Data Science-Based Battery Reutilization Management

Abstract

This chapter focuses on the data science-based technologies for battery reutilization management, which is the third stage of battery full-lifespan and crucial for the sustainable development of batteries. Battery reutilization mainly includes battery echelon utilization (secondary utilization) and material recycling. During the long-term service of the battery in EVs, the consistency of the battery cell is enlarged and its health would deteriorate. Therefore, the batteries retired from EVs cannot be directly used for secondary utilization. It is necessary to evaluate the residual value of these batteries by using the historical data or the test data, and then sort and regroup them for safe secondary utilization.

This chapter focuses on the data science-based technologies for battery reutilization management, which is the third stage of battery full-lifespan and crucial for the sustainable development of batteries. Battery reutilization mainly includes battery echelon utilization (secondary utilization) and material recycling. During the long-term service of the battery in EVs, the consistency of the battery cell is enlarged and its health would deteriorate. Therefore, the batteries retired from EVs cannot be directly used for secondary utilization. It is necessary to evaluate the residual value of these batteries by using the historical data or the test data, and then sort and regroup them for safe secondary utilization.

In this chapter, the basic process of battery echelon utilization and material recycling is first introduced. Then, the sorting and regrouping methods based on historical or/and test data for echelon utilization are proposed and verified. Finally, some classical battery material recycling methods are discussed. Echelon utilization and material recycling are important links in the full-lifetime cycle of Li-ion battery, while building a data-based traceability platform becomes essential for efficient battery reutilization management.

6.1 Overview of Battery Echelon Utilization and Material Recycling

According to the EV world sales database, 2.65 million EVs were sold worldwide in the first half of 2021, an increase of 168% compared with 2020 (as shown in Fig. 6.1), indicating vehicle electrification is fast-growing in the world. The International Energy Agency released a report named “Global Electric Vehicle Outlook” in 2020 and predicts that the global EV ownership will continue to grow in the next 10 years. Under current policies, the number of global EVs will reach 145 million in 2030. Due to the rapid growth of EVs, the global Li-ion battery shipments reached 294.5GWh in 2020, and the global demand for Li-ion batteries will increase to 1.3 TWh in 2030 [1]. According to the estimation of global EVs and power batteries in 2030, it is estimated that the total retired Li-ion batteries from global EVs will reach 460 GWh in 2030, and the total mass of retired batteries will reach 12.85 million tonnes from 2021 to 2030, equivalent to 1285 Eiffel towers [2].

Fig. 6.1

Global sales of electric vehicles in the past 3 years

The massive increase in demand for Li-ion batteries has brought two serious problems [3,4,5,6,7]: (1) the shortage of supply chain for the battery raw material. There is a shortage of some rare metals in the upstream materials of Li-ion batteries after the large-scale mining, such as lithium, cobalt, nickel, and manganese. From 2021 to 2030, the cumulative use of cobalt in batteries will exceed 30% of the world's recoverable amount. Moreover, these rare metals are unevenly distributed in the world. For example, more than 66% of cobalt is distributed in the Democratic Republic of the Congo. Therefore, cross-regional transportation also brings risks and challenges to the supply chain. (2) The retired Li-ion batteries contain electrolytes and heavy metals. If they cannot be properly handled, great pollution will be caused to the environment, soil, and drinking water. For example, a cell with a mass of 0.2 kg can pollute 1 km² of land for about 50 years and enters the bodies of people and animals through the food chains [8, 9]. Therefore, the safe disposal of these retired Li-ion batteries has become a global problem that needs to be solved urgently.

In fact, the batteries retired from EVs have high economic and environmental values. On the one hand, these batteries are rich in precious metals (such as lithium, cobalt, nickel, and manganese) [7, 10], and their content in the batteries is much higher than their raw ore. If these materials are recycled from the retired Li-ion batteries, not only can the environmental pollution caused by the disposal of batteries and land burial be reduced, but more importantly, the recycled materials can be used to remanufacture batteries and reduce the mining of raw ores. Therefore, the retired Li-ion batteries can be called urban ore, which has great economic and environmental values. On the other hand, the battery retired from the EVs still has around 80% of the initial capacity. Although it does not meet the safety requirements of EVs, it can be used in other scenarios where the safety requirements are lower than that of EVs, such as energy storage stations, power supply for 5G base stations, low-speed vehicles. This method of changing application scenarios to extend the service life of the battery is called echelon utilization. It can be seen that the echelon utilization and material recycling can maximize the full lifecycle value of Li-ion batteries, reduce the application cost of Li-ion batteries, and promote the sustainable development of Li-ion batteries. The above two technical roadmaps for the disposal of retired Li-ion batteries can be described in Fig. 6.2.

Fig. 6.2

Two technical roadmaps for the disposal of retired Li-ion batteries. a Material recycling; b Echelon utilization, reprinted from [11], with permission from Elsevier

6.1.1 Echelon Utilization

The best way to dispose of the retired Li-ion batteries is to perform echelon utilization first and then recycle materials, which can maximize the value of Li-ion batteries and promote the healthy and sustainable development of Li-ion batteries. Figure 6.2b shows the technical route of the echelon utilization route, which conforms to the 4R principle in concept very well: reuse, resell, update, and recycle. Therefore, this route has attracted enough attention and has also been extensively studied in recent years. With the advent of the large-scale retired Li-ion batteries and the improvement of relevant laws and regulations in various countries, the industrial chain of echelon utilization is gradually being established, and its prospects are broad.

Generally, batteries in EVs have three levels in structure: cells, modules, and packs. Therefore, the first outstanding issue in the echelon utilization is which level should be chosen for the secondary utilization, as it would directly affect the technical difficulty and costs. Figure 6.3 illustrates the advantages and disadvantages of echelon utilization at different levels, which can be summarized as:

Fig. 6.3

Advantages and disadvantages of the echelon utilization at different levels, reprinted from [11], with permission from Elsevier

• (1) If the battery pack is disassembled to the cell level, the workload is large, the cost is high, and the battery may be damaged. Moreover, cell regrouping will generate new material costs and may bring new safety risks. Therefore, from the perspective of economy and safety, it is not recommended to use cell as the basic unit for echelon utilization. In the future, with the development and maturity of robot automatic disassembly technologies, cells become more possible to be regrouped, which is worthy of recommendation due to its high flexibility.

• (2) Echelon utilization at the pack level is the most economical solution because it does not require disassembly. However, some of the cells in the retired battery pack have poor consistency, and the safety of secondary use is poor. Finally, the utilization at the pack level is limited by the application scenarios.

• (3) The echelon utilization at the module level is the best choice at present, due to the advantages of economy and flexibility. It can screen out the poor-quality modules without the requirements of deep disassembly for improving the safety of the regrouping system, further presenting good regrouping flexibility.

Then, the second outstanding problem is how to disassemble battery pack to the module. Battery disassembly is a complicated and dangerous task. First, the battery is a high-energy and chemical carrier; if it is not handled properly, it will cause various safety issues such as short circuits and liquid leakage. What's more serious is that it may cause an explosion or fire, further leading to casualties and property damage. Second, the toxic and carcinogenic substances in battery electrolyte and electrode materials may cause chemical hazards during disassembly. Third, batteries from different manufacturers may have different external structures, module connection methods, and process technologies, further bringing difficulties to the automatic disassembly. Therefore, battery disassembly is mostly manual or semi-automatic at present. The establishment of flexible and intelligent battery disassembly equipment is an urgent need for development in the future. In summary, the disassembly of large-scale retired Li-ion batteries faces the following technical challenges: (1) Lack of skilled dismantling workers and professional demolition tools to ensure the safety, integrity, and speed of disassembly; (2) The disassembly efficiency is low, further directly affecting the economic value; (3) The high reliability of battery component connection increases the difficulty and time cost of battery disassembly. A feasible solution is to standardize the design of cells, modules, and packs, and to design a battery connection method that is conducive to disassembly. In addition, the development of automatic disassembly robots is essential. However, due to the complexity of battery structure and connectivity, and rich wires and battery management system, complete automatic disassembly becomes a great challenge, and man–machine integration is a short-term solution.

The third outstanding issue is how to evaluate the residual value of the retired Li-ion batteries, and how to sort and regroup them for secondary use. The most critical issue for sorting the retired Li-ion batteries is how to ensure its accuracy and efficiency, which would directly determine the economy and safety of the echelon utilization. Regrouping should not only be achieved based on the sorting results, but also based on actual application scenarios. For example, high-capacity retired Li-ion batteries are suitable for energy applications, while low internal resistance Li-ion batteries are more suitable for power applications. For the retired Li-ion batteries without historical data, it is very time-consuming and energy-consuming to test batteries one by one. Obviously, the traditional sorting method based on capacity or internal resistance test is not applicable. An important challenge for the echelon utilization of the large-scale retired Li-ion batteries is to design the sorting methods and devices with excellent sorting speed and accuracy, which is suitable for different application scenarios. For the retired Li-ion batteries with historical data, the technical challenge here becomes how to quickly sort through massive amounts of data. At present, most retired Li-ion batteries have no historical data or the quality of historical data is poor. In this case, how to sort Li-ion batteries is challenging, while some data science-based methods will be described in detail in Sect. 6.2.

6.1.2 Material Recycling

With the innovation of battery material systems and the improvement of environmental protection requirements, battery recycling technology is facing major opportunities and severe challenges. First, since the cathode material is the most valuable one in a battery, the “dissolution–precipitation–recycling” process of valuable metals (such as cobalt) has formed a certain economic recovery-driven business model. However, to reduce battery costs and increase energy density, cathode materials are constantly being updated, and the research on waste Li-ion battery recycling technology is relatively backward, making it difficult for the corresponding recycling technology to keep up with the pace of material updates. If recyclers cannot recycle pure and high-quality materials, the recyclable value will become very low. Moreover, ternary batteries are developing in the direction of high nickel and low cobalt, which will weaken the existing battery recycling business model. Therefore, new recycling technologies need to be continuously updated and iterated to meet market demand.

Second, most of the current recycling technologies focus on the recycling of valuable metals such as lithium, nickel, cobalt, and copper, and little attention is paid to other low-value components. However, the recycling of negative electrode materials and electrolytes is an unavoidable issue. Here the electrolyte is volatile and generates gases such as HF and PF5, which are harmful to both the human body and the environment. In recent years, the recycling and reuse of electrolyte have become a research hotspot. However, these methods still have problems such as long recycling process, high recycling cost, low recycling rate, making them become unsuitable for large-scale industrial production, and lack market driving force. Third, the current recycling process of Li-ion batteries usually involves battery disassembly, smelting and/or acid leaching, chemical precipitation separation, and decomposition. In these processes, a large amount of energy and chemicals will be consumed, causing greenhouse gas emissions, secondary waste, and other environmental problems. Therefore, the development of green, efficient, and full-component recycling methods becomes a future trend.

The material recycling and resource utilization of waste Li-ion batteries can help recover the valuable materials, realize the recycling of valuable resources, reduce the impact of waste treatment on the environment, and reduce the development and consumption of natural resources. Li-ion batteries usually consist of a cathode, an anode, an electrolyte, a separator, a casing, and other components. A typical Li-ion battery contains about 25–30% cathode, 15–30% anode, 10–15% electrolyte, 18–20% shell, 3–4% independent components, and 10% other components. Typical cathode material is composed of 80–85% metal oxide powder, about 10% polyvinylidene fluoride binder, and 5% acetylene black. Graphite is commonly used as a negative electrode material, and it consists of hexagonal carbon atoms arranged in thin sheets. The separator is usually made of microporous polypropylene or polyethylene. The commonly used electrolytes in Li-ion battery include LiClO₄, LiAsF₆, LiPF₆, Li(CF₃SO₃), Li[N(CF₃SO₂)₂]. It can be seen that the cathode material is the most valuable part. Table 6.1 lists the chemical components of several typical Li-ion batteries. Obviously, the metal content of Li-ion batteries is even better than that of natural raw materials. According to the statistics from London Metal Exchange (LME), the average price of major metal materials in Li-ion batteries in 2020 is shown in Table 6.2, indicating that the metal materials in Li-ion batteries have a high recycling value, of which cobalt has the highest value. Therefore, high-cobalt Li-ion batteries have excellent recycling value. Ternary batteries have a higher recycling value than lithium iron phosphate (LFP) batteries.

Table 6.1 Chemical composition of some typical Li-ion batteries

Table 6.2 Price of main metal materials in Li-ion batteries (LMC, 2020)

There are many topics about battery material recycling, such as recycling mode and industrial chain, recycling methods. In this chapter, it should be noted that in the study of echelon utilization, the battery is called retired battery, while in the discussion of battery recycling, it is called waste or spent battery.

6.2 Sorting of Retired Li-Ion Batteries Based on Neural Network

The sorting and regrouping of the retired Li-ion batteries are based on one or more battery performance criteria, named sorting criteria. The purpose of battery sorting is to evaluate the residual value of the battery through these criteria. As most retired Li-ion batteries do not have full historical data, the evaluation of battery residual value can only be done through testing. In this section, the following two outstanding issues for data science-based battery sorting are addressed: (1) What kind of criteria should be constructed to truly reflect the ageing characteristics of the retired battery; (2) How to quickly and accurately obtain these criteria based on the test data for the sorting of the large-scale retired Li-ion batteries.

6.2.1 Data Science-Based Sorting Criteria

The consistency of Li-ion batteries retired from EVs is usually poor. Therefore, it is necessary to sort the retired Li-ion batteries before echelon utilization. In other words, Li-ion batteries with the same or similar performance need to be classified into the same category. The retired Li-ion battery may be in an accelerated period of performance degradation, and the inconsistency of cells or modules may increase more dramatically, further seriously affecting battery safety. Therefore, improving the accuracy and efficiency of sorting is of great significance to enhance the safety and economy of echelon utilization. The first basic but key problem of the data science-based sorting is to select one or more sorting criteria that can accurately reflect the true state of the retired Li-ion batteries.

There are many performance parameters that can characterize the residual value and health status of retired Li-ion batteries. Figure 6.4 lists the common sorting criteria at different scales. Theoretically, the performance degradation of retired Li-ion batteries is caused by the microscopic changes in the structure or material morphology at the molecular scale. Therefore, the microscale sorting criteria become most direct and accurate. However, it has a cumbersome process and requires high-precision equipment. Thus, it becomes unsuitable for the sorting of large-scale retired Li-ion batteries. The sorting criteria widely used in literature include battery appearance [12, 13], capacity or life [14,15,16,17], internal resistance [17], impedance spectrum [18, 19], charge and discharge curve characteristics [12, 13], or their combination. However, these macro characteristics can only reflect the battery external macro-characteristics and cannot reflect the internal state of the battery. Therefore, there is a contradiction between precision and complexity in the selection of sorting indicators, and how to balance them is a key issue.

Fig. 6.4

Sorting criteria for the retired Li-ion batteries at different scales, reprinted from [11], with permission from Elsevier

To build suitable criteria for the large-scale retired Li-ion batteries, the following issues need to be considered:

• (1) The multidimensional sorting criteria need to be constructed to evaluate battery status more comprehensively. The current single sorting index cannot fully reflect the ageing characteristics of Li-ion batteries. For example, classifying the retired Li-ion batteries just based on capacity may result in the batteries with different internal resistances being classified into the same category, which is obviously unreasonable. More importantly, the capacity and internal resistance cannot fully characterize the safety characteristics of a battery, and safety becomes most important in secondary utilization. In the sorting process, the typical side reactions (such as lithium plating, thickening of SEI film) or typical failures (such as internal short circuits) that seriously threaten the battery safety should be considered. Moreover, the secondary battery is in a period of accelerated decline in life, the inconsistency between battery modules or cells may be enlarged gradually, and capacity diving may occur, which will threaten the battery safety. Therefore, the prediction of battery life trajectory becomes significantly important. In summary, the ideal battery residual value evaluation requires the construction of multidimensional indicators, such as the typical side effects (representing the past states of Li-ion batteries), capacity and internal resistance (representing the current states), and life trajectory (representing the future states). These indicators can comprehensively evaluate the status of Li-ion batteries from different dimensions and improve the safety of echelon utilization. However, how to quickly obtain the multidimensional criteria has become an outstanding problem. The modern sensing and detection technologies, network technologies, big data science, and modelling technologies are powerful solutions.

• (2) The available historical data of the battery is very valuable, which contains an abundance of battery status information. Using historical data to evaluate battery residual value is an economical and accurate method. However, most of the existing historical data is unlabelled, further bringing challenges to battery state estimation and is also an issue to be solved.

• (3) The sorting efficiency is directly related to the economy and must be considered for the large-scale retired Li-ion batteries. Currently, most of the retired Li-ion batteries do not have available historical data, the sorting criteria can only be obtained through the test data of Li-ion batteries in this case. The traditional measurement method of the battery capacity and internal resistance requires about 3 h, which is not allowed for the state estimation of the large-scale retired Li-ion batteries. How to quickly obtain the criterion based on the test data becomes important. In Sect. 6.2.2, a valuable data science-based technical route of the fast capacity estimation is introduced.

However, the capacity and internal resistance are the external performance of the battery and cannot well characterize battery internal states. Electrochemical impedance spectroscopy (EIS) is a non-destructive measurement method that has been widely used to evaluate the performance of Li-ion batteries. EIS can reflect the internal characteristics of a battery with the minute level of test time. Therefore, it is a promising method for evaluating the residual value of retired Li-ion batteries. However, EIS is susceptible to factors such as SoC, SoH, charge/discharge rate, and temperature. In this context, further exploration is needed in the sorting process and evaluation. An effective data science-based method is proposed in Sect. 6.2.3.

6.2.2 Case 1: Sorting Criteria Estimation Based on Charging Data

To evaluate SoH of reutilized Li-ion batteries, capacity and internal resistance testing is the most direct method. However, this testing is very time-consuming and energy-consuming. In this section, a fast data science method of estimating reutilized battery capacity and internal resistance based on charging curves and machine learning is introduced.

Figure 6.5 illustrates the principle of the proposed data science method. It can be expressed as follows: First, a large quantity of retired Li-ion batteries (battery modules or cells) to be sorted are connected in parallel and rest for a long time to ensure that the battery terminal voltage becomes consistent. This process can be carried out in large batches and is obviously efficient. Then, these batteries are randomly selected for series charging, and a series of short-term charging voltage curves are obtained with rapid speed. These curves can be used to extract internal resistance and capacity. Afterwards, the capacity of a small number of samples is obtained through the standard capacity test, and a machine learning algorithm (such as neural network) will be used to establish a correlation model among the charge–discharge curve, internal resistance, and capacity. Finally, the established correlation model can be used to estimate the internal resistance and capacity characteristics of the remaining massive batteries. In this method, states of massive batteries will be estimated from a small sample test using machine learning, which greatly improves the sorting efficiency. However, how to establish a black-box model for accurately correlating charging/discharging curves with battery capacity has become a key issue.

Fig. 6.5

A fast estimation method of capacity and internal resistance based on machine learning, reprinted from [11], with permission from Elsevier

Take the sorting of battery cells as an example to illustrate the process of obtaining capacity and internal resistance. The process of obtaining capacity and internal resistance is as follows: First, a large number (N) of cells are selected for the parallel equalization to ensure that the terminal voltage of the cells remains the same. Then, M cells (M ≤ N) are selected for the series constant-current charging. Figure 6.6 shows the voltage curves of M cells in the serial-charging process. Simply, the internal resistance of each cell can be calculated as follows:

$$ R_{{\text{cha}},\Delta t} = \frac{{U_k \left( {t_1 + \Delta t} \right) - U_k \left( {t_1 } \right)}}{{I_{{\text{cha}}} }} $$

(6.1)

where $R_{{\text{cha}},\Delta t}$ is the charging internal resistance at time $\Delta t$; $I_{{\text{cha}}}$ is charging current.

Fig. 6.6

Schematic diagram of series charging curves, reprinted from [17], with permission from Elsevier

In addition, the charging curves of each cell will be affected by internal resistance. Therefore, the residual voltage after excluding the influence of internal resistance can be expressed as:

$$ \tilde{U}_k = U_k \left( {t_2 } \right) - \left( {U_k \left( {t_1 + \Delta t} \right) - U_k \left( {t_1 } \right)} \right) $$

(6.2)

where $\tilde{U}_k$ is the terminal voltage of cell $k \, \left( {k = 1,2, \ldots ,M} \right)$ excluding the influence of internal resistance; $U_k \left( {t_1 } \right)$,$U_k \left( {t_1 + \Delta t} \right)$, and $U_k \left( {t_2 } \right)$ are the terminal voltage at $t_1$,$t_1 + \Delta t$, and $t_2$, respectively.

Then, a small number of cells from M cells are randomly selected for the standard capacity test, and then the test capacities ($C_k$) and $\tilde{U}_k$ are used as samples to train the capacity model based on machine learning. In this study, a back-propagation neural network (BPNN) is applied to get the capacity model. Since the BPNN belongs to a frequently reported algorithm, it will not be described in detail here for brevity.

To verify the effectiveness of the proposed method, 108 retired cells are used to carry out a fast capacity estimation. After retiring a battery from EVs, its SoC would be generally discharged below 30% to ensure its safe transportation. Therefore, the SoCs of the experimental cells would be adjusted to different SoC levels for constructing the following three cases: Case 1: SoC = 5%, Case 2: SoC = 20%, Case 3: SoC = 30%. Then, the 108 cells are connected in series for the constant-current charging to get a series of charging curves. In the above three cases, the methods described in Sect. 6.2.1 are used to estimate battery capacity with the results shown in Fig. 6.7. Note that the capacity model is trained based on 36 battery cells, while another 72 cells are used to verify the accuracy of capacity estimation.

Fig. 6.7

Capacity estimation results in different cases. a Capacity estimation results; b capacity estimation error, reprinted from [17], with permission from Elsevier

It can be observed that the capacity estimation errors in these three cases are all within 5%. It can be inferred that the accuracy of capacity estimation will be further improved as the number of sample batteries increases. In the proposed data science method, a machine learning algorithm is used to establish a black-box model between battery capacities and charging curves, which greatly improves the efficiency of capacity estimation. It is very valuable for the sorting of large-scale retired Li-ion batteries.

6.2.3 Case 2: Sorting Criteria Estimation Based on EIS

6.2.3.1 Methodology

To accurately estimate the ageing characteristics of the retired battery, EIS is used for the non-destructive testing of Li-ion batteries. Here EIS test belongs to an efficient in-situ/ex-situ electrochemical characterization technology [20], which has been widely used in the electrochemical measurement and characterization of battery ageing [21, 22]. The EIS information can be expressed in the form of a Nyquist plot combining the real and imaginary parts of the impedance. The Nyquist plot of the impedance spectrum of a Li-ion battery cell is shown in Fig. 6.8, in which the horizontal axis is the real part of the impedance, and the vertical axis is the imaginary part of the impedance. The physical meanings of each parameter in Fig. 6.8 are listed in Table 6.3 [23, 24].

Fig. 6.8

Frequency distribution at Nyquist impedance spectroscopy for a cell

Table 6.3 Physical meaning of parameters in EIS diagram

Reference [25] pointed out that if the high-frequency part of the impedance spectrum, that is, the second semicircle part, is directly fitted, the data distortion may be caused, and the distribution of relaxation times (DRT) method can effectively avoid this problem. The AC impedance of Li-ion batteries is written as follows:

$$ z\left( f \right) = R_0 + \int\limits_0^\infty {\frac{\gamma \left( \tau \right)}{{1 + j2\pi f\tau }}{\text{d}}\tau + \frac{1}{{j2\pi fc_{{\text{in}}} }}} $$

(6.3)

where $f$ is the frequency, $\gamma$ is the time distribution of polarization losses, $\tau$ is the time constant of the corresponding impedance, and $c_{{\text{in}}}$ is the intercalation capacitance. The main purpose of DRT is to determine $\gamma \left( \tau \right)$ through $z\left( f \right)$.

Figure 6.9a shows the EIS test results of 35 retired cells with 20% SoC at 25 °C. Moreover, as shown in Fig. 6.9b, the DRT curve could be divided into four intervals [25], which is denoted as S₁, S₂, S₃, and S₄. Here the DRT curve can reflect different impedance types, as listed in Table 6.4. The magnitude of different impedances can be calculated by the upper and lower limits of the time constants of different impedances. For each interval, the polarization resistance $R_p$ could be calculated as follows:

$$ R_p = \int\limits_{\tau_L }^{\tau_U } {\gamma \left( \tau \right){\text{d}}\tau } $$

(6.4)

where $\tau_U$ and $\tau_L$ are the upper and lower limits of time constant, respectively.

Fig. 6.9

EIS test results of 35 retired cells with 20% SoC at 25 °C. a Nyquist plots; b DRT plots

Table 6.4 Types of impedance reflected in the different DRT intervals of Li-ion batteries

Battery capacity is one of the key criteria for the sorting of retired Li-ion batteries. However, traditional battery capacity is generally obtained by the standard capacity test, which requires a battery to be fully charged and discharged three times. Therefore, this test process is very time-consuming with the hour level, becoming obviously not conducive to the sorting of large-scale retired Li-ion batteries. The EIS test is very fast at the minute level, and the test results contain rich electrochemical characteristics. However, there is no direct relationship between capacity and EIS. In this study, an EIS-capacity correlation model is built using BPNN. As shown in Fig. 6.10, the brief process can be summarized as: First, a small number of sample cells are selected for the standard capacity test. Then, the capacity obtained from the capacity test and the DRT characteristics are used to train the BPNN model, and the EIS-capacity correlation model is obtained. Finally, the EIS-capacity correlation model is used to estimate the capacity of a large number of remaining cells. In this process, the capacity of cells can be obtained quickly only by inputting DRT characteristics. It can be seen that the advantage of the proposed method is that the capacity of most cells can be estimated quickly and accurately based on the capacity test of a small number of cells. In addition, other valuable battery ageing characteristics can be quickly extracted from EIS, such as S₁, S₂, S₃, and S₄. These electrochemical characteristics can reflect the internal state of the battery, which becomes useful for evaluating the residual value of the retired Li-ion batteries.

Fig. 6.10

Establishment process of the correlation model between capacity and EIS

6.2.3.2 Experimental Verification

In this study, the EIS tests are conducted on 35 retired cells. Then, the capacity of these cells is tested for the reference of the fast capacity estimation, and the results are shown in Fig. 6.11. In general, the retired Li-ion batteries are transported and stored at low SoC for safety, and the previous studies [26, 27] showed that the Nyquist plots of the same Li-ion batteries with different ageing degrees at low SoC are more distinct than those at high SoC. Therefore, the SoC of each retired cell is adjusted to 20% for the EIS tests.

Fig. 6.11

Capacity distribution of the 35 retired cells

In this study, the DRT (${\text{DRT}}_i$) and capacity $C_i^* \left( {i = 1,2, \ldots ,30} \right)$ of 30 cells are randomly selected as the training set of the BPNN model, and the rest five cells are selected as the validation set. Owing to the DRT of each cell having the same time constant, ${\text{DRT}}_i$ can be expressed as follows:

$$ \widetilde{{{\text{DRT}}}}_i = \gamma_i \left( \tau \right)\quad \left( {i = 1,2, \ldots 30} \right) $$

(6.5)

Owing to the randomness of the initial value, the training results of the BPNN model are generally unequal. In this study, the EIS-capacity model is trained three times, and the results of the verification set are shown in Fig. 6.12. It can be observed that the capacity estimation errors are less than 4% under the three training models. Therefore, it can be concluded that the EIS-capacity correlation model is accurately established using BPNN. Moreover, the EIS test time of each cell can be controlled within 10 min, while the traditional capacity test requires 3 h. It can be clearly seen that the efficiency of capacity acquisition of the proposed method is ten times higher than that of the traditional capacity test method with satisfactory accuracy, which greatly improves the sorting efficiency of the large-scale retired Li-ion batteries.

Fig. 6.12

Capacity estimation results based on the EIS-capacity correlation model

6.3 Regrouping Methods of Retired Li-Ion Batteries

6.3.1 Overview of Regrouping Methods

After obtaining the performance evaluation criteria of the retired Li-ion batteries, the next key question is how to regroup these Li-ion batteries based on the sorting criteria. The following issues need to be considered: (1) The performance evaluation of the battery involves multiple criteria, such as capacity, internal resistance, and EIS characteristics. Mathematically, the battery regrouping is a clustering issue. There are many current data science-based clustering algorithms, such as K-means clustering [28, 29], mean-shift clustering [30], density-based spatial clustering with noise [31], expectation–maximization clustering [32], cohesive hierarchical clustering [33], support vector machine regression analysis [34]. However, most of them are suited to binary classification for two-dimensional data. How to cluster batteries in a multidimensional space thus becomes the first key issue. (2) The purpose of battery regrouping is to ensure the safe and long-term reuse of batteries. Therefore, the echelon utilization scenario must be considered in regrouping. For example, energy-based and power-based scenarios have different requirements for batteries. The second key issue is how to construct the constraints of echelon utilization scenarios in battery clustering.

To address these two problems, an effective data science-based solution is provided in this study, as shown in Fig. 6.13. Assuming that batteries are regrouped based on four sorting criteria including the typical side reactions, residual life, internal resistance, and capacity, this becomes a four-dimensional clustering problem. Frequently, the multidimensional clustering can be solved by the hierarchical clustering method. Here the side reactions are related to battery safety, while the life, internal resistance, and capacity characteristics are related to battery functionality. For the echelon utilization of retired Li-ion batteries, battery safety is the priority. Therefore, the first level of battery clustering can use the side reaction characteristics as a criterion, which is a one-dimensional clustering problem. The purpose of the clustering is to classify Li-ion batteries with the same or similar side reaction characteristics. Li-ion batteries with the same side reactions are then clustered at the second level based on the residual life, capacity, and internal resistance, making it become a three-dimensional clustering problem. Furthermore, three-dimensional clustering can be further transformed into two-dimensional clustering by using constraints of echelon utilization scenarios. Generally, echelon utilization scenarios include energy-based and power-based scenarios. The typical application of the former is energy storage power station, while that of the latter is low-speed vehicles. Obviously, energy-based scenarios require high-capacity batteries, while power-based scenarios require low-resistance batteries. In this context, batteries used in energy-based application scenarios can be clustered by capacity and residual life, while batteries used in power-based scenarios can be clustered by internal resistance and residual life. They are two-dimensional clustering problems, which are easy to implement through algorithms and easy to understand.

Fig. 6.13

Multidimensional clustering method under echelon utilization scenario constraints, reprinted from [11], with permission from Elsevier

It should be noted that the four sorting criteria listed above are just an example. There are many criteria for clustering retired Li-ion batteries, such as capacity, internal resistance, and EIS characteristics. Two cases are given below to illustrate the application of the data science-based clustering method in the regrouping of the retired Li-ion batteries.

6.3.2 Case 1: Hard Clustering of Retired Li-Ion Batteries Using K-means

In this study, the K-means algorithm is adopted to cluster and regroup modules to improve the overall consistency of the entire battery pack. K-means is a well-known data science-based clustering algorithm in which the data items are clustered into K clusters such that each item only blogs to one cluster. Moreover, various echelon utilization scenarios have different requirements for capacity and internal resistance consistency. For example, greater capacity consistency is required than internal resistance consistency in an energy-type echelon application (e.g. energy storage power station), whereas more attention should be paid to the internal resistance consistency in a power-type application (such as the low-speed EVs). The energy-type and power-type echelon applications are two representative echelon utilization scenarios. Therefore, a reasonable data science-based clustering algorithm should consider the utilization scenario constraints.

The improved K-means algorithm is developed to describe the constraints of the echelon utilization scenarios. The proposed algorithm flow is listed in Table 6.5, in which the echelon utilization scenario factor $\delta$ is introduced, as shown in Eq. (6.9). Specifically, the closer the $\delta$ tends to 1, the higher the capacity consistency of the battery, which is accordingly clustered as being more suitable for an energy-type scenario with strict capacity requirements. The closer $\delta$ tends to 0, the higher the power density consistency of the battery, which is accordingly clustered as being more suitable for a power-type scenario with high-power requirements. By setting different $\delta$, various echelon utilization scenarios can be described, and more reasonable clustering results can be obtained.

Table 6.5 Process of the improved K-means algorithm

To verify the effectiveness of the proposed regrouping method, clustering is performed using both capacity and internal resistance of 108 cells obtained in Sect. 6.2.2. To highlight the advantages of the proposed method, the conventional K-means and the improved K-means algorithms are used to cluster these cells, and the results are shown in Figs. 6.14 and 6.15, respectively. It can be observed that cells are divided into six categories, and different values of $\delta$ will lead to different clustering results. Specifically, the smaller the value of $\delta$, the more consistent the internal resistances of the cells in the same group, and the less consistent the battery capacities. Therefore, the cluster results for a small $\delta$ are beneficial for a power-type echelon utilization scenario. In contrast, the closer the value of $\delta$ to 1, the more consistent the capacities of the cells in the same cluster, and the less consistent the battery internal resistances. Therefore, the cluster results for a large $\delta$ are beneficial for an energy-type echelon utilization scenario. Simply stated, cluster results can be obtained with emphasis on either capacity or internal resistance by setting the echelon utilization scenario coefficient $\delta$ to provide cluster results that are more suitable for the intended echelon utilization scenarios.

Fig. 6.14

Cluster result based on the conventional K-means algorithm, reprinted from [29], with permission from IEEE

Fig. 6.15

Cluster results under different coefficient δ, reprinted from [29], with permission from IEEE

It should be noted that the number of clusters is determined before clustering, which directly affects the clustering results. In the actual echelon utilization, the number of clusters can be determined according to the number of batteries, capacity, and internal resistance distribution. The regrouping of retired Li-ion batteries based on the cluster results can improve the adaptability and safety of cascade utilization scenarios.

6.3.3 Case 2: Soft Clustering of Retired Li-Ion Batteries Based on EIS

The traditional data science-based linear clustering methods, such as K-means, hierarchical clustering, are hard-clustering methods [35, 36], which means each sample is only assigned to a specific cluster. For the retired Li-ion batteries, the clustered cells only belong to one cluster, which is obviously inflexible for the regrouping of retired Li-ion batteries. For example, it is obviously unreasonable for the cells on the boundary of the clustering results to be strictly restricted to a fixed cluster. In this study, Gaussian mixture model (GMM) algorithm [37, 38] is used for the soft clustering of retired Li-ion batteries, which gives the possibility that one clustered cell belongs to each cluster. That is, a battery could not belong to a certain cluster, but may be shared by several clusters, further making the clustering results more flexible.

The GMM is a probabilistic data science model, which assumes that all data points are generated by a mixture of Gaussian distributions with a finite number of unknown parameters. It is usually used for unsupervised learning or soft clustering of unlabelled data. The probability distribution function of GMM can be defined as follows:

$$ \left\{ {\begin{array}{*{20}l} {\pi \left( x \right) = \sum\limits_{{k = 1}}^{K} {p_{k} } N\left( {\left. x \right|\mu _{k} ,\sum\limits_{k} {} } \right)} \\ {N\left( {\left. x \right|\mu _{k} ,\sum\limits_{k} {} } \right) = \frac{1}{{\left( {2\pi } \right)^{{\frac{d}{2}}} \left| {\sum\limits_{k} {} } \right|^{{\frac{1}{2}}} }}e^{{\left[ { - \frac{1}{2}\left( {x - \mu _{k} } \right)^{T} \sum\limits_{k}^{{ - 1}} {\left( {x - \mu _{k} } \right)} } \right]}} } \\ {\sum\limits_{{k = 1}}^{K} {\pi _{k} = 1} } \\ \end{array} } \right. $$

(6.11)

where x is the data points, K is the number of clusters, N is the Gaussian density function, $\pi$ is the mixing probability, $\mu_k$ and $\sum {_k }$ are the mean and covariance for the Gaussian k ($k \in \left\{ {1, \ldots ,K} \right\}$), respectively, and d is the data dimension.

Generally, the expectation–maximization algorithm is an iterative algorithm that is used to find the maximum likelihood estimation of the GMM when the parameters cannot be found directly, and it can be simply divided into the expectation step (E-step) and the maximization step (M-step). In the E-step, the available data is used to estimate the value of missing variable. Based on the estimated value, the parameters are updated with the complete data. Suppose the parameter to be determined is $\theta = \left[ {\pi ,\mu ,\Sigma } \right]$, and the detailed process of the GMM are listed in Table 6.6.

Table 6.6 Process of the GMM algorithm

The silhouette value $s\left( {x_i } \right)$ and silhouette coefficient $C_{SC}$ are used to evaluate the clustering effects, and they can be expressed as follows [39]:

$$ s\left( {x_i } \right) = \frac{{b\left( {x_i } \right) - a\left( {x_i } \right)}}{{{\text{max}}\left( {a\left( {x_i } \right),b\left( {x_i } \right)} \right)}} $$

(6.18)

$$ C_{SC} = \frac{1}{m}\sum\limits_{i = 1}^m {s\left( {x_i } \right)} $$

(6.19)

where $a\left( {x_i } \right)$ is the average distance between $x_i$ and other data points in the cluster, and $b\left( {x_i } \right)$ is the average distance from $x_i$ to all other points in the cluster.

The range of $s\left( {x_i } \right)$ is distributed from −1 to 1. Specifically, if $- 1 \le s\left( {x_i } \right) < 0$, the clustering is unreasonable, and if $s\left( {x_i } \right) = - 1$, the clustering becomes the worst. If $0 \le s\left( {x_i } \right) \le 1$, the clustering is reasonable, and if $s\left( {x_i } \right) = 1$, the clustering becomes the best.

To fully demonstrate the advantages of the soft clustering method, 200 retired cells with different ageing degrees are tested by EIS and DTR described in Sect. 6.2.2, and the results are shown in Fig. 6.16. According to the above soft clustering method using GMM, the 200 cells are soft clustered into five groups, which are labelled C1, C2, C3, C4, and C5. Figure 6.17 shows the clustering results, in which the abscissa is the cell number, and the ordinate is the clustering probability. P (CX) (X = 1, 2, 3, 4, 5) is the probability that one sample belongs to groups, and the P is distributed from 0 to 1. Specifically, the green, red, and white represent the weak, strong, and median probability of a battery belonging to a cluster, respectively. It can be observed that some cells can be grouped into different clusters at the same time. These cells with middle colours can be flexibly shared by multiple groups during regrouping, which means that these cells are soft clustered. Furthermore, if a cell with the probability of $0.3 < P < 0.7$ is defined as a cell with soft clustering characteristics, then there are 17 cells belonging to multiple categories that can be flexibly regrouped. It can be inferred that cells with soft clustering characteristics will increase as the number of cells increases, which greatly increases the regrouping flexibility of the large-scale retired Li-ion batteries.

Fig. 6.16

EIS and DRT plots of 200 retired cells

Fig. 6.17

Soft clustering results of the 200 cells using GMM

Figure 6.18 shows the distribution of the six sorting criteria for each cell after clustering, where the abscissa is the cell label and the ordinate is the normalized value of the six criteria. The five groups are marked with different colours. The cells in the solid colour group do not have soft clustering characteristics, while the cells in the shaded colour group have soft clustering characteristics. It can be observed that the six sorting criteria of the same group of cells all have the good consistency. For example, C1 has a large and consistent capacity, while C2 and C4 have large and consistent internal resistances. It shows that clustering based on the constructed multidimensional criteria can express the key characteristics of the battery, and then the refined regrouping can be carried out. In addition, the SC value calculated according to Eq. (6.12) is 0.478, indicating that the proposed algorithm has a good clustering effect.

Fig. 6.18

Consistency of the sorting criteria within each cluster

It should be noted that the cells in the shaded colour group of Fig. 6.19 are shared by multiple clusters. The value in the grey grid is the number of non-soft clustered cells in each group, and the white grid is the number of soft clustered cells in each group. For example, three cells are shared between C1 and C2, and six cells are shared between C1 and C5. It can be concluded that the proposed clustering method based on multidimensional criteria can improve the accuracy and rationality of battery regrouping, and the soft clustering algorithm based on GMM can improve the flexibility of large-scale battery regrouping.

Fig. 6.19

Battery distribution with the soft clustering characteristics

6.4 Material Recycling Method of Spent Li-Ion Batteries

6.4.1 Main Recycling Methods

The material recycling and reuse of spent Li-ion batteries is an important part of the closed-loop of the Li-ion battery full-lifespan cycle, which can realize the recycling of valuable resources, reduce the impact of waste treatment on the environment, and reduce the consumption of natural resources. Therefore, the recycling of spent Li-ion batteries has received widespread attention in recent years. Big data technology plays an increasingly important role in the lifecycle management of Li-ion batteries. The battery recycling traceability comprehensive management system is under construction and improvement in many countries around the world. During the use of the battery, the information of battery maintenance or battery retirement is transmitted to the platform by the vehicle manufacturer. After the battery is retired from EVs, the battery disassembly, echelon utilization, and recycling enterprises will submit the battery-related information. In all processes, the material flow in the full-lifespan cycle of batteries will be submitted to the traceability integrated management platform, and this submitted information forms the big data of batteries. Through the platform, we can understand the information of the waste batteries in a certain area, and guide the establishment of battery recycling enterprises and the application of battery recycling methods.

The recycling process of the spent Li-ion batteries is to separate the useful components in the batteries by using their physical and chemical properties to realize the reuse of resources. The recycling methods of the spent Li-ion batteries based on battery data can be roughly divided into physical and chemical methods, as shown in Fig. 6.20. At present, physical methods are mostly used as the pretreatment before the chemical methods. The physical method includes mechanical separation, heat treatment, mechanochemical treatment, and dissolution treatment. Chemical methods are an effective solution to improve material recycling rate and purity. The chemical treatment process includes acid leaching, biological leaching, solvent extraction, chemical precipitation, and electrochemical treatment. At present, the main chemical recycling methods are pyrometallurgy, hydrometallurgy, bioleaching, and mixed treatment.

Fig. 6.20

Main recycling methods for the spent Li-ion batteries, reprinted from [40], with permission from Elsevier

Due to the complexity of the Li-ion battery structure, pretreatment is required to achieve the maximum recovery rate. The discharge process before disassembly will reduce the energy of the waste battery, thereby preventing spontaneous combustion. The heat treatment removes and decomposes the electrolyte by thermochemically degrading organic compounds, thereby deactivating the battery. The battery pack includes many large components such as housing, battery management system, or cooling components. Therefore, they were first disassembled manually and then sorted according to size and chemical composition. The mechanical treatment can reduce material volume and separate individual battery materials. To release the positive and negative materials, Li-ion batteries must be crushed, grounded, and then sieved. High-energy grinding is used to reduce the particle size and increase the specific surface area, further simplifying the leaching process.

In pyrometallurgical processing, battery components are smelted at high temperatures to obtain a metal alloy composed of metals Cu, Ni, Co, and Fe, and then purified and separated by hydrometallurgy. Mn and Ti are usually not recycled as metals, but oxidized and form slag. In hydrometallurgical recycling, the cathode material is dissolved in acid, and individual metals are separated by solvent extraction. Inorganic acids are used to dissolve metal components during the leaching process. Subsequently, the metal is concentrated and purified by chemical precipitation, ion exchange, or solvent extraction. Compared with pyrometallurgical processes, hydrometallurgy has higher recovery efficiency, lower energy consumption, and lower emissions. However, hydrometallurgical technology has complicated process steps, high consumption of chemical reagents, and environmental pollution. Direct recycling is to recycle the negative and positive electrode materials as an integral part, and thus, they can be directly reused in Li-ion battery manufacturing. Since complicated purification processes and active material synthesis are avoided, direct recycling has economic advantages and is environmental friendly. However, its recycling efficiency largely depends on the health of the used Li-ion battery. In addition, the physical recovery method is a developing method, and it is expected to have a good development prospect.

Owing to there are many ways to recycle battery materials, some are still in their infancy. Moreover, the battery material is updated very quickly, resulting in the corresponding material recycling methods often lag behind the development of battery materials. Therefore, battery recycling methods are complex and diverse. Here are some cases to illustrate several classical battery recycling methods based on battery data.

6.4.2 Case 1: Physical Recycling Technologies

The physical recycling method uses the physical separation to recycle materials from the spent Li-ion batteries. The common physical recycling methods mainly include comminution and physical separation as:

• (1) Comminution

Before recycling battery materials by chemical or physical methods, the disassembled cells or modules need to be crushed. The common rotating comminution methods include hammer crushing [41], wet crushing [42, 43], shear crushing [44], impact crushing [45], and cutting milling. The main principles of these comminution methods are shown in Fig. 6.21a. Different crushing processes will produce different sizes and shapes of materials, which will seriously affect the subsequent separation processes [46, 47].

Fig. 6.21

Schematic illustration of the commonly used physical recycling methods. a Rotating comminution; b size separation; c magnetic separation, reprinted from [40], with permission from Elsevier

• (2) Separation

The physical separation is a commonly used technology to facilitate subsequent material recycling. It uses the physical properties of the mixture (e.g. colour, density, magnetic properties, particle size, and surface physical properties) to separate components from waste Li-ion batteries as much as possible. The commonly used physical separation methods are summarized as follows:

Size separation: Size separation is a common process for the preliminary separation of the crushed spent batteries. It is usually realized by the vibrating screen, and its main principle is shown in Fig. 6.21b. The comminuted mixture can be divided into fine particles (<1 mm) and coarse particles (>1 mm) by size. Generally, the coarse particles mainly consist of plastic, separator, aluminium foil, and copper foil, while the fine particles mainly consist of positive and negative materials. In Ref. [46], the mixed particles were divided into five categories: ultrafine particles (<0.5 mm), fine particles (0.5–1 mm), medium particles (1–2.5 mm), coarse particles (2.5–6 mm), and ultra-coarse particles (>6 mm). The separation results indicated that 82% Co was obtained in the ultrafine particles and 68% Co was obtained in the fine particles. For the NCM batteries, Co is the most valuable metal, while Ni is abundant. Therefore, the potential recycling value of Ni in cathode materials would exceed that of Co.

Magnetic separation: Magnetic separation is an effective method to separate metals from non-metallic components. It can be divided into dry-magnetic and fluid-magnetic separations. The schematic illustration of their working principle is shown in Fig. 6.21c. The process of the dry-magnetic separation can be described as follows: the mixed particles are conveyed to the magnetic roller by the conveyor belt. With the rotation of the magnetic roller, the non-magnetic particles fall into the non-magnetic collector, and the magnetic particles move to the baffle with the magnetic roller, and then fall into the magnetic collector. For the fluid-magnetic separation, the material and water are mixed based on a certain solid–liquid ratio and stirred at a little speed. The uniformly stirred suspension flows to the magnetic plate at a constant flow rate, and the magnetic particles are tightly adsorbed on the collecting mat under the action of magnetic force, while non-magnetic particles are taken away and collected by the collector under the action of water flow [48]. This method provides a new idea for separating the micromagnetic particles from mixed particles. To solve the problem that the existing mechanical separation technologies can only separate particles larger than 0.075 mm, Ref. [49] proposed a technology of ultrasonic dispersion and waterflow-magnetic separation to recover micromagnetic particles from mixed microparticles. Moreover, other metal and non-metal separation technologies were developed in recent years, such as electrostatic separation [50], eddy current separation [51].

Gravity separation: There is an obvious density difference among the mixture components obtained by crushing the spent batteries, which makes the gravity separation possible. The gravity separation can be achieved using shaker tables, vibrating screens, a fluid of intermediate density, or air separation. Reference [52] used different airflow rates to spray and clean mixed components for separation. It is shown that when the air velocity is 10.2–10.5 m s⁻¹, the smaller diameter polymers were separated, such as Cu and Al; when the air velocity is 10.6–13 m s⁻¹, the Cu and Al with larger diameter were separated directly. Moreover, the falcon centrifugal classifier is an efficient gravity separator, which is widely used to separate cathode and anode materials [53]. However, the gravity separation is not suitable for the separation of fine electrode materials [54].

Flotation separation: Flotation separation is an efficient process to separate fine particles based on the difference in surface hydrophilicity of mixtures [55]. For example, the cathode materials in spent Li-ion batteries are hydrophilic, while the anode materials are hydrophobic [56], which provides a basis for their flotation separation. However, the electrode material of spent Li-ion batteries is wrapped by organic matter, further reducing the difference in hydrophilicity between the positive electrode and negative electrode materials. Therefore, surface modification is a necessary step to improve flotation efficiency. Some surface modification methods, such as aerobic coasting [56], anaerobic pyrolysis [57], Fenton high-order oxidation [58], mechanical grinding [59], and cryogenic grinding [60], are applied to remove organic matters on the surface of the electrode materials. Owing to the low recycling rate of resources and serious environmental pollution, the aerobic coasting is less recommended. Although the Fenton high-order oxidation can remove organic matter from electrode materials, Fe²⁺ is introduced into the solution to activate the reaction [61]. Fe²⁺ remains on the surface of electrode materials, which complicates the subsequent metallurgical process. The mechanical grinding can remove some organic matter on the surface of electrode materials. However, the organic binder and electrolyte remain on the surface of electrode particles, resulting in a low recycling rate of cathode materials. Reference [62] proved that pyrolysis-assisted surface modification is an effective method to improve flotation efficiency.

It should be noted that the above physical separation methods may be combined to improve the recycling rate. First, the mixture can be separated by size after the battery is crushed. Then, the steel shell and ferromagnetic material can be removed by magnetic separation. Third, the separator and packaging can be recovered by density or electrostatic separations. Furthermore, the separation of plastics is realized by density separation. In recent years, some innovative methods have been proposed for physical recycling. Reference [63] reports a method of delaminating Li-ion battery electrode by using high-power ultrasonic generator. The adhesion between active material and collector can be quickly broken by the high-power ultrasonic. When the electrode is directly under the high-power ultrasonic generator, the stratification time of the electrode is less than 10 s, and the recycling efficiency of the proposed method is 100 times higher than that of the traditional methods. Reference [64] proposed a battery physical recycling method in which the battery is fully charged and then is placed it in water. This method can easily separate the negative electrode material and negative collector to obtain lithium salt. The physical recycling methods have the advantages of short process and being environment-friendly. The development of innovative physical recycling methods has great significance and bright prospects.

6.4.3 Case 2: Chemical Recycling Technologies

The current chemical recycling technologies of the spent Li-ion batteries mainly include pyrometallurgical, hydrometallurgical, and biometallurgical technologies. The pyrometallurgical technology generally does not require pretreatment, and the technical complexity is low. However, lithium and aluminium are easily discharged with the slag and cannot be recovered, and the metallurgical process has high-energy consumption and serious environmental pollution, which does not meet the requirements of environmentally friendly industries. Biometallurgy is the use of microorganisms to secrete inorganic or organic acids to recover metal substances in spent Li-ion batteries. This technology has the advantages of low energy consumption and low secondary pollution, but the microbial cultivation conditions are harsh and the leaching cycle is long. In comparison, hydrometallurgical technology has attracted more and more attention due to its advantages of high recycling efficiency, low cost, low energy consumption, and low secondary pollution. The mature technology of pyrometallurgy and hydrometallurgy is summarized as follows:

• (1) Pyrometallurgical technologies

Conventional pyrometallurgical technologies can be classified into pyrolysis and reduction roasting. The basic principle of pyrolysis is that the electrode materials will be converted into relatively stable oxidation or metal states under a high-temperature environment. This method is generally used for the recycling of cathode materials. Reference [65] revealed a unique phase transition behaviour of Li_1/3Ni_1/3Co_1/3Mn_1/3O₂ cathode during heating: the initial layer structure first transformed into an Li₂O₄-type from 236 to 350 _°C, and then to a M₃O₄-type spinel from 350 to 441 °C. Reference [66] indicated that the content of Ni, Co, and Mn in NCM cathode materials significantly affects the structural changes during heating, and the more Ni and less Co and Mn, the lower the temperature of the phase transition, as shown in Fig. 6.22a. Reference [67] proposed a new method for predicting the thermodynamics of thermal degradation of the cathode materials of Li-ion batteries. In summary, cobalt, nickel, copper, and other metals are melted and recovered as alloys in pyrolysis, which needs the subsequent treatment. In addition, lithium and other components will be discarded in the form of slag and gas at high temperatures, resulting in the loss of valuable metals. Furthermore, the high-energy consumption and toxic gas emission are the other defects of the pyrolysis method for the recycling of retired Li-ion batteries.

Fig. 6.22

Cases of pyrometallurgy methods. a Temperature region of the phase transitions for NMC; b collapsing model of recycling metals from LiCoO₂ by roasting, reprinted from [40], with permission from Elsevier

In the reduction roasting method, coke, carbon monoxide, and active metals are used as reducing agents to reduce metals from their compounds. Generally, the anode materials of battery are usually used as high-temperature reducing agents for the recycling of cathode materials. In Ref. [68], the NCM cathode materials were reduced to Li₂CO₃, MnO, NiO, Ni, and Co by calcining with coke as a reducing agent at 650 °C for 30 min. In Ref. [69], the LCO cathode material was reduced from the crystal structure of the cathode material to form Li₂CO₃ at high temperature. In Ref. [70], the cathode and anode materials of LCO were calcined together, and the following coupling reactions have occurred:

$$ \left\{ {\begin{array}{*{20}l} {4{\text{LiCoO}}_{2} + {\text{C}} \to 2{\text{Li}}_{2} {\text{O}} + {\text{CO}}_{2} + 4{\text{CoO}}} \\ {4{\text{LiCoO}}_{2} + 2{\text{C}} \to 2{\text{Li}}_{2} {\text{O}} + 2{\text{CO}} + 4{\text{CoO}}} \\ {{\text{Li}}_{2} {\text{O}} + {\text{CO}} + {\text{CoO}} \to {\text{Co}} + {\text{Li}}_{2} {\text{CO}}_{3} } \\ {{\text{Li}}_{2} {\text{O}} + {\text{C}} + 2{\text{CoO}} \to 2{\text{Co}} + {\text{Li}}_{2} {\text{CO}}_{3} } \\ \end{array} } \right. $$

(6.21)

From the point of view of crystal structure, the stronger attraction of graphite to oxygen than lithium and cobalt would make oxygen octahedrons inside lithium cobalt oxide unstable to break down. The whole process is illustrated in Fig. 6.22b. In addition, binders, separators, electrolytes, aluminium shells, plastics, and by-products of chemical plants (e.g. sulphur-containing tail gas, and slag) can also be the reducing agents for cathode materials. However, with the development of cobalt-free electrode materials in Li-ion batteries, high-temperature roasting technology is facing challenges.

• (2) Hydrometallurgical technologies

Hydrometallurgy is a common recycling method, which mainly uses the acid or alkali systems as the leaching agent. Under the combined action of the reducing agent, the waste cathode material is dissolved, so that the elements of Li, Ni, Co, and Mn are transferred to the liquid phase, and a multi-element mixed solution is formed to achieve the purpose of further recycling. The recycling process is shown in Fig. 6.23, and it can be observed that it has the characteristics of low environmental pollution and high recovery efficiency. The core step of hydrometallurgy recycling is the leaching process, which is mainly divided into two types: acid leaching and alkali leaching. Among them, acid leaching is the more commonly used method. In the acid leaching process, it is usually necessary to use a reducing agent as auxiliary material to reduce the high-valent transition metal elements in the waste cathode material to a low-valent state, thereby accelerating the leaching process. The acid leaching agents mainly include sulphuric acid, hydrochloric acid, nitric acid, phosphoric acid, and hydrofluoric acid, and the alkali leaching agents mainly include ammonia and ammonium sulphate. The main reducing agents include hydrogen peroxide, sodium sulphite, sodium bisulphite, sodium thiosulphate, ammonium chloride, etc. From the literature, the acid leaching method has absolute advantages in the process of cathode waste from solid state to ionic state. Generally, organic acid or inorganic acid is used as leaching agent, the M–O (M=Ni, Co, Mn) bond in the cathode material is destroyed by H+ structure, and then the metal is leached in ionic state. Inorganic acid has strong acidity, which can dissolve most valuable metals into ionic state and enter the solution, but its corrosion to the reaction vessel is also very serious; organic acid has weak acidity, has special spatial structure and binding site, and is widely used because of its advantages of the high recovery rate of valuable metals, low pollution, and easy control Pan application. The following two leaching methods are briefly introduced.

Fig. 6.23

Hydrometallurgy recycling process of spent Li-ion batteries

Inorganic acid leaching methods: The inorganic acids commonly used for solvent leaching of valuable metals in electrode waste generally include strong acids such as HCl, H₂SO₄, HNO_3, and H₃PO₄, among which HCl has the highest leaching efficiency. Under the optimum leaching conditions of HCI, the leaching rates of Ni, Co, Mn, and Li can reach more than 99%. However, HCl is volatile and easy to produce harmful gas, which requires special anti-corrosion equipment. The recycling process would cause great pollution to the environment, resulting in increased recycling costs.

In recent years, inorganic acid leaching has been widely concerned for its high leaching efficiency, high selectivity of leaching agent, and mature technology. Some representative leaching results of different cathode materials in different inorganic acids are listed in Table 6.7. Generally, the leaching effect is affected by the concentration of leaching agent, the amount of reducing agent, leaching temperature, leaching time, and solid–liquid ratio. Reference [71] proposed a two-step leaching method to extract valuable metals selectively from LiNi_xCo_yMn_1−x−yO₂ cathode materials: First, Ni, Co, and Li were leached from lixivium either as complexes or metallic ion by employing ammoniacal solution as the leaching agent and sodium sulphite as reductant. Second, manganese was deposited from Mn₃O₄ to (NH₄)₂Mn(SO₃)₂·H₂O as sodium sulphite was added. The loose and porous Mn₃O₄ is more favourable for ion diffusion and leaching reaction, as shown in Fig. 6.24.

Table 6.7 Summary of leaching results for the spent Li-ion batteries in inorganic different acids

Fig. 6.24

Process of a two-step leaching method, reprinted from [40], with permission from Elsevier

Moreover, the use of alkaline systems in hydrometallurgy recycling can also achieve an excellent leaching effect. For example, under the combined action of the leaching agent of ammonia and ammonium carbonate, and the reducing agent of ammonium sulphite, Co and Cu can be selectively leached and recovered from waste LiMn₂O₄ and ternary materials. Through the above analysis, it can be concluded that a suitable leaching system can successfully achieve the effective leaching and recycling of metal elements in the cathode materials of the waste Li-ion batteries. The recycling product is a mixed solution of multiple elements such as Li, Ni, Co, and Mn. In the subsequent processing steps, on the one hand, the metal salt compounds can be extracted from the leachate by selective separation and used as industrial raw materials; at the same time, the leachate can be directly used as the raw material of the electrode material regeneration process to improve the recovery of valuable metals.

From the literature, the leaching efficiency of the inorganic acid is very high. However, the wastewater, waste residue, and harmful gases (e.g. SO₂, SO₃, NO_x) will be produced in the leaching process, which poses a great threat to the ecological environment and human health. Therefore, green environmental protection, high efficiency, low-cost leaching agent leaching method are the future trend and direction.

Organic acid leaching methods: Since it is inevitable that volatile and toxic gases will be generated during the use of inorganic acids, which are harmful to human health and pollute the environment, and the current research focuses on the use of acid system recovery methods is also on some natural organic acids in recent years. Although the acidity of organic acids is lower than that of inorganic acids, some organic acids still show quite a good leaching rate in the leaching process, which is mainly due to the formation of complexes between organic acid radical ions and valuable metal cations [90]. According to the different leaching mechanisms, organic acids are divided into chelating organic acids, reducing organic acids, precipitating organic acids, and other organic acids [91]. The common chelating organic acids are citric acid, malic acid, platinum succinate, and aspartic acid, and the common reducing organic acids are ascorbic acid (C₆H₈O₆) and lactic acid (C₃H₆O₃). The typical precipitating organic acid is oxalic acid (H₂C₂O₄). In Ref. [92], C₃H₆O₃ was chosen as the leaching and chelating agent to recycle the cathode materials from spent Li-ion batteries. Table 6.8 lists some leaching effects of different cathode materials in different organic acids. After leaching valuable metals with organic acids, the leaching solution or recovered products are used to replace the raw materials for the resynthesis of electrode materials, to realize the closed-loop recycling of the spent Li-ion batteries.

Table 6.8 Summary of leaching results for the spent Li-ion batteries in organic acid

The waste cathode material recovered by the above method is usually a mixed system of several different metal ions. If a certain single metal ion is required in the subsequent process, the next step of separation operation is required, such as solvent extraction, chemical precipitation. The solvent extraction uses a two-phase system (usually an organic phase and an aqueous phase) to achieve separation through the uneven distribution of different ions in the two phases. The separation mechanism of chemical precipitation is the different solubility of metal compounds at a certain pH. In general, the solubility of transition metal hydroxides and oxalates is much lower compared to the corresponding lithium compounds. Thus, different metal ions can be used for cascaded precipitation separation at different pH values to efficiently separate metal ions such as Ni, Co, Mn, and Li. The commonly used precipitation agents include NaOH, H₂C₂O₄, (NH₄) ₂C₂O₄, Na₂CO₃, Na₃PO₄, etc. The subsequent processes are diverse and not be specifically introduced here.

The waste Li-ion batteries contain a large number of valuable metal resources, such as Ni, Co, Mn, Li, which have good prospects for recycling and are gradually being valued. At the same time, the recycling of large amounts used Li-ion batteries poses new challenges to environmental protection and sustainable use of resources and puts tremendous pressure on the development of appropriate recycling technologies. The pyrometallurgical recycling process has been widely studied due to its short process and high efficiency. The hydrometallurgical recycling process has gradually become a research hotspot due to its good selectivity to valuable metals and mild reaction conditions. It is relatively mature and worthy of promotion. The recycling mechanism of hydrometallurgy should be studied continuously, and the battery industry chain, the related recycling processes, and equipment should be perfected. Under the continuous improvement of battery recycling policies, a comprehensive utilization system should be established, a market recycling system should be improved. More importantly, cascade utilization and battery dismantling and recycling should be more effectively integrated and developed, and the waste Li-ion battery companies and other energy companies should coexist and develop in harmony.

6.5 Summary

This chapter describes the data science-based battery reutilization management for retired Li-ion batteries. First, the echelon utilization and material recycling of retired Li-ion batteries are briefly introduced. They can maximize the full-lifespan value of Li-ion batteries and alleviate the pressure of lithium, cobalt, manganese, and other resources, which is of great significance to the sustainable development of Li-ion batteries. Second, aiming at the sorting problem in the echelon utilization of large-scale Li-ion batteries, a data-based sorting criterion is constructed, the test data is used to train the neural network model, and the machine learning algorithm is used to complete the rapid sorting of large-scale Li-ion batteries. In addition, two cases are used to verify the proposed fast sorting method: one case is used to describe the fast estimation of battery capacity and internal resistance using partial charging data based on the proposed method; the other case describes a fast estimation method of battery capacity and other important characteristic parameters using EIS, and the experimental results show that the speed of obtaining battery capacity by the proposed method is 10 times higher than that by the traditional method. Third, aiming at the problem of battery regrouping in echelon utilization, a hard-clustering method based on K-means and a soft clustering method based on the GMM algorithm are proposed. Finally, the material recycling methods are comprehensively summarized and discussed. The establishment of a traceability management platform for material and information flows of Li-ion batteries based on big data in the full-lifespan cycle can greatly facilitate the echelon utilization and material recycling of Li-ion batteries.