神经网络下电气火灾预警算法设计  

Design of Electrical Fire Warning Algorithm Based on Improved Fuzzy Neural Network

作　　者：于兰[1,2] 贾振国 胡冬梅[3] YU Lan;JIA Zhen-guo;HU Dong-mei(School of Energy and Power,Changchun Institute of Technology,Changchun Jilin 130012,China;Jilin Construction Energy Supply and Indoor Environment Control Engineering Research Center,Changchun Jilin 130012,China;College of Electrical and Information Engineering,Beihua University,Jilin Jilin 132021,China)

机构地区：[1]长春工程学院能源动力工程学院,吉林长春130012 [2]吉林省建筑能源供应及室内环境控制工程研究中心,吉林长春130012 [3]北华大学电气信息工程学院,吉林吉林132021

出　　处：《计算机仿真》2025年第1期524-528,共5页Computer Simulation

基　　金：吉林省教育厅科学技术研究项目(JJKH20210677KJ);2021年吉林省教育科学规划课题(GH21366);2022年度教育部产学合作协同育人项目(220500383190819)。

摘　　要：由于电气火灾前出现的信号相对较弱,且受电气设备本身的复杂性和环境噪声等因素的影响,使得预警信号采集难度增大,从而导致预警效率与精度过低。为提高预警准确性,提出改进模糊神经网络下电气火灾预警算法。采用ICA-CEEMD小波阈值去噪算法对数据展开处理,利用3σ原则提取信号的细节信息,重构多源传感器数据;将传感器采集的数据转变为时间序列形式,通过N-TCD算法剔除数据中的异常值;将传感器采集的温度数据、烟雾浓度数据和一氧化碳浓度数据输入模糊神经网络中,输出孤立火灾概率,建立温度时序模型,计算时序火灾概率,加权融合处理孤立火灾概率和时序火灾概率获得最终的电气火灾概率,完成电气火灾的预警。仿真结果表明,所提算法具有较高的预警精度和预警效率。Due to the weak signals that appear before electrical fires,as well as the complexity of electrical equipment and environmental noise,the difficulty of collecting the warning signals increses,resulting in low warning efficiency and accuracy.To improve the accuracy of early warning,this paper proposed an electrical fire warning algorithm based on improved fuzzy neural network.At first,ICA-CEEMD wavelet threshold denoising algorithm was applied to data processing.According to 3o principle,we extracted signal details and reconstructed multi-source sensor data.Moreover,we transformed the data collected by the sensor into time series,and eliminated the outlier in the data by N-TCD algorithm.Furthermore,we input the temperature data,smoke concentration data,and carbon monoxide concentration data collected by sensors into the fuzzy neural network to output the probability of isolated fire and construct a temperature time series model.Meanwhile,we calculated the sequential probability of fire.After that,we obtained the electrical fire probability through weighted fusion of isolated fire probability and time series fire probability.Finally,we completed the early warning of electrical fires.The experimental results show that the proposed algorithm has high warning accuracy and efficiency.

关 键 词：模糊神经网络 异常数据剔除 温度时序模型 电气火灾预警 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]