Battery life prediction method and device in battery replacement cabinet

Abstract

The application is suitable for the technical field of data recognition, and provides a battery life prediction method and device in a battery swap cabinet. The battery life prediction method in the battery swap cabinet comprises: collecting multi-dimensional battery characteristic data when a to-be-recognized battery completes a charging and discharging cycle and is connected to the battery swap cabinet; based on the multi-dimensional battery characteristic data, a dynamic stress spectrum corresponding to the to-be-recognized battery is counted; wherein the dynamic stress spectrum is a statistical characteristic reflecting the use intensity in different time windows, which is calculated based on historical cycle data of the to-be-recognized battery; and based on the dynamic stress spectrum, the remaining life corresponding to the to-be-recognized battery is calculated. The dynamic stress spectrum calculated based on the historical cycle data can effectively capture the use intensity of the battery in different time windows. The introduction of this statistical characteristic makes the prediction process not only consider static parameters, but also dynamically reflect the stress change of the battery in the use process, thereby improving the adaptability and accuracy of the model.

Description

Technical Field

[0001] This invention belongs to the technical field of data recognition, and in particular relates to a method and device for predicting battery life in a battery swapping cabinet. Background Technology

[0002] With the rapid development of electric vehicles and their supporting infrastructure, battery swapping technology has gradually become an important means of battery management. As the core equipment for battery exchange, battery swapping cabinets not only provide electric vehicles with a convenient way to replenish energy, but also play a crucial role in battery management and maintenance. Battery performance and lifespan directly affect the efficiency of electric vehicle use and user experience; therefore, accurately predicting the remaining battery lifespan is particularly important.

[0003] Traditional battery life prediction methods rely heavily on static indicators such as battery capacity and internal resistance, which typically fail to fully reflect performance changes during actual use. These methods neglect the dynamic characteristics of batteries under different usage conditions, thus limiting the accuracy and reliability of the prediction results. Summary of the Invention

[0004] In view of this, embodiments of the present invention provide a method and apparatus for predicting battery life in a battery swapping cabinet, in order to solve the technical problem that traditional methods ignore the dynamic characteristics of batteries under different usage conditions, which limits the accuracy and reliability of the prediction results.

[0005] A first aspect of this invention provides a method for predicting battery life in a battery swapping cabinet, the method comprising: When the battery to be identified completes one charge-discharge cycle and is connected to the battery swapping cabinet, multi-dimensional battery characteristic data is collected; wherein, the multi-dimensional battery characteristic data includes at least the depth of discharge, the average discharge rate, and the average ambient temperature during the cycle; Based on the multidimensional battery feature data, the dynamic stress spectrum corresponding to the battery to be identified is statistically analyzed; wherein, the dynamic stress spectrum is a statistical feature reflecting the usage intensity within different time windows, calculated based on the historical cycle data of the battery to be identified. Based on the dynamic stress spectrum, the remaining lifespan of the battery to be identified is calculated.

[0006] Furthermore, the step of statistically analyzing the dynamic stress spectrum corresponding to the battery to be identified based on the multidimensional battery feature data includes: Based on the multidimensional battery characteristic data, an instantaneous stress coefficient characterizing the degree of accelerated aging in this cycle is calculated; wherein, the instantaneous stress coefficient is obtained by combining the depth of discharge stress coefficient, rate stress coefficient and temperature stress coefficient. The instantaneous stress coefficient, the depth of discharge stress coefficient, the rate stress coefficient, and the temperature stress coefficient are added as new records in chronological order to the historical stress sequence for the maintenance of the battery to be identified. In the historical stress sequence, based on a preset first rolling window, the first mean and standard deviation of the instantaneous stress coefficient are statistically analyzed, and the proportion of the discharge depth stress coefficient greater than the first depth threshold and the proportion of the rate stress coefficient greater than the first rate threshold are statistically analyzed. In the historical stress sequence, based on a preset second rolling window, the second mean value of the instantaneous stress coefficient and the cumulative duration for which the temperature stress coefficient is greater than the temperature threshold are statistically analyzed; wherein, the window size of the second rolling window is larger than the window size of the first rolling window; The first mean, the second mean, the standard deviation, the proportion of values ​​greater than the first depth threshold, the proportion of values ​​greater than the first magnification threshold, and the cumulative duration of values ​​greater than the temperature threshold are spliced ​​together in a predetermined order to generate a one-dimensional dynamic stress spectrum. For each element in the dynamic stress spectrum, normalization is performed based on a preset nominal range to obtain the normalized dynamic stress spectrum.

[0007] Furthermore, the step of calculating the instantaneous stress coefficient characterizing the degree of accelerated aging in this cycle based on the multidimensional battery feature data includes: Divide the discharge depth by the preset rated discharge depth reference value to obtain k. d The power is used as the discharge depth stress coefficient; where k d The first empirical index is greater than 1; Divide the average discharge rate by the preset rated discharge rate reference value to obtain k. c The power is used as the magnification stress coefficient; where k c It is the second empirical index, which is greater than 1; Substituting the average ambient temperature during the cycle into the Arrhenius model, we obtain the temperature stress coefficient. The instantaneous stress coefficient is obtained by weighted summation of the discharge depth stress coefficient, the rate stress coefficient, and the temperature stress coefficient.

[0008] Furthermore, the step of calculating the remaining life of the battery based on the dynamic stress spectrum includes: Obtain the current battery health state feature vector of the battery to be identified at the current detection time; wherein, the current battery health state feature vector includes at least electrochemical features and thermodynamic features; The current battery health status feature vector is input into the pre-trained first regression model, and the first regression model outputs a value that represents the current actual capacity retention rate of the battery, which is used as the current health status score. Based on the dynamic stress spectrum, the future aging acceleration factor is calculated. Calculate the current health stress index based on the current health status score and future aging acceleration factors; From the historical database, retrieve the historical health stress index values ​​of the battery to be identified at the most recent M detection times to form a time series; Based on the time series, linear regression is used to calculate the decay rate of the health stress index relative to the number of detections; wherein, the number of detections is related to the number of charge-discharge cycles of the battery; Multiply the decay rate by the future aging acceleration factor to obtain the corrected decay rate; If the current health stress index is not greater than the health stress index threshold, it is determined that the battery to be identified has reached the end of its lifespan, and the remaining lifespan cycle count is set to 0. If the decay rate exceeds the limit, the decay rate is determined to be abnormal, and the remaining lifetime cycle count is set to a preset maximum value or an invalid flag. Subtract the current health stress index from the health stress index threshold, and divide by the decay rate to obtain the number of cycles of remaining lifespan.

[0009] Furthermore, the step of obtaining the battery health state feature vector corresponding to the battery to be identified at the current detection time includes: Calculate the slope of the linear fit of the voltage-time curve during the constant current charging stage, as a characteristic of the voltage rise rate. During the constant voltage charging stage, the current decay curve is subjected to exponential fitting, and its time constant is extracted as a polarization dissipation characteristic. After the battery to be identified has been left to stand for a preset time, a short-time pulse discharge is applied, the voltage transient response is recorded, and the pulse internal resistance and the relaxation time for the voltage to recover to a stable value are calculated. After entering a low-current constant-voltage float charging state at the end of the charging process, the voltage drop during the first preset time period is recorded, and the float charging self-discharge rate is calculated. The voltage rise rate characteristic, the polarization dissipation characteristic, the relaxation time, and the float charge self-discharge rate are used as electrochemical characteristics. During the constant current charging phase, the temperature rise rate at the highest temperature point on the surface of the battery to be identified is calculated. At the same time point, the variance of all measured surface temperatures of the batteries to be identified is calculated as a characteristic of temperature distribution uniformity. After charging is completed and the battery has been left to stand for a second preset period of time, the equilibrium temperature difference between the average surface temperature of the battery to be identified and the ambient temperature is calculated. The temperature rise rate, the temperature distribution uniformity, and the equilibrium temperature difference are taken as thermodynamic characteristics.

[0010] Furthermore, before the step of inputting the battery health state feature vector into a pre-trained first regression model, and having the first regression model output a value representing the current actual capacity retention rate of the battery as a current health state score, the method further includes: Obtain a historical battery sample set. For each historical battery in the sample set, collect the initial battery health state feature vector corresponding to the detection point at multiple detection points in its life cycle. The true capacity retention rate of the historical battery at this test point is obtained through laboratory standard capacity testing. Use the initial battery health state feature vector and the actual capacity retention rate as a training sample pair; Using all training sample pairs from all historical batteries, with the initial battery health state feature vector as input features and the true capacity retention rate as output labels, train a support vector regression model. During training, grid search and cross-validation methods are used to adjust the core hyperparameters of the support vector regression model; wherein, the core hyperparameters include at least the penalty coefficient C, the kernel function coefficient gamma, and the kernel function type; After training is complete, the model structure and parameters are saved as the first regression model in the pre-training process.

[0011] Furthermore, the step of calculating the future aging acceleration factor based on the dynamic stress spectrum includes: The normalized dynamic stress spectrum is input into the pre-trained second regression model, which outputs an initial predicted value of the future aging acceleration factor. The second regression model is a feedforward neural network with the same input layer dimension as the dynamic stress spectrum dimension and a single neuron as the output layer. The activation function is ReLU to ensure that the output is non-negative. Extract all instantaneous stress coefficients located within the first rolling window from the historical stress sequence to form the recent stress sequence; Calculate the standard deviation and coefficient of variation of the recent stress sequence; Multiply the coefficient of variation by the preset stability sensitivity coefficient, and substitute the product into the exponential function to obtain the mode stability coefficient; The first value is obtained by multiplying the predicted value of the future aging acceleration factor by the model stability coefficient; Subtracting 1 from the mode stability coefficient yields the second value; The second value is multiplied by a preset baseline acceleration factor to obtain the future aging acceleration factor.

[0012] Furthermore, the step of calculating the current health stress index based on the current health status score and future aging acceleration factors includes: Obtain the preset health status score threshold at the end of life and the preset high health status threshold; When the current health status score is greater than the high health status threshold, 1 is used as the stress health status coupling coefficient; When the current health status score is between the end-of-life health status score threshold and the high health status threshold, the difference between the high health status threshold and the current health status score is multiplied by the sensitivity coefficient and then incremented by 1 to obtain the stress health status coupling coefficient; where the high health status threshold is greater than the end-of-life health status score threshold. The effective aging acceleration factor is obtained by multiplying the future aging acceleration factor with the stress-health state coupling coefficient. Subtract the current health status score from the health status score threshold at the end of life to obtain the health status margin; The baseline health stress index is obtained by dividing the health status margin by the effective aging acceleration factor. The mean value T_stress_avg of the temperature stress coefficient within the second rolling window is extracted from the historical stress sequence that generated the dynamic stress spectrum. The temperature fine-tuning factor is calculated based on the mean value T_stress_avg; where, the temperature fine-tuning factor = 1 - θ*max(0, T_stress_avg - T_ref); where T_ref is the preset reference temperature stress coefficient, θ is the temperature influence coefficient, 0 < θ < 1; The final current health stress index is calculated based on the temperature fine-tuning factor and the baseline health stress index; where the current health stress index = max(HPI_min, baseline health stress index * temperature fine-tuning factor), and HPI_min is the preset minimum health stress index.

[0013] A second aspect of the present invention provides a battery life prediction device for a battery swapping cabinet, comprising: The data acquisition unit is used to acquire multi-dimensional battery characteristic data when the battery to be identified completes one charge-discharge cycle and is connected to the battery swapping cabinet; wherein, the multi-dimensional battery characteristic data includes at least the depth of discharge, the average discharge rate, and the average ambient temperature during the cycle; The statistical unit is used to statistically analyze the dynamic stress spectrum corresponding to the battery to be identified based on the multidimensional battery feature data; wherein the dynamic stress spectrum is a statistical feature reflecting the usage intensity within different time windows, calculated based on the historical cycle data of the battery to be identified. The calculation unit is used to calculate the remaining life of the battery to be identified based on the dynamic stress spectrum.

[0014] A third aspect of the present invention provides a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps in the battery life prediction method in the battery swapping cabinet described in the first aspect.

[0015] A fourth aspect of the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in the battery life prediction method in the battery swapping cabinet described in the first aspect.

[0016] The beneficial effects of this invention compared to existing technologies are as follows: By collecting multi-dimensional characteristic data of the battery (including depth of discharge, average discharge rate, and average ambient temperature during cycling), this method can comprehensively reflect the dynamic state of the battery during actual use. This comprehensive data collection and analysis makes battery life prediction more accurate, thus better reflecting the actual aging of the battery. The dynamic stress spectrum calculated based on historical cycle data can effectively capture the usage intensity of the battery within different time windows. The introduction of this statistical feature allows the prediction process to not only consider static parameters but also dynamically reflect the stress changes of the battery during use, thereby improving the adaptability and accuracy of the model. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A schematic flowchart of a battery life prediction method in a battery swapping cabinet provided by the present invention is shown. Figure 2 This diagram illustrates a battery life prediction device in a battery swapping cabinet according to an embodiment of the present invention. Figure 3 A schematic diagram of a terminal device provided in an embodiment of the present invention is shown. Detailed Implementation

[0019] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of the invention. However, those skilled in the art will understand that the invention can be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods are omitted so as not to obscure the description of the invention with unnecessary detail.

[0020] This invention provides a method and apparatus for predicting battery life in a battery swapping cabinet, which solves the technical problem that traditional methods ignore the dynamic characteristics of batteries under different usage conditions, thus limiting the accuracy and reliability of the prediction results.

[0021] First, this invention provides a method for predicting battery life in a battery swapping cabinet. Please refer to... Figure 1 , Figure 1 A schematic flowchart of a battery life prediction method in a battery swapping cabinet provided by the present invention is shown. Figure 1 As shown, the battery life prediction method in this battery swapping cabinet may include the following steps: Step 101: When the battery to be identified completes one charge-discharge cycle and is connected to the battery swapping cabinet, collect multi-dimensional battery characteristic data; wherein, the multi-dimensional battery characteristic data includes at least the depth of discharge, average discharge rate and average ambient temperature during the cycle; Multidimensional battery characteristic data includes, but is not limited to, depth of discharge, average discharge rate, and average ambient temperature during cycling.

[0022] The collected data are not simply voltage and current measurements, but stress-related parameters directly correlated with battery aging mechanisms. Depth of discharge reflects the energy throughput during a single cycle and is a key factor affecting battery capacity decay. Average discharge rate reflects the intensity of the discharge current, directly impacting internal heat generation and polarization losses. Average ambient temperature during cycling is the primary environmental stress that accelerates almost all electrochemical side reactions. The selection of these three dimensions together constitutes the fundamental data describing the damage intensity of a single cycle to the battery, laying the foundation for subsequent stress analysis.

[0023] Step 102: Based on the multidimensional battery feature data, calculate the dynamic stress spectrum corresponding to the battery to be identified; wherein, the dynamic stress spectrum is a statistical feature reflecting the usage intensity within different time windows, calculated based on the historical cycle data of the battery to be identified. The core of this step, from single-data collection to historical patterns, is to aggregate the discrete data collected in each cycle into a comprehensive feature that reflects the long-term usage history of the battery—the dynamic stress spectrum—through statistical methods.

[0024] The statistics are not a simple average of the entire history, but rather a calculation of recent (e.g., the past 50 cycles), medium-term, or specific phases (e.g., the high-temperature period of summer). This captures dynamic patterns such as changes in user habits and seasonal influences. The spectrum contains statistical characteristics of usage intensity calculated based on stress parameters such as depth of discharge, average discharge rate, and average ambient temperature.

[0025] The dynamic stress spectrum is a multidimensional feature vector, where each dimension represents a stress intensity index statistically derived from different time windows and perspectives. It is a digital and structured summary of the complex usage history of a battery.

[0026] Specifically, step 102 includes steps 1021 to 1026: Step 1021: Based on the multidimensional battery characteristic data, calculate the instantaneous stress coefficient characterizing the degree of accelerated aging in this cycle; wherein, the instantaneous stress coefficient is obtained by combining the depth of discharge stress coefficient, rate stress coefficient and temperature stress coefficient; This is the foundation of the entire feature engineering process. It doesn't view raw data such as depth of discharge, rate capability, and temperature in isolation, but rather transforms them into coefficients that directly affect the battery aging rate. This represents a shift from phenomenological data to mechanistic influence.

[0027] Specifically, step 1021 includes steps A1 to A4: Step A1: Divide the discharge depth by the preset rated discharge depth reference value to obtain k. d The power is used as the discharge depth stress coefficient; where k d The first empirical index is greater than 1; Discharge depth stress coefficient = (Discharge depth / Reference value of rated discharge depth) kd .

[0028] The depth of discharge / rated depth of discharge reference value normalizes the data. The rated depth of discharge reference value is typically a design or testing standard value (e.g., the battery's nominal recommended depth of discharge, such as 80%). A ratio greater than 1 indicates a deeper discharge than the reference, while a ratio less than 1 indicates a shallower discharge.

[0029] Index k d >1 is the key here. It means that the effect of discharge depth on aging is non-linearly accelerated.

[0030] For example: Suppose k d =2 (square relationship), when the depth of discharge is 1.2 times the reference value, the stress coefficient is 1.2. 2=1.44. When it is 1.5 times, the coefficient is 2.25. This indicates that the additional damage caused by deep discharge is far greater than the linear proportion of the increase in discharge depth. This is entirely consistent with the known mechanism in battery chemistry that deep discharge exacerbates the structural stress of electrode materials and accelerates the loss of active materials.

[0031] Step A2: Divide the average discharge rate by the preset rated discharge rate reference value to obtain k. c The power is used as the magnification stress coefficient; where k c It is the second empirical index, which is greater than 1; Rate stress coefficient = (Average discharge rate / Reference value of rated discharge rate) kc .

[0032] The rated discharge rate reference value (e.g., 1C) is used to measure the relative magnitude of the discharge current. The exponent k c > 1 also reflects the nonlinear effect of high-rate discharge on accelerated aging. High-rate discharge leads to greater polarization, more ohmic heat, and more uneven current distribution, thus accelerating aging. k c The intensity of this acceleration was quantified.

[0033] Step A3: Substitute the average ambient temperature during the cycle into the Arrhenius model to obtain the temperature stress coefficient; The temperature stress coefficient is calculated using the Arrhenius model (average ambient temperature during cycling). The Arrhenius equation is a physicochemical cornerstone describing the relationship between chemical reaction rates and temperature. Its form is typically exp(-Ea / (R*T)), where Ea is the activation energy, R is the gas constant, and T is the absolute temperature. Using this model implies acknowledging that battery aging is essentially an electrochemical side reaction, and temperature is the most critical factor affecting the rates of these reactions. This model accurately characterizes the effect of a small increase in temperature leading to an exponential increase in the aging rate (generally, empirically, the aging rate doubles for every 10°C increase in temperature).

[0034] Step A4: Weighted summation of the discharge depth stress coefficient, rate stress coefficient, and temperature stress coefficient to obtain the instantaneous stress coefficient.

[0035] Instantaneous stress coefficient = w1 * depth of discharge stress coefficient + w2 * rate stress coefficient + w3 * temperature stress coefficient The weights w1, w2, and w3 reflect the relative importance of the three stress modes in contributing to the overall aging of the battery. These weights can be obtained by fitting bench aging test data of the battery and are adjustable empirical parameters in the scheme. For example, w1, w2, and w3 can be set to 0.22, 0.46, and 0.32, respectively.

[0036] In the embodiments corresponding to steps A1 to A4, by comprehensively considering multiple dimensions such as depth of discharge, rate, and temperature, the instantaneous stress coefficient of the battery in a specific cycle is calculated, providing a quantitative indicator for battery health status assessment and remaining life prediction. This method is designed to take into account nonlinear relationships and the influence of environmental factors, and can more accurately reflect the true aging condition of the battery.

[0037] Step 1022: Add the instantaneous stress coefficient, the depth of discharge stress coefficient, the rate stress coefficient, and the temperature stress coefficient as a new record to the historical stress sequence for the maintenance of the battery to be identified in chronological order; This step establishes a unique, time-ordered health profile for each battery. This profile not only records the overall damage (instantaneous stress coefficient) but also retains the original coefficients of each individual stress component, providing a complete data foundation for subsequent multi-dimensional, multi-scale statistical analysis.

[0038] Step 1023: In the historical stress sequence, based on the preset first rolling window, the first mean and standard deviation of the instantaneous stress coefficient are statistically analyzed, and the proportion of the discharge depth stress coefficient greater than the first depth threshold and the proportion of the rate stress coefficient greater than the first rate threshold are statistically analyzed. The first mean reflects the average damage intensity over a recent period (e.g., the past 20 cycles). The standard deviation reflects the variability in recent usage intensity. Larger variability may indicate more demanding usage conditions.

[0039] We statistically analyze the frequency of recent extreme operating conditions, such as deep discharge or high-rate discharge exceeding the threshold. This is a very valuable characteristic because occasional extreme use can have a far greater impact on lifespan than mild, everyday use.

[0040] The first window is relatively small and is used to capture recent, high-frequency usage behavior characteristics and intensity fluctuations.

[0041] Step 1024: In the historical stress sequence, based on a preset second rolling window, calculate the second mean value of the instantaneous stress coefficient and the cumulative duration for which the temperature stress coefficient is greater than the temperature threshold; wherein, the window size of the second rolling window is larger than the window size of the first rolling window; The average damage intensity (second mean)* is calculated over a larger time window (e.g., the past 200 cycles) to reflect the average level of use in the medium to long term.

[0042] Temperature has a strong cumulative effect on aging over time. Statistically measuring the cumulative duration of high temperatures (high stress coefficients) is a more accurate way to quantify the contribution of thermal aging than simply counting the number of high-temperature events.

[0043] The second window is larger and is used to capture medium- to long-term usage patterns and environmental exposure histories with cumulative effects.

[0044] Step 1025: The first mean, the second mean, the standard deviation, the proportion greater than the first depth threshold, the proportion greater than the first magnification threshold, and the cumulative duration greater than the temperature threshold are spliced ​​together in a predetermined order to generate a one-dimensional dynamic stress spectrum. The six (or more) features identified above are concatenated into a one-dimensional vector in a predetermined order. This is the final form of the dynamic stress spectrum—a fixed-length, digitized feature vector. It integrates information from multiple dimensions, including recent intensity, fluctuations, frequency of extreme events, medium- to long-term average intensity, and thermal accumulation effects, comprehensively encoding the battery's historical stress patterns.

[0045] Step 1026: For each element in the dynamic stress spectrum, normalize it based on a preset nominal range to obtain the normalized dynamic stress spectrum.

[0046] For each element in the dynamic stress spectrum, normalization is performed based on its predefined nominal range. This is a standard preprocessing step for machine learning model inputs. Normalization (such as scaling to the [0,1] interval) eliminates differences in the dimensions and orders of magnitude of different features, accelerates model convergence, and improves prediction stability and accuracy. It enables subsequent lifetime prediction models to more fairly weigh the importance of each feature.

[0047] In the embodiments corresponding to steps 1021 to 1026, a dynamic stress spectrum is generated by calculating and statistically analyzing various aspects of the stress experienced by the battery during charging and discharging. This dynamic stress spectrum will provide an important basis for subsequent prediction of the battery's remaining lifespan, helping to improve the reliability and safety of battery use.

[0048] Step 103: Calculate the remaining lifespan of the battery to be identified based on the dynamic stress spectrum.

[0049] This step is the final output of the method. The input to the prediction model is no longer the raw, single-shot battery data, but the highly generalized dynamic stress spectrum generated in step 102.

[0050] The remaining lifespan of a battery depends primarily on the stress patterns it has experienced throughout its history, rather than the instantaneous state of its last cycle. Dynamic stress spectra are a quantitative expression of these historical stress patterns. By establishing a mapping between stress spectra and remaining lifespan, the accuracy and robustness of predictions are expected to be significantly higher than predictions based on single-point data.

[0051] Specifically, step 103 includes steps B1 to B10: Step B1: Obtain the current battery health state feature vector of the battery to be identified at the current detection time; wherein, the current battery health state feature vector includes at least electrochemical features and thermodynamic features; Specifically, step B1 includes steps B11 to B19: Step B11: Calculate the slope of the linear fit of the voltage-time curve during the constant current charging stage as a characteristic of the voltage rise rate. During constant current charging, the voltage rise is mainly determined by the battery's internal impedance and polarization voltage. As the battery ages, its internal impedance (especially ohmic resistance and charge transfer impedance) increases, resulting in a faster voltage rise at the same charging current.

[0052] Linear fitting of the voltage-time curve yields a stable, noise-resistant slope that is more robust than directly taking the voltage difference between two points. This is a macroscopic feature that is highly sensitive to increases in internal resistance.

[0053] Step B12: During the constant voltage charging stage, the current decay curve is subjected to exponential fitting, and its time constant is extracted as a polarization dissipation characteristic. Towards the end of constant-voltage charging, the current decays exponentially to near zero. The time constant of this decay process reflects the rate at which the internal polarization (concentration polarization, electrochemical polarization) dissipates within the battery. In aged batteries, the structure of the electrode active materials degrades, and the lithium-ion diffusion coefficient decreases, leading to more severe polarization and slower dissipation. Therefore, an increase in the time constant is an important indicator of aging. This characteristic deeply reflects the battery's kinetic properties and is a good complement to the voltage rise rate characteristic.

[0054] Step B13: After the battery to be identified has been left to stand for a preset time, apply a short-time pulse discharge, record the voltage transient response, and calculate the pulse internal resistance and the relaxation time for the voltage to recover to a stable value. The pulse internal resistance is calculated based on ΔV / ΔI. It directly and dynamically measures the ohmic internal resistance of the battery at a specific SOC and is one of the most direct indicators of aging (aging leads to SEI film thickening, contact degradation, etc., increasing internal resistance). The pulse also measures the time required for the voltage to recover to a stable value. This reflects the diffusion process and charge redistribution rate within the battery. In aged batteries, ion diffusion capacity deteriorates, and relaxation time becomes longer.

[0055] Step B14: After entering the low-current constant-voltage float charging state at the end of the charging process, record the voltage drop during the first preset time period and calculate the float charging self-discharge rate. After a full charge, maintain a small float charge, and the battery voltage should remain stable. If the voltage continues to drop, it indicates self-discharge. Increased self-discharge rate is usually associated with accelerated aging side reactions such as internal micro-short circuits, SEI film instability, or electrolyte decomposition. This is a sensitive characteristic reflecting the internal chemical stability of the battery. Slight aging may not immediately manifest in capacity, but it will first lead to changes in self-discharge characteristics.

[0056] Step B15: Use the voltage rise rate characteristic, the polarization dissipation characteristic, the relaxation time, and the float charge self-discharge rate as electrochemical characteristics; Step B16: During the constant current charging phase, calculate the temperature rise rate at the highest temperature point on the surface of the battery to be identified; The heat generated during charging mainly comes from ohmic heat and reaction heat. As the internal resistance of an aging battery increases, the ohmic heat (I0.05) increases under the same current. 2 An increase in R (resistance) leads to a faster rate of temperature rise. This is a characteristic associated with both internal resistance and the risk of thermal runaway. An abnormally rapid rate of temperature rise is an early signal of battery performance degradation and safety hazards.

[0057] Step B17: At the same time point, calculate the variance of all surface temperature measurements of the batteries to be identified as a characteristic of temperature distribution uniformity. Internal aging of the battery (such as localized lithium deposition or uneven electrode degradation) or poor external contact can lead to uneven heat generation and dissipation, manifesting as an uneven surface temperature field. Temperature variance is a powerful indicator of the battery's internal consistency and external connectivity. In battery packs with poor consistency, the temperature variance of individual cells will increase significantly with aging.

[0058] Step B18: After charging is completed and the battery has been left to stand for a second preset time period, calculate the equilibrium temperature difference between the average surface temperature of the battery to be identified and the ambient temperature. After charging is complete, the battery temperature will gradually reach equilibrium with the ambient temperature. If a significant temperature difference persists after equilibrium, it indicates the presence of a continuous, low-rate exothermic reaction within the battery (such as a persistent side reaction), or that the battery's insulation / heat dissipation design results in an extremely long thermal time constant. An abnormal equilibrium temperature difference may indicate an internal micro-short circuit or other aging mechanisms involving continuous exothermic processes, serving as an important indicator of safety and health status.

[0059] Step B19: The temperature rise rate, the temperature distribution uniformity, and the equilibrium temperature difference are taken as thermodynamic characteristics.

[0060] In the embodiments corresponding to steps B11 to B19, the battery health status feature vector obtained through the above steps integrates the battery's electrochemical and thermodynamic characteristics, providing rich and accurate data support for subsequent lifetime prediction. This multi-dimensional feature extraction method helps to comprehensively assess the battery's health status, thereby improving the accuracy of remaining lifetime prediction.

[0061] Step B2: Input the current battery health status feature vector into the pre-trained first regression model, and the first regression model outputs a value representing the current actual capacity retention rate of the battery as the current health status score; The first regression model is an independent model for state estimation. Its task is to accurately deduce the current actual capacity retention of the battery based on rapid health check data. This step provides an accurate starting point for subsequent lifetime prediction.

[0062] As an optional embodiment of this application, steps C1 to C6 are further included before step B2: Step C1: Obtain a historical battery sample set. For each historical battery in the sample set, collect the initial battery health state feature vector corresponding to the detection point at multiple detection points during its life cycle. Step C2: Obtain the true capacity retention rate of the historical battery at this test point through laboratory standard capacity testing; The historical battery sample set needs to include a sufficient number of batteries of different types (such as those from different batches or different usage scenarios) to ensure the model's generalization ability.

[0063] Multiple testing points throughout the lifecycle mean longitudinally tracking each battery, sampling at multiple stages from its brand-new state (SOH=100%), through moderate aging (e.g., SOH=80%), to the end of its lifespan (e.g., SOH=70%). This ensures that the model learns the characteristic changes in the battery throughout its aging trajectory.

[0064] Standardized laboratory capacity testing is crucial for obtaining accurate and reliable labels. This typically involves conducting a complete charge-discharge cycle under strictly controlled laboratory conditions (constant temperature, standard charge-discharge regime) to measure the battery's actual remaining capacity and calculate its ratio to the initial capacity (capacity retention rate). This data serves as the ground truth label for supervised learning and is the cornerstone of the model's ability to make accurate predictions.

[0065] Step C3: Use the initial battery health state feature vector and the true capacity retention rate as a training sample pair; Each sample pair establishes a mapping relationship between observable, non-invasive features and crucial performance metrics that cannot be directly and quickly obtained. The goal of the model is to learn this complex mapping function.

[0066] Step C4: Using all training sample pairs from all historical batteries, with the initial battery health state feature vector as the input feature and the true capacity retention rate as the output label, train a support vector regression model. Support Vector Regression (SVR) is a classic model for handling small-sample, nonlinear regression problems. In battery aging prediction, the amount of accurately labeled laboratory data available early on may be limited. In this case, SVR often performs more stably and is less prone to overfitting compared to more complex models such as deep neural networks. Furthermore, SVR handles the nonlinear relationship between features and labels very well.

[0067] The model learns a function f such that for an input feature vector X, its output f(X) is as close as possible to the true capacity retention ratio Y.

[0068] Step C5: During training, the core hyperparameters of the support vector regression model are adjusted using grid search and cross-validation; wherein the core hyperparameters include at least the penalty coefficient C, the kernel function coefficient gamma, and the kernel function type; The penalty coefficient C controls the model's tolerance for errors in the training data. The larger the C value, the more the model tends to fit all training points (potentially overfitting); the smaller the C value, the smoother the model (potentially underfitting).

[0069] The kernel function coefficient gamma affects the range of influence of the kernel function (such as the RBF kernel). The larger the gamma, the smaller the range of influence for each sample point, and the more complex the model; the smaller the gamma, the wider the range of influence, and the smoother the model.

[0070] The type of kernel function determines how the model maps data to a high-dimensional space to find linear relationships. Common kernel types include linear kernels, polynomial kernels, and radial basis function (RBF) kernels. RBF kernels are frequently used due to their strong nonlinear fitting capabilities.

[0071] Grid search systematically traverses and attempts within a predefined range of hyperparameter combinations. The training data is divided into multiple parts, with one part used as the validation set and the rest as the training set in turn. This process of training and validation is repeated multiple times to evaluate the performance of each hyperparameter set. This effectively avoids overfitting caused by the randomness of data partitioning and finds the hyperparameter combination with optimal generalization ability.

[0072] Step C6: After training is complete, save the model structure and parameters as the first regression model in the pre-training.

[0073] The trained SVR model (including its kernel function, support vectors, coefficients, and all other parameters) is serialized and saved. In the battery swapping cabinet's prediction system, simply loading this pre-trained model file allows for rapid forward inference of new, unknown battery feature vectors, outputting their health status score. This enables a "train once, use many times" deployment model.

[0074] In the embodiments corresponding to steps C1 to C6, the above steps construct a regression model trained on historical data, enabling it to effectively map the battery's health state feature vector to the actual capacity retention rate. This process ensures the model's accuracy and reliability, providing a solid foundation for subsequent battery health state scoring.

[0075] Step B3: Based on the dynamic stress spectrum, calculate the future aging acceleration factor; Specifically, step B3 includes steps B31 to B37: Step B31: Input the normalized dynamic stress spectrum into the pre-trained second regression model, and the second regression model outputs an initial predicted value of the future aging acceleration factor; wherein, the second regression model is a feedforward neural network, the dimension of its input layer is the same as the dimension of the dynamic stress spectrum, the output layer is a single neuron, and the activation function is ReLU to ensure that the output is non-negative; The normalized dynamic stress spectrum is input into a pre-trained second regression model (feedforward neural network), which outputs an initial predicted value for the future aging acceleration factor. A feedforward neural network is used because the dynamic stress spectrum, as a multi-dimensional feature vector, may have a complex and non-linear mapping relationship with the aging acceleration factor. Neural networks are well-suited for learning and fitting such complex patterns.

[0076] The input layer dimension is the same as the dynamic stress spectrum dimension: ensuring complete reception of feature information. The output layer is a single neuron because the output is a single acceleration factor value. The activation function is ReLU (Rectified Linear Unit), guaranteeing a non-negative output (the aging acceleration factor cannot be negative). This is a very reasonable and crucial constraint. Feedforward neural networks are a traditional technique and will not be elaborated upon here.

[0077] This neural network is a complex mapper from historical stress patterns to future aging trends. It learns through training what combinations of stress spectra (such as prolonged high temperatures and frequent high-rate discharges) will lead to faster future aging.

[0078] Step B32: Extract all instantaneous stress coefficients located within the first rolling window from the historical stress sequence to form a recent stress sequence; The recent instantaneous stress coefficients (within the first rolling window) are extracted from the historical stress sequence, and their standard deviation and coefficient of variation are calculated. The recent stress sequence focuses on the instantaneous damage intensity over a recent period (e.g., the past 20 cycles).

[0079] Step B33: Calculate the standard deviation and coefficient of variation of the recent stress sequence; The standard deviation measures the absolute volatility of the sequence. The coefficient of variation is the standard deviation divided by the mean. This is a relative volatility indicator that eliminates the influence of different mean levels on volatility assessment. The coefficient of variation better reflects the instability or randomness of usage patterns.

[0080] Step B34: Multiply the coefficient of variation by the preset stability sensitivity coefficient, and substitute the product into the exponential function to obtain the mode stability coefficient; The stability coefficient is calculated as exp(-stability sensitivity coefficient * coefficient of variation). This is an exponentially decaying function. When the coefficient of variation is 0 (historical usage patterns are extremely stable), the stability coefficient is 1 (completely stable). As the coefficient of variation increases (historical usage patterns become more volatile), the stability coefficient decreases exponentially, approaching 0.

[0081] The stability sensitivity coefficient is an adjustable parameter used to control how sensitive the coefficient of variation is to the influence of the stability coefficient. The larger the coefficient, the more sensitive the stability is to fluctuations.

[0082] The pattern stability coefficient is a value between 0 and 1 that quantifies the regularity / predictability of historical usage patterns. The closer the value is to 1, the more stable the user habits or environment, and the higher the confidence level in predicting the future based on historical patterns.

[0083] Step B35: Multiply the predicted value of the future aging acceleration factor by the model stability coefficient to obtain the first value; Step B36: Subtract 1 from the mode stability coefficient to obtain the second value; Step B37: Multiply the second value by the preset baseline acceleration factor to obtain the future aging acceleration factor.

[0084] Future aging acceleration factor = (mode stability coefficient * neural network prediction) + ((1 - mode stability coefficient) * baseline acceleration factor).

[0085] The first term is: Pattern stability coefficient * Neural network prediction. When the historical pattern is stable (coefficient close to 1), the final result highly trusts the neural network's prediction.

[0086] The second term is: (1 - model stability coefficient) * baseline acceleration factor. When the historical model is extremely unstable (coefficient close to 0), the final result depends almost entirely on a preset baseline acceleration factor.

[0087] The baseline acceleration factor can be understood as a default or conservative estimate when reliable historical data is lacking. For example, it can be set to 1 (standard aging rate), or an empirical value can be set based on factors such as battery type and season.

[0088] Using the stability of historical data as a weight, a smooth interpolation is performed between personalized predictions and conservative default values. When user behavior is regular, a highly personalized prediction is given; when user behavior is erratic, a safe and conservative estimate is used instead.

[0089] In the embodiments corresponding to steps B31 to B37, through the above steps, this method integrates dynamic stress data, historical stress fluctuations, and model predictions to form a multi-level aging acceleration factor calculation process. Each step aims to ensure that the assessment of the future health status of the battery is as accurate as possible, thereby providing a scientific basis for battery management and maintenance.

[0090] Step B4: Calculate the current health stress index based on the current health status score and future aging acceleration factors; The future aging acceleration factor is a key bridge in this embodiment. Based on a dynamic stress spectrum characterizing historical usage intensity patterns, it predicts the acceleration factor of the aging rate relative to standard operating conditions over a future cycle. For example, if the factor is 1.2, the expected future aging rate is 1.2 times that of standard conditions. This enables inferences from past usage patterns to future aging rates.

[0091] This index combines the battery's current physical condition (health score) with future stresses (accelerators of aging). A healthy battery that will face high-intensity use in the future may have a stress score similar to that of a less healthy battery that will face moderate use in the future. This index provides a more comprehensive reflection of the battery's buffer space before the end of its lifespan.

[0092] Specifically, step B4 includes steps B41 to B49: Step B41: Obtain the preset end-of-life health status score threshold and the preset high health status threshold; Step B42: When the current health status score is greater than the high health status threshold, use 1 as the stress health status coupling coefficient; Step B43: When the current health status score is between the end-of-life health status score threshold and the high health status threshold, multiply the difference between the high health status threshold and the current health status score by the sensitivity coefficient and add 1 to obtain the stress health status coupling coefficient; wherein, the high health status threshold is greater than the end-of-life health status score threshold. Step B44: Multiply the future aging acceleration factor by the stress-health coupling coefficient to obtain the effective aging acceleration factor; The pressure-health state coupling coefficient is intended to quantify the sensitivity of the current health status of the battery to future pressure. A healthier battery is less sensitive to future pressure (with sufficient buffer); a battery closer to the end of its life is more sensitive to future pressure (with less buffer).

[0093] In the high health state zone (SOH > high health state threshold), the coefficient is fixed at 1. This means that when the battery is very healthy, the future aging acceleration factor will fully act on its life prediction, without additional sensitivity amplification.

[0094] In the decay-sensitive zone (end-of-life threshold < SOH ≤ high health state threshold), the coefficient = 1 + sensitivity coefficient * (high health state threshold - current SOH).

[0095] As the SOH decreases from the high health state threshold to the end-of-life threshold, this coefficient increases linearly from 1. The sensitivity coefficient controls the rate of increase, which simulates the accelerated decay phenomenon in the late stage of battery aging. For an already aged battery, the same pressure (such as high-rate discharge) will cause greater relative damage to it. This coefficient amplifies the effect of future pressure on the aged battery.

[0096] Step B45: Subtract the end-of-life health state score threshold from the current health state score to obtain the health state margin; Step B46: Divide the health state margin by the effective aging acceleration factor to obtain the reference health pressure index; Reference health pressure index = (current SOH - end-of-life threshold) / (future aging acceleration factor * pressure-health state coupling coefficient).

[0097] (Current SOH - end-of-life threshold) is the absolute remaining capacity buffer of the battery from failure. Future aging acceleration factor * pressure-health state coupling coefficient is the future expected aging rate corrected by the current health state sensitivity.

[0098] Reference health pressure index = remaining buffer / corrected consumption rate. It intuitively represents how long the remaining buffer of the battery can last (in relative units) when future pressure and current sensitivity are taken into account. A larger value indicates a longer remaining life.

[0099] Step B47: Extract the mean value T_stress_avg of the temperature stress coefficient within the second rolling window from the historical stress sequence that generates the dynamic stress spectrum; Step B48: Calculate the temperature fine-tuning factor based on the mean T_stress_avg; where, temperature fine-tuning factor = 1 - θ*max(0, T_stress_avg - T_ref); where T_ref is the preset reference temperature stress coefficient, θ is the temperature influence coefficient, 0 < θ < 1; Step B49: Calculate the final current health stress index based on the temperature fine-tuning factor and the baseline health stress index; where, the current health stress index = max(HPI_min, baseline health stress index * temperature fine-tuning factor), and HPI_min is the preset minimum health stress index.

[0100] High temperature is the most significant environmental stress causing battery aging. Even if the usage pattern (dynamic stress spectrum) remains the same in the future, prolonged exposure to higher ambient temperatures will accelerate the actual aging of the battery.

[0101] The mean value T_stress_avg of the temperature stress coefficient within the second rolling window (reflecting the medium- to long-term trend) is obtained from the historical data of the generated dynamic stress spectrum. This represents the average thermal environment of the battery in the near future.

[0102] Temperature fine-tuning factor = 1 - θ * max(0, T_stress_avg - T_ref).

[0103] T_ref is a preset reference temperature stress coefficient, corresponding to a standard or ideal ambient temperature (such as 24 degrees Celsius).

[0104] `max(0,...)` means that a penalty (downward adjustment) is only applied when the average temperature stress is higher than the reference value. If the temperature is lower than the reference temperature, it is considered to have no additional negative impact.

[0105] θ is the temperature influence coefficient, which controls the intensity of temperature penalty.

[0106] If a battery is exposed to high temperatures for an extended period (T_stress_avg > T_ref), its health stress index will be discounted (multiplied by a factor less than 1) because the high temperature continuously depletes its health buffer. This reflects the cumulative effect of environmental stress.

[0107] Current Health Stress Index = max(HPI_min, Baseline Health Stress Index * Temperature Adjustment Factor).

[0108] The baseline exponent is adjusted downwards using a temperature fine-tuning factor. The `max(HPI_min,...)` operation sets a minimum value `HPI_min` (e.g., a very small positive number). This prevents the calculated exponent from being too low or even negative (due to calculation errors or extreme cases), which could lead to mathematical anomalies in subsequent lifetime calculations (such as division by a very small decay rate). This enhances the numerical stability and robustness of the algorithm.

[0109] In the embodiments corresponding to steps B41 to B48, through the above steps, the method comprehensively considers the battery's health status, aging rate, and the influence of external temperature to calculate a more accurate current health stress index. This index not only reflects the battery's current health status but also predicts its future performance, thus providing an important basis for battery management and maintenance.

[0110] Step B5: Retrieve the historical health stress index values ​​of the battery to be identified at the most recent M detection times from the historical database, forming a time series; Step B6: Based on the time series, use linear regression to calculate the decay rate of the health stress index relative to the number of detections; wherein the number of detections is related to the number of charge-discharge cycles of the battery; Step B7: Multiply the decay rate by the future aging acceleration factor to obtain the corrected decay rate; By analyzing the decreasing trend of the health stress index over time (number of tests), its historical degradation rate can be obtained. This reflects the aging rate of the battery under past usage patterns. Multiplying the historical degradation rate by the future aging acceleration factor yields the corrected degradation rate. This means that this embodiment does not simply use past aging rates to predict the future, but rather scales the historical aging rate according to the predicted future usage intensity (reflected by the dynamic stress spectrum). If the predicted future usage is more demanding (acceleration factor > 1), the degradation rate is increased; otherwise, it is decreased. This allows the prediction to dynamically adapt to changes in battery usage patterns.

[0111] Step B8: If the current health stress index is not greater than the health stress index threshold, it is determined that the battery to be identified has reached the end of its lifespan, and the remaining lifespan cycle count is set to 0. Step B9: If the decay rate is greater than the limit value, the decay rate is determined to be abnormal, and the remaining lifetime cycle number is set to a preset maximum value or an invalid flag. Step B10: Subtract the current health stress index from the health stress index threshold, and divide by the decay rate to obtain the number of cycles of remaining lifespan.

[0112] If the current health stress index is less than or equal to the threshold, then the remaining lifespan is 0. This is a reasonable stop-loss judgment to prevent meaningless predictions about a failed battery.

[0113] If the corrected degradation rate exceeds the limit, then the remaining lifetime = maximum value / invalid flag. This is an important robust design feature. When the calculated degradation rate is abnormally fast (potentially due to data noise, model error, or sudden battery drop), the system does not output an unreliable short lifetime but instead marks it as abnormal and requires review, or provides a conservative maximum value, thus avoiding misjudgment and improving the system's usability.

[0114] Remaining lifespan cycles = (Current health stress index - threshold) / Corrected decay rate. This is a classic "remaining distance / velocity = remaining time" model.

[0115] (Current Health Stress Index - Threshold) represents the remaining buffer distance of the battery from its current state to the end of its life (threshold).

[0116] The corrected decay rate represents the rate at which the health stress index decreases under expected future usage intensity.

[0117] The model is intuitive and computationally efficient. Furthermore, since the inputs (pressure index and decay rate) have undergone the aforementioned complex and meticulous calculations and corrections, the prediction results have a high degree of reliability.

[0118] In the embodiments corresponding to steps B1 to B10, the remaining battery life is effectively predicted by comprehensively considering the current battery health status, dynamic stress spectrum, and historical health stress index. Each step is logically interconnected, ensuring a comprehensive assessment and accurate prediction of the battery health status.

[0119] In the embodiments corresponding to steps 101 to 103, by collecting multi-dimensional characteristic data of the battery (including depth of discharge, average discharge rate, and average ambient temperature during cycling), this method can comprehensively reflect the dynamic state of the battery during actual use. This comprehensive data collection and analysis makes battery life prediction more accurate, thus better reflecting the actual aging of the battery. The dynamic stress spectrum calculated based on historical cycling data can effectively capture the usage intensity of the battery in different time windows. The introduction of this statistical feature allows the prediction process to not only consider static parameters but also dynamically reflect the stress changes of the battery during use, thereby improving the adaptability and accuracy of the model.

[0120] like Figure 2 This invention provides a battery life prediction device for battery swapping cabinets. Please refer to [link / reference]. Figure 2 , Figure 2 A schematic diagram of a battery life prediction device in a battery swapping cabinet provided by the present invention is shown, as follows: Figure 2 The battery life prediction device in the battery swapping cabinet shown includes: The acquisition unit 21 is used to acquire multi-dimensional battery characteristic data when the battery to be identified completes one charge-discharge cycle and is connected to the battery swapping cabinet; wherein, the multi-dimensional battery characteristic data includes at least the depth of discharge, the average discharge rate, and the average ambient temperature during the cycle. The statistical unit 22 is used to statistically analyze the dynamic stress spectrum corresponding to the battery to be identified based on the multidimensional battery feature data; wherein, the dynamic stress spectrum is a statistical feature reflecting the usage intensity within different time windows, calculated based on the historical cycle data of the battery to be identified. The calculation unit 23 is used to calculate the remaining life of the battery to be identified based on the dynamic stress spectrum.

[0121] This invention provides a battery life prediction device for battery swapping cabinets. By collecting multi-dimensional characteristic data of the battery (including depth of discharge, average discharge rate, and average ambient temperature during cycling), this method can comprehensively reflect the dynamic state of the battery during actual use. This comprehensive data collection and analysis makes battery life prediction more accurate, thus better reflecting the actual aging of the battery. The dynamic stress spectrum calculated based on historical cycling data can effectively capture the usage intensity of the battery in different time windows. The introduction of this statistical feature allows the prediction process to not only consider static parameters but also dynamically reflect the stress changes of the battery during use, thereby improving the adaptability and accuracy of the model.

[0122] Figure 3 This is a schematic diagram of a terminal device provided in an embodiment of the present invention. Figure 3 As shown, a terminal device 3 in this embodiment includes: a processor 30, a memory 31, and a computer program 32 stored in the memory 31 and executable on the processor 30, such as a battery life prediction program for a battery swapping cabinet. When the processor 30 executes the computer program 32, it implements the steps in the various embodiments of the battery life prediction method for a battery swapping cabinet described above, for example... Figure 1 Steps 101 to 103 are shown. Alternatively, when the processor 30 executes the computer program 32, it implements the functions of each unit in the above-described device embodiments, for example... Figure 2 The function of the unit shown.

[0123] For example, the computer program 32 can be divided into one or more units, which are stored in the memory 31 and executed by the processor 30 to complete the present invention. The one or more units can be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program 32 in the terminal device 3. For example, the specific functions of each unit of the computer program 32 can be divided as follows: The data acquisition unit is used to acquire multi-dimensional battery characteristic data when the battery to be identified completes one charge-discharge cycle and is connected to the battery swapping cabinet; wherein, the multi-dimensional battery characteristic data includes at least the depth of discharge, the average discharge rate, and the average ambient temperature during the cycle; The statistical unit is used to statistically analyze the dynamic stress spectrum corresponding to the battery to be identified based on the multidimensional battery feature data; wherein the dynamic stress spectrum is a statistical feature reflecting the usage intensity within different time windows, calculated based on the historical cycle data of the battery to be identified. The calculation unit is used to calculate the remaining life of the battery to be identified based on the dynamic stress spectrum.

[0124] The terminal device includes, but is not limited to, a processor 30 and a memory 31. Those skilled in the art will understand that... Figure 3 This is merely an example of a terminal device 3 and does not constitute a limitation on a terminal device 3. It may include more or fewer components than shown, or combine certain components, or different components. For example, the terminal device may also include input / output devices, network access devices, buses, etc.

[0125] The processor 30 can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor.

[0126] The memory 31 can be an internal storage unit of the terminal device 3, such as a hard disk or memory of the terminal device 3. The memory 31 can also be an external storage device of the terminal device 3, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc., equipped on the terminal device 3. Furthermore, the memory 31 can include both internal and external storage units of the terminal device 3. The memory 31 is used to store the computer program and other programs and data required by the roaming control device. The memory 31 can also be used to temporarily store data that has been output or will be output.

[0127] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0128] It should be noted that the information interaction and execution process between the above-mentioned devices / units are based on the same concept as the method embodiments of the present invention. For details on their specific functions and technical effects, please refer to the method embodiments section, which will not be repeated here.

[0129] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this invention. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0130] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps described in the various method embodiments above.

[0131] This invention provides a computer program product that, when run on a mobile terminal, enables the mobile terminal to implement the steps described in the above-described method embodiments.

[0132] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include at least: any entity or device capable of carrying the computer program code to a photographing device / terminal device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks.

[0133] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0134] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0135] In the embodiments provided by this invention, it should be understood that the disclosed apparatus / network devices and methods can be implemented in other ways. For example, the apparatus / network device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0136] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units.

[0137] It should be understood that, when used in this specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0138] It should also be understood that the term “and / or” as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0139] As used in this specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if [the described condition or event] is detected" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once [the described condition or event] is detected," or "in response to detection of [the described condition or event]."

[0140] Furthermore, in the description of this invention and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0141] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of the invention include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0142] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for predicting battery life in a battery swapping cabinet, characterized in that, The battery life prediction method in the battery swapping cabinet includes: When the battery to be identified completes one charge-discharge cycle and is connected to the battery swapping cabinet, multi-dimensional battery characteristic data is collected; wherein, the multi-dimensional battery characteristic data includes at least the depth of discharge, the average discharge rate, and the average ambient temperature during the cycle; Based on the multidimensional battery feature data, the dynamic stress spectrum corresponding to the battery to be identified is statistically analyzed; wherein, the dynamic stress spectrum is a statistical feature reflecting the usage intensity within different time windows, calculated based on the historical cycle data of the battery to be identified. Based on the dynamic stress spectrum, the remaining lifespan of the battery to be identified is calculated.

2. The battery life prediction method in the battery swapping cabinet as described in claim 1, characterized in that, The step of statistically analyzing the dynamic stress spectrum corresponding to the battery to be identified based on the multidimensional battery feature data includes: Based on the multidimensional battery characteristic data, an instantaneous stress coefficient characterizing the degree of accelerated aging in this cycle is calculated; wherein, the instantaneous stress coefficient is obtained by combining the depth of discharge stress coefficient, rate stress coefficient and temperature stress coefficient. The instantaneous stress coefficient, the depth of discharge stress coefficient, the rate stress coefficient, and the temperature stress coefficient are added as new records in chronological order to the historical stress sequence for the maintenance of the battery to be identified. In the historical stress sequence, based on a preset first rolling window, the first mean and standard deviation of the instantaneous stress coefficient are statistically analyzed, and the proportion of the discharge depth stress coefficient greater than the first depth threshold and the proportion of the rate stress coefficient greater than the first rate threshold are statistically analyzed. In the historical stress sequence, based on a preset second rolling window, the second mean value of the instantaneous stress coefficient and the cumulative duration for which the temperature stress coefficient is greater than the temperature threshold are statistically analyzed; wherein, the window size of the second rolling window is larger than the window size of the first rolling window; The first mean, the second mean, the standard deviation, the proportion of values ​​greater than the first depth threshold, the proportion of values ​​greater than the first magnification threshold, and the cumulative duration of values ​​greater than the temperature threshold are spliced ​​together in a predetermined order to generate a one-dimensional dynamic stress spectrum. For each element in the dynamic stress spectrum, normalization is performed based on a preset nominal range to obtain the normalized dynamic stress spectrum.

3. The battery life prediction method in the battery swapping cabinet as described in claim 2, characterized in that, The step of calculating the instantaneous stress coefficient characterizing the degree of accelerated aging in this cycle based on the multidimensional battery feature data includes: Divide the discharge depth by the preset rated discharge depth reference value to obtain k. d The power is used as the discharge depth stress coefficient; where k d The first empirical index is greater than 1; Divide the average discharge rate by the preset rated discharge rate reference value to obtain k. c The power is used as the magnification stress coefficient; where k c It is the second empirical index, which is greater than 1; Substituting the average ambient temperature during the cycle into the Arrhenius model, we obtain the temperature stress coefficient. The instantaneous stress coefficient is obtained by weighted summation of the discharge depth stress coefficient, the rate stress coefficient, and the temperature stress coefficient.

4. The battery life prediction method in the battery swapping cabinet as described in claim 1, characterized in that, The step of calculating the remaining life of the battery based on the dynamic stress spectrum includes: Obtain the current battery health state feature vector of the battery to be identified at the current detection time; wherein, the current battery health state feature vector includes at least electrochemical features and thermodynamic features; The current battery health status feature vector is input into the pre-trained first regression model, and the first regression model outputs a value that represents the current actual capacity retention rate of the battery, which is used as the current health status score. Based on the dynamic stress spectrum, the future aging acceleration factor is calculated. Calculate the current health stress index based on the current health status score and future aging acceleration factors; From the historical database, retrieve the historical health stress index values ​​of the battery to be identified at the most recent M detection times to form a time series; Based on the time series, linear regression is used to calculate the decay rate of the health stress index relative to the number of detections; wherein, the number of detections is related to the number of charge-discharge cycles of the battery; Multiply the decay rate by the future aging acceleration factor to obtain the corrected decay rate; If the current health stress index is not greater than the health stress index threshold, it is determined that the battery to be identified has reached the end of its lifespan, and the remaining lifespan cycle count is set to 0. If the decay rate exceeds the limit, the decay rate is determined to be abnormal, and the remaining lifetime cycle count is set to a preset maximum value or an invalid flag. Subtract the current health stress index from the health stress index threshold, and divide by the decay rate to obtain the number of cycles of remaining lifespan.

5. The battery life prediction method in the battery swapping cabinet as described in claim 4, characterized in that, The step of obtaining the battery health status feature vector of the battery to be identified at the current detection time includes: Calculate the slope of the linear fit of the voltage-time curve during the constant current charging stage, as a characteristic of the voltage rise rate. During the constant voltage charging stage, the current decay curve is subjected to exponential fitting, and its time constant is extracted as a polarization dissipation characteristic. After the battery to be identified has been left to stand for a preset time, a short-time pulse discharge is applied, the voltage transient response is recorded, and the pulse internal resistance and the relaxation time for the voltage to recover to a stable value are calculated. After entering a low-current constant-voltage float charging state at the end of the charging process, the voltage drop during the first preset time period is recorded, and the float charging self-discharge rate is calculated. The voltage rise rate characteristic, the polarization dissipation characteristic, the relaxation time, and the float charge self-discharge rate are used as electrochemical characteristics. During the constant current charging phase, the temperature rise rate at the highest temperature point on the surface of the battery to be identified is calculated. At the same time point, the variance of all measured surface temperatures of the batteries to be identified is calculated as a characteristic of temperature distribution uniformity. After charging is completed and the battery has been left to stand for a second preset period of time, the equilibrium temperature difference between the average surface temperature of the battery to be identified and the ambient temperature is calculated. The temperature rise rate, the temperature distribution uniformity, and the equilibrium temperature difference are taken as thermodynamic characteristics.

6. The battery life prediction method in the battery swapping cabinet as described in claim 4, characterized in that, Before the step of inputting the battery health state feature vector into a pre-trained first regression model, and having the first regression model output a value representing the current actual capacity retention rate of the battery as a current health state score, the method further includes: Obtain a historical battery sample set. For each historical battery in the sample set, collect the initial battery health state feature vector corresponding to the detection point at multiple detection points in its life cycle. The true capacity retention rate of the historical battery at this test point is obtained through laboratory standard capacity testing. Use the initial battery health state feature vector and the actual capacity retention rate as a training sample pair; Using all training sample pairs from all historical batteries, with the initial battery health state feature vector as input features and the true capacity retention rate as output labels, train a support vector regression model. During training, grid search and cross-validation methods are used to adjust the core hyperparameters of the support vector regression model; wherein, the core hyperparameters include at least the penalty coefficient C, the kernel function coefficient gamma, and the kernel function type; After training is complete, the model structure and parameters are saved as the first regression model in the pre-training process.

7. The battery life prediction method in the battery swapping cabinet as described in claim 4, characterized in that, The step of calculating the future aging acceleration factor based on the dynamic stress spectrum includes: The normalized dynamic stress spectrum is input into the pre-trained second regression model, which outputs an initial predicted value of the future aging acceleration factor. The second regression model is a feedforward neural network with the same input layer dimension as the dynamic stress spectrum dimension and a single neuron as the output layer. The activation function is ReLU to ensure that the output is non-negative. Extract all instantaneous stress coefficients located within the first rolling window from the historical stress sequence to form the recent stress sequence; Calculate the standard deviation and coefficient of variation of the recent stress sequence; Multiply the coefficient of variation by the preset stability sensitivity coefficient, and substitute the product into the exponential function to obtain the mode stability coefficient; The first value is obtained by multiplying the predicted value of the future aging acceleration factor by the model stability coefficient; Subtracting 1 from the mode stability coefficient yields the second value; The second value is multiplied by a preset baseline acceleration factor to obtain the future aging acceleration factor.

8. The battery life prediction method in the battery swapping cabinet as described in claim 4, characterized in that, The steps for calculating the current health stress index based on the current health status score and future aging acceleration factors include: Obtain the preset health status score threshold at the end of life and the preset high health status threshold; When the current health status score is greater than the high health status threshold, 1 is used as the stress health status coupling coefficient; When the current health status score is between the end-of-life health status score threshold and the high health status threshold, the difference between the high health status threshold and the current health status score is multiplied by the sensitivity coefficient and then incremented by 1 to obtain the stress health status coupling coefficient; where the high health status threshold is greater than the end-of-life health status score threshold. The effective aging acceleration factor is obtained by multiplying the future aging acceleration factor with the stress-health state coupling coefficient. Subtract the current health status score from the health status score threshold at the end of life to obtain the health status margin; Divide the health status margin by the effective aging acceleration factor to obtain the baseline health stress index; The mean value T_stress_avg of the temperature stress coefficient within the second rolling window is extracted from the historical stress sequence that generated the dynamic stress spectrum. The temperature fine-tuning factor is calculated based on the mean value T_stress_avg; where, the temperature fine-tuning factor = 1 - θ*max(0, T_stress_avg - T_ref); where T_ref is the preset reference temperature stress coefficient, θ is the temperature influence coefficient, 0 < θ < 1; The final current health stress index is calculated based on the temperature fine-tuning factor and the baseline health stress index; where the current health stress index = max(HPI_min, baseline health stress index * temperature fine-tuning factor), and HPI_min is the preset minimum health stress index.

9. A battery life prediction device for a battery swapping cabinet, characterized in that, The battery life prediction device in the battery swapping cabinet includes: The data acquisition unit is used to acquire multi-dimensional battery characteristic data when the battery to be identified completes one charge-discharge cycle and is connected to the battery swapping cabinet; wherein, the multi-dimensional battery characteristic data includes at least the depth of discharge, the average discharge rate, and the average ambient temperature during the cycle; The statistical unit is used to statistically analyze the dynamic stress spectrum corresponding to the battery to be identified based on the multidimensional battery feature data; wherein the dynamic stress spectrum is a statistical feature reflecting the usage intensity within different time windows, calculated based on the historical cycle data of the battery to be identified. The calculation unit is used to calculate the remaining life of the battery to be identified based on the dynamic stress spectrum.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps in the battery life prediction method in the battery swapping cabinet as described in any one of claims 1 to 8.

Images