由多目標函數識別鋰離子電池的電化學模型參數

Parameter Identification of an Electrochemical Model for Lithium-ion Batteries by Using Multi-Objective Function

林韋任(Wei-Jen Lin)

指導教授 ： 陳國慶

國立臺灣大學/工學院/應用力學研究所/碩士(2021年)

https://doi.org/10.6342/NTU202101759

摘要

鋰離子電池為現在世界主要所用的儲能元件，未來經過使用後替換下來的電池數量會日益增加，延續電池使用壽命並且避免資源的浪費，因此本研究建立全新識別鋰離子電池所用的電化學模型參數識別模型基準，使用放電曲線與ICA曲線作為目標函數識別，電池使用狀態下的電化學模型參數。
本研究使用NMC商用柱狀電池進行充放電循環實驗，由實驗所獲得的資料作為參數識別模型逼近擬合的目標值。因每個參數對目標函數的影響程度不同，採用定量增減的方式量化參數的敏感程度，研究中發現每個目標函數的高敏感參數皆不同。使用NSGA演算法探討不同目標函數的識別結果，研究中表明了ICA曲線的峰谷特徵的目標函數較整段ICA曲線的識別效果差。使用NSGA-III識別15個電化學參數時，放電曲線最低誤差可達到0.06588(V)，ICA曲線最低誤差可達到0.31817(Ah/V)。
為解決最佳化演算法耗時的參數識別時間，提出另一個參數識別模型，將深度學習與電化學模型結合，透過實驗取得的電壓、電流與容量訓練該深度學習模型，最終識別結果表明，該深度學習模型仍保有一定的低識別誤差水準，並且能節省大量可觀的計算時間。

關鍵字

鋰離子電池 ； 電化學模型 ； 多目標函數 ； NSGA ； 機器學習

並列摘要

Lithium-ion batteries are currently the main energy storage components used in the world. In the future, the substantial number of used batteries are eliminated. To avoid waste of resources and extending battery life, this study established a new electrochemical model to determine the benchmark of the lithium-ion battery parameter identification model. The model use the discharge curve and the ICA curve as the objective function to determine the electrochemical model parameters under battery usage.
In this study, NMC commercial cylindrical batteries were used in the charge/discharge cycle experiment, and the data obtained from the experiment was used as a parameter identification model to approximate the target value. Since each parameter has a different degree of influence on the objective function, quantitative increase and decrease methods are used to quantify the sensitivity of the parameter. In the research, it is found that the highly sensitive parameters of each objective function are different. The NSGA algorithm is used to explore the recognition results of different objective functions. Studies have shown that the objective function of the peak and valley characteristics of the ICA curve is not as effective as the entire ICA curve. When using NSGA-III to identify 15 electrochemical parameters, the lowest error of the discharge curve can reach 0.06588 (V), and the lowest error of the ICA curve can reach 0.31817 (Ah/V).
In order to solve the time-consuming parameter identification time of the optimization algorithm, another parameter identification model is proposed. This model combines deep learning with an electrochemical model, and trains the deep learning model through the voltage, current and capacity obtained in the experiment. . The final recognition results show that the deep learning model still maintains a certain low recognition error level and can save a lot of considerable computing time.

並列關鍵字

Lithium-ion battery ； Electrochemical model ； Multi-objective function ； NSGA ； Machine learning

參考文獻

[1] Ephrem Chemali, Phillip J. Kollmeyer, Matthias Preindl, Ali Emadi, State-of-charge estimation of Li-ion batteries using deep neural networks: A machine learning approach, Journal of Power Sources, Volume 400, Pages 242-255, 2018.

[2] Jungsoo Kim, Jungwook Yu, Minho Kim, Kwangrae Kim, Soohee Han, Estimation of Li-ion Battery State of Health based on Multilayer Perceptron: as an EV Application, IFAC-PapersOnLine, Volume 51, Issue 28, Pages 392-397, 2018.

[3] M. Doyle, T.F. Fuller, and J. Newman, Modelling the Galvanostatic Charge and Discharge of the Lithium/Polymer/Insertion Cell, Journal of the Electrochemical Society, 140(6), 1992.

[4] Liqiang Zhang, Lixin Wang, Gareth Hinds, Chao Lyu, Jun Zheng, Junfu Li, Multi-objective optimization of lithium-ion battery model using genetic algorithm approach, Journal of Power Sources, Volume 270, Pages 367-378, 2014.

[5] Minho Kim, Huiyong Chun, Jungsoo Kim, Kwangrae Kim, Jungwook Yu, Taegyun Kim, Soohee Han, Data-efficient parameter identification of electrochemical lithium-ion battery model using deep Bayesian harmony search, Applied Energy, Volume 254, 113644, 2019.