High-safety high-specific-energy hybrid solid-liquid lithium ion battery health early warning test system and method

Abstract

The application discloses a high-safety high-specific-energy hybrid solid-liquid lithium ion battery health early warning test system and method, and belongs to the technical field of lithium ion battery health management. The system comprises six modules of a multi-parameter sensor array, data acquisition and preprocessing, health state evaluation, abnormality detection and fault identification, digital twin and life prediction, and hierarchical early warning and control. The method comprises the following steps: collecting multi-dimensional parameters such as electrochemical impedance spectroscopy, temperature, pressure, acoustic emission and gas concentration in real time; calculating a solid-liquid interface degradation index and performing multi-scale health evaluation; adopting an adaptive threshold and a cross-parameter coupling feature to identify a fault mode; predicting a remaining useful life through a physical information neural network digital twin model; and triggering a single-body level, a module level and a battery pack level three-level early warning according to a predicted value. The application is suitable for hybrid solid-liquid lithium ion batteries of different positive electrode materials and solid electrolyte types, and is effective in scenarios such as electric vehicles, energy storage power stations and aviation power.

Description

Technical Field

[0001] This invention relates to the health management of lithium-ion batteries, and more particularly to a health early warning testing system and method for high-safety, high-specific-energy hybrid solid-liquid lithium-ion batteries. Background Technology

[0002] Lithium-ion batteries are widely used in electric vehicles and energy storage systems due to their advantages such as high energy density, long cycle life, and low self-discharge rate. Pure liquid lithium-ion batteries are limited by the flammability and insufficient thermal stability of organic electrolytes, posing a risk of thermal runaway under conditions of high temperature, overcharging, or mechanical damage. All-solid-state batteries, by replacing liquid electrolytes with solid electrolytes, fundamentally improve safety, but face technical bottlenecks such as high solid-solid interface contact resistance, low room-temperature ionic conductivity, and complex manufacturing processes.

[0003] Hybrid solid-liquid lithium-ion batteries employ a dual-electrolyte system combining a solid electrolyte layer and a liquid electrolyte layer. This approach balances high energy density with safety, making it a crucial technology in the current power battery field. In these batteries, the solid electrolyte layer functions as a separator and suppresses lithium dendrite formation, while the liquid electrolyte layer maintains good interfacial wettability between the electrodes and the electrolyte. However, the introduction of the dual-electrolyte system complicates the battery's aging mechanism. The stability of the solid-liquid interface, the rate of liquid component consumption, and the mechanical degradation behavior of the solid electrolyte all significantly impact the battery's health.

[0004] Existing battery health management systems are mostly designed for pure liquid or pure solid-state batteries, making it difficult to adapt to the unique aging characteristics of hybrid solid-liquid systems. For example, Chinese invention patent application CN121748586A discloses an energy storage management and safety protection method for hybrid solid-liquid electrolyte energy storage batteries. It generates interface health and thermal risk indices by fusing multi-dimensional data, constructing a safe operating window to achieve adaptive safety boundary control. However, this method focuses on system-level charge and discharge management control, lacking in-depth characterization of the solid-liquid interface degradation mechanism and single-cell-level fault diagnosis capabilities. Chinese invention patent application CN119667492A discloses a model-based battery thermal runaway early warning system and method. It establishes an electrochemical-thermal and electrochemical-expansion force coupling model to determine multi-level safety thresholds, achieving multi-level early warning of thermal runaway. However, this scheme is designed for liquid batteries and does not consider the impact of solid electrolyte layer mechanical performance degradation on the safety boundary. Chinese invention patent application CN122238920A discloses an intelligent detection system and method for the health status of lithium batteries. It uses recursive Bayesian estimation to filter and track the peak position of the incremental capacity analysis curve and combines it with innovation anomaly detection to achieve micro-short circuit early warning. However, this method does not distinguish the differences in degradation characteristics of different types of battery systems.

[0005] Therefore, there is an urgent need for a health early warning test system and method specifically designed for hybrid solid-liquid lithium-ion batteries, which can characterize the degradation features of the solid-liquid interface, realize multi-scale health assessment and fault identification, and accurately predict the remaining service life. Summary of the Invention

[0006] Purpose of the invention: The purpose of this invention is to provide a health early warning test system and method for high-safety, high-specific-energy hybrid solid-liquid lithium-ion batteries. Through multi-parameter collaborative sensing, multi-scale health assessment, cross-parameter coupled fault identification, and physical information neural network digital twin, the system can achieve health status monitoring and graded early warning of hybrid solid-liquid lithium-ion batteries throughout their entire life cycle.

[0007] Technical Solution: A health early warning test method based on a high-safety, high-specific-energy hybrid solid-liquid lithium-ion battery, applied to a health early warning test system comprising a multi-parameter sensor array and a temperature-controlled test chamber, wherein the hybrid solid-liquid lithium-ion battery adopts a dual-electrolyte system combining a solid electrolyte layer and a liquid electrolyte layer, comprising the following steps:

[0008] Step one involves real-time acquisition of multi-dimensional parameter data of the battery during the charge-discharge cycle of the hybrid solid-liquid lithium-ion battery using a multi-parameter sensor array. The multi-parameter sensor array includes: an electrochemical impedance spectroscopy module for online measurement of electrochemical impedance spectroscopy with an amplitude of 10mV in the frequency range of 0.1Hz to 10kHz; a pressure sensor with a range of 0 to 100kPa and a resolution of 0.1kPa, used to monitor changes in the expansion force on the battery casing surface; an acoustic emission sensor that acquires acoustic emission signals in the frequency band of 20kHz to 500kHz with a sampling rate of 2MHz; a temperature sensor array consisting of multiple thermocouples distributed on the battery surface for measuring surface temperature distribution; and a gas sensor that detects the concentrations of CO, CO2, and H2 gases with a detection limit of 10ppm.

[0009] Step two involves a multi-scale health status assessment based on electrochemical impedance spectroscopy (EIS) data. The ohmic impedance Rs (real intercept in the high-frequency region), charge transfer impedance Rct (half-circular fitting in the mid-frequency region), and Warburg diffusion impedance Zw (45° slope segment in the low-frequency region) characteristics are extracted from the impedance spectrum. The solid-liquid interface degradation index DI = Rct / Rct0 - Rs / Rs0 is calculated, where Rct0 and Rs0 are the initial values ​​of the charge transfer impedance and ohmic impedance of the new battery, respectively. This index characterizes the relative increase in the solid-liquid interface contact resistance and can distinguish between two different degradation mechanisms: solid electrolyte layer degradation (mainly contributed by Rs) and decreased solid-liquid interface wettability (mainly contributed by Rct). Simultaneously, peak identification is performed on the incremental capacity curve (IC curve) during the constant current charging stage, and the peak position offset ΔV_peak and peak height decay rate ΔIC_peak / IC_peak0 are extracted. The solid-liquid degradation index DI, ΔV_peak, and ΔIC_peak / IC_peak0 are fused to construct a health feature vector, which is then input into a health status assessment model based on a long short-term memory network. The output includes a multi-dimensional health assessment result that includes capacity retention rate SOH_cap, internal resistance growth rate RIR, and power degradation rate PDR.

[0010] Step 3: Perform individual-level anomaly signal detection on the multi-dimensional parameter data. A baseline state vector B=[μ,σ,dr / dt] is constructed based on the statistical characteristics of the multi-dimensional parameters over the most recent N sampling periods, where μ is the mean, σ is the standard deviation, and dr / dt is the rate of change. The baseline state vector is dynamically updated using an exponentially weighted moving average method: B(t+1)=α·B_raw(t)+(1-α)·B(t), where α is the smoothing coefficient, ranging from 0.1 to 0.3. An adaptive threshold is set as TH=μ+k·σ, where k is the confidence coefficient, ranging from 2 to 3. Any parameter exceeding its adaptive threshold is considered an anomaly signal.

[0011] Multi-parameter fusion analysis was performed on the abnormal signals to construct a weighted feature matrix W = Σw_i·F_i, where the weights w_i were determined using the analytic hierarchy process (AHP) based on the sensitivity and reliability of each parameter to different fault modes. Principal component analysis was used to extract cross-parameter coupling features, which were then input into a trained random forest classifier (200 decision trees) to output the probability distribution of each fault mode category. The fault modes included: solid-liquid interface degradation (characterized by abnormal Rct growth and increased high-frequency components in the acoustic emission signal), lithium dendrite growth (characterized by sudden pressure changes and abnormal imaginary part of impedance), internal short circuit (characterized by abnormal self-discharge rate and localized temperature rise), and gas production expansion (characterized by increased gas concentration and shell expansion).

[0012] Step 4: Input the multi-dimensional health assessment results and fault identification results into a pre-constructed battery digital twin model. The model describes the internal physical and chemical processes of the battery with an electrochemistry-thermal-mechanics coupled partial differential equation system, which serves as the physical constraint term. The equation system includes: the Butler-Volmer kinetic equation describing the electrode reaction rate, the heat conduction equation describing the temperature distribution, and the stress equilibrium equation describing the mechanical behavior of the solid electrolyte interphase. A data-driven neural network model is constructed by taking historical multi-dimensional parameter data and their corresponding life labels as training data. The physical constraint term is embedded into the loss function of the neural network to form a hybrid driving model of Physics-Informed Neural Network (PINN), where the loss function is L=L_data+λ·L_physics, λ is the weight coefficient of physical constraint, with a value ranging from 0.01 to 0.1. The model outputs the predicted value of remaining useful life (RUL).

[0013] Step 5: Generate a corresponding level of early warning signal according to the predicted value of remaining useful life. The hierarchical early warning includes three levels: Level 1 early warning (corresponding to the single cell level), which is triggered when RUL<T_cell, and a single cell level power derating operation strategy is implemented: the charging current is reduced to 50% of the rated value, and temperature monitoring is strengthened; Level 2 early warning (corresponding to the module level), which is triggered when RUL<T_module, and a module level isolation strategy is implemented: the early warning cell is bypassed from the module circuit and local heat dissipation is started; Level 3 early warning (corresponding to the battery pack level), which is triggered when RUL<T_pack, and a battery pack level power-off protection strategy is implemented: the main circuit of the battery pack is disconnected and a fire preparation procedure is started. Where T_cell>T_module>T_pack, and the typical values are 30 days, 15 days and 7 days respectively.

[0014] The present invention further provides a health early warning test system based on high-safety high-specific-energy hybrid solid-liquid lithium-ion batteries, comprising: a multi-parameter sensor array module installed on the inner wall of a temperature-controlled test cavity, configured to collect multi-dimensional parameter data of the battery in real time; a data acquisition and preprocessing module configured to perform filtering, denoising and standardization processing on the raw data collected by the multi-parameter sensor array; a health state assessment module configured to perform the multi-scale health state assessment in Step 2; an anomaly detection and fault identification module configured to perform the anomaly signal detection and fault mode identification in Step 3; a digital twin and life prediction module configured to perform the remaining useful life prediction in Step 4; and a hierarchical early warning and control module configured to perform the hierarchical early warning in Step 5 and output control instructions.

[0015] Beneficial effects:

[0016] (1) For the dual electrolyte system unique to hybrid solid-liquid lithium-ion batteries, a solid-liquid interface degradation index (DI) is proposed, which can quantitatively distinguish between two degradation mechanisms: solid electrolyte degradation and decreased wettability of the solid-liquid interface, filling the gap in the characterization ability of existing health assessment methods for hybrid systems.

[0017] (2) Through multi-parameter collaborative sensing and cross-parameter coupling feature extraction, the accurate identification of various fault modes of hybrid solid-liquid batteries is realized. Compared with single-parameter or simple threshold judgment methods, the fault identification accuracy is higher and the false alarm rate is lower.

[0018] (3) A digital twin model is constructed using a physical information neural network. The physical constraints of the electrochemical-thermal-mechanical coupling equations are embedded into the loss function of the data-driven model. This ensures that the model follows the physical and chemical processes of the battery and leverages the fitting ability of deep learning to complex nonlinear relationships, resulting in higher accuracy in predicting the remaining service life.

[0019] (4) The three-level graded early warning mechanism is adapted to the safety management level of hybrid solid-liquid batteries from single cells to modules to packs. Each level corresponds to a different response strategy, which can take targeted measures at different deterioration stages to avoid over-response or under-response. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the overall process of the health early warning testing method of the present invention;

[0021] Figure 2 This is a schematic diagram of the health early warning testing system of the present invention;

[0022] Figure 3 This is a schematic diagram of the digital twin model construction and lifetime prediction process of the present invention. Detailed Implementation

[0023] To make the technical solution of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0024] Example 1:

[0025] like Figure 1 and Figure 2 As shown, this embodiment is applied to an NCM811 positive electrode / silicon-carbon negative electrode hybrid solid-liquid lithium-ion battery (rated capacity 100Ah, nominal voltage 3.7V). The solid electrolyte layer is a Li6PS5Cl sulfide solid electrolyte with a thickness of 30μm; the liquid electrolyte layer is a 1mol / L LiPF6 EC / DMC / EMC (volume ratio 1:1:1) solution, accounting for 15% of the total electrolyte volume. The battery adopts a stacked soft-pack structure.

[0026] Step 1: Install the battery in a temperature-controlled test chamber, with the chamber temperature set to 25℃±1℃. The multi-parameter sensor array includes: an electrochemical impedance spectroscopy module (frequency range 0.1Hz~10kHz, amplitude 10mV), a pressure sensor (range 0~50kPa, resolution 0.1kPa), an acoustic emission sensor (center frequency 150kHz, sampling rate 2MHz), four K-type thermocouples (distributed on the four corners of the battery surface), and a gas sensor (detecting H2 and CO, lower limit 10ppm). Cyclic testing is performed using a 1C constant current constant voltage charging and 1C constant current discharging regime, with a data acquisition cycle of 10 seconds.

[0027] Step 2: After every 10 charge-discharge cycles, allow the device to rest for 10 minutes after charging to 50% SOC, and then perform online electrochemical impedance spectroscopy (EIS) measurements. Extract the ohmic impedance Rs (real intercept at 1 kHz), charge transfer impedance Rct (mid-frequency semi-circular diameter), and Warburg diffusion impedance Zw (low-frequency slope). Using initial values ​​Rs0 = 8 mΩ and Rct0 = 25 mΩ as a baseline, calculate the solid-liquid interface degradation index DI = Rct / Rct0 - Rs / Rs0. When the DI value reaches 1.5, significant solid-liquid interface degradation is considered to have begun. Simultaneously, peak identification is performed on the IC curve during the charging phase, and the offset ΔV_peak and peak height attenuation rate ΔIC_peak / IC_peak0 at the third peak position (corresponding to approximately 3.6 V) are extracted. The DI, ΔV_peak, and ΔIC_peak / IC_peak0 are fused into a three-dimensional health feature vector, which is then input into the LSTM health status assessment model (input layer dimension 3, hidden layer dimension 64, output layer dimension 3). The outputs are capacity retention rate SOH_cap, internal resistance growth rate RIR, and power degradation rate PDR.

[0028] Step 3: Construct a baseline state vector using pressure, temperature, and acoustic emission data from the most recent 50 sampling periods (500 seconds). An exponentially weighted moving average (α=0.15) is used to dynamically update the baseline, with a confidence coefficient k=2.5. After 320 cycles, the pressure signal exceeds the threshold, and the proportion of high-frequency components (>200kHz) in acoustic emission energy increases from 12% of the baseline to 28%, indicating an abnormal signal. The weighted feature matrix is ​​analyzed using principal component analysis to extract the first three principal components (cumulative variance explained 92.3%), which are then input into a random forest classifier. The output fault mode probability distribution is: solid-liquid interface degradation 68%, lithium dendrite growth 22%, internal short circuit 3%, and gas production expansion 7%. The primary fault mode is determined to be solid-liquid interface degradation.

[0029] Step 4: Input the health assessment result and the fault identification result into the digital twin model. In the electrochemistry-thermal-mechanics coupling model, the Butler-Volmer equation describes the electrode reaction, the heat conduction equation adopts a two-dimensional axisymmetric model, and the stress balance equation takes into account the expansion stress of the solid electrolyte layer. The PINN model adopts a 4-layer fully connected network (128 neurons in each layer, Tanh activation function), and the physical constraint weight λ=0.05. The training data is obtained from full life cycle test data of 50 batteries of the same batch. The model predicts that the remaining useful life of the battery is 28 days.

[0030] Step 5: Set T_cell=30 days, T_module=15 days, T_pack=7 days. The current RUL=28 days<T_cell=30 days, so a first-level early warning is triggered, and a single-cell level power derating operation strategy is implemented: the charging current is reduced to 50A (50% of the rated 100A), and temperature data is collected every 5 minutes. When the maximum temperature exceeds 45°C, the current is further reduced to 30A.

[0031] The test results of Example 1 are shown in the following table:

[0032]

[0033] Example 2:

[0034] This example is applied to the scenario of an energy storage power station, and the battery is a hybrid solid-liquid lithium-ion battery with LFP positive electrode / graphite negative electrode (rated capacity 280Ah, nominal voltage 3.2V). The solid electrolyte layer is LLZTO (Li7La3Zr2O 12 Ta-doped) oxide solid electrolyte with a thickness of 50μm; the liquid electrolyte layer is a DOL / DME (volume ratio 1:1) solution of 1.2mol / L LiTFSI, and the volume accounts for 10% of the total volume of the electrolyte. The battery adopts a prismatic aluminum case structure, and 12 single cells are connected in series to form a module.

[0035] The operating condition of the energy storage power station is 1.5 complete charge-discharge cycles per day (0.3C charging, 0.5C discharging), and the ambient temperature fluctuates from 15°C to 35°C. The temperature-controlled test chamber is set at 30°C±2°C.

[0036] In Step 1, since the energy storage scenario has extremely high requirements for safety, the multi-parameter sensor array adds redundant configuration: two sets of electrochemical impedance measurement modules are configured as mutual backups, the measuring range of the pressure sensor is adjusted to 0 to 80kPa, and a CO₂ detection channel is added to the gas sensor. The data acquisition period is 30 seconds.

[0037] In step 2, the impedance characteristics of the oxide solid electrolyte system are different from those of the sulfide system: Rs0=15mΩ (higher than that of the sulfide system), Rct0=18mΩ (lower than that of the sulfide system). A temperature sensitivity coefficient term is added to the health feature vector, that is, the change rate of DI value at different temperatures dDI / dT, which is used to compensate for the influence of temperature fluctuations in energy storage scenarios on impedance characteristics. The input dimension of the LSTM model is adjusted to 4.

[0038] In step 3, since the module is formed by connecting 12 cells in series, the construction of the baseline state vector takes the parameter differences of each cell in the module as a reference. When the DI value of a cell deviates from the average value of the module by more than 2 times the standard deviation, an abnormal mark is triggered. After 480 cycles of operation, the pressure change rate dr / dt of the No. 7 cell exceeds the baseline by 2.8 times the standard deviation, and the H2 gas concentration rises from the baseline <10ppm to 35ppm, which is determined as an abnormal signal. The weighted feature matrix extracts the first 4 principal components through PCA (cumulative variance interpretation rate is 94.1%), and the random forest classifier outputs the failure mode probabilities: gas generation and expansion 72%, solid-liquid interface degradation 18%, lithium dendrite growth 5%, internal short circuit 5%, and the main failure mode is determined as gas generation and expansion.

[0039] In step 4, the training data of the PINN model comes from full life cycle test data of 80 same-batch batteries under different temperature conditions. Due to large temperature fluctuations in energy storage scenarios, the temperature input dimension of the model is expanded to a two-dimensional input including ambient temperature and battery surface temperature. The model predicts that the remaining useful life of the No. 7 cell is 21 days.

[0040] In step 5, RUL=21 days of the No. 7 cell <T_cell=30 days, triggering a first-level early warning. After receiving the early warning signal, the energy storage management system limits the charge and discharge current of this cell to 40% of the rated value, and arranges module-level inspection in the next maintenance window.

[0041] Example 3:

[0042] This example is applied to the power battery scenario of electric aircraft. The battery is a NCM622 positive electrode / lithium metal negative electrode hybrid solid-liquid lithium-ion battery (rated capacity 45Ah, nominal voltage 3.6V). The solid electrolyte layer is LATP (Li 1.3 Al 0.3 Ti 1.7 (PO4)3) oxide solid electrolyte, with a thickness of 20μm; the liquid electrolyte layer is a FEC / DEC (volume ratio 3:7) solution of 0.8mol / L LiFSI, and the volume accounts for 20% of the total volume of the electrolyte. The battery adopts a cylindrical full-shell structure.

[0043] Special operating conditions are required for aviation power batteries: high charge / discharge rates (1C charging, 2C to 3C discharging), wide temperature range (-10℃ to 50℃), and complex vibration environment. A vibration simulation platform is added to the test chamber, with a vibration frequency range of 5Hz to 2000Hz and an acceleration range of 0.5g to 3g.

[0044] In step one, due to the weight sensitivity of aviation batteries, the sensor array employs a miniaturized configuration: a MEMS pressure sensor (range 0~60kPa), a thin-film thermocouple (response time <50ms), and a miniature electrochemical impedance chip (frequency range 100Hz~5kHz). The acoustic emission sensor utilizes a fiber Bragg grating (FBG) sensor, combining acoustic emission detection and temperature measurement functions to meet the electromagnetic compatibility requirements of the aviation environment. The data acquisition cycle is 5 seconds.

[0045] In step two, the IC curve characteristics change rapidly under high-rate conditions. IC curve measurements are performed every 20 cycles under low-rate calibration at 0.5C. During high-rate (2C / 3C) operation, impedance characteristics are the primary health indicator. The weight of the solid-liquid interface degradation index (DI) increases to 1.5 times under high-rate conditions because concentration polarization at the solid-liquid interface is more significant at high rates, and changes in the DI value are a stronger indicator of health status.

[0046] In step three, vibration and noise under aviation operating conditions significantly impact the signal quality of the pressure sensor and acoustic emission sensor. To address this, a signal preprocessing step is added before anomaly signal detection: wavelet packet decomposition (db4 basis function, 5-level decomposition) is used to filter out vibration and noise, retaining low-frequency components (<1kHz) related to the battery's internal state. The smoothing coefficient α of the baseline state vector is adjusted to 0.2 to adapt to high-dynamic conditions. The confidence coefficient k is increased to 2.8 to reduce the false alarm rate.

[0047] In step four, the PINN model adds a mechanical vibration load term to the physical constraints to describe the effect of vibration on the propagation of microcracks in the solid electrolyte layer: crack propagation rate da / dN = C·(ΔK)^m, where C and m are material constants, and ΔK is the stress intensity factor amplitude. The training data includes full lifecycle data for both static and vibration conditions. The model predicts that the remaining lifespan of the battery is 12 days under continuous 3C discharge conditions.

[0048] In step 5, the safety threshold of aviation batteries is more stringent: T_cell=20 days, T_module=10 days, T_pack=5 days. Currently, RUL=12 days < T_cell=20 days, which triggers a first-level early warning, and a power derating operation strategy is implemented: the maximum discharge rate is limited to 1.5C. When RUL further drops to 10 days, a second-level early warning is triggered, and the system recommends replacing the module after the current flight mission. When RUL drops to 5 days, a third-level early warning is triggered, and the system forcibly prohibits the continued use of the battery pack.

[0049] The above three embodiments show that the health early warning test method proposed by the present invention can be adapted to hybrid solid-liquid lithium-ion batteries with different cathode material systems (NCM811, LFP, NCM622), different solid electrolyte types (sulfide, oxide) and different application scenarios (electric vehicles, energy storage power stations, aviation power), and achieve targeted health monitoring and hierarchical early warning through parameter adjustment and module configuration.

[0050] The above described embodiments only illustrate several embodiments of the present invention. The description thereof is relatively specific and detailed, but it should not be construed as a limitation on the scope of the patent of the present invention. It should be noted that, for those skilled in the art, without departing from the concept of the present invention, several modifications and improvements can be made, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the appended claims.

Claims

1. A health warning test method based on a high-safety, high-specific-energy hybrid solid-liquid lithium-ion battery, applied to a health warning test system including a multi-parameter sensor array and a temperature-controlled test chamber, wherein the hybrid solid-liquid lithium-ion battery adopts a dual-electrolyte system composed of a solid electrolyte layer and a liquid electrolyte layer, characterized in that... Includes the following steps: S1. During the charge-discharge cycle of the hybrid solid-liquid lithium-ion battery, the multi-dimensional parameter data of the battery are collected in real time through the multi-parameter sensor array. The multi-dimensional parameter data includes electrochemical impedance spectroscopy data, temperature data, pressure data, acoustic emission signal data, and gas concentration data. S2. Based on the electrochemical impedance spectroscopy data, perform multi-scale health status assessment, specifically: extract ohmic impedance, charge transfer impedance and diffusion impedance features from the impedance spectrum, calculate the solid-liquid interface degradation index; fuse the solid-liquid degradation index with the incremental capacity curve features to construct a health feature vector, input it into the health status assessment model, and output a multi-dimensional health assessment result including capacity retention rate, internal resistance growth rate and power decay rate. S3. Perform individual-level abnormal signal detection on the multi-dimensional parameter data, and use an adaptive threshold detection method to determine the abnormal signal, wherein the adaptive threshold is dynamically updated based on the baseline features of the multi-dimensional parameters. The abnormal signal is subjected to multi-parameter fusion analysis. By constructing a weighted feature matrix and extracting cross-parameter coupling features, the coupling features are mapped to the fault mode category. The fault mode includes at least one of solid-liquid interface degradation, lithium dendrite growth, internal short circuit and gas production expansion. The fault identification result is output. S4. Input the multi-dimensional health assessment results and fault identification results into the pre-built battery digital twin model. The digital twin model is constructed by fusing an electrochemical-thermal-mechanical coupling model with a physical information neural network to predict the remaining service life of the battery and output the predicted value of the remaining service life. S5. When the predicted remaining service life is lower than the preset graded warning threshold, a warning signal of the corresponding level is generated and output. The graded warning includes three levels of warning corresponding to the single cell level, module level and battery pack level. Each level of warning corresponds to a different response strategy.

2. The health early warning test method for high-safety, high-specific-energy hybrid solid-liquid lithium-ion batteries according to claim 1, characterized in that, The multi-parameter sensor array includes: an electrochemical impedance spectroscopy module for measuring electrochemical impedance spectroscopy in the frequency range of 0.1 Hz to 10 kHz; a pressure sensor for monitoring changes in the expansion force on the surface of the battery casing, with a range of 0 to 100 kPa and a resolution of 0.1 kPa; an acoustic emission sensor for acquiring acoustic emission signals in the frequency band of 20 kHz to 500 kHz; a temperature sensor array for measuring the temperature distribution on the battery surface at multiple points; and a gas sensor for detecting the concentrations of CO, CO2, and H2 gases.

3. The health early warning test method for high-safety, high-specific-energy hybrid solid-liquid lithium-ion batteries according to claim 1, characterized in that, The method for calculating the solid-liquid interface degradation index in step S2 is as follows: the charge transfer impedance Rct is obtained from the semi-circular arc fitting in the mid-frequency region of the electrochemical impedance spectrum, and the ohmic impedance Rs is obtained from the intercept in the high-frequency region. The solid-liquid interface degradation index DI = Rct / Rct0 - Rs / Rs0 is then calculated, where Rct0 and Rs0 are the initial values ​​of the charge transfer impedance and ohmic impedance of the new battery, respectively. The incremental capacity curve feature is obtained by identifying the peak value of the incremental capacity curve during the constant current charging stage and extracting the peak position offset and peak height decay rate.

4. The health early warning test method for high-safety, high-specific-energy hybrid solid-liquid lithium-ion batteries according to claim 1, characterized in that, The adaptive threshold update method in step S3 is as follows: a baseline state vector is constructed using the statistical characteristics of multi-dimensional parameters in the most recent N sampling periods, including the mean, standard deviation, and rate of change; the baseline state vector is dynamically updated using an exponentially weighted moving average method. The adaptive threshold is set as the sum of the mean of the baseline state vector and k times the standard deviation, where k is the confidence coefficient, with a value ranging from 2 to 3.

5. The health early warning test method for high-safety, high-specific-energy hybrid solid-liquid lithium-ion batteries according to claim 1, characterized in that, The specific process of multi-parameter fusion analysis in step S3 is as follows: construct a weighted feature matrix for the abnormal signal, where the weights are allocated according to the sensitivity and reliability of each parameter to the fault mode; extract cross-parameter coupling features through principal component analysis, where the coupling features characterize the correlation changes between different parameters; input the coupling features into a random forest classifier, and output the probability distribution of each fault mode category.

6. The health early warning test method for high-safety, high-specific-energy hybrid solid-liquid lithium-ion batteries according to claim 1, characterized in that, The digital twin model in step S4 is constructed as follows: the internal physicochemical processes of the battery are described by a set of electrochemical-thermal-mechanical coupled partial differential equations as physical constraints; historical multi-dimensional parameter data and their corresponding lifetime labels are used as training data to construct a data-driven neural network model; the physical constraints are embedded into the loss function of the neural network to form a hybrid driven model of physical information neural network, with the loss function being L=L_data+λ·L_physics, where λ is the physical constraint weight coefficient.

7. The health early warning test method for high-safety, high-specific-energy hybrid solid-liquid lithium-ion batteries according to claim 1, characterized in that, The specific judgment rules for the graded early warning in step five are as follows: when the predicted remaining service life is lower than the single cell early warning threshold T_cell, a first-level early warning is triggered and a single cell-level power reduction operation strategy is executed; when the predicted remaining service life is lower than the module early warning threshold T_module, a second-level early warning is triggered and a module-level isolation strategy is executed; when the predicted remaining service life is lower than the battery pack early warning threshold T_pack, a third-level early warning is triggered and a battery pack-level power failure protection strategy is executed; where T_cell > T_module > T_pack.

8. The health early warning test method for high-safety, high-specific-energy hybrid solid-liquid lithium-ion batteries according to claim 1, characterized in that, The solid electrolyte layer of the hybrid solid-liquid lithium-ion battery is at least one of sulfide solid electrolyte, oxide solid electrolyte or polymer solid electrolyte, with a thickness of 10~100μm; the liquid electrolyte layer is an organic solvent electrolyte containing lithium salt, accounting for 5%~30% of the total volume of the electrolyte inside the battery.

9. The health early warning test method for high-safety, high-specific-energy hybrid solid-liquid lithium-ion batteries according to claim 1, characterized in that, The method further includes: inputting the multi-dimensional health assessment results and fault identification results into the maintenance strategy decision module, and determining a maintenance plan based on the fault mode type and the remaining service life prediction value. The maintenance plan includes at least one of continued monitoring, reduced power operation, module replacement, or battery pack retirement.

Images