Enhancing battery capacity estimation accuracy using the bald eagle search algorithm

• Published in: Energy reports 2023-11, Vol.10, p.2710-2724
• Main Authors: El Marghichi, Mouncef, Loulijat, Azeddine, Dangoury, Soufiane, Chojaa, Hamid, Abdelaziz, Almoataz Y., Mossa, Mahmoud A., Hong, Junhee, Geem, Zong Woo
• Format: Article
• Language: English
• Subjects: Bald eagle search optimization algorithm; Battery aging; Battery management systems; Capacity; Electric vehicles; Lithium-ion
• ISSN: 2352-4847; 2352-4847
• DOI: 10.1016/j.egyr.2023.09.082

Accurate assessment of metrics such as state of health (SOH) is of paramount importance in effective battery management systems (BMS), given the propensity of batteries to experience capacity degradation with aging. This research endeavors to significantly enhance the precision of battery capacity estimation by effectively mitigating the inherent uncertainties associated with state of charge (SOC) estimation and measurement. To address this challenge, we introduce an innovative approach leveraging the bald eagle search algorithm (BES), a method inspired by the systematic hunting behavior of bald eagles. BES strategically navigates the search space, identifying and selecting promising solutions through fitness evaluations. Our principal aim, utilizing the inherent capabilities of BES, is to pinpoint the optimal candidate that minimizes a designated cost function, while ensuring real-time cell capacity updates facilitated by the incorporation of a memory forgetting factor. The distinctiveness of this study is twofold: firstly, the strategic integration of the BES algorithm within the context of battery capacity optimization, and secondly, the inclusion of a memory forgetting factor to enhance real-time capacity estimations. The efficacy of our approach is rigorously substantiated through validation using NASA’s Prognostic Data, along with three battery scenarios for plug-in hybrid and electric vehicles. BES consistently outperformed four aggressive algorithms, demonstrating heightened accuracy with a peak error rate of only 1.06% in the most demanding scenario. Furthermore, the predictive performance measures remained consistently below 0.41%, underscoring the robustness of our proposed methodology.