Joint SOC-SOH Estimation for Li-Ion Batteries Based on Multi-Method Fusion

The state of health (SOH) estimation is of guiding significance for the practicality and economy of battery. To overcome the difficulty of battery SOH estimation and its result susceptibility to noise, considering the better estimation effect of fusion algorithm and anti-jamming capability to noise, a multi-algorithm fusion was proposed to carry out SOC- SOH joint estimation for the state of charge (SOC) and health of lithium-ion battery. Based on a convolution neural network (CNN) model, the fusion algorithm was arranged with particle swarm optimization deep confidence network and adaptive extended Kalman/Adaptive H_∞ filtering ((PSO-DBN)-AEKF/AHIFF). Firstly, the SOH data were pre-processed based on wavelet transform to make significant noise removal. Secondly, the denoised data were input to the trained CNN model for SOH estimation, and then the ((PSO-DBN)-AEKF/AHIFF) algorithm was fused for SOC estimation. Finally, a Typhoon HIL602+ hardware-in-the-loop platform was built for Matlab/Simulink/Python environment and DST and UDDS operating conditions. The results show that the estimation error of SOH is within 1% and the estimation error of SOC is within 2%, proving the effectiveness of the joint estimation of SOC-SOH with multi-method fusion, and excellent estimation accuracy and robustness.