In battery management systems, online estimation for state of charge (SOC), state of health (SOH), capacity fade, etc., are based on experimentally-derived equivalent circuit models (ECM), and using lookup tables via publicly available datasets for lithium-ion (Li-ion) based battery cells. Extraction of these parameters for estimating SOC, SOH, etc., undergo extensive testing, which are incorporated into known definitions or through the use of machine learning (ML), deep learning (DL), or neural network (NN) based algorithms for parameters of interest (POI). In this paper, an effort to develop a hybrid ensemble model with a gated recurrent unit (GRU) for time-series computations, and a convolutional neural network (CNN) for feature extraction, is made for estimation of SOH and capacity fade under dynamic electric vehicle (EV) driving conditions. The LG18650HG2 dataset[1] will be used for parameter extraction along training and testing of the algorithm with mixed drive cycles. The SOC data is incorporated into the model to estimate SOH and capacity fade over the entire battery life cycle. Conclusively, the algorithm is tested on the generated dataset and scored using the RMSE performance metric.
Online State-of-Health and Capacity Fade Estimation Using Machine-Learning-, Deep-Learning-, and/or Neural-Network-Based Algorithms
19.06.2024
1584818 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Attenuation prediction for fade mitigation using neural network with in situ learning algorithm
Online Contents | 2012
|Deep Learning and Neural Network Algorithms
Springer Verlag | 2021
|SAE Technical Papers | 2023
|