Robust circuit parameter estimator in noise-outliers environments based on RANSAC merged with RLS for reliable LiFePO4 state-of-charge

Abstract

As the electric vehicles (EVs) market continues to expand and evolve, ensuring the optimal performance, safety, and efficiency of EVs becomes increasingly critical. In real-world EV driving scenarios, the accuracy of state-of-charge (SoC) estimation, based on equivalent circuit models, often suffers from outliers and measurement noise. To address this challenge, A novel approach that integrates the random sample consensus (RANSAC) algorithm with recursive least squares (RLS) estimation is proposed. This integrated RANSAC-RLS method robustly determines the optimal model, even in environments where outliers and noise occur. Furthermore, it adeptly manages anomalies, which pose challenges for conventional methods while facilitating efficient and precise appraisal in practical driving environments. Comparative analysis shows that the RANSAC-RLS method is effective against spurious data and yields accurate SoC estimation under real-world conditions. Additionally, analysis of the computational trade-off revealed that although RANSAC-RLS requires approximately twice the execution time of standard RLS, it maintains a root mean square error (RMSE) below 1.725 % even in noisy and outlier-contaminated environments, significantly enhancing robustness.