应用神经网络与声纹识别的锂电池运行状态预警

Early Warning System of Lithium Battery Running State Based on Neural Network and Voiceprint Recognition

为了提高锂电池储能预警能力,结合边缘计算技术设计了一种锂电池储能预警系统。采用多模态锂电池声纹识别装置采集锂电池运行过程中发出的声音特征,并对声纹特征参数进行提取,计算出声纹信号的功率谱。在锂电池声纹识别装置中使用了FPGA芯片和硬件加速器,通过构建BP-SNN融合神经网络的故障预警模型结合反向传播神经网络与脉冲神经网络,利用特定的神经元模型进行时间信息的计算与传递,从而对锂电池监测数据的时间序列进行有效处理,提高锂电池故障预警能力。

In order to improve the early warning ability of lithium battery energy storage,an early warning system of lithium battery energy storage is designed by combining edge computing technology.A multi-mode voiceprint recognition device for lithium battery is used to collect the sound features emitted during the running of lithium battery,extract the voiceprint feature parameters,and calculate the power spectrum of the voiceprint signal.The FPGA chip and hardware accelerator are used in the voice print recognition device of lithium battery.By constructing the fault warning model of BP-SNN fusion neural network,combining the back-propagation neural network and pulse neural network,the time information is calculated and transmitted by the specific neuron model,so that the time series of lithium battery monitoring data can be processed effectively,and improve the lithium battery fault warning capability.