基于双层机器学习的非对称超级电容器跨表征EIS参数估计OA

Electrochemical impedance spectroscopy(EIS)is a crucial technique for characterizing the dynamic properties of asymmetric supercapacitors.However,for electrode materials with high noise levels,the traditional nonlinear least-squares method often fails to yield accurate fits owing to the difficulty in determining reliable initial impedance parameters.To address this challenge,this study proposes an intelligent EIS parameter identification method based on cross-characterization fusion and a two-layer machine learning architecture.This approach extracts physicochemical information from cyclic voltammetry and constant-current charge-discharge curves to construct a feature matrix,establishing an interpretable mapping between these features and key EIS parameters(e.g.,internal resistance and constant phase element parameters).After dimensionality reduction and feature enhancement through latent variable extraction and time-series normalization,a two-layer machine learning architecture integrating XGBoost,random forest,ridge regression,and elastic net regression is developed for training,enabling high-precision cross-characterization-based EIS parameter estimation.Experimental results demonstrate that the predicted parameters closely align with those derived directly from EIS,with high stability across different material systems,including cobalt-manganese-based electrodes.This study provides a reliable strategy for initializing EIS fitting in asymmetric supercapacitors,significantly improving convergence and accuracy,and offers a practical approach for identifying electrochemical impedance parameters under high-noise conditions.