Risks of mineral resources in the supply of renewable energy batteries

Abstract

Renewable energy batteries play a crucial role in the stable storage of clean energy. However, the supply risks associated with critical mineral raw materials closely related to renewable energy batteries - namely lithium, manganese, cobalt, and nickel - significantly threaten the safety and stability of these batteries. Therefore, this study selects representative factors from four aspects: resources, market, international relations, and technology, and employs the SMAA-TRI method to assess the supply risks of critical minerals required for renewable energy storage batteries. The results indicate that: (1) From 2006 to 2022, the supply risk of lithium resources for renewable energy batteries in China evolved from medium-high to high, while the risks for manganese, nickel, and cobalt resources remain within the high-risk range; (2) Predictions from the BP neural network model suggest that lithium, manganese, nickel, and cobalt would continue to be in the high-risk range over the next three years; (3) Sensitivity analysis reveals that environmental safety, resource recovery rates, substitution rates, external dependencies, and production concentration would become significant factors constraining the supply risks of renewable energy storage batteries.

Introduction

With the advancement of global climate governance, clean energy has experienced large-scale development within the electricity sector, gradually becoming the primary force driving the low-carbon energy transition¹. However, it is noteworthy that renewable energy sources such as solar, wind, and hydroelectric power present challenges of unstable power supply and intermittency. This necessitates the utilization of energy storage technologies to consolidate and stabilize the generated electricity. Lithium-ion batteries have emerged as the mainstream solution for renewable energy storage due to their high energy density, long lifespan, and environmental friendliness². Lithium, manganese, nickel, and cobalt are the four most critical mineral raw materials in current renewable energy storage batteries, particularly lithium-ion batteries. Lithium-ion batteries are extensively employed in renewable energy storage systems, and their performance is significantly dependent on the critical materials within the batteries. Lithium, serving as the core anode material, directly influences the battery’s energy density and cycle life. Cobalt, nickel, and manganese, as cathode materials, enhance the battery’s stability and safety, contribute to increased energy density, and help reduce battery costs while maintaining favorable performance by providing higher voltage and greater specific capacity. These four mineral resources are essential components of large-scale energy storage systems. Therefore, investigating the supply risks associated with these mineral resources is crucial for ensuring the sustainable development of renewable energy storage systems³.

With the widespread application of lithium batteries, the supply of the mineral resources they rely on—particularly critical raw materials such as lithium, manganese, cobalt, and nickel—faces numerous risks. These supply risks may lead to disruptions in the battery production supply chain, thereby constraining the stable development of renewable energy storage batteries. Firstly, the scarcity of important mineral resources like lithium, cobalt, and nickel, combined with geopolitical risks, results in an unstable raw material supply⁴. Secondly, the rapid growth in global demand within the renewable energy industry has caused a supply-demand imbalance in the market, leading to price volatility, increased production costs, and supply chain uncertainties⁵. Furthermore, technical bottlenecks in areas such as energy density, safety, and recyclability limit the widespread application and cost reduction of lithium batteries⁶. In addition, uncertainties in international relations and trade policies have significantly impacted the global supply chain of renewable energy storage batteries. Therefore, accurately identifying and assessing these resource supply risks is vital for ensuring the stable development of the renewable energy industry.

Current scholars often use matrix methods, cluster analysis, and radar methods to explore the supply risks of critical mineral resources. The matrix method uses risk matrices and two-dimensional tables for semi-qualitative risk analysis but has the drawback of difficult parameter selection⁷. Cluster analysis aggregates the supply risk index by calculating each sub-indicator, but it performs poorly in evaluating uncertain indicators⁸. In order to avoid these limitations, a qualitative and quantitative research method represented by Stochastic Multi-Criteria Acceptability Analysis (SMAA) has been proposed and applied to the risk assessment of minerals such as copper, cobalt, chromium, and nickel⁹. This method effectively addresses the issue of inconsistent opinions among decision-makers and resolves the multi-attribute decision-making problem with missing weight information.

To sum up, existing research on the sources of supply risks for critical minerals in renewable energy storage batteries is relatively scattered, notably lacking the integration of international relations into a unified analytical framework encompassing resources, market, and technology. Moreover, current assessment methods fail to meet the differing standards of qualitative and quantitative criteria simultaneously, and combining these methods lacks multi-criteria consideration. To address these research gaps, this study employs the SMAA-TRI method to evaluate the supply risks of four critical mineral resources (lithium, manganese, nickel, and cobalt) in China, conducting a comprehensive analysis of future primary supply risk points across four dimensions: resources, market, international relations, and technology. Furthermore, this study utilizes neural network models and sensitivity analysis to forecast the key factors that would influence the supply risks of these critical mineral resources in the future. This research aims to provide scientific evidence for enhancing the security and stability of the renewable energy storage batteries supply chain and offer theoretical support for formulating more effective resource management and international cooperation strategies.

The structure of this study is as follows. Section “Literature review” reviews the literature on assessing critical mineral risks and provides an overview of the current state of research. Section “Methods” analyzes the critical methods for developing the mineral supply risk index and the primary data sources. Section “Results analysis and discussion” discusses the evaluation results of the supply risks for critical mineral resources and identifies the sources of these risks. Section “Conclusion” summarizes the findings of the entire study and offers policy recommendations.

Literature review

Related research on renewable energy storage batteries

With the continuous expansion of demand in the renewable energy market, scholars have noticed that the safety of critical mineral supply may constrain the development of batteries¹⁰. Existing studies on the supply risk of critical minerals involve different dimensions of risk assessment indicators, such as resources, markets, and technology¹¹. Althaf et al. conducted a risk assessment of critical metals and minerals supply from the perspectives of resources and markets, including concentration of production, price fluctuations, and external dependence as risk assessment indicators¹². Zhao et al. enriched the dimensions of markets and resources based on risk management and vulnerability theory, adding ore extraction ratio and resource security in the resource dimension and production growth rate in the market dimension, and compared the supply risk of chromium and gallium in China, the United States, and India¹³, effectively evaluating the supply risk of chromium and gallium. In addition to selecting risk indicators from the dimensions of resources and markets, some scholars have added the dimension of technology to assess the supply risk of critical mineral resources. Palomino et al. selected indicators of recyclability and substitutability from the technology dimension, the concentration of supply from the market dimension, and external dependence from the resource dimension, discussing the supply risk of platinum group metals in spintronics¹⁴. Yuan et al. quantitatively assessed the supply risk of platinum from the perspective of technology, using resource recovery rate and resource substitutability as risk assessment indicators¹⁵.

Besides the dimensions of resources, markets, and technology, internal political risks of countries can affect the output of natural resources, thereby impacting the supply security of mineral resources¹⁶. Anish et al. used the GeoPolRisk method to analyze the impact of concentration of production and political stability on geopolitical supply risks of some metals and fossil fuels in different OECD countries¹⁷. Zhang et al. studied the impact of political instability in producing countries on the supply chain risks of some critical metals in solar panels¹⁸. Meanwhile, some scholars have noted that the environmental conditions of mining countries will also affect resource supply⁹. In addition to the above indicators of resources, markets, technology, political stability, and environmental security, the impact of fluctuations in international relations on bilateral trade between countries has become increasingly significant, especially in critical areas. Wang and Tao used gravity models to empirically analyze the impact of fluctuations in China-ASEAN bilateral relations on bilateral trade, concluding that the improvement of country relations is conducive to trade between China and ASEAN countries¹⁹. Conversely, deteriorating relations between countries lead to a significant reduction in trade volume²⁰. Table 1 summarizes and organizes the critical mineral evaluation factors involved in existing literature.

Table 1 Critical mineral supply risk assessment indicators in the existing literature.

Research on risk assessment of critical mineral resources

With the advancement of battery technology, some scholars have focused on studying the critical raw materials of lithium-ion batteries and evaluating their supply risks²⁵. Some scholars use qualitative methods for risk assessment of critical mineral resources, such as expert questionnaires and the Analytic Hierarchy Process^26,27. This type of method usually uses expert consultation or comprehensive evaluation to evaluate the weight of indicators, which has a certain degree of subjectivity²⁸. Recently, more and more scholars have adopted quantitative methods to assess the risks of critical minerals, such as complex network methods and supply risk quantification methods. For example, Li et al. utilized complex network methods to comprehensively assess the supply risk of China’s copper industry chain²⁹. Zhang et al. quantified the supply risk of critical minerals to assess the supply risk of minerals used in clean energy technologies³⁰. Quantitative methods generally involve using complex mathematical models to quantify the supply risk of critical minerals but lack subjective consideration in obtaining results, making it difficult to solve complex system problems and prone to practical biases³¹.

To overcome the limitations of previous methods, scholars have begun integrating quantitative and qualitative approaches to assess the supply risk of critical mineral resources. One such method is the Stochastic Multi-Criteria Acceptability Analysis (SMAA), pioneered by Lahdelma et al. in 1978. Tervonen et al. later introduced the SMAA-TRI method, which incorporates the ELECTRE TRI method with imprecise and arbitrarily distributed weight values³². This approach effectively manages inaccurate data and uncertain parameters, visualizing the robustness of ELECTRE TRI results. This method has been widely used in ranking classification³³ and risk assessment of mineral resource supply⁹. However, in using the SMAA-TRI method to assess the supply risk of critical mineral resources, Monte Carlo iterative simulation of indicator weights is often applied, and there is little research on using combined subjective and objective weights as risk indicator weights.

Prospective prediction of risks associated with critical mineral resources

The Backpropagation (BP) artificial neural network³⁴ is a widely used multilayer feedforward neural network employing the error backpropagation algorithm³⁵. Renowned for its reliability and ease of implementation, BP neural networks are extensively utilized in predictive research across diverse domains, such as air quality index prediction, enterprise management efficiency forecasting, and power load fluctuation prediction. Chen et al. applied a BP neural network model to assess meteorological conditions’ impact on haze weather, achieving promising predictive accuracy³⁶. Liu et al. integrated a model-driven DEA method with a data-driven BP neural network approach to evaluate coal enterprises’ safety management efficiency, highlighting the adaptability of their models³⁷. Shen et al. utilized BP neural network models and support vector machine models to forecast the randomness of renewable energy and power consumption plans in microgrids, demonstrating superior predictive performance compared to SVM³⁸. Notably, BP neural network models excel in predicting errors and trends^39,40 and exhibit robustness to input and output variable distributions⁴¹, making them suitable for prospective forecasting tasks concerning critical mineral resources. However, limited research remains on the prospective prediction of supply risks associated with critical mineral resources.

After reviewing existing literature, it was found that existing research on risk assessment of critical mineral resources mainly emphasizes resource, market, and technological aspects, neglecting the impact of international relations. In terms of analytical methods, combining quantitative and qualitative methods can reduce the subjectivity of evaluation and consider the interconnection of systems. In addition, research on incorporating prospect prediction into risk assessment is still a blank spot.

To fill this research gap, this study evaluated the supply risk of critical mineral resources for lithium-ion batteries from four dimensions and made forward-looking risk predictions. The innovations include the following three points: (1) Based on integrating dimensions of resources, markets, and technology, this paper innovatively incorporates indicators related to international relations and political stability into the construction of the indicator system, greatly enriching the perspective of existing assessments of supply risk for critical mineral resources. (2) Application of the SMAA-TRI method combining quantitative and qualitative approaches and adopting a unique weighting method that combines subjective weights for complex qualitative issues with objective weights considering the intensity and conflict of indicators. (3) We employ artificial intelligence backpropagation neural networks to predict future supply risk situations and identify key influencing factors constraining the supply risk of critical mineral resources for lithium-ion batteries.

Methods

Construction of indicator system

The security of critical mineral resource supply needs to consider supply stability, sustainability, timeliness, and economy. Based on this, this study constructed a risk assessment index system for the supply of critical mineral resources in lithium-ion batteries for renewable energy storage batteries. The indicator system includes four dimensions and 14 sub-indicators, as shown in Fig. 1.

Fig. 1

Indicator system for supply risk of critical mineral resources.

• (1) Resource dimension

This dimension includes five sub-indicators: ratio to global reserves, ratio of reserves to production, ratio to global production, resource security, and external dependence. Among them, the ratio to global reserves reflects China’s relative position in the world’s resource endowment³⁰. The ratio of reserves to production can measure the sustainability of mineral resources per unit of time⁴². The proportion of global production reflects China’s relative position in the global supply of mineral resources⁴³. Resource security is the ratio of critical mineral resource reserves to consumption, reflecting the domestic resource security capability when external resource supply is completely interrupted. External dependence refers to the ratio of net imports of specific critical mineral resources to domestic apparent consumption, reflecting the degree of dependence of domestic mineral resource consumption on foreign resources⁴⁴, where positive values indicate import dependence and negative values indicate export dependence. The higher the degree of external dependence, the greater the dependence on the supply of mineral resources from other countries.

• (2) Market dimension

The market dimension belongs to external risk factors. It includes four sub-indicators: production growth rate, concentration of production, consumption growth rate, and price fluctuation. Specifically, the production growth rate refers to the ratio of the current year’s production of specific critical mineral resources to the previous year, reflecting the stability of mineral supply¹³. The concentration of production reflects the proportion of production from major producing countries of specific critical minerals to global production. The Herfindahl-Hirschman Index (HHI) measures the concentration of mineral production, representing the monopoly risk in foreign mineral resource markets⁴⁵. A higher concentration of production implies a higher risk of resource monopolization. The consumption growth rate indicates the degree of increased or decreased domestic consumption of specific critical mineral resources. A higher consumption growth rate suggests a faster increase in market demand for the mineral resource, increasing the probability of supply-demand imbalance and thus elevating the risk of mineral supply⁴⁶. Price fluctuation refers to the magnitude of annual average price changes of specific critical mineral resources, reflecting the supply-demand relationship in the mineral market. Greater price fluctuation indicates supply-demand imbalances, signifying higher supply risk for the mineral resource⁴⁷.

• (3) International relations dimension

The international relations dimension falls under external risk factors, reflecting the diplomatic relations of the primary source countries of critical mineral resources and the influence of domestic political stability or changes in environmental policies in source countries, posing risks of insufficient or interrupted supply. Critical mineral resources’ depletable and scarce nature has transformed them from ordinary commodities into particular political assets, becoming significant leverage in international relations and geopolitical maneuvering. Therefore, this dimension is divided into three sub-indicators: political stability, environmental security, and country relations. Political stability reflects the level of political stability in mineral-source countries⁴⁸, typically measured using the Political Stability and Absence of Violence/Terrorism (PV) value from the Global Governance Index. Environmental security refers to changes in resource extraction policies in critical mineral-source countries due to environmental pollution, reducing the production of critical minerals and affecting their supply. Consequently, the worse the environmental conditions, the greater the risk to mineral supply⁴⁹. Country relations refer to the diplomatic relations between China and major trading partners of critical minerals. When relations between countries are tense, interruptions in the supply of mineral resources are highly likely. Following the research, a combined qualitative and quantitative approach is used to assign values to events and accumulate the impact of events over time. This ultimately yields a score for country relations. Lower event scores indicate poorer country relations, leading to higher supply risks for critical mineral resources⁵⁰.

• (4) Technological dimension

Technological advancements can mitigate dependency on critical mineral resources by exploring economically viable substitutes. Advancements in recycling technology can also enhance resource recovery rate and lower recycling costs. Consequently, this dimension encompasses two sub-indicators: resource substitution rate and resource recovery rate. The resource substitution rate evaluates the feasibility of replacing minerals without cost increases or performance reductions²³. The substitutability index from the European Commission report measures this rate⁹. Meanwhile, the resource recovery rate assesses the impact of waste consumption on resource supply pressure by comparing recovered resources with critical mineral consumption¹⁴. The supplementary Table S1 provides a detailed calculation method for evaluation indicators.

SMAA-TRI method

The SMAA-TRI method is an extension of the Stochastic Multi-criteria Acceptability Analysis (SMAA) approach, integrating the Elimination and Choice Translating Reality (ELECTRE) technique. As an effective multi-criteria decision analysis tool, the SMAA-TRI method allows for the inclusion of non-deterministic parameters within the model and provides decision-makers with more comprehensive outputs beyond parameter inference, thereby proving to be a suitable tool for addressing complex decision-making problems⁵¹. The SMAA-TRI method possesses comprehensive multi-criteria evaluation capabilities and flexible adaptability. Establishing a multi-criteria evaluation framework that incorporates uncertainty analysis, it effectively identifies the supply risk levels of mineral resources. This method has been successfully applied in complex decision-making environments such as environmental management, energy management, and risk assessment¹³. Therefore, this study employs the SMAA-TRI method to assess the supply risks of critical mineral resources for renewable energy storage batteries. The methodological workflow is illustrated in Fig. 2.

Fig. 2

Process flowchart of the SMAA-TRI method.

The critical parameters of the SMAA-TRI method are:

• (1) Weights of Indicators (w): Represent the importance of each indicator.

• (2) Threshold Profile (Pr_h): Defines each risk level h.

• (3) Preference Threshold (p-value) and Indifference Threshold (q-value): Fuzzified the boundaries of the profile Pr_h to measure uncertainty in risk levels.

• (4) Cut Level (λ): Minimum cumulative weight supporting the next class of indicators, with a range of [0.5,1] indicating most decision criteria.

• (5) Data and Parameters of ELECTRE TRI: Include the veto threshold (v), all with deterministic values.

• (6) Category Acceptability Index$\pi _{i}^{h}$: Describes the probability of assigning alternative solutions to category C_h, defined through a categorization function that calculates the category index h for the alternative solution x_i as assigned by ELECTRE TRI.

  $$h=K(i,\Lambda ,\phi ,w,T)$$

  (1)

A category membership function:

$$m_{i}^{h} (\lambda ,\phi ,w,T) = \left\{ \begin{gathered} 1,\quad if\;K(i,\Lambda ,\phi ,w,T) = h \hfill \\ 0,\quad otherwise \hfill \\ \end{gathered} \right.$$

(2)

Applied as a multidimensional integral on a finite parameter space to compute the category acceptability index, as shown below:

$$\pi _{i}^{h} = \int\limits_{{0.5}}^{1} {f_{L} (\Lambda )} \int\limits_{{\phi \in \Phi }} {f_{\Phi } (\phi )} \int\limits_{{w \in W}} {f_{W} (w)m_{i}^{h} (\Lambda ,\phi ,w,T)} dwd\phi d\Lambda$$

(3)

The acceptability index represents the probability of category levels, with the highest probability indicating the corresponding category for the alternative solution. If parameters are stable, the acceptability index for each alternative should be 1 for one grade and 0 for others.

Fuzzy analytic hierarchy process with the objective CRITIC

This study combines the Subjective Fuzzy Analytic Hierarchy Process (FAHP) with the Objective CRITIC method to achieve scientifically reliable indicator weights. While the SMAA-TRI method can operate without weights, its reliance on Monte Carlo simulations compromises results stability and accuracy. Thus, indicator weights are essential. The combined weighting method integrates expert knowledge with data objectivity. The framework is shown in Fig. 3.

Fig. 3

Flow chart of indicator weighting combining CRITIC and FAHP.

The steps of the CRITIC method are as follows:

• (1) Data standardization

$${\text{For}}\;{\text{positively}}\;{\text{oriented}}\;{\text{indicators}}:x^{\prime}_{{ij}} = \frac{{x_{{ij}} - \min (x_{j} )}}{{\max (x_{j} ) - \min (x_{j} )}}$$

(4)

$${\text{For}}\;{\text{inversely}}\;{\text{oriented}}\;{\text{indicators}}:x^{\prime}_{{ij}} = \frac{{\max (x_{j} ) - x_{{ij}} }}{{\max (x_{j} ) - \min (x_{j} )}}$$

(5)

where $\hbox{max} ({x_j})$ and $\hbox{min} ({x_j})$are the maximum and minimum values in the j-th indicator, respectively.

• (2) Information calculation

Information is divided into contrast intensity and conflict. Contrast intensity is represented as standard deviation, as shown in Eq. (6).

$$\left\{ \begin{gathered} \overline{x} _{j} = \frac{1}{n}\sum\limits_{{i = 1}}^{n} {x_{{ij}} } \hfill \\ S_{j} = \sqrt {\frac{{\sum\nolimits_{{i = 1}}^{n} {(x_{{ij}} - \overline{x} _{j} )^{2} } }}{{n - 1}}} \hfill \\ \end{gathered} \right.$$

(6)

where S_j denotes the j-th indicator’s standard deviation and is the j-th indicator’s mean.

Indicator conflict is represented in correlation coefficients, as shown in Eq. (7), where r_ij represents the correlation coefficient between evaluation indicators i and j.

$$R_{j} = \sum\limits_{{i = 1}}^{p} {(1 - r_{{ij}} )}$$

(7)

The formula for information content C_j and weight calculation are shown in Eqs. (8) and (9).

$$C_{j} = S_{j} \sum\limits_{{i = 1}}^{p} {(1 - r_{{ij}} )} = S_{j} \times R_{j}$$

(8)

$$W_{j} = \frac{{C_{j} }}{{\sum\nolimits_{{j = 1}}^{p} {C_{j} } }}$$

(9)

The fuzzy analytical hierarchy process (FAHP) combines fuzzy mathematics with the analytical hierarchy process (AHP) to resolve complex qualitative problems. FAHP employs fuzzy mathematics to determine the weights of indicators, effectively evaluating complex, multi-factor, multi-level issues, thus enhancing the scientific reliability of decision outcomes. FAHP enables decision-makers to derive precise weight arrangements from fuzzy complementary judgment matrices under uncertainty⁵². The specific steps are as follows:

• (1) Establishing the fuzzy complementary judgment matrix:

$$R = \left[ {\begin{array}{*{20}c} {\begin{array}{*{20}c} {r_{{11}} } & \cdots & {r_{{1n}} } \\ \end{array} } \\ {\begin{array}{*{20}c} \vdots & \ddots & \vdots \\ \end{array} } \\ {\begin{array}{*{20}c} {r_{{n1}} } & \cdots & {r_{{nn}} } \\ \end{array} } \\ \end{array} } \right]$$

(10)

where the notation r_ij represents the degree of membership

• (2) Single-layer weight calculation.

The weights of the elements (a₁, a₂ ,…, a_nd) the fuzzy consistency matrix R = (r_ij)_n×n are denoted as w₁, w₂,…, w_n, respectively. The following relationship holds and single-layer weights are as follows:

$$r_{{ij}} = 0.5 + a(w_{i} - w_{j} ),\quad i,j = 1,2, \ldots ,n$$

(11)

$$w_{i} = \frac{1}{n} - \frac{1}{{2a}} + \frac{1}{{na}}\sum\limits_{{k = 1}}^{n} {r_{{ik}} } ,\quad i,k = 1,2, \ldots ,n$$

(12)

• (3) Consistency check of fuzzy complementary judgment matrix

The compatibility index between the fuzzy judgment matrix $A={({a_{ij}})_{n \times n}}$ and its characteristic matrix ${W^*}={({W_{ij}})_{n \times n}}$ serves as a consistency test indicator, with the calculation formula as follows:

$$I(A,W^{*} ) = \frac{1}{{n^{2} }}\sum\limits_{{i = 1}}^{n} {\sum\limits_{{j = 1}}^{n} {\left| {a_{{ij}} + W_{{ij}} - 1} \right|} }$$

(13)

where W = (w₁, w₂, …., w_n)^T is the vector of importance weights of the fuzzy judgment matrix.

• (4) Calculation of Total Hierarchy Weights.

Within the indicator system, under the A level, the B level consists of m factors with weights denoted as b_jm. The relative total ranking weight of the j-th indicator in the B level to the A level is:

$$w_{j} = \sum\limits_{{k = 1}}^{m} {{\text{a}}_{k} } b_{{kj}} ,\;\;(j = 1,2, \ldots m)$$

(14)

The backpropagation neural network (BPNN) model

The Backpropagation Neural Network (BPNN) is a supervised learning algorithm often used to train multilayer perceptions. It is a multilayer feedforward network based on error backpropagation and employs the Widrow-Hoff learning algorithm and a nonlinear differentiable transfer function. The BP network model typically consists of three layers: the input layer, one or multiple hidden layers, and the output layer. Neurons between layers are fully connected, while neurons within layers are not connected. Sigmoid activation functions are commonly used for nodes in hidden layers.

Before prediction with the BP neural network, network training is conducted to adjust its weights and thresholds, enabling the network to possess associative memory and prediction capabilities. The training process of the BP network involves the following steps:

• (1) Data Preprocessing: Preprocessing is performed using normalization methods to enhance the training and prediction performance of the neural network.

• (2) Design and Parameter Setting of the Prediction Network: The optimal number of nodes in the hidden layer is determined using the Equation $g<\sqrt {m+n} +a$, where g is the number of hidden layer nodes, m is the number of input layer nodes, n is the number of output layer nodes, and a is a constant ranging from 1 to 10. Additionally, parameters such as learning rate, training error, and number of training iterations are set.

• (3) Accuracy Testing of the Neural Network: Following network training, the trained model is subjected to accuracy testing to assess the applicability of the prediction model.

Results analysis and discussion

This section provides a comprehensive assessment and prospect prediction of the supply risks of critical mineral resources. In addition, sensitivity analysis is also used to determine future risk sources and propose policy recommendations.

Indicator weighting

This study used the CRITIC weighting method to objectively allocate weights to indicators. Firstly, using the SPSSPRO platform, calculate the target weights for four critical minerals based on indicator data (W_Li, W_Mn, W_Ni, W_Co). Secondly, the fuzzy analytic hierarchy process is used to determine the subjective weights of critical mineral supply risk indicators. Based on the fuzzy complementary judgment matrix of each institution, the results of the three institutions are averaged and synthesized to obtain the final subjective total weight. Finally, according to the principles of game theory, the subjective weights determined by the FAHP method and the objective weights determined by the CRITIC weighting methods (${\text{W}}_{{{\text{Li}}}}^{*}$,${\text{W}}_{{{\text{Mn}}}}^{*}$,${\text{W}}_{{{\text{Ni}}}}^{*}$ and${\text{W}}_{{{\text{Co}}}}^{*}$,) are combined.

The values of β₁ and β₂ were calculated based on game theory methods. The final weight results for each step are shown in Table 2.

Table 2 FAHP-CRITIC weighting calculation results.

In Table 2, the objective weights, subjective weights, and the subjective-objective combined weights based on game theory are presented. Firstly, the indicator data of the four critical minerals differ, resulting in different objective weights. Among the objective weights, lithium resources have the highest weight for C22 (resource substitution rate), which can be attributed to the significant fluctuations in the value of indicator C22 from 2006 to 2022, as well as its low correlation with other indicators. This makes C22 the most influential indicator in the decision-making process, thus receiving the highest weight. Similarly, the different objective weights of manganese, nickel, and cobalt resources are also determined by the degree of fluctuation and inter-correlation of their respective indicators. A higher weight indicates greater volatility in the indicator values, although it is possible that the weight may not fully correspond to the actual importance. Secondly, the subjective weights of critical mineral indicators are derived from the extensive experience of experts and a comprehensive consideration of various external influencing factors, thus carrying a strong degree of subjectivity. Finally, to mitigate the limitations of both subjective and objective weighting methods, a combined weighting approach is used to determine the final indicator weights. For example, in the combined weight of lithium resources, the highest weight shifts from the objective weight of C22 to C21 (resource recovery rate), indicating that the combination of the CRITIC and FAHP methods not only reduces the subjectivity of the FAHP method but also decreases the weight fluctuations caused by changes in indicator data, thereby making the indicator weights more reliable.

Supply risk assessment of critical mineral resources

After that, we apply the SMAA-TRI method to evaluate the supply risk of critical mineral resources for lithium-ion batteries in Chinese electric vehicles. Each mineral is classified into four risk levels: Low (L), Low-Medium (LM), Medium-High (MH), and High (H). The cutting level λ ranges from 0.65 to 0.85, and the values of p and q blur the boundaries of risk levels, with p being twice the value of q⁵³. Considering the data characteristics and relevant literature, this study determines the risk level thresholds for each indicator, detailed in Table 3.

Table 3 Supply risk categories and thresholds for critical mineral resources.

The basis for the delineation of the thresholds for each indicator in Table 3 is as follows:

Firstly, prior scholarly research serves as a crucial basis for classification. The production growth rate (C11) is categorized based on the work of Zhao et al., which employs a 10% increment in historical production growth rates as a level for risk classification¹³. The concentration of production(C12) is classified with a 0.05 increment for determining risk levels and differences, as informed by studies from Yu et al. and Zhao et al.^9,13. The classification of country relations (C31) follows threshold divisions based on research by Yan et al.⁵⁰. Political stability (C33) is classified according to the study of Yu et al. and Dominik et al., using a mean standard error of 0.2^9,53. The classification of the reserve to production ratio (C41) is drawn from Yu et al. with the difference value set to 5, following the work of Jinhua, Cheng et al.^9,54. The risk levels and difference values for the ratio of global reserves (C43) are derived from the studies of Yu et al. and Yang Xin^9,55.

Secondly, the historical data characteristics of the indicators in previous research also provide significant thresholds for classification. The risk levels of the consumption growth rate (C13) are classified based on indicator data, referring to the work of Zhao et al., with a 10% increment and a difference value set at 0.05¹³. Price fluctuation (C14) is classified based on historical price data and the research of Liu et al., with the risk level set at a 5% increment⁵⁶. Resource recovery rate (C21) classification is based on data results and the recycling levels of countries such as the United States and Japan, with a 10% increment per level. Compared to the study by Yu, the average difference is 0.05⁹. Resource substitution rate (C22) is classified by risk level based on the indicator results, with a 25% increment per level. Environmental security (C32) classification is based on studies^13,53, with an average change of 3.43 derived from the 10-year Environmental Performance Index (EPI) published by Yale University across 180 countries. The classification of ratio to global production (C42) risk levels is determined by historical production ratios and major mineral-producing countries, with a 5% increment per level. The external dependence (C44) is classified based on historical data on the external dependency for mineral products, with a 10% increment for the first level. Compared to the study of Yu, the difference is approximately 0.05⁹. Finally, the resource security (C45) classification combines the findings of Lu and Liu et al. with indicator data, with a difference value of 1^56,57.

In evaluating supply risks for four critical mineral resources used in lithium-ion batteries for renewable energy storage batteries in China using the SMAA-TRI method, the JSMAA software is employed to facilitate implementation. After applying the SMAA-TRI method, probability values for the supply risk levels of four critical mineral resources are derived. Figure 4 presents a visual analysis of these probabilities from 2006 to 2022: green denotes low probability (L), yellow denotes medium-low probability (ML), orange denotes medium-high probability (MH), and red denotes high probability(H).

Fig. 4

Supply risk probability values of four critical mineral resources.

According to Fig. 4, among the four crucial minerals for lithium-ion batteries of renewable energy storage batteries in China, the supply risk of lithium resources fluctuated between moderate to high-risk levels from 2006 to 2022, with intervals of moderate to high risk observed in 2006–2010, 2012–2014, and 2017, while the remaining years remained consistently at a high-risk level. Factors such as increasing external dependence, high consumption growth rate, declining resource security, low resource recovery rate, and strained relationships with source countries have continuously elevated the supply risk of lithium to moderate to high-risk levels and beyond. Notably, external dependence is a significant factor influencing the supply risk of lithium resources. In 2011, China’s external dependence on lithium resources increased from 40% to over 70% in 2012 and has since remained above 70% until 2021 and 2022, albeit experiencing slight declines, remaining above 50%. Peak levels of external dependence were reached in 2015 and 2016, exacerbating the supply risk. Furthermore, with the rapid development of renewable energy industry in China, the demand for lithium resources has continuously expanded, posing challenges to the security of lithium resource supply. Between 2006 and 2022, the average annual consumption of lithium resources in China increased by 28%, while the overall resource reserves showed no significant changes, resulting in a continuous decrease in resource security from 142 in 2006 to 21 in 2022. Additionally, although the recycling rate of lithium-ion batteries in China has increased, the current rate is less than 20%, far below that of developed countries such as the United States and Japan. This low recycling rate is detrimental to resource supply security. Finally, the tense diplomatic relations between China and major mineral source countries have intensified the supply risk of lithium resources. In 2022, China’s diplomatic relations with Australia, the largest source of minerals, deteriorated, potentially resulting in a reduction or even a ban on lithium ore exports to China, severely impacting the security of the lithium resource supply.

From 2006 to 2022, manganese metal remained consistently within the high-risk zone. The high external dependence of China’s manganese resources, coupled with low resource substitution rates and inadequate resource security, have perpetuated the supply risk of manganese resources at elevated levels. Firstly, overreliance on external sources of manganese is detrimental to supply security, with China’s average external dependence reaching as high as 86% over 17 years. Security of supply is heavily influenced by the countries of origin, as indicated by the environmental security indices of the source countries from 2006 to 2022, all of which were below 57.5, exacerbating the supply risk. Secondly, despite being the world’s largest consumer of manganese resources, China’s reserves only account for approximately 4% of the global total, resulting in a manganese resource security rating of below 8, which provides inadequate security in the event of supply interruptions. Furthermore, according to the Mineral Commodity Summaries reports by the United States Geological Survey between 2006 and 2022, satisfactory substitutes for manganese metal are nonexistent. The indispensable role of manganese metal and its near-zero substitution rate exacerbate the high-risk nature of manganese resource supply.

During the period spanning from 2006 to 2022, nickel metal consistently remained within the high-risk zone. The primary factors contributing to this status include China’s high external dependence on nickel resources, low resource security, and significant price volatility. Firstly, from 2006 to 2022, China’s annual average external dependence on nickel resources reached a staggering 91%, indicating an excessive reliance on external sources. This heightened dependency renders import stability susceptible to the domestic political stability, environmental security, and diplomatic relations of major supplier nations. Secondly, environmental security poses a high-risk factor, as source countries may implement policies restricting or even banning nickel ore exports, impacting China’s import supply of nickel ores. Additionally, over 17 years, China’s consumption of nickel resources has steadily increased, positioning the country as the world’s largest consumer of nickel resources. However, despite this demand, nickel reserves in China represent only 3% of the global total, resulting in extremely low resource security. By 2022, the resource security rating for nickel stood at a mere 1.7, indicating that domestic nickel resources would be insufficient to support domestic consumption without access to external sources.

The overall supply risk of cobalt metal remains at a high-risk level. The supply risk of cobalt resources is influenced by external dependence, resource security, and concentration of production, resulting in cobalt resources consistently facing high risk. Primarily, China’s demand for cobalt resources relies heavily on imports, with China’s annual average external dependence on cobalt resources reaching 90% from 2006 to 2022. Additionally, the high concentration of cobalt production is detrimental to the stability of the cobalt resource supply, as over half of the world’s cobalt mines are in the Democratic Republic of Congo (DRC), holding a monopolistic position and serving as a primary source of cobalt ore imports for China. Furthermore, the DRC’s poor domestic political stability, internal unrest, and environmental conditions result in the susceptibility of cobalt mining production and export trade to disruptions, thereby rendering China’s cobalt resource supply stability fragile. Lastly, China’s cobalt resources are scarce, accounting for only 1% of the total global consumption, while China ranks as the top consumer of cobalt worldwide. Consequently, over the 17 years, China’s cobalt resource security has continuously declined from 4.7 in 2006 to 1.6 in 2022. In supply interruptions, domestic cobalt resources would be inadequate to support domestic consumption, perpetuating the high level of cobalt resource supply risk.

Prediction of critical mineral resource supply risks

The BP neural network model, also known as the Backpropagation Neural Network, was utilized in this study and implemented using Matlab software. The BP neural network comprises four input layer nodes, one output layer node, and ten hidden layer nodes. The transfer function for hidden layer neurons is “tansig” and for output layer neurons, it is “logsig”. The training utilizes the “trainlm” function with a minimum error threshold of 1 × 10^− 7, 1000 epochs, and a learning rate step size of 0.001.

In this study, mean absolute percentage error (MAPE) was employed to evaluate the predictive performance of the neural network model, the expression is $MAPE=\sum\limits_{{i - 1}}^{n} {\frac{{\left| {{y_i} - {x_i}} \right|}}{{{x_i}}}} \times \frac{{100}}{n}$ where y_i is the fitted value, and x_i is the actual value. Due to the BP neural network’s inherent volatility, repeated training is necessary to obtain ideal results. The MAPE of the predicted values of each indicator data after training is shown in Fig. 5. The MAPE of the 14 indicators data for lithium, manganese, nickel, and cobalt are all less than 5%. This indicates that after undergoing network training, the BP neural network model exhibits a high level of predictive accuracy, enabling the prediction of indicator data.

Fig. 5

Mean absolute percentage error for four critical mineral indicator data.

With the help of the BP neural network model, we predicted the trajectory of risk supply indicators from 2025 to 2027. After evaluating these indicators using the SMAA-TRI method, the supply risks of the four critical minerals in the future are shown in Fig. 6. As shown in Fig. 6, the results of applying the SMAA-TRI method to evaluate the supply risk of four crucial mineral resources. The X-axis represents the timeline spanning from 2025 to 2027. The Y-axis denotes four supply risk levels: Low (L), Medium-Low (ML), Medium-High (MH), and High (H). The Z-axis indicates the probability values of critical mineral resources falling into the four risk level categories, with the highest column representing the maximum probability and corresponding risk level for each mineral’s supply risk each year.

Fig. 6

Acceptability of supply risk for four critical mineral resources.

Firstly, in the lithium resource risk probability chart in Fig. 6, it is observed that from 2025 to 2027, the probability of lithium resources being categorized as high risk (H) in 2025 slightly exceeds that of being medium-high risk (MH), with the probability of high risk for the remaining years being one and other levels being 0. Therefore, in the coming three years, the supply risk of lithium resources in Chinese renewable energy storage batteries is classified as high risk. The prediction results of risk indicators indicate that the decline in resource security, tense relationships with source countries, and low resource recovery rates contribute to the high supply risk of lithium resources.

Secondly, as indicated in the manganese resource risk probability chart in Fig. 6, the probability values of manganese resources being categorized as high risk (H) are mostly 1, with other levels being 0. Thus, manganese metal remains in the high-risk zone from 2025 to 2027. This is primarily due to the anticipated expansion of manganese consumption driven by the rapid growth of the renewable energy market in the future. Additionally, high external dependency and low resource security, attributed to insufficient manganese reserves and low-grade ores in China, contribute to the high supply risk of Chinese manganese resources.

Thirdly, according to the nickel resource risk probability chart in Fig. 6, the probability values of nickel resources being categorized as high risk (H) are all 1, with other levels being 0, indicating that the supply risk of nickel metal remains consistently high from 2025 to 2027. Predictive indicators suggest that China’s consumption of stainless steel and renewable energy storage batteries will continue to rise, supporting increased nickel metal prices. Moreover, due to domestic resource scarcity, declining resource security, and persistently high external dependency, the short-term supply risk of nickel remains high.

Lastly, the cobalt resource risk probability chart in Fig. 6 reveals that the probability values of cobalt resources being categorized as high risk (H) are the highest, indicating that cobalt metal remains in the high-risk zone from 2025 to 2027. The main reason for the high supply risk lies in China’s scarcity of cobalt resources. Predictive results of indicators indicate that as a critical material for storage batteries, the consumption of cobalt resources will continue to grow in the future, leading to high external dependency. Additionally, as the primary import source, the Democratic Republic of Congo is characterized by poor political stability and environmental safety, which may lead to interruptions in cobalt resource supply. Moreover, with low nickel recovery rates in China, these factors collectively contribute to the high supply risk level of Chinese cobalt resources. Overall, the supply risk of the four crucial mineral resources during 2025–2027 is consistently high, indicating a potential critical mineral resource supply shortage.

Sensitivity analysis

Based on the predictive results, from 2024 to 2027, the supply risk of four essential mineral resources, including lithium, manganese, nickel, and cobalt, remains within the moderate to high-risk and high-risk zones, indicating a high likelihood of supply shortages. Therefore, it is imperative to conduct a single-factor sensitivity analysis of supply risk indicators to identify the key factors influencing the supply risk of critical mineral resources and propose targeted supply risk management strategies to effectively reduce the supply risk of these four essential mineral resources in the short term.

The single-factor sensitivity analysis of supply risk indicators primarily involves evaluating the impact of variations in different risk indicators on the supply risk of critical minerals. Firstly, by iteratively changing the data of each risk indicator, it is assumed that the Chinese government, in accordance with national conditions, implements relevant measures to uniformly reduce the indicator values to a low-risk level while keeping the remaining indicator data unchanged.

The indicator data for the four critical mineral resources from 2006 to 2022 under the scenario assumptions remain consistent over the seventeen years. For the sake of brevity, “-” is used to indicate that the indicator values remain unchanged. The details are presented in Table 4.

Table 4 Assumed values for four critical mineral resources period 2006–2022.

In Table 4, under the hypothetical scenario, taking C11 (Production Growth Rate) in Lithium as an example, the value of C11 is set to 0.2 for the period from 2006 to 2022 (according to Table 3, 0.2 is the threshold for C11 to be considered low risk), while the values of other indicators remain unchanged. The threshold values of all indicators are kept constant. The SMAA-TRI method is then employed to calculate the supply risk for Lithium from 2006 to 2022. The results are compared with the supply risk scenario where C11 is left unmodified, allowing for an assessment of the impact of the C11 indicator on the overall supply risk of Lithium. Similarly, the same procedure is applied to other indicators.

Subsequently, the modified indicator data are subjected to calculations using the SMAA-TRI method, resulting in the probability values of critical minerals being classified into different risk levels after the modification of indicator data. Finally, the difference between the original probability values of risk levels for critical minerals and the modified probability values is computed, and the differences across the 14 indicators are compared. Ultimately, the sensitivity results of single indicators for supply risk are derived. The single-factor sensitivity results for critical minerals are summarized in Fig. 7 after organizing the data.

Fig. 7

Single-factor sensitivity of supply risk for four critical mineral resources.

From the findings presented in Fig. 7, it is evident that among the 14 risk indicators, changes in environmental security, resource recovery rate, resource substitution rate, external dependence, and concentration of production significantly impact the supply risk of critical mineral resources. Therefore, these key influencing indicators for the supply risk of critical mineral resources in line with China’s national conditions have been identified. To reduce the supply risk of China’s crucial mineral resources in the short term, the Chinese government could focus on these key risk assessment indicators and mitigate supply risk. To mitigate supply risks in the short term, the following policy recommendations are proposed:

Firstly, the following policy recommendations are proposed at the governmental level:

• (1) Enhance the Security of Critical Mineral Resources: Strengthen the security of critical mineral resources and reduce external dependency by increasing exploration investments, accelerating the discovery of new mineral deposits, enhancing proven reserves, and advancing lithium extraction technologies from salt lakes. Additionally, prioritize the research of alternative materials for renewable energy storage batteries.

• (2) Establish a Comprehensive and Efficient Lithium-Ion Battery Recycling System: Develop a comprehensive and efficient recycling system for lithium-ion batteries by adopting environmentally friendly recycling technologies. This will improve the recovery rates of critical minerals, thereby enhancing overall resource recycling efficiency.

• (3) Strengthen Resource Diplomacy: Import diversified minerals from countries with high environmental standards to avoid over-reliance on a single source nation. This strategy aims to minimize the impact of supply disruptions, reduce production concentration, and enhance environmental safety.

Secondly, the following specific measures should be adopted by industry stakeholders (e.g., battery manufacturers, mineral extraction companies) to address the supply risks of critical minerals:

• (1) Invest in the Development of Alternative Materials and Novel Battery Technologies: Allocate more resources towards researching alternative materials and innovative battery technologies such as sodium-ion batteries, magnesium-ion batteries, and solid-state batteries to decrease dependence on high-risk minerals like lithium, cobalt, and nickel. For instance, sodium-ion batteries utilize sodium ions instead of lithium as charge carriers. Sodium resources are more abundant, widely distributed, and cost-effective, presenting significant potential for renewable energy integration. Magnesium-ion batteries employ magnesium ions as charge carriers, offering advantages over lithium-based systems, including abundant resource availability, lower costs, and enhanced safety features. These attributes make magnesium-based storage materials promising candidates for next-generation energy storage solutions characterized by environmental sustainability and high energy density. Solid-state batteries replace the conventional liquid electrolytes in lithium batteries with solid electrolytes, providing benefits such as higher energy density, improved safety, and extended cycle life. Furthermore, solid-state batteries can utilize not only lithium but also more affordable and environmentally friendly metals like sodium, magnesium, and aluminum, thereby reducing reliance on high-risk critical minerals such as lithium, cobalt, and nickel. In summary, although these novel battery technologies are still in the research and early commercialization stages, they exhibit substantial potential in enhancing battery performance and decreasing dependence on critical mineral resources.

• (2) Secure Stable Supply of Critical Raw Materials: Enter into long-term contracts with mineral suppliers from various countries and regions globally to ensure a stable supply of critical raw materials. Additionally, establish domestic mineral resource extraction and processing facilities to reduce dependence on imported minerals.

• (3) Promote Recycling and Reuse of Renewable Energy Storage Systems: Actively engage in the recycling and circular utilization of renewable energy storage systems to increase the recovery rates of critical minerals from used batteries, thereby decreasing the demand for newly mined resources.

Additionally, for researchers engaged in the study of mineral resource supply security, the following actions are recommended to mitigate supply risks:

• (1) Actively participate in the formulation and revision of relevant government policies by providing professional industry insights and recommendations. This includes contributing to the development and promotion of industry standards for lithium-ion batteries and their recycling processes.

• (2) Further explore the dynamic changes of various mineral resources within global supply chains and their long-term impacts on battery production, offering more forward-looking risk assessments.

Finally, through these comprehensive measures, the tripartite collaboration among policymakers, industry stakeholders, and researchers will collectively promote the stability and sustainable development of the renewable energy storage battery industry.

Conclusion

This study takes China as an example and uses the SMAA-TRI method to evaluate the supply risks of lithium, manganese, nickel, and cobalt in renewable energy storage batteries. Furthermore, the BP neural network model was used to predict its supply risk from 2025 to 2027. Finally, the key indicators that constrain supply risk were determined using single-factor sensitivity analysis. There are several research conclusions.

• (1) This study utilized the combination weighting method and the SMAA-TRI method to investigate the supply risk of four crucial mineral resources, namely lithium, manganese, nickel, and cobalt, in Chinese renewable energy storage batteries across four dimensions. The results indicate that during the period from 2006 to 2022, the supply risk of lithium resources in China ranged from moderate to high-risk zones, while manganese, nickel, and cobalt resources were consistently within the high-risk zone, indicating a significant likelihood of supply shortages, primarily due to the risky nature of the indicator values.

• (2) The risk prediction results indicate that from 2025 to 2027, the supply risk of lithium, manganese, nickel, and cobalt will all remain within the high-risk zone, primarily due to unfavorable values of risk indicators. Therefore, in the foreseeable future, an insufficient supply of critical mineral resources may occur, potentially constraining the development of renewable energy storage batteries in China.

• (3) The sensitivity analysis revealed that environmental risk, resource recovery rate, substitution rate, external dependence, and market concentration significantly influence supply risk. Proposed policy measures include increased investment in exploration, battery technology, mining, establishing a comprehensive battery recycling system, and strengthening mineral resource diplomacy.

Currently, the SMAA-TRI method can assess the supply risk of crucial mineral resources and predict the future of supply risk using a neural network prediction model. However, there are certain shortcomings and opportunities for development with this study’s proposed technique. The results of the SMAA-TRI method are contingent upon the quality and comprehensiveness of the input data. Additionally, the complexity of defining the parameter space restricts the method’s adaptability to changing environments. Due to the limitations of the indicator data, the indicators utilized in this study are based on the national level, examining the supply risks of critical minerals across the entire country. Consequently, there may be biases in the supply risk assessments of important domestic minerals. Furthermore, the prediction outcomes might be affected by the small sample size of the indicator data. In future work, our focus will be on incorporating the interrelationships among various domestic industries into the assessment of supply risks for critical minerals. In subsequent research, we aim to develop more precise small-sample forecasting algorithms to enhance the reliability of risk predictions and provide valuable insights.