Scholarlink

<p>Power batteries exhibit significant time varying non linearity, and there are significant differences in the State of Charge (SOC) estimates obtained from different equivalent circuit models. A method for estimating the SOC of lithium - batteries based on multi - model probability fusion is proposed. First, according to model research, the Thevenin model, DP model, and fractional order model (FOM) are selected, and the rationality of these models is verified within the proposed framework. Second, the FFRLS algorithm is optimized into the VFFRLS algorithm. Considering the computational accuracy and efficiency of multi - model parameter identification, the VFFRLS algorithm is optimized to reduce the computational amount. Regarding the low utilization rate of historical data when using AUKF for SOC estimation, as well as the problem that the algorithm may fail in simulation due to the non positive definiteness of the covariance matrix, the SVD - AMIUKF algorithm is proposed by combining the multi innovation (MI) theory, singular value decomposition (SVD), and AUKF algorithm. The advantages and disadvantages of the three equivalent circuit models studied in different intervals are experimentally verified, and the results show that they can be adapted to the multi - model framework structure. In the multi - model fusion framework, the SOC estimates obtained from these three models are dynamically weighted and allocated based on the Bayesian probability method to complete the SOC estimation. Finally, the proposed model is verified under different experimental conditions. The experimental results show that, under the conditions of 15&#xa0;°C, 25&#xa0;°C, and 40&#xa0;°C in daylight saving time, the Mean Absolute Percentage Error (MAPE) of this model is 0.12%, 0.753%, and 1.12% respectively. Compared with the estimation results of the Thevenin model, the DP model, and the FOM, the accuracy increases by 94.2%, 90%, and 89.1% respectively at 15&#xa0;°C; at 25&#xa0;°C, the accuracy increases by 69.8%, 12.9%, and 55% respectively; at 40&#xa0;°C, the accuracy increases by 69.9%, 12.9%, and 27.3% respectively. Under the FUDS conditions at 15&#xa0;°C, 25&#xa0;°C, and 40&#xa0;°C, the MAPE of this model is 0.15%, 0.28%, and 0.37% respectively. Compared with the above three models, the accuracy increases by 85%, 64.3%, and 80.5% respectively at 15&#xa0;°C; at 25&#xa0;°C, the estimation accuracy increases by 70.5%, 69.8%, and 74.5% respectively; at 40&#xa0;°C, the estimation accuracy increases by 70.6%, 81.3%, and 55.9% respectively, verifying the effectiveness of this method.</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

State of charge estimation of power batteries based on a multi-model fusion method

  • 期刊论文

  • 发表日期：

    2026年5月23日

查看全文
  • Jiabo Li,
  • Yufei Luo,
  • Di Tian,
  • Zhonglin Sun,
  • Yuanjin Zhang

摘要

Power batteries exhibit significant time varying non linearity, and there are significant differences in the State of Charge (SOC) estimates obtained from different equivalent circuit models. A method for estimating the SOC of lithium - batteries based on multi - model probability fusion is proposed. First, according to model research, the Thevenin model, DP model, and fractional order model (FOM) are selected, and the rationality of these models is verified within the proposed framework. Second, the FFRLS algorithm is optimized into the VFFRLS algorithm. Considering the computational accuracy and efficiency of multi - model parameter identification, the VFFRLS algorithm is optimized to reduce the computational amount. Regarding the low utilization rate of historical data when using AUKF for SOC estimation, as well as the problem that the algorithm may fail in simulation due to the non positive definiteness of the covariance matrix, the SVD - AMIUKF algorithm is proposed by combining the multi innovation (MI) theory, singular value decomposition (SVD), and AUKF algorithm. The advantages and disadvantages of the three equivalent circuit models studied in different intervals are experimentally verified, and the results show that they can be adapted to the multi - model framework structure. In the multi - model fusion framework, the SOC estimates obtained from these three models are dynamically weighted and allocated based on the Bayesian probability method to complete the SOC estimation. Finally, the proposed model is verified under different experimental conditions. The experimental results show that, under the conditions of 15 °C, 25 °C, and 40 °C in daylight saving time, the Mean Absolute Percentage Error (MAPE) of this model is 0.12%, 0.753%, and 1.12% respectively. Compared with the estimation results of the Thevenin model, the DP model, and the FOM, the accuracy increases by 94.2%, 90%, and 89.1% respectively at 15 °C; at 25 °C, the accuracy increases by 69.8%, 12.9%, and 55% respectively; at 40 °C, the accuracy increases by 69.9%, 12.9%, and 27.3% respectively. Under the FUDS conditions at 15 °C, 25 °C, and 40 °C, the MAPE of this model is 0.15%, 0.28%, and 0.37% respectively. Compared with the above three models, the accuracy increases by 85%, 64.3%, and 80.5% respectively at 15 °C; at 25 °C, the estimation accuracy increases by 70.5%, 69.8%, and 74.5% respectively; at 40 °C, the estimation accuracy increases by 70.6%, 81.3%, and 55.9% respectively, verifying the effectiveness of this method.