A Lithium-ion Battery Fault Diagnosis Method Based on Deep Belief Network and Extreme Learning Machine

Xuan Zhang ； Kui Shao ； Meng Huang ； Bo Zhang ； Jing Wu ； Chaolong Zhang ； Shanhe Jiang

《World Scientific Research Journal》 7卷5期 (2021/05) Pp. 197-207

https://doi.org/10.6911/WSRJ.202105_7(5).0021

摘要

Lithium-ion battery is widely used as energy storage unit in electric vehicles, mobile base stations and new energy sources. The safe operation of lithium-ion power battery is very important to ensure the normal use of the system. High internal resistance, small capacity and low state of charge (SOC) are progressive faults that often occur in the use of lithium-ion battery. A fault feature extraction method based on DBN is proposed to diagnose the three faults in this work. The learning rate of RBM composed DBN is generated by particle swarm optimization (PSO). After the feature data are extracted, a fault diagnosis model is established by extreme learning machine (ELM) to identify different faults. The experimental results show that battery fault diagnosis method proposed in the work can correctly identify each fault, and the diagnosis accuracy is 100%, which is obviously better than other feature extraction methods in fault diagnosis.

關鍵字

Battery ； Fault diagnosis ； Deep belief network

參考文獻

Chen Z , Xiong R , Tian J , et al. Model-based fault diagnosis approach on external short circuit of lithium-ion battery used in electric vehicles[J]. Applied Energy, 2016, 184:365-374.

Cui Z Q , Xu M Q , Wang R X . Study on Intelligence Fault Diagnosis Approach Based on Digraph Model and Application to Satellite Battery[J]. Advanced Materials Research, 2014, 889-890:929-932.

Daubechies I , Sweldens W . Factoring wavelet transforms into lifting steps[J]. Journal of Fourier Analysis & Applications, 1998, 4(3):247-269.

Strickland R N , Hahn H I . Wavelet transform methods for object detection and recovery[J]. IEEE Transactions on Image Processing, 1997, 6(5):P.724-735.

Rhee I , Bachman K E , Park B H , et al. DNMT1 and DBNMT3b cooperate to silence genes in human cancer cells[J]. Nature, 2002, 416(6880):552-556.