Method and device for evaluating surplus value of power battery

Abstract

The invention provides a method for evaluating the surplus value of a power battery. The method comprises the following steps: S1, solving the surplus value score of the power battery based on at least one influence index; s2, solving a residual value prediction result of the power battery based on the historical residual value time sequence of the power battery by means of a machine learning model; and S3, based on the at least one surplus value score and the surplus value prediction result, determining the surplus value of the power battery. The invention further provides a device for evaluating the surplus value of the power battery and a computer program product. According to the method, trend prediction based on historical data and immediate influence of static indexes are ingeniously fused, a more comprehensive power battery value evaluation system is constructed, and the economic benefits of battery pack recovery and reutilization are optimized.

Description

Technical Field

[0001] The present application relates to a method for evaluating the residual value of a power battery, and also relates to an apparatus for evaluating the residual value of a power battery and a computer program product. Background Art

[0002] With the rapid development of the electric vehicle industry, the issues of battery retirement and recycling have become particularly urgent. Currently, the accuracy of battery evaluation technology is insufficient and real-time data is not fully utilized, which limits the precise monitoring of battery status and the intelligent prediction of market value. These limitations directly lead to poor economic benefits in the recycling and reuse of out-of-warranty battery packs.

[0003] Although existing technologies have attempted to estimate battery residual value based on historical data through machine learning algorithms, these methods have not fully integrated the actual state of the battery and market change factors, resulting in inaccurate evaluation of the immediate value of the battery. Therefore, there are still deficiencies in the evaluation of the value of power battery packs in the prior art. Summary of the Invention

[0004] The purpose of the present application is to provide a method for evaluating the residual value of a power battery, an apparatus for evaluating the residual value of a power battery, and a computer program product, so as to solve at least some problems in the prior art.

[0005] According to a first aspect of the present application, there is provided a method for evaluating the residual value of a power battery, the method comprising the following steps:

[0006] Step S1, obtaining a residual value score of the power battery based on at least one influencing index;

[0007] Step S2, obtaining a residual value prediction result of the power battery based on the historical residual value time series of the power battery by means of a machine learning model; and

[0008] Step S3, determining the residual value of the power battery based on at least one residual value score and the residual value prediction result.

[0009] The present application particularly includes the following technical concepts: The present application ingeniously combines the trend prediction based on historical data with the immediate impact of static indicators to construct a more comprehensive power battery value evaluation system. By integrating valuation information from multiple data sources, this solution ensures a detailed consideration of various factors affecting battery value. This combination not only greatly improves the accuracy and prediction ability of battery value evaluation, but also enhances the accuracy and interpretability of the valuation results at the current moment. Generally speaking, the present application optimizes the recycling and reuse process of battery packs, bringing higher economic benefits to the battery recycling market.

[0010] In an exemplary embodiment, in step S3, a pre-established valuation model is used to weight the remaining value scores of multiple items and the remaining value prediction results to determine the remaining value of the power battery.

[0011] In an exemplary embodiment, a machine learning algorithm is used to assign weight parameters to the remaining value scores of multiple items and the remaining value prediction results in the following manner: the weight parameters are determined by minimizing the model valuation error using a deep learning regression model; the weight parameters are determined by feature importance analysis using an interpretability tool; and / or, the weight parameters at different times are dynamically determined using an attention mechanism.

[0012] In an exemplary embodiment, in step S2, a deep neural network model is used as the machine learning model to obtain the remaining value prediction result of the power battery. The deep neural network model includes:

[0013] - An input layer for receiving the historical remaining value time series composed of the historical remaining values BVF of the power battery at the past n moments t-1 , BVF t-2 , …, BVF t-n constituting the historical remaining value time series;

[0014] - At least one hidden layer including a long short-term memory layer and a fully connected layer, and used to capture patterns and trends in the historical remaining value time series; and

[0015] - An output layer for outputting the predicted remaining value BVF of the power battery at moment t t , as the remaining value prediction result of the power battery at moment t.

[0016] In an exemplary embodiment, the influencing indicators include battery state indicators. In step S1, real-time monitoring data related to the power battery is collected from in-vehicle sensors, in-vehicle systems, and / or a background server, and the remaining value score of the power battery based on the battery state indicators is obtained according to the real-time monitoring data.

[0017] In an exemplary embodiment, the battery state indicators include battery health indicators, charging behavior indicators, and / or usage characteristic indicators. Among them, step S1 includes: determining the remaining value score based on the battery health indicators according to the voltage, current, temperature, charge-discharge cycle times, DC internal resistance, and / or cell voltage consistency of the power battery; determining the remaining value score based on the charging behavior indicators according to the charging session data, charging power, and / or fast charging ratio of the power battery; and / or, determining the remaining value score based on the usage characteristic indicators according to the usage cycle, usage mode, and / or environmental conditions where the power battery is located.

[0018] In an exemplary embodiment, the impact metrics include economic status metrics. In step S1, open-source battery market data is obtained from the Internet, government open data platforms, industry research reports, and / or enterprise credit platforms, and the residual value score of power batteries based on the economic status metrics is calculated according to the open-source battery market data.

[0019] In an exemplary embodiment, the economic status metrics include battery market price metrics and / or raw material price metrics. Among them, step S1 includes: determining the battery market price score according to the battery price data from at least one market price source and market influencing factors; and / or determining the residual value score based on the raw material price metrics according to the price of at least one raw material required for manufacturing power batteries and market influencing factors.

[0020] According to a second aspect of the present application, there is provided a device for evaluating the residual value of power batteries. The device includes a processor and a memory. The memory stores computer program instructions. When the computer program instructions are executed by the processor, the processor is capable of executing the method according to the first aspect of the present application.

[0021] According to a third aspect of the present application, there is provided a computer program product, which includes computer program instructions. When the computer program instructions are executed by a processor, the processor is capable of executing the method according to the first aspect of the present application. Description of the Drawings

[0022] Hereinafter, the present application will be described in more detail by referring to the drawings, and the principles, features, and advantages of the present application can be better understood. The drawings include:

[0023] Figure 1 A flowchart showing a method for evaluating the residual value of power batteries according to an exemplary embodiment of the present application;

[0024] Figure 2 A schematic diagram showing the principle of determining the residual value of power batteries by means of a pre-established valuation model according to an exemplary embodiment of the present application;

[0025] Figure 3 A schematic diagram showing the structure of a deep neural network model adopted in an exemplary embodiment of the present application; and

[0026] Figure 4 A block diagram showing a device for evaluating the residual value of power batteries according to an exemplary embodiment of the present application. Detailed Embodiments

[0027] In order to make the technical problems, technical solutions and beneficial technical effects to be solved by the present application clearer and more understandable, the present application will be further described in detail below in conjunction with the accompanying drawings and multiple exemplary embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, rather than to limit the protection scope of the present application.

[0028] Figure 1 The flowchart of the method for evaluating the residual value of a power battery according to an exemplary embodiment of the present application is shown. The method includes step S1, step S2 and step S3.

[0029] The present application mainly focuses on the evaluation of the residual value of power batteries, covering various battery types, including lithium-ion batteries, nickel-metal hydride batteries, lithium iron phosphate batteries, ternary lithium batteries, lithium cobalt oxide batteries, and solid-state batteries, etc. These batteries are widely used in various types of electric vehicles, such as battery electric vehicles (BEV), hybrid electric vehicles (HEV), plug-in hybrid electric vehicles (PHEV), extended-range electric vehicles (EREV), and fuel cell electric vehicles (FCEV).

[0030] For example, the method can be automatically triggered when the battery warranty expires to evaluate the residual value of the battery. At the same time, it can also respond to user needs and be manually triggered at any appropriate time to provide users with instant battery value information. This flexibility makes the solution of the present application not only applicable to the automatic evaluation after the battery warranty expires, but also applicable to the battery residual value query initiated by users, so as to provide battery value evaluation services throughout the entire life cycle of the battery.

[0031] In step S1, the residual value score of the power battery is obtained based on at least one influencing index.

[0032] Here, for example, a variety of influencing indexes are predefined to specifically evaluate the instant value of the battery at a specific moment (such as at the current moment t or at the expected future moment t + 1). The data sources of the influencing indexes are diverse, covering multiple key aspects of battery performance and market conditions. For example, these can be mainly divided into battery state indexes and economic state indexes. The battery state indexes are used to measure the current performance of the battery, and can also reflect the trend of battery performance deterioration in a future period of time. The economic state indexes reflect the current market conditions, such as the market transaction prices of batteries or raw materials and macroeconomic factors, etc. These factors jointly determine the economic score of the battery.

[0033] In context, the residual value of a power battery can also be referred to as "salvage value", "recycling value", "reuse value", etc. The manifestation of this value is diverse and is not limited to the resale price in monetary terms, but can also include other forms of benefits. For example, in the presence of a trade-in program, the residual value of the battery can be manifested as a discount, a discount rate, or as a basis for redeeming prizes, maintenance services, or repair services.

[0034] In addition, the residual value of a power battery can also be interpreted from a usage perspective. For example, the assessment of the residual value can be translated into specific usage suggestions, such as "It is recommended to recycle the battery as soon as possible", "It is recommended to recycle the battery within 3 months", or "It is recommended to continue using the battery". These suggestions are based on the current performance status of the battery and the expected service life, providing practical guidance for users on battery usage and recycling.

[0035] In step S2, with the help of a machine learning model, based on the historical residual value time series of the power battery, the predicted result of the residual value of the power battery is obtained.

[0036] Here, "historical residual value" can refer to two situations, namely the historical data of this battery and the historical data of other batteries.

[0037] - Historical data of this battery: This represents the valuation records of the battery to be evaluated at various past time points. To construct this historical data set, electric vehicles can be sent to a certified battery evaluation and recycling agency regularly, and real-time value data of the battery can be obtained through professional evaluations.

[0038] - Historical data of other batteries: In addition to the historical data of individual batteries, the machine learning model can also utilize the historical residual value information of other batteries. This includes the resale price or transaction price of the same type or model of batteries with similar usage conditions in the market.

[0039] In practical applications, using the historical data of the target battery itself may provide more accurate prediction results because these data directly reflect the usage status and aging process of the battery. However, if the historical data of the individual battery is insufficient or unavailable, using the historical data of other batteries is also an effective alternative, especially when these batteries are similar to the target battery in terms of type, usage conditions, environmental factors, and the vehicle models they are assembled in. In addition, the individual battery data and the general historical data of other batteries can be considered in combination when necessary.

[0040] In this step, for example, a deep neural network (DNN) can be used as the machine learning model to predict the residual value of the power battery. This prediction process can be expressed by the following formula:

[0041] DNN(BVF past )=Φ(BVFt-1 , BVF t-2 , …, BVF t-n )

[0042] Among them, DNN(BVF past ) represents the predicted remaining value based on the historical remaining value time series. Φ represents the prediction function of the deep neural network, which receives the historical remaining values BVF t-1 , BVF t-2 , …, BVF t-n at the past n moments as inputs and outputs the prediction of the remaining value BVF t at the future moment t.

[0043] In addition to the deep neural network, other types of machine learning models such as linear regression, decision tree, Bayesian network, support vector machine (SVM), etc. can also be considered to adapt to different data characteristics and prediction requirements. The specific architecture of the deep neural network will be described in detail below in combination with Figure 3 and will not be elaborated here for the sake of brevity.

[0044] In step S3, based on at least one remaining value score and the remaining value prediction result DNN(BVF past ), the remaining value BVF(t) of the power battery is determined.

[0045] In this process, the static remaining value score immediately maps the current state of the battery and the market conditions, while the deep neural network model predicts the future value trend by analyzing historical data, capturing the dynamic changes over time. This method ensures the forward-looking and current accuracy of the valuation by integrating the immediate static score and the dynamic time series prediction, providing a more comprehensive perspective for battery value assessment.

[0046] In one embodiment, a pre-established valuation model can be used to weight multiple remaining value scores and the remaining value prediction result BVF past to determine the remaining value BVF(t) of the power battery. Exemplarily, weighted summation can be implemented through the following formula:

[0047] BVF(t) = αH(t) + βM(t) + γP(t) + δR(t) + ∈D(t) + DNN(BVF past )

[0048] Among them, H(t), M(t), P(t), R(t), D(t) respectively represent the remaining value scores based on different impact indicators, and DNN(BVF past ) represents the remaining value prediction result obtained by using the machine learning model. α, β, γ, δ, ∈ respectively represent weight parameters, and these weight parameters reflect the influence degree of each factor on the battery value.

[0049] In the above formula, t can represent either the current moment or a future moment starting from the current moment. When evaluating the remaining value of the battery at a future moment, the scores based on each influencing indicator should not only reflect the current state of the battery or the market, but also integrate the predictions of future battery and market states as well as technological trends. To this end, for example, machine learning models can be constructed separately for each influencing indicator, and these models will use historical data to predict future score changes, thereby providing a more accurate assessment of the remaining value of the battery.

[0050] It should be noted that although no weight parameter is assigned to the machine learning-based prediction term DNN(BVF past ) in the above formula, it should be understood that this is only exemplary. In fact, DNN(BVF past ) can also participate in the weighting process through appropriate weight parameters. This weight parameter will adjust the proportion of the prediction result in the final calculation of the remaining value, ensuring that, together with the score results based on other influencing indicators, it jointly and reasonably determines the remaining value of the battery.

[0051] In one embodiment, for example, machine learning algorithms can be used to assign weight parameters to multiple remaining value scores and remaining value prediction results. The following are several commonly used methods:

[0052] - Data-driven training: Train a deep learning regression model, such as LSTM (Long Short-Term Memory Network) or Transformer (Transformer Model), by using historical data and relevant indicators, and use optimization algorithms such as gradient descent to adjust the model weights to minimize the valuation error between the model prediction value and the actual value, so that the model can learn the optimal weight parameters.

[0053] - Feature importance analysis: Use interpretive tools, such as SHAP (Shapley Additive Explanations) or LIME (Local Interpretable Model-Agnostic Explanations), to evaluate the contribution of each feature to the model prediction result. These tools help determine the importance of the features by explaining the model's prediction results, thereby indirectly assigning weights to each indicator.

[0054] - Adaptive weight mechanism: Introduce an attention mechanism during the model training process to dynamically calculate the indicator weight parameters at different time points to adapt to the time-varying characteristics of the data. This method allows the model to automatically adjust the weights according to the changes in the data to better capture new data patterns.

[0055] In addition, the determination of the weight parameters may also include user input, allowing the user to adjust the importance of each influencing indicator according to factors such as specific application scenarios, personal preferences, or experience.

[0056] In addition to the weighted summation method, a weighted product or other more complex mathematical expressions can also be used to construct the model to more comprehensively consider the influence of various factors.

[0057] In one embodiment, in order to improve the accuracy of the evaluation, a customized residual value model can be constructed for different vehicle models. Thus, the residual value of the power batteries of various vehicle models can be calculated more accurately.

[0058] In one embodiment, the visual output of the residual value can also be implemented in a step not shown. Such visual output may include: the estimated residual value of the battery at the current moment, the future trend of the estimated value change, a multi-dimensional comparison chart based on different vehicle models and usage conditions, and an interactive analysis (the user can understand how factors such as charging behavior and usage frequency affect the residual value evaluation of the battery through an interactive interface).

[0059] Figure 2 A schematic diagram showing the principle of determining the residual value of a power battery with the aid of a pre-established evaluation model according to an exemplary embodiment of the present application is shown.

[0060] The evaluation model 20 evaluates the residual value of the battery from two key aspects: 1) considering the immediate influence of each key influencing indicator on the battery residual value at the current moment; 2) predicting the future change trend of the battery value through time series analysis.

[0061] As Figure 2 shown, these influencing indicators can be generally classified into two categories:

[0062] - Battery state indicators 21: including battery health indicators, charging behavior indicators, and usage characteristic indicators. These indicators reflect the current performance and usage status of the power battery.

[0063] - Economic state indicators 22: including battery market price indicators and raw material price indicators. These indicators reflect the influence of market factors on the battery value.

[0064] To accurately evaluate the remaining value score of a battery based on battery state indicator 21, it is first necessary to collect real-time monitoring data related to the battery state. This data can be obtained through in-vehicle sensors of vehicles equipped with power batteries, including key performance parameters such as battery voltage, current, and temperature. At the same time, in-vehicle systems such as the battery management system (BMS) and real-time monitoring module (RTM) can also be considered to monitor and record the operating state and health status of the battery. In addition, if the vehicle is a new energy vehicle and uses the RTM mechanism to regularly report data to the automaker's backend or the national regulatory platform, real-time monitoring signals can also be collected from these backend servers (such as cloud servers).

[0065] Exemplarily, the collected real-time monitoring data includes:

[0066] - Battery voltage (V);

[0067] - Battery current (I);

[0068] - Battery temperature (T);

[0069] - Number of charge-discharge cycles (C), which represents the total number of charge-discharge cycles of the battery and is an important indicator for evaluating the degree of battery aging;

[0070] - Calculated value based on the cell spread of individual battery voltages, which helps to understand the voltage consistency of individual batteries inside the battery;

[0071] - Charging session data, which is located based on the start and end times of vehicle charging. For example, it includes charging start and end times, charging power, charging mode, charging efficiency, charging current and voltage, etc.;

[0072] - DC internal resistance (DCR)

[0073] - Normalized charge-discharge rate of individual batteries (Cell Normalized Rate)

[0074] After successfully collecting this real-time monitoring data, this data can be further used to calculate the score of the battery based on battery state indicator 21. The following shows an exemplary calculation method:

[0075] Health status score

[0076] The state of health score H(t) of the power battery (i.e., the remaining value score based on the state of health indicator) reflects the current health and performance status of the battery. This score can be calculated through the following formula:

[0077] H(t) = ω 1 V + ω 2 I + ω 3T + ω 4 C + ω 5 Cel Spread + ω 6 DCR

[0078] + ω 7 Consistency

[0079] Wherein:

[0080] - H(t) represents the battery health state score of the power battery at time t.

[0081] - ω 1 , ω 2 , ω 3 , ω 4 , ω 5 , ω 6 , ω 7 represent weight coefficients, corresponding to the importance of voltage, current, temperature, capacity, cell voltage distribution, DC internal resistance, and cell voltage consistency respectively. These weight coefficients can be adjusted according to the actual situation to reflect the influence degree of different factors on the health state score.

[0082] - V represents the voltage of the battery.

[0083] - I represents the current of the battery.

[0084] - T represents the temperature of the battery.

[0085] - C represents the capacity of the battery.

[0086] - Cell Spread represents the cell voltage distribution of the battery.

[0087] - DCR represents the DC internal resistance.

[0088] - Consistency represents the cell voltage consistency.

[0089] A higher health state score indicates that the battery currently has a good health state, and thus a higher residual value can be obtained.

[0090] Charging behavior score

[0091] The charging behavior score R(t) of the power battery (i.e., the residual value score based on the charging behavior index) can reflect the potential influence of different charging habits on the battery health state and the deterioration speed. The charging behavior score can be calculated by the following formula:

[0092] R(t) = λ 1 × Charging Sessions + λ 2 × Charging Power

[0093] +λ 3 ×Fast Charging Ratio

[0094] Where:

[0095] -R(t) represents the charging behavior score of the power battery at time t.

[0096] -λ 1 ,λ 2 ,λ 3 are weighting coefficients corresponding to the importance of the number of charging sessions, charging power, and fast charging ratio respectively. These weighting coefficients can be adjusted according to the actual situation to accurately reflect the impact of different charging behaviors on the charging behavior score.

[0097] -Charging Sessions represents the charging session data.

[0098] -Charging Power represents the charging power of the battery. High-power charging may increase the heat inside the battery, thus affecting its health status.

[0099] -Fast Charging Ratio represents the proportion of the battery charged in the fast charging mode. Frequent fast charging may have a negative impact on the battery life.

[0100] A higher charging behavior score means good charging habits (such as avoiding overcharging, setting a maximum charging power limit, reducing the fast charging ratio, etc.), which can slow down battery aging, extend its service life, and thus help maintain a higher residual value of the battery.

[0101] Usage characteristic score

[0102] The usage characteristic score D(t) of the power battery (i.e., the residual value score based on usage characteristic indicators) can reflect the potential impact of different usage behaviors on the battery health status and degradation rate. The usage characteristic score can be calculated by the following formula:

[0103] D(t) = μ 1 ×Usage Cycles + μ 2 ×Usage Patterns

[0104] + μ 3 ×Environmental Conditions

[0105] Where:

[0106] -D(t) represents the usage characteristic score of the power battery at time t.

[0107] -μ 1 , μ 2 , μ 3 are weight coefficients, corresponding to the importance of usage cycle, usage pattern, and environmental conditions respectively. These weight coefficients can be adjusted according to the actual situation to reflect the influence degree of different factors on the battery usage characteristic score.

[0108] -Usage Cycles represents the usage cycle of the battery. This indicator is directly related to the cycle life of the battery. The increase in the number of cycles usually leads to a gradual decrease in the battery capacity.

[0109] -Usage Patterns represents the usage patterns of the battery, such as depth of charge and discharge, usage time distribution, and driving habits, etc. These patterns reflect the usage intensity and frequency of the battery. For example, frequent rapid acceleration and rapid braking will cause the battery to withstand a large current impact, putting the battery in an extreme usage state, thus accelerating the aging process.

[0110] -Environmental Conditions represents the environmental conditions where the battery is located, such as temperature, humidity, and vibration, etc. These conditions have a significant impact on the chemical stability and physical integrity of the battery.

[0111] Extreme temperature conditions will accelerate the attenuation of the battery capacity.

[0112] A higher usage characteristic score indicates that good usage characteristics (such as avoiding using under extreme temperatures, reducing rapid acceleration and rapid braking) can reduce the damage to the battery and extend its service life, which helps to maintain a relatively high residual value of the battery.

[0113] Generally speaking, the health status score directly reflects the current health condition of the battery, while the charging behavior score and the usage characteristic score reveal the impact of user behavior on the future aging and value decay of the battery. Even if the health status scores of different batteries are the same, different charging habits and usage behaviors will also lead to obvious differences in the battery aging speed and value decline trend.

[0114] To accurately evaluate the residual value score of power batteries based on economic state indicators, first, open-source battery market data can be collected from multiple channels. These channels include but are not limited to the Internet, government open data platforms, industry research reports, enterprise credit investigation platforms, etc. The open-source battery market data collected can include key price information such as the transaction price, resale price / recycling price of the battery, the selling price of new batteries, and raw material prices. In addition, economic factors such as supply and demand relationship, supply chain information, economic policies, and industry development situation can also be considered, which can provide more comprehensive background information for the market value of the battery.

[0115] Next, the collected open-source battery market data can be used to calculate the remaining value score of power batteries based on economic state indicators. The following shows an exemplary calculation method:

[0116] Battery market price score

[0117] The battery market price score M(t) of power batteries (i.e., the remaining value score based on the battery market price indicator) can reflect the impact of battery price fluctuations in the market on the evaluation of battery remaining value. The battery market price score can be calculated by the following formula:

[0118]

[0119] Where:

[0120] - M(t) represents the battery market price score of power batteries at time t.

[0121] - w i represents the weight coefficient assigned to the i-th market price source. These weight coefficients can be set according to factors such as the quality, timeliness, and stability of data sources, or can also be set according to the scale, level, and activity of the market. Different market price sources may include:

[0122] Battery manufacturers, used car markets, professional battery recycling institutions, online trading platforms, and industry analysis reports, etc. These sources provide information reflecting the battery market value from different perspectives.

[0123] - Market Price i (t) represents the battery price data of the same model from the i-th market price source at time t. According to actual needs, this can either refer to the selling price of brand-new batteries or the resale price of second-hand recycled battery packs, reflecting the real-time dynamic changes in the battery market price.

[0124] - n represents the total number of market price sources.

[0125] A higher battery market price score directly reflects the high-value positioning of power batteries in the current market environment, which is usually positively correlated with the higher remaining value of the batteries.

[0126] Optionally, in order to more comprehensively evaluate the remaining value of the battery, economic factors can also be taken into account. This can be achieved by introducing an economic adjustment factor as a market influencing factor into the existing model. This factor can, for example, reflect the impact of the economic environment and market trends (such as changes in market demand, supply chain stability, government incentive policies, and development trends of the new energy industry) on the battery market price.

[0127] Raw material price score

[0128] The raw material price score M(t) of the power battery (i.e., the residual value score based on the raw material price index) can reflect the impact of changes in battery manufacturing costs on the assessment of the battery's residual value. The raw material price score can be calculated by the following formula:

[0129]

[0130] Where:

[0131] - P(t) represents the raw material price score of the power battery at time t.

[0132] - Raw Material Price j (t) represents the price of the jth raw material required for manufacturing the power battery at time t.

[0133] - Market Indicator k (t) represents the value of the kth market influencing factor at time t, which includes, for example, economic indicators (such as GDP growth rate, inflation rate, industrial production index), supply chain information (supply chain stability, supply-demand relationship), availability of alternative materials, policy changes (tax incentive policies, environmental protection regulations), etc.

[0134] - v j represents the weight of the jth raw material, reflecting the proportion of this raw material in the total battery cost.

[0135] - u k represents the weight of the kth market influencing factor, reflecting the degree of influence of this market influencing factor on the raw material price fluctuation.

[0136] - m represents the total number of raw materials considered.

[0137] - p represents the total number of market influencing factors.

[0138] A higher raw material price score reflects higher manufacturing costs of the battery. Therefore, the raw materials in used batteries may increase the overall residual value of the battery due to their higher recycling value.

[0139] By integrating these factors, the valuation model 20 can provide an accurate and immediate prediction of the residual value of the power battery. In addition, the valuation model 20 not only integrates these static indicators 21, 22, but also incorporates the prediction results 23 of the machine learning model to enhance the accuracy and forward-looking nature of the prediction. The final battery valuation formula is as follows:

[0140] BVF(t) = αH(t) + βM(t) + γP(t) + δR(t) + ∈D(t) + DNN(BVF past )

[0141] It should be noted that the above-listed residual value scoring formulas are only exemplary and can have various different expressions. In the valuation model 20, not only can the weight parameters of each influencing index be determined through machine learning, but also the weight coefficients of the constituent elements within each scoring model can be optimized using machine learning algorithms.

[0142] Figure 3 The structural schematic diagram of the deep neural network model adopted in the exemplary embodiments of the present application is shown.

[0143] In this embodiment, a deep neural network model (hereinafter simply referred to as the DNN model) is used as the machine learning model to obtain the predicted result of the residual value of the power battery. The type of the DNN model can be differently selected according to its structure and application scenarios. Common DNN models include: Feedforward Neural Network, Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), Long Short-Term Memory Network (LSTM), Generative Adversarial Network (GAN), Deep Belief Network (DBN), Residual Network (ResNet), and Transformer model.

[0144] As Figure 3 exemplarily shown, the DNN model is composed of three types of key layers: the input layer 301, the hidden layers 302, 303, and the output layer 304.

[0145] The input layer 301 is responsible for receiving the historical residual value sequence of the power battery at the past n moments, that is, BVF t-1 , BVF t-2 , …, BVF t-n . Each node of the input layer 301 corresponds to a specific historical data point.

[0146] The hidden layers 302, 303 include multiple Long Short-Term Memory layers (hereinafter simply referred to as LSTM layers) and fully connected layers. The LSTM layer 302 is particularly suitable for processing and predicting time series data because it can capture long-term dependencies in historical data. The fully connected layer 303 is used to further process features and learn complex patterns in the data. The states of the hidden layers 302, 303 not only depend on the data of the current input layer 301 but also are affected by the state of the hidden layer 302 at the previous moment, which enables the network to remember and utilize previous information. The weights and biases in the network are trainable parameters, and they are adjusted in the deep learning framework to optimize the model performance.

[0147] The function of the output layer 304 is to predict and output the residual value BVF of the power battery at time t t, that is, the predicted result DNN(BVF of the remaining value of the power battery at time t past ).

[0148] This DNN model has been trained using a large amount of historical remaining value data, for example. This training process mainly involves adjusting the internal parameters of the network. Specifically, the network weights and biases can be randomly initialized first to set the starting point for the training process. Then, in the forward propagation process, the input data passes through each layer of the DNN model, and the output of each layer is used as the input of the next layer until the predicted remaining value is generated. Next, the loss function is calculated by comparing the model prediction result with the actual remaining value, and the network parameters are adjusted according to the loss function through the backpropagation algorithm to minimize the prediction error. By continuously repeating the forward propagation and backpropagation processes, the model gradually optimizes its parameters in multiple iterations until the prediction result reaches the expected accuracy.

[0149] In addition, it can also be tested under different environmental conditions and battery states to ensure the universality and robustness of the model in various battery types and usage scenarios.

[0150] Figure 4 Fig. shows a block diagram of an apparatus 10 for evaluating the remaining value of a power battery according to an exemplary embodiment of the present application.

[0151] As Figure 4 shown, the apparatus 10 includes a processor 11 and a memory 12. The memory 12 stores computer program instructions. When the computer program instructions are executed by the processor, the processor 11 can, for example, execute a method for evaluating the remaining value of a power battery (which has been described in detail above with the help of Figure 1 , and will not be elaborated here). The computer program instructions can be stored in a computer-readable storage medium. The computer-readable storage medium can include, for example, a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card, a secure digital card, a flash memory card, at least one magnetic disk storage device, a flash memory device, or other volatile solid-state storage devices. The processor 11 can be a central processing unit, or can also be other general-purpose processors, digital signal processors, application-specific integrated circuits, off-the-shelf programmable gate arrays, or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0152] In an optional embodiment, the device 10 may further include or be connected to a human-machine interface, such as a display. The human-machine interface is connected to the processor 11 and the memory 12 through a bus to visualize the estimated remaining value of the battery. For example, with the help of the human-machine interface, the immediate remaining value of the power battery can be displayed in the form of a recycling price, or the changing trend of the remaining price of the battery over a period of time can be displayed, so as to facilitate vehicle users and recycling operators to quickly understand the value of the battery pack.

[0153] The device 10 has multiple deployment possibilities. It can be embedded inside the vehicle and directly obtain real-time monitoring data from in-vehicle sensors or systems to evaluate the remaining value of the battery; it can also be deployed on a server and transmit the evaluation results to the user's mobile device, such as a smart phone, through a communication interface, enabling the user to keep track of the remaining value of the battery at any time. This diverse deployment strategy enables the device 10 to meet different application scenarios and user requirements.

[0154] Although specific embodiments of the present application are described in detail herein, they are given for purposes of explanation only and should not be considered as limiting the scope of the present application. Various substitutions, changes, and modifications can be conceived without departing from the spirit and scope of the present application.

Claims

1. A method for evaluating the residual value of a power battery, the method comprising the following steps: Step S1, obtaining a residual value score of a power battery based on at least one influencing indicator; Step S2, using a machine learning model to obtain a residual value prediction result of the power battery based on the historical residual value time series of the power battery; as well as Step S3, determining the residual value of the power battery based on at least one residual value score and a residual value prediction result.

2. According to the method of claim 1, in step S3, a plurality of residual value scores and residual value prediction results are weighted by means of a pre-established valuation model to determine the residual value of the power battery.

3. The method according to claim 2, wherein: With the help of machine learning algorithms, weight parameters (α, β, γ, δ, ∈) are assigned to multiple residual value scores and residual value prediction results in the following way: Determine the weight parameters (α, β, γ, δ,∈) by minimizing the model valuation error with the help of a deep learning regression model; Determine the weight parameters (α, β, γ, δ, ∈) through feature importance analysis with the help of interpretive tools; and / or The weight parameters (α, β, γ, δ,∈) at different times are dynamically determined with the help of the attention mechanism.

4. The method according to any one of claims 1 to 3, wherein: In step S2, a deep neural network model is used as a machine learning model to obtain a residual value prediction result of a power battery, and the deep neural network model includes: - Input layer (301), which is used to receive the historical residual value (BVF t-1 ,BVF t-2 ,…,BVF t-n ) constitutes the historical surplus value time series; - at least one hidden layer (302, 303) comprising a long short-term memory layer and a fully connected layer and configured to capture patterns and trends in the historical surplus value time series; and - Output layer (304), which is used to output the predicted residual value (BVF t ), as the prediction result of the residual value of the power battery at time t.

5. The method according to any one of claims 1 to 4, wherein: The influencing index includes a battery status index. In step S1, real-time monitoring data related to the power battery is collected from the vehicle-mounted sensor, the vehicle-mounted system and / or the background server, and the residual value score of the power battery based on the battery status index is obtained according to the real-time monitoring data.

6. The method according to claim 5, wherein: The battery status indicator includes a battery health indicator, a charging behavior indicator and / or a usage characteristic indicator, wherein step S1 includes: Determine the residual value score based on the battery health index according to the voltage, current, temperature, number of charge and discharge cycles, DC internal resistance and / or single cell voltage consistency of the power battery; Determine a residual value score based on a charging behavior indicator according to the charging session data, charging power and / or fast charging ratio of the power battery; and / or The residual value score based on the usage characteristic index is determined according to the usage cycle, usage mode and / or environmental conditions of the power battery.

7. The method according to any one of claims 1 to 6, wherein: The influencing indicators include economic status indicators. In step S1, open source battery market data is obtained from the Internet, government open data platforms, industry research reports and / or corporate credit reporting platforms, and the residual value score of the power battery based on the economic status indicator is obtained according to the open source battery market data.

8. The method according to claim 7, wherein: The economic status indicator includes a battery market price indicator and / or a raw material price indicator, wherein step S1 includes: Determine a battery market price score based on battery price data from at least one market price source and market influencing factors; and / or The residual value score based on the raw material price indicator is determined based on the price of at least one raw material required to manufacture power batteries and market influencing factors.

9. A device for evaluating the residual value of a power battery, wherein: The device (10) comprises a processor (11) and a memory (12), wherein the memory (12) stores computer program instructions, and when the computer program instructions are executed by the processor (11), the processor (11) is capable of performing the method according to any one of claims 1 to 8.

10. A computer program product comprising computer program instructions, wherein: The computer program instructions, when executed by a processor, enable the processor to perform a method according to any one of claims 1 to 8.

Images