SOC Errors at Room Temperature: Lithium-Ion Battery State-of-Charge Estimation Results

Huipin Lin published experimental results on June 4, 2026, comparing a hybrid time-series convolutional network with Unscented Kalman Filter against other deep learning models for battery state-of-charge estimation. The dataset likely contains error metrics from tests using the University of Maryland's Dynamic Stress Test (DST), US06, and Federal Urban Driving Scheme (FUDS) datasets. The proposed method achieved mean absolute error values between 1.015% and 1.470% under different driving conditions.

Use Cases

• Benchmarking SOC estimation algorithms based on reported MAE and RMSE values.
• Training or validating battery models using the DST, US06, and FUDS test condition results.
• Comparing the performance of convolutional networks, LSTMs, and GRUs for time-series battery data analysis.

Strengths

• Includes specific error metrics: MAE values of 1.305%, 1.470%, and 1.015% under three test conditions.
• Results are directly comparable, with the proposed method reducing SOC estimation error by approximately 7.6%–39.6% compared to baseline models.

Limitations

• Row count is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• The 9.5 KB file size suggests a very limited scope, likely containing summary results rather than raw time-series data.

Provenance

Source

Huipin Lin via figshare.

Collection Method

Experimental results from applying a hybrid machine learning framework to battery test datasets.

Freshness

Last updated 2026-06-04 17:34:30; freshness should be verified.