Deep Learning Framework for Lithium-ion Battery State of Charge Estimation: Recent Advances and Future Perspectives

Accurate state of charge (SOC) constitutes the basis for reliable operations of lithium-ion batteries. The deep learning technique, a game changer in many fields, has recently emerged as a promising solution to accurate SOC estimation, particularly in the era of battery big data consisting of field and testing data. It enables end-to-end SOC estimation using raw battery operating data as input for various battery chemistries under different operating conditions. This article first identifies SOC estimation problems and introduces a general framework of deep learning-based SOC estimation and then reviews the recent applications of deep learning in SOC estimation with a focus on the model structure. Three kinds of prevalent deep neural networks (DNNs) are explained, including the fully connected neural network, recurrent neural network and convolutional neural network. Furthermore, advanced applications such as transfer learning and the combination of deep learning with other methods are discussed. Finally, challenges and future opportunities regarding data collection, model development and real-world applications are systematically examined to give insights into this area. Apart from SOC estimation, the present study is also promising to inspire advances in other battery management tasks.

Automated summary

Accurate state of charge (SOC) is crucial for the reliable operations of lithium-ion batteries. Deep learning technique has recently emerged as a promising solution for accurate SOC estimation, especially in the era of battery big data. This article reviews the deep learning-based SOC estimation framework and the recent applications of deep learning in SOC estimation, focusing on the model structure. It also discusses advanced applications like transfer learning and the combination of deep learning with other methods. Finally, it examines the challenges and future opportunities in data collection, model development, and real-world applications in this area.