Enhanced strain assistance for SOC estimation of lithium-ion batteries using FBG sensors

Abstract

The accurate estimation of the state of charge (SOC) is essential to guarantee the safe and reliable operation of battery systems. Recently, more and more studies and applications have adopted optic fiber sensors to aid SOC estimation. However, it faces challenges such as limited performance and high costs. To address these challenges, this work proposed using a novel multi-position strain to enhance strain assistance for SOC estimation. Three fiber Bragg grating (FBG) sensors are arranged near the negative electrode, near the positive electrode, and in the middle of the battery, respectively. Strains at multiple positions are utilized as input features for the SOC estimation model, either individually, in dual combination, or triple combination. The impact of the number and placement of FBG sensors on SOC estimation is assessed. Temporal Convolutional Network (TCN), Convolutional Neural Network (CNN), and Gated Recurrent Unit (GRU) were employed to evaluate the effectiveness of multi-position strain. Furthermore, an FBG demodulation system based on a tunable Fabry-Perot (FP) filter was built to obtain strain information from wavelength signals. Compared to the commercially demodulation systems, the proposed demodulation system achieves a cost reduction of over 90 %. Experimental results verify that, compared to a traditional single strain, the dual strains significantly improve SOC estimation accuracy. In static tests, the root mean squared error (RMSE) and mean absolute error (MAE) are reduced by up to 73.66 % and 71.72 %, respectively. In dynamic tests, RMSE and MAE reductions reach up to 72.49 % and 74.01 %, respectively.