State of Charge Estimation Using Data-Driven Models for Inverter-Based Systems

Lithium-ion batteries are playing a critical role in many applications nowadays, from small-scale electronic devices to grid-scale storage systems. To maintain its continuous operation and increase its life span, the state of charge of the battery should be determined to ensure safe operating conditions. Among the existing charge estimation methods, data-driven models are flourishing these days. This work presents a state of charge estimation for the eFlex 52.8V/5.4 kWh lithium iron phosphate battery pack at the Energy Systems Research Laboratory (ESRL) at FIU. Three different machine learning models were implemented and trained through Python code to achieve the most accurate SoC estimation. Although the three proposed models can efficiently estimate the battery’s SoC with an acceptable error percentage, the random forest regression model has proven its outperformance among the selected models with a percentage mean square error less than 0.01.