Lithium-ion battery state of charge prediction based on machine learning approach

With the extensive utilization of lithium ion batteries as renewable energy source in electronics devices, smart network and electric vehicles, supplementary enhancements in the performance of lithium-ion batteries and accurate prediction of state of charge (SOC) are still a great challenge to battery research and innovation community. Machine learning (ML), which is one of the essential tools of artificial intelligence, is promptly changing many areas with its capability to learn from provided data and solve multifaceted tasks, and it has emerged as a new method used to solve research issues in the area of lithium ion batteries. In this paper, we investigate the relationship between input factors including current, voltage and temperature, and predicted SOC of lithium ion battery. The effectiveness of three ML models - linear regression, Gaussian process regression (GPR) and support vector machine (SVM) were assessed and compared. It was found that the predictions made by these models accurately matched the data from experiments.