Use of Machine Learning to Predict Lithium-Ion Cell Voltages During Charging and Discharging

This paper uses machine learning to predict either the voltage, current, or state of charge (SOC) during a charging/discharging of a Li Ion cell. Due to the non-linear characteristics of a Li Ion cells, machine learning is used to create a model for prediction. Predicting the Li Ion cell characteristics is useful for products in determining the end of discharge (EOD) levels under different loading conditions. The model used is a decision tree regressor which is implemented using Python. This model was also compared against a random forest regressor model. The decision tree regressor model has a faster response time and 100 times smaller model size when compared to the random forest regressor model. To train the model approximately 2.8 million data points of charging and discharging data of a Sanyo 18650 3.7V 2.6Ah Li-Ion cell is used. It was found that the accuracy of the model is limited due to the limited ranges of the training data. This model can be used to predict the behavior of Lithium-Ion cells/batteries under different loading conditions.