Kumar Anjan, Jasim Laith Hussein, Vijaya Padmanabha, Patel Dipak, Gowrishankar J, Sivaranjani R, Srivastava Ankur, Kundlas Mayank, Hota Sarbeswara, Hamidi Banafshe
Department of electronics and communication engineering, GLA University, Mathura, 281406, India.
Department of computers Techniques engineering, College of technical engineering, The Islamic University of Al Diwaniyah, Al Diwaniyah, Iraq.
Sci Rep. 2025 Aug 13;15(1):29616. doi: 10.1038/s41598-025-15359-0.
Addressing a key research gap in the lack of unified AI-based approaches that ensure both high predictive accuracy and informed design trade-offs, this study presents a synergistic methodology that integrates advanced intelligent techniques to optimize the thermal and hydraulic performance of liquid-cooled Li-ion battery thermal management systems (TMS). A TMS with asymmetric U-shaped channels was used as a case study to validate the intelligent proposed framework. Minimizing the maximum temperature (T), temperature difference (ΔT), and pressure drop (ΔP) were considered key design objectives. In the first phase, predictive modeling was performed using multilayer perceptron neural networks (MLPNN) optimized by three metaheuristic algorithms: cheetah optimizer (CO), grey wolf optimizer (GWO), and marine predators algorithm (MPA). The results showed outstanding accuracy, with the CO-MLPNN achieving R > 0.9999 for predicting T, while the GWO-MLPNN models performed best for ΔT (R > 0.9969) and ΔP (R > 0.9999). In the second phase, the multi-objective salp swarm algorithm (MOSSA) was benchmarked against MOPSO, with the latter producing more diverse and convergent Pareto fronts. The optimal solutions spanned T = 32.3-38 °C, ΔT = 3.7-5.1 °C, and ΔP = 15-50 Pa. Design trends indicated a preference for higher mass flow rates and longer channels, enhancing thermal regulation. The third phase employed the VIKOR method to generate 21 decision-making scenarios reflecting various stakeholder priorities, facilitating robust, application-specific design strategies. This novel framework not only improves the accuracy and comprehensiveness of battery TMS design but also promotes sustainability by supporting efficient, adaptive, and intelligent engineering decisions.
针对缺乏统一的基于人工智能的方法以确保高预测准确性和明智的设计权衡这一关键研究空白，本研究提出了一种协同方法，该方法集成了先进的智能技术，以优化液冷锂离子电池热管理系统（TMS）的热性能和水力性能。以具有不对称U形通道的TMS为例进行研究，以验证所提出的智能框架。将最小化最高温度（T）、温差（ΔT）和压降（ΔP）视为关键设计目标。在第一阶段，使用由三种元启发式算法优化的多层感知器神经网络（MLPNN）进行预测建模：猎豹优化器（CO）、灰狼优化器（GWO）和海洋捕食者算法（MPA）。结果显示出卓越的准确性，CO-MLPNN在预测T时R＞0.9999，而GWO-MLPNN模型在预测ΔT（R＞0.9969）和ΔP（R＞0.9999）方面表现最佳。在第二阶段，将多目标樽海鞘群算法（MOSSA）与多目标粒子群优化算法（MOPSO）进行基准测试，后者产生了更多样化且收敛的帕累托前沿。最优解范围为T = 32.3 - 38°C，ΔT = 3.7 - 5.1°C，ΔP = 15 - 50 Pa。设计趋势表明倾向于更高的质量流率和更长的通道，从而增强热调节。第三阶段采用VIKOR方法生成21种决策场景，反映不同利益相关者的优先级，促进稳健的、特定应用的设计策略。这个新颖的框架不仅提高了电池TMS设计的准确性和全面性，还通过支持高效、自适应和智能的工程决策促进了可持续性。
