Federated mutual learning enables consistent hierarchical sorting of retired batteries across heterogeneous testing nodes

Abstract Accurate sorting is a prerequisite for the echelon utilization of retired power batteries, yet testing data are scattered across geographically distributed nodes that cannot share raw data, and the incoming battery populations differ substantially among nodes, i.e. the data are non-independent and identically distributed (non-IID). Here we propose a hierarchical sorting method based on federated mutual learning. Each node builds a local hierarchical model that combines prior-feature clustering with cluster-wise decision trees; model-output knowledge is exchanged through a shared unlabeled reference set, and distribution-discrepancy weighting together with confidence filtering enables bidirectional knowledge transfer without raw data leaving any node. Validation on 1,080 simulated retired batteries across four non-IID nodes shows that the global average sorting accuracy reaches 91.10%, 12.83 percentage points above local training alone, with a cross-node consistency of 97.13%; across 11 random-seed repeats the average accuracy is 90.17% ± 2.53%. The proposed method improves sorting accuracy and cross-node consistency under heterogeneous multi-node conditions.

Authors

Institutions

Publication Details

Journal
Research Square
Published
2026-11-16
DOI
https://doi.org/10.21203/rs.3.rs-11100592/v1
Primary Topic
Battery Modeling and Management
Type
preprint

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Federated mutual learning enables consistent hierarchical sorting of retired batteries across heterogeneous testing nodes

Lingrong Pang, Qiang Fan, Ye Huayang¹, Xiankui Wen et al.
3 links
Research Square
Battery Modeling and Management
preprint

Federated mutual learning enables consistent hierarchical sorting of retired batteries across heterogeneous testing nodes

Lingrong Pang, Qiang Fan, Ye Huayang¹, Xiankui Wen, Qiangqiang Liao
preprint en

Abstract

Abstract Accurate sorting is a prerequisite for the echelon utilization of retired power batteries, yet testing data are scattered across geographically distributed nodes that cannot share raw data, and the incoming battery populations differ substantially among nodes, i.e. the data are non-independent and identically distributed (non-IID). Here we propose a hierarchical sorting method based on federated mutual learning. Each node builds a local hierarchical model that combines prior-feature clustering with cluster-wise decision trees; model-output knowledge is exchanged through a shared unlabeled reference set, and distribution-discrepancy weighting together with confidence filtering enables bidirectional knowledge transfer without raw data leaving any node. Validation on 1,080 simulated retired batteries across four non-IID nodes shows that the global average sorting accuracy reaches 91.10%, 12.83 percentage points above local training alone, with a cross-node consistency of 97.13%; across 11 random-seed repeats the average accuracy is 90.17% ± 2.53%. The proposed method improves sorting accuracy and cross-node consistency under heterogeneous multi-node conditions.

Research Square
Shanghai University of Electric Power (CN), Guizhou Power Grid Co., Ltd. (China) (CN)
National Key Research and Development Program of China
Affordable and clean energy, Industry, innovation and infrastructure
Battery Modeling and Management
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon