AI sa Rasan sa Konfidansyalite sa Anviwònman sa Ka Tounen Wouy Batri Nan Gran Pwofi

Authors: Shengyu Tao Haizhou Liu Chongbo Sun Haocheng Ji Guanjun Ji Zhiyuan Han Runhua Gao Jun Ma Ruifei Ma Yuou Chen Shiyi Fu Yu Wang Yaojie Sun Yu Rong Xuan Zhang Guangmin Zhou Hongbin Sun Abstract Unsorted retired batteries with varied cathode materials hinder the adoption of direct recycling due to their cathode-specific nature. The surge in retired batteries necessitates precise sorting for effective direct recycling, but challenges arise from varying operational histories, diverse manufacturers, and data privacy concerns of recycling collaborators (data owners). Here we show, from a unique dataset of 130 lithium-ion batteries spanning 5 cathode materials and 7 manufacturers, a federated machine learning approach can classify these retired batteries without relying on past operational data, safeguarding the data privacy of recycling collaborators. By utilizing the features extracted from the end-of-life charge-discharge cycle, our model exhibits 1% and 3% cathode sorting errors under homogeneous and heterogeneous battery recycling settings respectively, attributed to our innovative Wasserstein-distance voting strategy. Economically, the proposed method underscores the value of precise battery sorting for a prosperous and sustainable recycling industry. This study heralds a new paradigm of using privacy-sensitive data from diverse sources, facilitating collaborative and privacy-respecting decision-making for distributed systems. Introduction Lithium-ion batteries (LIBs), serving as energy storage devices, have gained widespread utilization across various domains, from industry production to daily life as an accepted technical route. Projections suggest that the global production scale of LIBs will surpass 1.3 TWh by 2030 when the escalating demand for batteries will far outpace the availability of vital metal resources like lithium and cobalt , . However, the current average lifespan of LIB products stands at 5–8 years, leading to an imminent surge in retired batteries in many countries. If not appropriately managed, retired batteries will result in unsustainable resource wastage and environmental harm. Given these circumstances, the development of battery recycling technology assumes crucial importance as we confront the impending tide of LIB retirements . 1 2 3 4 Recent advances in battery recycling research have been focused on the pyrometallurgical, hydrometallurgical, and direct recycling approaches . In contrast to the pyrometallurgical and hydrometallurgical methods, direct recycling stands apart as a distinct approach. This process does not inflict secondary damage on the material structure, enabling more efficient structural repair and performance restoration. Moreover, direct recycling exhibits higher profitability, as it entails lower energy consumption, reduced greenhouse gas emissions, and lighter environmental footprints , . In actual production, however, battery recyclers frequently encounter LIBs comprising unknown components or battery modules that consist of a mixture of different cathode material types. Considering that direct recycling can be heavily cathode-specific, such a complexity renders the application of direct recycling infeasible for achieving value conversion of the retired batteries . It is crucial to emphasize that even if the vital metals from mixed cathode material types can be extracted using conventional recycling strategies, the interplay between different cathode materials during the recycling process can adversely impact product quality . Therefore, understanding cathode material type information on the recycling side markedly impacts the direct recycling route choice and ultimately improves product quality, profitability, and sustainability. 5 6 7 8 9 Human-assisted direct recycling has been proposed to identify retired battery cathode material type information in the pre-treatment link, which is still not financially viable when the recycling industry is scaling up . To effectively retrieve the retired battery cathode type information, the scientific and industrial community has recently initiated a battery lifetime tracing system and emerging concepts like battery passport and battery data genome . Although substantial batteries have been utilized before those initiatives, there is a growing consensus that battery information should be accessible throughout the life chain to facilitate second-life decision-making . This is notably the case for the battery recycling sector, the last station of the battery’s second life, as the recycling route can be heavily cathode-specific. However, battery lifetime tracing systems or battery passports are enabled by electronic gadgets like bar codes and near-field communications, which could introduce intensive investment and could be widely incompatible with different battery designers. Furthermore, electronic gadgets remain challenging to consistently manage throughout their lifespan, leading to worn-out devices and inaccessibility at the recycling stage since the modern manufacturing process of LIBs is still not production-to-recycling integrated . Hence, more breakthroughs are urgently needed to achieve an efficient battery cathode type sorting only using easy-to-access field information , , opposite to the historical data recorded or the human-assisted manner, facilitating the adoption of direct recycling to improve the quality and profitability of recycled products. 1 10 11 12 13 14 15 16 In the past few years, machine learning has emerged as a viable tool to tackle open questions in all battery fields. In other battery-related topics, machine learning has recently allowed us to automatically discover complex battery mechanisms , , , predict remaining useful life , , , , , evaluate the state of health , , , optimize the cycling profile , , approximate the failure distribution , even to guide the battery design , , and predict life-long performance immediately after manufacturing . In the case of battery recycling, few works have investigated machine learning regarding cathode materials , , which blames the scarce battery data, especially for those cycled to the end-of-life stage. The vast majority of published studies showcase very limited sample sizes and are even more limited in battery cathode diversity . The scarcity is attributed to the intensive cost, the long testing time , and, most importantly, the data privacy due to commercial or interest concerns. Consequently, the privacy issue rigidifies the dilemma where the existing battery data, though substantial in volume and diversity from multiple parties such as battery manufacturers, practical applications, academic institutions, and third-party platforms, cannot be shared. Such a dilemma calls for studying the cathode material sorting to optimize battery recycling route choice in a collaborative while privacy-persevering fashion. 17 18 19 20 21 22 23 24 19 25 26 27 28 29 30 31 32 33 34 35 36 37 Federated machine learning, as a distributed and privacy-preserving paradigm, has the potential to resolve both multi-party collaboration (equivalently, the battery data volume and diversity) and privacy issues through collaborative machine learning , , . In each training iteration, the distributed data owners perform local training with their local computational power, encrypt the as-trained model parameters/results, and upload them to a central coordinator for aggregation. Facts that raw datasets never leave their respective data owners and that transferred parameters/results are properly encrypted to protect data privacy. Federated machine learning has been extensively investigated in numerous applicative fields, including public health , , clinical diagnosis , , , e-commerce , Internet of Things , mobile computing , and smart grid , , . This approach can revolutionize the data-driven research paradigm in wide energy sectors by enabling privacy-preserving collaboration, especially for those with limited data access. Regarding the battery recycling sector, federated machine learning assumes promising possibilities for leveraging the giant amount of battery data that already exists but cannot be shared due to privacy concerns. With such a collaborative while privacy-preserving paradigm, retired battery sorting can be implemented with high accuracy, efficiency, scalability, and generalization, optimizing the quality and profitability of recycled products. To our knowledge, federated machine learning studies focused on battery recycling have never been reported. 38 39 40 41 42 43 44 45 46 47 48 49 50 51 In this study, we perform a cathode material sorting of the retired batteries, leveraging the existing battery data from multiple collaborators, such as battery manufacturers, practical application operators, academic research institutions, and third-party platforms, in a collaborative while privacy-preserving machine learning fashion as illustrated in Fig. . Our federated machine learning model was trained using only one cycle of field testing data via a standardized feature extraction process, without any prior knowledge of the historical operation conditions. We compare the predictive power of our federated machine learning model with that of independently learned local models based on local data under both homogeneous and heterogeneous battery recycling circumstances. The heterogeneity issue is resolved by our proposed Wasserstein-distance voting strategy. An economic evaluation of retired battery recycling using our proposed federated machine learning framework is conducted, highlighting the relevance and necessity of accurate sorting of retired batteries. We comprehensively discuss the model interpretability, battery recycling implications, and broader prospects of the future battery recycling practice integrated with federated machine learning. 1 Multiple data sources, such as battery manufacturers (Image courtesy Addionics), practical application operators (battery pack in the floor pan of a Tesla. Image courtesy of Tesla), academic research institutions, and third-party platforms, can be data contributors. The battery data are neither exchanged between contributors nor uploaded to the battery recycler. Instead, the data contributors train local models and share model parameters with the battery recycler to build a global model. The proposed Wasserstein-distance voting technique fuses the local models into the global model, which is robust to data imbalance and noise. Battery recyclers can use the jointly-built model for battery sorting, combined with the easy-to-access field testing data. Our federated machine learning framework encourages collaborators to sharing the data while preserving data privacy as apposed to the traditional data islanding paradigm. a b Results Data collection and standardization The unique battery kinetics in different battery types are often high-dimensional and hard to characterize due to divergent operating cases, manufacturing variability, and historical usages . To find a solution to this dilemma, we collected and standardized 130 retired batteries with 5 cathode material types from 7 manufacturers to construct an out-of-distribution, equivalently heterogeneous dataset. Given different historical usages, the capacities of the collected batteries are below 90% of the nominal capacity. The battery cathode materials are lithium cobalt oxide (LCO), nickel manganese cobalt (NMC), lithium ferrophosphate (LFP), nickel-cobalt-aluminum oxide (NCA), and NMC-LCO blended types, which are further grouped into 9 classes based on the manufacturers (Supplementary Table ). We intentionally include batteries with divergent historical usages, from laboratory testing to electric vehicle driving profiles, to train a generalized model for the battery recycler independent of historical usages and battery types. 52 1 For standardization, all data required from the recycler are the currently-probed (field-testing) cycle with one charging and discharging test, which is easy to implement in practical cases. The as-probed data are first denoised by filling in missing values, replacing outliers, and performing median filtering. Human-induced and cathode-heterogeneity-induced noises are deliberately retained, though, to make the model robust to imperfect inputs. The data are then linearly interpolated for curve filling (Supplementary Fig. ) and feature engineered for dimensionality reduction, with a shared set of standardization parameters (Supplementary Note ). Features extracted from the standardization pipeline are well interpretable, a concern of significant commercial interest. To the best of our knowledge, it is the first time that heterogeneous battery data from multiple sources and historical usages are utilized to assist in the strategy design of battery recycling. 1 1 Figure a, demonstrate the feature engineering process. We focus on the charging and discharging curve of the retired batteries in the last cycle, i.e., one charging and one discharging cycle (Supplementary Figs. – ). In the charging cycle, 15 features are extracted from the voltage-capacity and dQ/dV curves, where V and Q refer to the voltage and capacity values, respectively. The same set of features are extracted for the discharging cycle. As a result, 30 features are extracted in total, as indicated from F1 to F30. Refer to Supplementary Table and Supplementary Note for a detailed explanation of the features. Figure c showcases the absolute and relative feature values of the selected batteries from each class. Most relative feature values in different classes overlap in the −1 to 0 region (with the light green color) and are indistinguishable, illustrating the difficulty in classifying battery type using one cycle of battery data. The difficulty is expected because the divergent historical operation conditions can influence the charging-discharging kinetics of the batteries so that the extracted features can be largely correlated despite the different battery types (Supplementary Fig. ). Rather than directly interpreting the extracted features using expert knowledge, we employ an alternative data-driven approach that automatically leverages the latent patterns across various battery types. 2 b 2 5 2 2 2 6 For the charging process, 15 features are extracted from the voltage-capacity (left) and dQ/dV curve (right). The same set of features are for the discharging process as F16 to F30. Features are visualized by classes, following the format CxBn, indicating the nth battery from class x. The size of a circle maps the absolute feature value. Source data are provided as a Source Data file. a b c Retired battery sorting with homogeneous data access We first consider a setting where the battery data are homogeneously distributed across the collaborators (namely, the clients). The homogeneity means that each client offers to share the battery data across all 9 classes, even though the specific number of batteries is not restricted (Supplementary Table ). We train our federated machine learning model without requiring information on the historical use of the retired batteries. In our work, the recycler and the clients only need to test the retired batteries at the current (field-testing) cycle, specifically, with a complete charging-discharging cycle for a standard feature engineering process initiated by the recycler. Local models are trained based on features extracted from their private battery data. The federated machine learning framework aggregates the local model parameters, rather than the private battery data, for the recycler to classify the retired batteries. 3 Figure shows the sorting results when clients contribute homogeneous battery data. Figure compares two federated machine learning methods, i.e., the majority voting (MV) and our proposed Wasserstein distance voting (WDV), with the independent learning (IL) paradigm. It should be noted that the accuracy for the IL is averaged over all clients in a non-federated manner. Compared with the IL, the MV does not sacrifice sorting performance, with an average accuracy of 95%, while being capable of protecting data privacy and mitigating computational burden. However, 3 classes are missorted using the MV. For instance, 3 batteries in NMC (SNL, class 8, 15 in total) are missorted into NCA (SNL, class 7), resulting in a sorting accuracy of 80%. The sorting accuracy for NCA (UL-PUR, class 9) is 81%, with 2 batteries missorted into NMC (MICH_Form, class 4) and 1 battery missorted into NMC/LCO blended type (HNEI, class 2), respectively. In contrast, the WDV outperforms the MV since it only missorted one battery, resulting in a sorting accuracy of 99%. We also evaluate the prediction probability of each class for the MV and WDV, respectively. It turns out that the WDV makes a more confident sorting than the MV since the prediction probabilities of the WDV are generally right-skewed to a higher probability value. Therefore, our proposed WDV produces higher sorting accuracies across all classes, and the sorting is of richer probability confidence margins. 3 3a The confusion matrix for the majority voting (MV) and Wasserstein distance voting (WDV) methods, respectively. We consider the prediction probability distribution for each class. The sorting of independent learning (IL) is annotated. Sorting accuracy distribution and privacy budget (PB) of the IL, MV, and WDV in the presence of random noise. The PB value is referenced at a 90% accuracy level. Average F1-score of sorting results and PBs in each class using the IL, MV, and WDV. The PB values are all referenced at a 0.9 F1-score level. Data are presented as mean values ±1 standard deviation. Feature importance, in descending order. The subplot shows the feature space spanned by the first two most salient features. Data are presented as mean values + 1standard deviation. Source data are provided as a Source Data file. a b c d We also evaluate the privacy budget (PB, “Methods” section), considering that client data might be vulnerable to reverse engineering by eavesdropping on private data . In this regard, we add random Gaussian noise to the client data with different intensities. The intensity of the randomness is controlled by a noise-to-signal ratio (NSR), ranging from 1% to 10%. Figure shows the accuracy and privacy budget comparison when using IL, MV, and WDV, respectively. The sorting accuracy of the MV decreases from 95% to 82%, similar to that of the IL when the noise intensity increases from 1% to 10%. In this noise range, the median sorting accuracy of the MV and IL is 92% and 86%, respectively. In contrast, the sorting accuracy of the WDV is still above 90% in the presence of 10% noise, which is a stringent noise level in practical cases. The WDV has a median sorting accuracy of over 95% in the same noise range. Taking an average sorting accuracy of 90% as an acceptable reference level, the PB values of IL, MV, and WDV are 4, 6, and 10, respectively. Therefore, applying federated machine learning produces a more privacy-secure sorting than IL, hence being capable of preventing data eavesdropping. Furthermore, our proposed WDV is more accurate and performs much better (with nearly doubled PB values) in the privacy-accuracy trade-off than the MV. In addition, the robustness to stringent noise when using the WDV also implies a good tolerance of the battery measurement requirement, reducing the expensive battery testing disbursement. 53 3b Noticing that a high sorting accuracy does not necessarily imply an acceptable sorting for a specific class, we also consider within-class sorting performances. Figure shows the F1-score and privacy budget of the IL, MV, and WDV in each predicted class; note that the privacy setting is identical to that in Fig. . The result shows that the IL has smaller F1-scores than the federated machine learning manner in all the classes, making poor sortings. Regarding federated machine learning, WDV outperforms the MV in each predicted class by producing higher average F1 scores. The deviation range of the F1 scores for WDV is smaller than that of the MV, indicating that the WDV is more robust (Supplementary Fig. ). Therefore, our proposed WDV not only has a better overall sorting accuracy among all nine classes (Fig. ) but also within each class, compared with the MV. Regarding the privacy budget, the PB value when using the non-federated IL, referenced at a 0.9 F1-score level, is significantly lower than the federated way (Supplementary Table ) across all classes. This indicates a more severe risk of data leakage for IL compared with federated machine learning. When further applying our proposed WDV, the private budget can increase by 78% and 44% compared with the non-federated IL and the federated MV, respectively (Supplementary Table ). The results demonstrate that the WDV successfully leverages the battery-chemistry-related insights hidden in clients while effectively preserving client data privacy. 3c 3b 7 3b 4 4 We then interpret our federated machine learning model by evaluating the most salient features correlated with battery cathode chemistry. Figure shows the importance of the features in descending order. The error bar indicates the importance deviation. Features F1 and F16 rank the top two features regarding out-of-bag importance (“Methods” section). Interestingly, these two features have a clear physical interpretation of the battery dynamics, which we will further discuss in later sections. Here, we rationalize these two features by plotting the grouped battery samples in the feature space spanned by features F1 and F16. The subplot of Fig. shows that NMC/LCO blended type (HNEI, class 2), NMC (MICH_Expa, class 3), and LFP (SNL, class 6) (sharing the color with Fig. ) are clearly separable in the spanned feature space. For the remaining classes, the batteries are still separable (see the zoomed-in view), though in relatively more minor grains. On the contrary, the non-salient features have a relatively weaker sorting ability due to the non-separable feature space spanned 3d 3d 3a