融合模糊神经网络与时序模型的火灾预警算法

为有效解决传统火灾监测器误报率高的问题,提出一种结合模糊神经网络(FNN)模型和温度时序模型(TSM)的火灾预警算法,即FNN-TSM,其研究对象包括综合管廊中的3个火灾特征参数:温度,烟雾浓度和一氧化碳浓度。建立FNN模型,获得每个数据点的孤立火灾概率;建立TSM以探测特征参数的变化率,获得每个数据点的时序火灾概率;采用复合火灾决策方法,获得最终的火灾概率。实验结果表明,该算法比其它算法具有更高的火灾预警准确率和实时性。

Fire early warning based on fuzzy neural network and temperature time series model

To effectively solve the problem of high false alarm rate of traditional fire monitors,a fire warning algorithm combining fuzzy neural network(FNN)model and temperature time series model(TSM)was proposed,namely FNN-TSM.The research object of this method included three fire characteristic parameters of temperature,smoke concentration and carbon monoxide concentration in the integrated pipe gallery.The FNN model was established to obtain the isolated fire probability of each data point.The TSM was established to detect the rate of change of the feature parameters and obtain the time series fire probability of each data point.The composite fire decision method was used to obtain the final fire probability.Experimental results show that the proposed algorithm has higher fire warning accuracy and better real-time performance than other algorithms.