Metaheuristic Optimization of an Empirical Lithium-Ion Battery Aging Model for Mobile-Device Operating Conditions

Abstract

Lithium-ion battery degradation is influenced by operating temperature, electrical current, depth of discharge (DoD), and charge–discharge cycling. This study presents a computational framework for identifying operating conditions associated with a reduced value of an empirical battery-ageing objective function. Unlike an initial monotonic formulation, the objective function used here combines deviation-based penalty terms for temperature and current, reproducing the experimentally reported U-shaped ageing response around a moderate operating point, with a physically motivated coupling between cycle count and depth of discharge expressed as a fixed monthly charge-throughput demand (N . DoD = K). This coupling yields a genuine, analytically verifiable interior optimum, unlike a purely monotonic formulation whose minimum would trivially occur at the boundary of the search domain. A genetic algorithm (GA) searching 50 candidates over 100 generations (crossover probability 0.7, mutation probability 0.01) converged to T ≈ 35.0 ◦C, I ≈ 1.50 A, DoD ≈ 50.1% and N ≈ 79.8 cycles/month, with fmin ≈ 1.25 × 10−2, matching the closed-form analytical optimum to four significant figures. Simulated annealing (SA), under an identical search domain, converged to T ≈ 34.4 ◦C, I ≈ 2.74 A, DoD ≈ 67.0%, N ≈ 59.7 cycles/month, with fmin ≈ 1.54 × 10−2 (about 24% higher than GA). A response-surface heatmap of f(T, I) at the optimal N and DoD confirms an elliptical, moderate-condition minimum consistent with the model equation. These numerical results are conditional on the illustrative coefficients used, which are neither fitted to measured cycling data nor tied to a named cell chemistry, and do not constitute a chemistry-validated ageing optimum. Experimental calibration and further validation are required before the identified conditions can be interpreted as general recommendations for mobile-device batteries.