BazeBai/ECM_features

Equivalent Circuit Model(ECM) features

The problem to solve

Lithium-ion (Li-ion) batteries have been widely used in electric vehicles (EVs), smart grids, and portable electronics owing to their merits (e.g., high energy density, low self-discharge rate, long cycle life, etc.). However, the performance of Li-ion batteries gradually deteriorates with repeated usage, increasing the safety risk. State of health (SOH) is a crucial indicator for supervising the operation of Li-ion batteries, which is defined as the ratio of the capacity at present to its initial value.

Type of data

I use the parameters of elements in ECM as features, As a simplified form, ECM utilizes the common circuit components to simulate the characteristics of the battery. Additionally, ECM is easier to implement yet has a simple model structure, which makes it difficult to reflect the complex physical and chemical reactions of the Li-ion batteries.

Review previous datasets

There is so such a public dataset in the past. But there are some relevant datasets. Dataset provided by Ref., is obtained from 12 commercially available 45 mAh Eunicell LR2032 coin cells, where a continuous charge-discharge cycle is performed. The 12 cells are tested in three climate chambers at 25 °C (25C01–25C08), 35 °C (35C01-35C02) and 45 °C (45C01-45C02), respectively. During the aging experiment, the cells first undergo the constant current-constant voltage (CC-CV) charging at a 1C rate until the voltage reaches 4.2 V. Subsequently, the cells are discharged in constant current (CC) mode until the voltage decreases to 3V. EIS measurement is conducted by an excitation current of 5 mA, with the frequency ranging from 0.02 Hz to 20 kHz. And my dataset is generated based on this dataset. There is another EIS dataset, Dataset provided in paper "Lithium-ion battery aging dataset based on electric vehicle real-driving profiles." consists of ten INR21700-M50T cells at 23℃. These cells are charged in CC-CV mode at different C-rates ranging from 0.25C to 3C until 4.2V. Subsequently, CC discharge with 0.25C is used to bring the cells to 80% SOC and the Urban Dynamometer Driving Schedule (UDDS) is employed to cycle the SOC from 80% to 20%. After 25 or 50 cycles, the measurement of EIS is performed at 20%, 50%, and 80% SOC, with the frequency from 0.01Hz to 10kHz.

Power Analysis

Power analysis is used to determine the minimum sample size required for my study.

1.     Define the effect size as 0.18.
   

2.    Define alpha (α) as 0.05.
   

3.    Define power as 0.8.
   

4.    Choose t-test as the statistical method then calculate the required sample size with t-test.
   

Using tool G Power to calculate the sample size. After calculating, the sample size in each group is 245.

Data collection protocol

The batteries in the experiments should perform a continuous charge-discharge cycle. And the cells should be tested in three climate chambers at 25 C, 35 C, 45 C. During aging experiments, the cells first undergo the constant current-constant voltage (CC-CV) charging process at a 1C rate until the voltage reaches 4.2 V. Subsequently, the cells are discharged in constant current (CC) mode until the voltage decreases to 3V. The measurement of EIS is conducted in an excitation current of 5 mA, with the frequency ranging from 0.02 Hz to 20 kHz. The whole process must be constant, or the quality of this dataset will be innfluenced.

The model contains a series of elements such as equivalent resistance (Rs), capacitance (CPE) and Warburg impedance (Ws). These elements have different parameters. In the shown ECM, the parameters are (Rs, R1, CPE1-Q, CPE1-α, R2, CPE2-Q, CPE2-α, Ws-Z0, Ws-τ).

1. R Z=R
2. CPE(Q,α) Z=1/(Q(jω)^α )
3. W_s (Z_0,τ) Z=Z_0/√jωτ*tanh⁡√jωτ According to the formulas above, Li-ion batteries can be fitted by the EIS measurement using the least squares.

The repo link

https://github.com/Baze-Bai/baze_Dataset_project/tree/Dataset_project_baze

Data analysis of dataset

I use boxplot to analyze the features of battery 25C06 as an example, as the picture shown, there are some outliers in parameters R2 and Ws-τ. Some of them are too large, I can also use boxplot to strip out the outliers. I use scatterplot to observe the tendancy and relationship between features and SOH of batteries. Some of them show almost linear relationship, such as R2 and CPE2-Q. Some of them show weak relationship such as Ws-τ. I use heat map to analyze the Pearson correlation between features and SOH of batteries, The results are listed in heatmap where Pearson correlation coefficients between the nine extracted parameters and the battery SOH are visualized. Several features, such as Rs, R1, and CPE1-Q, show high negative correlation coefficients with SOH. CPE1-α, representing the constant phase element exponent, shows strong positive correlations with SOH in Dataset. Parameters such as R2, CPE2-Q, and Ws-τ demonstrate mixed correlations with SOH, varying between datasets. Ws-τ and CPE2-Q show stronger correlations in Dataset A. Besides, CPE1-α exhibits consistently low correlation coefficients the whole dataset, indicating its insensitivity to changes in SOH. It is noted that 25C01~45C02 are the cells in Dataset.

ethics statement

The dataset used consists of battery impedance spectroscopy data obtained from a publicly available collection. It does not contain any personally identifiable information, nor does it involve human subjects. The data collection process followed ethical guidelines and was carried out in compliance with all relevant regulations.

No informed consent from human participants was required, as the dataset does not contain sensitive personal data. All data were anonymized and de-identified prior to release.

The dataset is shared under a Creative Commons Attribution 4.0 International License (CC BY 4.0), and users are required to cite the source when using the data for further research or analysis.

There are no identified ethical concerns regarding the use of this dataset. However, researchers are advised to follow ethical guidelines when applying the dataset to further studies on energy storage systems.

open source license

Creative Commons Attribution 4.0 International License (CC BY 4.0)