"Full factorial design of experiments dataset for parallel-connected lithium-ion cells imbalanced performance investigation".

This paper shares an experimental dataset of lithium-ion battery parallel-connected modules. The campaign, conducted at the Stanford Energy Control Laboratory, employs a comprehensive full factorial Design of Experiment methodology on ladder-configured parallel strings. A total of 54 test conditions were investigated under various operating temperatures, cell-to-cell interconnection resistance, cell chemistry, and aging levels. The module-level testing procedure involved Constant Current Constant Voltage (CC-CV) charging and Constant Current (CC) discharge. Beyond monitoring total module current and voltage, Hall sensors and thermocouples were employed to measure the signals from each individual cell to quantify both current and temperature distribution within each tested module configuration. Additionally, the dataset contains cell characterization data for every cell (i.e. NCA Samsung INR21700-50E and NMC LG-Chem INR21700-M50T) used in the module-level experiments. This dataset provides valuable resources for developing battery physics-based, empirical, and data-driven models at single cell and module level. Ultimately, it contributes to advance our understanding of how cell-to-cell heterogeneity propagates within the module and how that affects the overall system performance.

Evaluating the feasibility of batteries for second-life applications using machine learning.

This article presents a combination of machine learning techniques to enable prompt evaluation of retired electric vehicle batteries as to either retain those batteries for a second-life application and extend their operation beyond the original and first intent or send them to recycling facilities. The proposed algorithm generates features from available battery current and voltage measurements with simple statistics, selects and ranks the features using correlation analysis, and employs Gaussian process regression enhanced with bagging. This approach is validated over publicly available aging datasets of more than 200 with slow and fast charging cells, with different cathode chemistries, and for diverse operating conditions. Promising results are observed based on multiple training-test partitions, wherein the mean of Root Mean Squared Percent Error and Mean Percent Error performance errors are found to be less than 1.48% and 1.29%, respectively, in the worst-case scenarios.

Lithium-ion battery aging dataset based on electric vehicle real-driving profiles.

This paper describes the experimental dataset of lithium-ion battery cells subjected to a typical electric vehicle discharge profile and periodically characterized through diagnostic tests. Data were collected at the Stanford Energy Control Laboratory, at Stanford University. The INR21700-M50T battery cells with graphite/silicon anode and Nickel-Manganese-Cobalt cathode were tested over a period of 23 months according to the Urban Dynamometer Driving Schedule (UDDS) discharge driving profile and the Constant Current (CC)-Constant Voltage (CV) charging protocol designed at different charging rates - ranging from C/4 to 3C. Ten (10) cells are tested in a temperature-controlled environment (23 ∘ C). A periodic assessment of battery degradation during life testing is accomplished via Reference Performance Tests (RPTs) comprising of capacity, Hybrid Pulse Power Characterization (HPPC), and Electrochemical Impedance Spectroscopy (EIS) tests. The dataset allows for the characterization of battery aging under real-driving scenarios, enabling the development of models and management strategies in electric vehicle applications.

Experimental data of lithium-ion batteries under galvanostatic discharge tests at different rates and temperatures of operation.

In this paper, we report data from lithium battery cells from: Panasonic NCR-18650B (3350 mAh), LG Chem INR21700-M50 (4850 mAh) and A123 Systems ANR26650m1-B (2500 mAh). They own the same anode composition, graphite-based, and different cathode chemistry: lithium-nickel-cobalt aluminum-oxide (NCA), lithium-nickel-manganese-cobalt-oxide (NMC) and lithium-iron-phosphate (LFP), respectively. In this study, six cell samples were tested for each chemistry. The experiments consist in fully discharging the cells from 100% state-of-charge until the cell cutoff discharge voltage. The discharge is performed under controlled temperature conditions, namely 5 °C, 25 °C and 35 °C, and subjecting the battery cells to galvanostatic discharge rates ranging from C/20 to 5C, for NCA and NMC, and from C/20 to 20C, for LFP chemistry. The IncuMax IC-500R thermal chamber provides the reference temperature to the cell. The input current profiles are configured via the MITS Pro-software, and transmitted through the TCP/IP connection to the Arbin measurement system and the Arbin LBT21024. Voltage, current and cell surface temperature are measured on each cell and for each experiment to characterize the cells in terms of discharge capacity, discharge efficiency, thermal robustness, specific energy and specific power. A comprehensive analysis of the data is found in [1].

Pushing the Eenvelope in Battery Estimation Algorithms.

Accurate estimation of lithium-ion battery health will (a) improve the performance and lifespan of battery packs in electric vehicles, spurring higher adoption rates, (b) determine the actual extent of battery degradation during usage, enabling a health-conscious control, and (c) assess the available battery life upon retiring of the vehicle to re-purpose the batteries for "second-use" applications. In this paper, the real-time validation of an advanced battery health estimation algorithm is demonstrated via electrochemistry, control theory, and battery-in-the-loop (BIL) experiments. The algorithm is an adaptive interconnected sliding mode observer, based on a battery electrochemical model, which simultaneously estimates the critical variables such as the state of charge (SOC) and state of health (SOH). The BIL experimental results demonstrate that the SOC/SOH estimates from the observer converge to an error of 2% with respect to their true values, in the face of incorrect initialization and sensor signal corruption.