Joelton Deonei Gotz

The advancement in automobile electrification has been causing concern about retired batteries. Since the battery cannot store and deliver energy for a high‐speed vehicle when its state of health (SOH) is below 80%, this battery must have another destination. A second application for energy storage from solar or wind farms can be a solution. However, a pack of second‐life batteries (SLB) is built in some cases from batteries from different first applications, which introduces the problem of the pack's heterogeneity—this makes applying a standard Battery Management System (BMS) difficult. As a solution, data‐driven and Machine Learning (ML) models have been used to help BMS for SLB, learning from the data. However, these models are specific to a single package. Therefore, this work applies Federated Machine Learning (FML) to build a global State‐of‐Charge (SOC) for SLB. The main intention is to develop a global model based on several learnings. This way, the proposed idea comprises three steps using 32 different SLBs with a capacity ranging from 1300 to 2500 mAh, which introduces a significant difficulty to the process. In the first step, a single BMS comprises a Raspberry Pi 3, and some sensors monitor the voltage, current, and surface temperature of a single SLB. The data are used to build a Long Short‐Term Memory network to predict the SOC. The process was done on 20 different SLB batteries. The results indicated an R ² above 99 and an RMSE below 3%. After that, each BMS sends the weight of each model to the server. In the second step, the server receives the weights and builds a global model using the ensemble learning approach. The model inserted an exclusion rule to avoid outliers in the decisions. The global model is tested with four different SLBs. The validation process shows an R ² of 94 and an RMSE of 5%. The global model has been downloaded to eight SLB batteries with BMS for testing. As the model has never seen the data of the eight batteries, the R ² was 96, and the RMSE was below 5.6% on average. The obtained results indicate that the proposed method is capable of achieving reliable SOC estimation under heterogeneous SLB operating conditions. This way, future investigations will involve inserting this idea into a pack of more than one battery.

The growing volume of retired batteries has created a new market for low-power applications with minimal requirements. This market enables the use of Second-life Batteries (SLB) with a remaining capacity below 80% and Third-life batteries (TLB) with a remaining capacity below 60%. However, the challenge lies in repurposing the pack, particularly in selecting the cells to compose the new pack. It happens because retired batteries have distinct characteristics that differ from one another, and this distinction cannot be observed in simple capacity experiments. Therefore, it is essential to understand battery characteristics during operation to classify their lifetime accurately. Therefore, this work reports 81 experiments on nine batteries, segmented by degradation stage (from first to third life). Each battery was discharged at constant current levels ranging from 0.45C A to 2.05C, with increments of 0.23C. The results indicated that, although the batteries are in the same group, they present different characteristics, especially when discharged at low or high current rates. Although capacity is commonly used as the primary parameter for grouping a new pack, this study demonstrates that cells with similar capacities exhibit distinct behaviors during discharge. Additionally, the operation of retired batteries should be optimized to avoid unnecessary capacity loss. When a battery with low remaining capacity is discharged at high current rates, the voltage discharge profile is influenced by polarization and internal resistance effects, and, as a result, the voltage cut-off reaches the standard 2.5V earlier, reducing the battery’s discharge time. Finally, the study presents a reclassification of the batteries based on empirical insights from the experiments conducted in this work.

When a lithium-ion battery’s state of health (SOH) falls below 80%, it must be retired from electric vehicle (EV) use and properly redirected. Two main options are available: recycling or repurposing for lower-demand applications. Since recycling remains costly and complex, repurposing is often the preferred option—resulting in what is known as a second-life battery (SLB). SLBs are potentially used in stationary energy storage systems or low-speed vehicles. However, SLBs require exceptional management to ensure safe operation, as their packs typically consist of heterogeneous batteries with varying characteristics. Therefore, intelligent control is essential for the charge/discharge process. Key indicators such as state of charge (SOC) and SOH are widely studied and critical for SLB operation. Additional indicators—state of energy (SOE), state of power (SOP), and state of temperature (SOT)—are also necessary for increasing the performance. Given the complexity and dynamic nature of SLB systems, advanced management strategies are needed, such as those based on machine learning (ML). In this work, ML was implemented using an ensemble learning (EL) approach to minimize overfitting and estimate the key parameters of SLBs. Thirty experiments were conducted on an SLB pack of four heterogeneous batteries to generate sufficient data for training and testing the EL models. The results showed low RMSE values for random forest and bagging models in estimating SOC, SOE, SOP, and SOT. This indicates that the approach is suitable for real-time SLB operation. Future work will improve SOP and SOT estimation and establish standard methods for calculating these parameters in SLBs.

Battery thermal management systems (BTMS) are crucial for regulating temperature in lithium-ion battery packs, especially in high-power applications like electric vehicles. This study presents a multi-objective optimization of a 1P7S LFP prismatic battery pack using phase change material (PCM) as a BTMS. Previous research has primarily focused on reducing maximum temperature rise; this study addresses the challenge of minimizing inter-cell temperature differences. The pack’s thermal behavior is modeled under various conditions, utilizing advanced multiphysics simulations and efficient optimization algorithms. The approach considers multiple objectives: minimizing temperature rise, reducing temperature gradients, and optimizing PCM utilization. Results demonstrate the effectiveness of the optimized PCM-based BTMS in maintaining uniform temperature distribution across prismatic cells during high-power discharge. A Pareto front of optimal solutions is presented to balance thermal performance, system complexity, and cost. This research contributes to developing safer and more reliable battery systems for high-power applications.

Lithium-ion batteries (LIB) are the mainstream technology for energy storage in several industrial segments, such as mobility and stationary systems for solar, wind, or other alternative energy source. This technology has a long lifetime, low self-discharge, high capacity and density, and can store energy longer. Despite that, LIB is suitable to supply power for electric mobility if its state of health (SOH) is higher than 80%. Then, an alternative for batteries with SOH below that is recycling or second-life batteries (SLB). The first is expensive and complex; only some companies retain the technology. Still, the SLB can be an excellent solution to maintaining LIBs in operation for slow vehicles or stationary systems. However, SLB requires intelligent battery management systems (BMS) because the packs are composed of cells with different characteristics, which makes the operation more difficult. This work presents a system consisting of two Machine Learning (ML) layers to automatically estimate the state of charge (SOC) of SLB independent of the battery’s capacity or age. In the first phase, a Random Forest (RF) model was built and trained to discover the curve capacity of different SLB characteristics and capacities. After the capacity curve selection, in the second phase, a new RF model was built and trained for each capacity curve to make SOC inferences of the batteries. The discharge data curve of one hundred batteries was used for the development, whereas eighteen were used for the training and eighty-two for tests. The results indicated a root square mean error (RSME) below 45 mAh for the capacity estimation (phase 1), and an RSME below 0.87% was found in the second phase for the SOC estimation. Finally, the capacity and SOC models have been inserted in a Raspberry system to measure the main parameters of SLB (voltage and current) and make inferences in real time independent of the cell’s age. The results showed an RSME below 100 mAh for the first layer and below 1% for the second layer of the system. The excellent results indicated that the proposed idea can be applied to SLB in a general way.