What question did this study set out to answer?

To optimize the design of liquid-cooled plates (LCPs) for efficient thermal management of Li-ion batteries.

February 28, 2026Case Studies in Thermal Engineering1 citationsOpen Access

Pareto-Based Multi-Objective Design of Battery Cold Plates with Leaf-Inspired Bifurcations Using Arithmetic Optimization Algorithm and WASPAS Technique

HGHatem Gasmi BLBorhen Louhichi ABAli Basem

Key Points

To optimize the design of liquid-cooled plates (LCPs) for efficient thermal management of Li-ion batteries.
Integrated AI workflow with four stages: data analysis, predictive modeling, optimization, and design ranking
Predictive modeling using GMDH-type artificial neural networks (ANN)
Pareto-based optimization via arithmetic optimization algorithm (MOAOA) and particle swarm optimization (MOPSO)
Final design ranking using WASPAS decision-making method
Achieved thermal resistance of 0.25–0.35 K/W with pressure drops of 10–25 Pa
Identified balanced designs at mass flow rates of 0.8–1.5 g/s
GMDH models showed strong predictive accuracy (R² > 0.99) for thermal resistance and pressure drop
Provided context-specific optimal designs ranging from low thermal resistance (0.1793 K/W) to energy-efficient solutions

Abstract

Efficient thermal management is essential for ensuring the safety, reliability, and longevity of Li-ion batteries, yet liquid-cooled plate (LCP) designs face persistent challenges in balancing cooling efficiency with hydraulic performance. This study presents an AI-based integrated workflow for the design and optimization of LCPs featuring leaf-inspired bifurcation geometries. The framework consists of four sequential stages: data analysis, predictive modeling using GMDH-type ANN, Pareto-based optimization via multi-objective arithmetic optimization algorithm (MOAOA) and multi-objective particle swarm optimization (MOPSO), and final design ranking with the WASPAS decision-making method. The GMDH models demonstrated high predictive accuracy, achieving R 2 > 0.99 for thermal resistance and pressure drop. Both MOAOA and MOPSO produced nearly identical Pareto fronts, confirming robustness in capturing trade-offs between thermal and hydraulic performance. Optimized inputs revealed balanced designs at mass flow rates of 0.8–1.5 g/s, channel widths of ∼3.7–3.9 mm, and heights of 2.4–2.5 mm, achieving thermal resistance of 0.25–0.35 K/W with pressure drops of 10–25 Pa, ensuring efficient cooling without excessive hydraulic penalties. Decision analysis revealed context-specific optimal designs, ranging from ultra-low thermal resistance (0.1793 K/W) at the cost of high pressure drop (106.81 Pa) to energy-efficient solutions with minimal pumping penalties (ΔP = 2.14 Pa).

问 AI

Bookmark

View Full Paper