The State-of-Charge Estimation of LFP batteries using Random Forest and XGBoost tuned by a Randomized Search Algorithm

This paper presents a method for estimating the state of charge (SoC) of lithium iron phosphate (LFP) batteries using extreme gradient boosting (XGBoost) and random forest (RF) models. Unlike other estimation techniques, this approach does not require prior knowledge of the battery’s model or parameters. Instead, it leverages variations in ambient temperature, previous battery voltage, and current data to accurately predict the SoC. The proposed method involves adjusting the hyperparameters, which are optimized using a Randomized Search Algorithm. We compare the predictions of both tuned and untuned models across various driving cycle test cases, including the Dynamic Stress Test (DST), Urban Dynamometer Driving Schedule (UDDS), and Federal Urban Driving Schedule (FUDS). The technique has been evaluated with an LFP battery, focusing on the SoC predictions from the tuned models of RF and XGBoost. Results demonstrate high accuracy, simplicity, and robustness of the proposed method.