Prediction and evaluation of health state for power battery based on Ridge linear regression model

Area of Science:

• * Power Systems Engineering
• * Data Science
• * Battery Technology

Background:

• * State of Health (SOH) is critical for power battery performance and cycle efficiency.
• * Accurate SOH prediction is essential for effective battery management systems.
• * Data-driven approaches offer a promising avenue for SOH estimation.

Purpose of the Study:

• * To apply data-driven algorithms for accurate power battery health estimation.
• * To develop and validate a methodology for SOH prediction using real-world vehicle data.
• * To compare the performance of different regression models for SOH prediction.

Main Methods:

• * Utilized actual operating vehicle data from a monitoring platform.
• * Implemented a charging event segmentation algorithm and K-means clustering for slow charging event extraction.
• * Performed feature engineering using Pearson/Spearman coefficients and one-hot encoding.
• * Employed the Ridge linear regression model for SOH prediction.

Main Results:

• * Achieved a Mean Absolute Error (MAE) of less than 5%, meeting practical application requirements.
• * Ridge regression demonstrated superior performance compared to Linear Regression, Lasso, and Elastic Net.
• * Identified Ridge regression as suitable for low-dimensional feature scenarios without cell-specific data.

Conclusions:

• * The proposed data-driven approach effectively predicts power battery State of Health.
• * Ridge linear regression is a viable and accurate model for SOH estimation in specific contexts.
• * The findings support the practical implementation of advanced algorithms for battery management.