State of charge estimation for energy storage lithium-ion batteries based on gated recurrent unit neural network and adaptive Savitzky-Golay filter
Jinbo Lu,Yafeng He,Huishi Liang,Miangang Li,Zinan Shi,Kui Zhou,Zhidan Li,Xiaoxu Gong,Guoqiang Yuan
DOI: https://doi.org/10.1007/s11581-023-05252-z
IF: 2.961
2023-10-19
Ionics
Abstract:The accurate estimation of lithium-ion battery state of charge (SOC) is the key to ensuring the safe operation of energy storage power plants, which can prevent overcharging or over-discharging of batteries, thus extending the overall service life of energy storage power plants. In this paper, we propose a robust and efficient combined SOC estimation method, GRU-ASG, which combines the gated recurrent unit (GRU) neural network and the adaptive Savitzky-Golay filter (ASG). Firstly, the "many-to-one" structure GRU is used to establish the mapping model between the battery-measured variables (voltage, current, temperature) and SOC, and achieve the SOC initial estimation. Then, the output SOC of the GRU network is smoothed online using the Spielman coefficient–based ASG filtering algorithm proposed in this paper to reduce the fluctuation of SOC. Finally, the accurate and stable estimated SOC is obtained. This paper uses six different operating condition datasets collected from an energy storage plant during the discharge process and uses four of them as training datasets and the remaining two as test datasets. The results show that the proposed method can select the optimal window length online adaptively to smooth the initial estimate of SOC. Moreover, the estimation accuracy of the proposed method is the highest compared to the single GRU network and the GRU network with a combination of other filtering algorithms. In particular, the mean square error (MSE) is less than 0.15% and the mean absolute error (MAE) is less than 3% for the two test sets.
chemistry, physical,physics, condensed matter,electrochemistry