Scholarlink

<p>Accurate forecasting of lithium-ion battery (LIB) retirement volumes is critical for optimizing recycling infrastructure and supporting China’s circular economy. However, existing studies typically rely on sales-based models that overlook battery chemistry differences and replacement dynamics. To address these limitations, this study proposes a novel forecasting framework that integrates Triad Feature Selection, Support Vector Regression, and the Weibull distribution. Triad Feature Selection identifies the principal determinants—namely the NEV market penetration rate and GDP—by integrating grey relational analysis, Pearson correlation analysis, and LASSO regression. SVR is then used to predict battery installation capacity, while the Weibull model simulates retirement behavior over time. The model achieves superior accuracy (MAE: 2.91 GWh for NMC, 17.35 GWh for LFP; R<sup>2</sup>: 0.9915 and 0.9724, respectively) compared to benchmark models. Scenario analysis reveals that under the medium scenario, NMC battery retirements peak at 140 GWh in 2042, while LFP batteries, driven by strong policy incentives and significant cost advantages, are projected to reach 852.35 GWh by 2050. This disaggregated and scenario-based forecasting framework provides robust insights for targeted recycling strategies, policy formulation, and sustainable resource management, aligning with China’s 2060 carbon neutrality goal.</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

Forecasting NMC and LFP battery retirement in China: a novel TFS-SVR-Weibull approach

  • 期刊论文

  • 发表日期：

    2025年11月18日

查看全文
  • Jian Zhang,
  • Xianzhong Mu

摘要

Accurate forecasting of lithium-ion battery (LIB) retirement volumes is critical for optimizing recycling infrastructure and supporting China’s circular economy. However, existing studies typically rely on sales-based models that overlook battery chemistry differences and replacement dynamics. To address these limitations, this study proposes a novel forecasting framework that integrates Triad Feature Selection, Support Vector Regression, and the Weibull distribution. Triad Feature Selection identifies the principal determinants—namely the NEV market penetration rate and GDP—by integrating grey relational analysis, Pearson correlation analysis, and LASSO regression. SVR is then used to predict battery installation capacity, while the Weibull model simulates retirement behavior over time. The model achieves superior accuracy (MAE: 2.91 GWh for NMC, 17.35 GWh for LFP; R2: 0.9915 and 0.9724, respectively) compared to benchmark models. Scenario analysis reveals that under the medium scenario, NMC battery retirements peak at 140 GWh in 2042, while LFP batteries, driven by strong policy incentives and significant cost advantages, are projected to reach 852.35 GWh by 2050. This disaggregated and scenario-based forecasting framework provides robust insights for targeted recycling strategies, policy formulation, and sustainable resource management, aligning with China’s 2060 carbon neutrality goal.