Scholarlink

<p>For several critical applications like electric vehicles and uncrewed aerial vehicles, the state of health (SoH) and remaining useful life (RUL) of the battery must be accurately estimated and thus play an important role. Typically, this problem has been addressed using machine learning methods, often those based on neural networks. These usually require massive training data and entail an elevating computational cost. In the specific context, this paper presents a novel approach based on visibility graphs for time series prediction and its application to the problem of SoH and RUL estimation in lithium-ion batteries. A case study based on the NASA PCoE (Prognostics Center of Excellence) dataset was developed to check its feasibility and applicability, and the results were compared to the results of a neural network-based method reported in the literature. It resulted in an excellent performance as it approximated a mean absolute percentage error (MAPE) value of around 6.24%, showcasing its accuracy. Using Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and Mean Averaged Error (MAE) metrics, performance was also evaluated and compared with other methods. The new method proposed achieved superior performance in many cases, showcasing its robustness and efficacy under various conditions, even when compared to computationally expensive methods.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Visibility Graph-Based Multi-Step Forecasting Method for State-of-Health and Remaining Useful Life Estimation of Lithium-Ion Battery

  • 期刊论文

  • 发表日期：

    2025年6月6日

查看全文
  • Luiz Rogério de Freitas Júnior,
  • Frederico Gadelha Guimarães

摘要

For several critical applications like electric vehicles and uncrewed aerial vehicles, the state of health (SoH) and remaining useful life (RUL) of the battery must be accurately estimated and thus play an important role. Typically, this problem has been addressed using machine learning methods, often those based on neural networks. These usually require massive training data and entail an elevating computational cost. In the specific context, this paper presents a novel approach based on visibility graphs for time series prediction and its application to the problem of SoH and RUL estimation in lithium-ion batteries. A case study based on the NASA PCoE (Prognostics Center of Excellence) dataset was developed to check its feasibility and applicability, and the results were compared to the results of a neural network-based method reported in the literature. It resulted in an excellent performance as it approximated a mean absolute percentage error (MAPE) value of around 6.24%, showcasing its accuracy. Using Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and Mean Averaged Error (MAE) metrics, performance was also evaluated and compared with other methods. The new method proposed achieved superior performance in many cases, showcasing its robustness and efficacy under various conditions, even when compared to computationally expensive methods.