State of Charge Estimation for Liquid Metal Batteries with Gaussian Process Regression Framework

This paper proposes a novel SoC estimation method for liquid metal batteries using gaussian process regression, which is featured with relatively small training dataset, the ability of uncertainty quantification and fast estimation. In the introduced framework, only battery voltage and current measurements are needed to estimate SoC instantly. First, given the training dataset, hyperparameters in the gaussian process regression model are optimized by maximizing the marginal likelihood function. Then, the trained gaussian process regression model can be utilized straightly to estimate SoC when new measurements come. Finally, the estimation result and its confidence interval can provide a comprehensive view of the true SoC together. Experimental data of liquid metal battery from constant current cycling process is used to train and verify the proposed gaussian process regression model, the accuracy of estimation is proved to be qualified for the employment in battery management systems.