Abstract
6G networks raise device energy demands to a new level. Meeting these demands requires better cathode materials and smarter battery management. This paper proposes a hybrid ANN NSGA-II framework for the multi-objective optimization of high-nickel NMC cathodes, where NMC denotes nickel, manganese, and cobalt. The approach combines an artificial neural network surrogate with the Non-dominated Sorting Genetic Algorithm II to search the composition and operating space efficiently. It maximizes initial capacity IC and capacity retention rate CRR while reducing cobalt usage and lithium excess. The method advances both learning and optimization. First, an extrapolation suppression strategy keeps the search inside the reliable prediction domain of the surrogate. Second, a Dirichlet initialization generates diverse yet feasible starting populations with a bias toward nickel-rich regions. Third, a bounded simplex repair enforces chemical composition constraints and preserves valid ratios among nickel, cobalt, manganese, and optional dopants. Together, these elements yield stable convergence and credible Pareto fronts. We also bridge materials and systems. A load-template mapping translates IC, CRR, voltage window, and rate descriptors into device-level indicators under representative 6G duty cycles, such as heavy, medium, and light profiles. The resulting Pareto-optimal designs show higher energy delivery at beginning of life and improved retention over cycling, which extends runtime for 6G devices and small cells. The framework links machine learning with evolutionary search and connects material choices to network-driven use cases. It supports battery management policies that align with 6G power bursts and standby periods, and it offers a practical path to nickel-rich, cobalt-lean cathodes with longer service life.Keywords: 6G; machine learning; ANN; ANN NSGA-II; battery management; multiobjective optimization.
| Date of Award | 15 Nov 2025 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | C.F. Kwong (Supervisor), Zheng Chu (Supervisor) & David Chieng (Supervisor) |