Sodium-ion battery cycle life and soc-soh collaborative estimation method

Abstract

The application discloses a sodium-ion battery cycle life and SOC-SOH cooperative estimation method, to solve the problem that the existing SOC and SOH estimation is difficult to accurately depict the two-stage nonlinear characteristics of sodium-ion battery capacity attenuation, and the estimation error is large. The method is based on the battery state of charge and the battery state of health, a double-time-scale state space model is constructed, and the battery terminal voltage is observed; a state vector of multi-state joint observation and an observation vector based on three channels are defined; based on the multi-innovation adaptive unscented Kalman filtering algorithm, the observation vector is observed in three channels, and the covariance is adaptively adjusted through the three-channel innovation, the state vector is updated under the double-time-scale, and the cooperative estimation value of the battery state of charge and the battery state of health is obtained. The method realizes the cooperative update and rapid correction of "second-level SOC-periodic SOH", and improves the estimation accuracy and response speed.

Description

Technical Field

[0001] This application relates to the field of secondary battery state estimation, and in particular to a method for synergistic estimation of cycle life and SOC-SOH of sodium-ion batteries. Background Technology

[0002] Sodium-ion batteries have become an important alternative system for large-scale energy storage due to their abundant resources and good low-temperature performance. However, the cathode materials of sodium-ion batteries suffer from lattice slip and transition metal dissolution during cycling, causing the battery's capacity decay path to exhibit a "fast-slow" two-stage nonlinear characteristic. This makes it difficult to estimate the battery's state of charge (SOC) and state of health (SOH).

[0003] Existing SOC estimations are mostly based on coulomb counting or a single Kalman filter, without considering the rated capacity drift caused by SOH decay, resulting in large errors under high-rate energy storage-charging switching conditions. Furthermore, existing SOH estimations mostly use offline capacity testing or simple impedance testing, which cannot form a closed loop with online SOC algorithms and do not utilize the long-cycle data of "full charge-deep discharge" inherent in integrated systems, resulting in long testing cycles and high costs. Summary of the Invention

[0004] This application provides a method for co-estimating the cycle life and SOC-SOH of sodium-ion batteries, which solves the problem that existing SOC and SOH estimations are difficult to accurately characterize the two-stage nonlinear characteristics of sodium-ion battery capacity decay and have large estimation errors when switching under high-rate conditions.

[0005] The method for co-estimating the cycle life and SOC-SOH of sodium-ion batteries provided in this application includes: Based on the battery state of charge and battery health state, a dual-time-scale state-space model is constructed to observe the battery terminal voltage; Define a state vector for multi-state joint observation and an observation vector based on three channels; wherein, the multi-state includes battery state of charge, battery health state, DC internal resistance, first polarization voltage and second polarization voltage, and the three channels include the terminal voltage, cumulative ampere-hours and constant current charging time under set conditions; Based on the multi-innovation adaptive unscented Kalman filter algorithm, the observation vector is observed in three channels, and the covariance is adaptively adjusted through the three-channel innovation to update the state vector under two time scales, thereby obtaining a joint estimate of the battery state of charge and battery health.

[0006] In one example, the dual-timescale state-space model includes a slow-timescale model and a fast-timescale model; The slow timescale model describes the degradation of battery health status through a double exponential term of cycle life, degradation coefficient, and temperature. The fast timescale model describes the changes in the battery's state of charge through battery health, rated capacity, coulombic efficiency, current, and process noise.

[0007] In one example, the slow time-scale model is represented as ,in, This indicates the battery health status in the k-th cycle. and Indicates the attenuation coefficient. and T represents cycle life, and T represents temperature. The fast timescale model is expressed as: ,in, This represents the state of charge of the battery at time t. Indicates Coulomb efficiency. Represents current. Indicates the rated capacity. This indicates process noise.

[0008] In one example, constructing a dual-timescale state-space model to observe the battery terminal voltage includes: By joint observation variables The battery terminal voltage, coupled with the battery's state of charge and state of health, is observed; among them, Indicates terminal voltage. The open-circuit voltage that represents the coupling between the battery's state of charge and its state of health. Represents current. Indicates DC internal resistance. and These represent the first polarization voltage and the second polarization voltage, respectively.

[0009] In one example, the multi-innovation adaptive unscented Kalman filter algorithm performs three-channel observations on the observation vector and adaptively adjusts the covariance through three-channel innovation to update the state vector at two time scales, obtaining a joint estimate of the battery state of charge and battery health state, including: Initialize the state vector; Using a preset scaling factor, Sigma points are generated through unscented transformation UT to update the battery state of charge and polarization voltage, thus obtaining preliminary state prediction values. Perform three-channel observation and prediction, and calculate the predicted observation values; Calculate the three-channel innovation sequence based on the actual observations and the predicted observations; Based on the three-channel information sequence, the observation noise covariance is adaptively adjusted, and the preliminary state prediction value is corrected and updated to obtain the corrected state prediction value, which is used as an estimate of the battery state of charge. Based on the estimated value of the battery's state of charge and the dual-time-scale state-space model, the estimated value of the battery's state of health is determined.

[0010] In one example, the three-channel observation prediction includes: pass Terminal voltage observations were conducted; among them, The open-circuit voltage that represents the coupling between the battery's state of charge and its state of health. Represents current. Indicates DC internal resistance. and These represent the first polarization voltage and the second polarization voltage, respectively. pass Hourly observations were performed, and the results were obtained after discretization. ;in, This represents the state of charge of the battery at time t. Indicates the rated capacity. This indicates the battery health status at time t; During the constant current charging phase, under the set condition of reducing the state of charge (SOC) from 20% to a terminal voltage of 3.6V, through... Constant current charging time was observed, and simplified to 0.5C constant current. ,in, Indicates the constant current charging time. Indicates constant current charging current. This represents the change in polarization voltage. Represents the polarization coefficient. This indicates the DC internal resistance.

[0011] In one example, the adaptive adjustment of the observation noise covariance based on the three-channel information sequence includes: Based on the three-channel information sequence and chi-square test, determine whether there is zero drift of the current sensor according to the values ​​of the ampere-hour information and the terminal voltage information; If so, adaptively adjust the observation noise covariance to modify the weights of the three channels.

[0012] In one example, prior to obtaining the joint estimate of the battery state of charge and battery state of health, the method further includes: Extract health factors related to battery health status; A Gaussian process regression (GPR) model was trained, using RBF and white noise kernel functions, and online hyperparameter optimization was performed to obtain the predicted battery health status. When the preset correction conditions are met, the battery health state estimate obtained by the multi-innovation adaptive unscented Kalman filter algorithm is replaced with the observation results of the GPR model to achieve battery health state correction.

[0013] In one example, after obtaining the joint estimate of the battery state of charge and battery state of health, the method further includes: Based on the collaborative estimation value, when it is determined that the battery meets the preset lifespan conditions, the charge / discharge rate is reduced and a lifespan termination warning is issued. The preset lifespan conditions include at least one of the following: the battery health status drops to 80%, and the single-cell voltage variance is greater than 150mV.

[0014] The sodium-ion battery cycle life and SOC-SOH co-estimation system provided in this application embodiment operates based on any of the sodium-ion battery cycle life and SOC-SOH co-estimation methods described above. The system includes: The module is used to build a dual-time-scale state-space model based on the battery's state of charge and state of health, and to observe the battery's terminal voltage. A definition module is used to define the state vector of multi-state joint observation and the observation vector based on three channels; wherein, the multi-state includes battery state of charge, battery health state, DC internal resistance, first polarization voltage and second polarization voltage, and the three channels include the terminal voltage, cumulative ampere-hours and constant current charging time under set conditions; The observation module is used to perform three-channel observation of the observation vector based on the multi-innovation adaptive unscented Kalman filter algorithm, and to update the state vector under dual time scales by adaptively adjusting the covariance through the three-channel innovation, thereby obtaining a joint estimate of the battery state of charge and battery health state.

[0015] This application provides a method for co-estimating the cycle life and SOC-SOH of sodium-ion batteries, which can achieve the following beneficial effects: To address the two-stage degradation characteristics of sodium-ion batteries, this paper considers the impact of battery health state degradation on rated capacity inaccuracies and subsequent capacity drift, which in turn affects battery state of charge (SOC). A coupled model of SOC and SOC is then implemented. An innovative adaptive unscented Kalman filter algorithm is proposed, extending the state vector to include SOC and DC internal resistance in the online-updated state vector. Furthermore, by introducing constant-current charging time into the observation vector, real-time correction of SOC and DC internal resistance is achieved, reducing estimation errors and improving accuracy. This enables online joint estimation of SOC and SOC. Additionally, a three-channel innovation approach with adaptive covariance matching effectively reduces estimation errors caused by sensor zero drift, ensuring the stability and accuracy of the joint estimation.

[0016] Furthermore, by organically combining multiple innovative adaptive unscented Kalman filtering algorithms with GPR, the collaborative updating and rapid correction of "second-level SOC-cycle-level SOH" are achieved, improving estimation accuracy and response speed. Cyclic life testing, state estimation, and life warning are completed simultaneously without additional downtime testing, realizing "evaluation while running". This significantly reduces testing costs and time, and significantly improves estimation accuracy and shortens the testing cycle. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. The accompanying drawings described herein are used to provide a further understanding of this application and constitute a part of this application. The illustrative embodiments of this application and their descriptions are used to explain this application and do not constitute an improper limitation of this application. In the accompanying drawings: Figure 1 This is a schematic diagram of the topology of the integrated operating condition cycle life test platform provided in the embodiments of this application; Figure 2 This is a flowchart of the method for co-estimating the cycle life and SOC-SOH of sodium-ion batteries provided in the embodiments of this application; Figure 3 This is a schematic diagram of a dual-timescale state-space model provided in an embodiment of this application; Figure 4 A flowchart of the multi-innovative adaptive unscented Kalman filter algorithm provided in the embodiments of this application; Figure 5 This is a schematic diagram of GPR health factor screening provided in an embodiment of this application; Figure 6 A schematic diagram showing the comparison between measured SOH and predicted SOH over 180 days, provided for an embodiment of this application; Figure 7This is a schematic diagram of the sodium-ion battery cycle life and SOC-SOH co-estimation system provided in an embodiment of this application. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0019] In this embodiment of the application, an integrated operating condition cycle life test platform needs to be constructed first in order to perform joint estimation of SOC and SOH.

[0020] Specifically, such as Figure 1 As shown, the integrated platform includes a grid / photovoltaic system, a 120kW bidirectional AC / DC module, a 200kW DC / DC bidirectional charging module, a container consisting of a 430kWh sodium-ion battery cluster, a battery management control unit (BMU), an energy management system (EMS), and a cloud server.

[0021] The core parameters of the experimental battery are: positive electrode NaNi1 / 3Fe1 / 3Mn1 / 3O2, negative electrode hard carbon, single cell capacity 100Ah, operating voltage 1.5–3.95 V, and internal resistance of approximately 0.55 mΩ at 25°C.

[0022] The set cycle conditions are: daily operation of "two charge and two discharge", 0.5 C constant current charging to 3.95 V, constant voltage to 0.05 C; rest for 30 min; 0.5 C discharge to 2.0 V; cycle for 24 h continuous operation.

[0023] Unlike the constant temperature conditions in the laboratory, the operating temperature of this platform can span 65 ℃. Natural temperature fluctuations can be used to further verify the robustness of the temperature coupling decay model without additional temperature control costs.

[0024] Figure 2 The flowchart of the method for co-estimating the cycle life and SOC-SOH of sodium-ion batteries provided in this application embodiment specifically includes the following steps: S201: Based on the battery state of charge and battery health state, a dual-time-scale state-space model is constructed to observe the battery terminal voltage.

[0025] To address the "fast-slow" two-stage nonlinear decay characteristics of sodium-ion batteries, a dual-timescale state-space model can be constructed to describe the two stages. Furthermore, based on the impact of battery health decay on battery state of charge, the battery state of charge and battery health are coupled and modeled.

[0026] Specifically, the dual-timescale state-space model includes a slow-timescale model and a fast-timescale model, which are used to describe the degradation characteristics of sodium-ion batteries in the slow and fast phases, respectively.

[0027] The slow timescale model describes the degradation of battery state of health (SOH) using two exponential terms: cycle life, degradation coefficient, and temperature. By introducing these two exponential terms, cycle life and degradation coefficient, the model can more accurately reflect the degradation of battery SOH. Furthermore, by coupling with temperature, it can simulate the degradation of battery SOH at different temperatures.

[0028] In one possible implementation, the slow time-scale model can be represented as: Formula 1 in, This represents the battery health status in the k-th cycle, where k is a cycle unit of days. and The attenuation coefficient refers to the attenuation rate per unit time, such as... =0.0042, =0.0018, meaning a decay of 0.42% and 0.18% per cycle. and Cycle life refers to the maximum number of times a battery can operate under specified conditions before its performance degrades to a specific threshold. =850 cycles, =4200 cycles, meaning that after the battery completes 850 and 4200 charge-discharge cycles, its capacity still remains above a certain threshold (usually 80% of the total capacity). T represents temperature. By introducing the above dual exponents, and through experimental fitting with eight sodium-ion batteries, the goodness-of-fit R was obtained. 2 =0.993, significantly better than the goodness-of-fit R of the single exponential model. 2 =0.921.

[0029] Fast timescale models describe the changes in battery state of charge (SOC) using battery health status, rated capacity, coulombic efficiency, current, and process noise. This model considers the impact of battery health status degradation on SOC changes and takes into account the drift caused by inaccurate rated capacity due to SOC degradation, incorporating rated capacity drift into the model to achieve a more accurate description of battery SOC.

[0030] In one possible implementation, the fast timescale model is represented as: Formula 2 in, This indicates the battery's state of charge at time t, where t is a cycle in seconds. Indicates Coulomb efficiency. Represents current. This represents the change in current over time. Indicates the rated capacity, such as This indicates the actual capacity obtained due to changes in battery health. This indicates process noise.

[0031] In one embodiment, after constructing a dual-timescale state-space model, the battery terminal voltage data is observed by jointly observing variables under the coupling of battery health state and battery state of charge.

[0032] In one possible implementation, the joint observation variables can be represented as: Formula 3 in, Indicates terminal voltage. The open-circuit voltage that represents the coupling between the battery's state of charge and its state of health. Represents current. Indicates DC internal resistance. and These represent the first polarization voltage and the second polarization voltage, respectively.

[0033] like Figure 3 As shown, the battery health state is described by a slow timescale model, and the actual battery capacity is updated accordingly. This data is then fed into a fast timescale model, which describes the battery state of charge. Finally, the battery terminal voltage is observed.

[0034] Furthermore, this application proposes a Diverse Multiple Innovation Adaptive Unscented Kalman Filter (DMI-AUKF) algorithm. This algorithm extends the existing Adaptive Unscented Kalman Filter (AUKF) algorithm to include multi-state joint observation and introduces a new information channel, which can effectively reduce the estimation error caused by sensor zero drift.

[0035] The following steps S202 and S203 will be explained in detail.

[0036] S202: Define the state vector for multi-state joint observation and the observation vector based on three channels; wherein, the multi-state includes battery state of charge, battery health state, DC internal resistance, first polarization voltage and second polarization voltage, and the three channels include the terminal voltage, cumulative ampere-hours and constant current charging time under set conditions.

[0037] The state vector can be represented as

[0038] As can be seen, the state vector establishes a five-state joint observation of "battery state of charge, battery state of health, DC internal resistance, first polarization voltage, and second polarization voltage". Existing AUKF only estimates SOC and two polarization voltages, neglecting SOH and... Perform offline calibration, i.e., SOH and With fixed parameters, the SOC estimation error accumulates with the number of cycles, leading to a large error in the final estimate. This method, however, expands the state variables by incorporating battery health and DC internal resistance into the state vector, performing online joint estimation via online recursion to address both SOH and... Real-time correction can further and effectively reduce the error in SOC estimation.

[0039] The observation vector can be represented as

[0040] Among them, constant current charging time The set conditions are 20% SOC → 3.6V. Compared with the existing AUKF, this algorithm can correct SOH and [other conditions] in real time by introducing a constant current charging time into the observation vector. The value of is reduced to decrease hysteresis, which further reduces the estimation error of SOC.

[0041] S203: Based on the multi-innovation adaptive unscented Kalman filter algorithm, the observation vector is observed in three channels, and the covariance is adaptively adjusted through the three-channel innovation to realize the update of the state vector under dual time scales, thereby obtaining the joint estimate of the battery state of charge and battery health state.

[0042] In this embodiment of the application, after defining the state vector and the observation vector, combined with Figure 4 The multi-innovative adaptive unscented Kalman filter algorithm proposed in this application includes the following steps: First, initialize the state vector and collect the current real-time physical quantities and feature quantities.

[0043] Second, using a preset scaling factor, Sigma points are generated through unscented transformation (UT) to update the battery state of charge and polarization voltage, thus obtaining preliminary state prediction values.

[0044] In one implementation, the preset scaling factor can be adopted... The parameter configuration generates 2n+1=11 Sigma points, and the state uncertainty is propagated through Cholesky decomposition to solve the error problem of nonlinear modeling.

[0045] Specifically, in the update of the state vector, a double exponential slow time scale model is used for SOH, which is updated once every cycle; for SOC, a fast time scale model is used, and coulomb counting and SOH are coupled and updated once every second. The slow-scale SOH is incorporated as a coupling term to solve the error caused by capacity drift in traditional SOC estimation; the polarization voltage is dynamically updated using an RC network to match the instantaneous polarization characteristics of the battery.

[0046] Third, conduct three-channel observation and prediction, and calculate the predicted observation values.

[0047] This step derives the predicted values ​​for the three observation channels based on the state vector, specifically including: For terminal voltage, through (i.e., Formula 3) is used for observation; where, It represents the open-circuit voltage that couples the battery's state of charge and state of health, rather than being a single function of SOC. This enables voltage feedback for capacity decay. Represents current. Indicates DC internal resistance. and These represent the first polarization voltage and the second polarization voltage, respectively.

[0048] For cumulative ampere-hours, ampere-hours are observed using the following formula four: Formula 4 Right now Formula 5 After discretization, we get: Formula Six in, This represents the state of charge of the battery at time t. Indicates the rated capacity. This indicates the battery health status at time t.

[0049] The constant current charging time, under the set condition of decreasing from 20% SOC to a terminal voltage of 3.6V during the constant current charging phase, is observed using the following formula six: Formula 7 At a constant current of 0.5C, it can be simplified to: Formula 8 in, Indicates the constant current charging time. Indicates constant current charging current. This represents the change in polarization voltage. Represents the polarization coefficient. This represents the DC internal resistance. This formula establishes the relationship between Tcc and SOH. The explicit analytical relationship allows for the estimation of SOH without waiting for a complete charge-discharge cycle.

[0050] Fourth, calculate the three-channel innovation sequence based on the actual observations and the predicted observations.

[0051] The absolute value of the innovation represents the degree of deviation between the observed and predicted values. The three-channel innovation... Figure 4 In Calculation can be expressed as ,in, Indicates terminal voltage information. This indicates the cumulative amount of new information. This indicates the information regarding constant current charging time.

[0052] SOH affects voltage prediction through OCV coupling terms. The voltage prediction is affected by the ohmic voltage drop term, and the terminal voltage information will inversely correct for SOH and The cumulative ampere-hour information reflects the instantaneous state, directly reflecting the capacity drift of the state of charge (SOH) and can reflect the integral constraint. Cross-validation of the two can detect sensor zero drift. In addition, the constant current charging time information specifically constrains the SOH and... .

[0053] Fifth, based on the three-channel information sequence, the observation noise covariance is adaptively adjusted, and the preliminary state prediction value is corrected and updated to obtain the state correction prediction value, which serves as an estimate of the battery state of charge.

[0054] Specifically, based on the calculated three-channel information and the chi-square test, the presence of current sensor zero drift is determined according to the values ​​of ampere-hour information and terminal voltage information. If so, it indicates that the observed cumulative ampere-hour or terminal voltage values ​​may have a large error. Therefore, it is necessary to adaptively adjust the observation noise covariance and modify the weights of the three channels respectively, reducing the weight of cumulative ampere-hour or terminal voltage and increasing the confidence in other channels to reduce capacity estimation error.

[0055] In one possible implementation, through Perform chi-square test, when and When the current sensor is found to have zero drift, the observation noise covariance is adjusted to... This reduces the ampere-hour weight by 44%, instead relying on the voltage and Tcc channels. Experiments show that the capacity estimation error has been reduced from 3.2% to 0.4%.

[0056] After adaptive covariance adjustment, the predicted values ​​of SOC, Up1, and Up2 updated over time are finally corrected to obtain the state-corrected predicted values.

[0057] Fifth, based on the estimated value of the battery state of charge and the dual-time-scale state-space model, determine the estimated value of the battery health state.

[0058] After each cycle, SOH is updated according to the slow time scale model, and R0 is updated synchronously to match the aging characteristics of the battery. The updated SOH and R0 are fed back to the SOC time update cycle in the next round, serving as coupling terms of the coulomb counting formula and the terminal voltage observation formula, thereby realizing deep linkage between SOC and SOH and solving the problem of decoupling between the two in traditional algorithms.

[0059] In this method, SOC and polarization voltage are updated every 1 second; SOH, It updates every 1 cycle (end of charging) and can achieve coordinated estimation of "second-level SOC-cycle-level SOH" without the need for additional pulse current.

[0060] In one embodiment, before obtaining a co-estimated value of the battery state of charge and battery state of health, the method can also perform a rapid correction of the state of charge (SOH) based on Gaussian process regression (GPR), specifically including the following steps: First, extract health factors related to battery health status. Combined with... Figure 5 Health factors such as Tcc, IC peak value, ΔV / ΔQ slope, temperature T, and cumulative throughput Qthr were used. By also introducing the Tcc health factor into GPR, the correlation with battery health status can be strengthened. Experiments showed that the correlation coefficient r between Tcc and SOH was -0.96, which was significantly higher than the correlation coefficient of IC peak value and SOH of -0.78.

[0061] Second, a Gaussian process regression (GPR) model was trained using RBF and white noise kernel functions, and online hyperparameter optimization was performed to obtain the predicted battery health status.

[0062] During training, aging data from 8 batteries in the same batch, spanning 0–600 cycles, can be used as the training set, with 2 batteries used for testing. Experiments showed that after GPR hyperparameter optimization, the predicted SOH RMSE was 0.47%, and MAE was 0.39%.

[0063] Third, when the preset correction conditions are met, the observation results of the GPR model are used to replace the battery health state estimate obtained by the multi-innovation adaptive unscented Kalman filter algorithm, thereby achieving battery health state correction. The preset correction conditions can be: number of cycles > 100 and ΔSOH > 1%. In this case, the prediction accuracy of DMI-AUKF may be problematic. In this case, the SOH value estimated by DMI-AUKF is replaced with the GPR result to prevent filter lag.

[0064] There may be a joint estimation framework of AEKF+LSTM (Long Short-Term Memory Network), but this framework is only applicable to lithium-ion batteries under laboratory conditions. For sodium-ion batteries in the high-frequency switching scenario of "energy storage-charging integration", sodium-ion decay exhibits nonlinear double exponential characteristics, and the hidden state of a single LSTM is difficult to accurately characterize. Furthermore, this framework does not make full use of the natural deep charge and deep discharge data of "two charge and two discharge" and still requires offline calibration. In addition, this framework requires PC-level computing power and cannot be embedded in existing BMUs.

[0065] Therefore, this application adopts GPR, with only 5-dimensional health factors as input, online hyperparameter optimization time of 90s, and memory usage of <12 kB. In contrast, the LSTM fully connected solution requires 1.8 MB of RAM and cannot embed a BMU. The GPR kernel function of this application uses RBF + white noise, and the hyperparameters are updated online by maximizing edge likelihood. The memory usage is <12 kB, the CPU usage is <3%, and it can run on a Cortex-M4 bare machine.

[0066] In one embodiment, after obtaining a joint estimate of the battery's state of charge and state of health, this method can further determine, based on the joint estimate, when the battery meets preset lifespan conditions, reduce the charge / discharge rate (e.g., 0.2 C) and issue a lifespan termination warning. The preset lifespan conditions include at least one of the following: the battery's state of health drops to 80%, or the single-cell voltage variance exceeds 150mV. This "dual criterion + automatic power reduction" strategy effectively reduces the false alarm rate.

[0067] In addition, this method can also simultaneously output a double exponential extrapolation curve to predict the remaining number of cycles, RUL. Experiments show that the error is ≤3 cycles.

[0068] In this embodiment, considering the two-stage characteristics of sodium-ion battery degradation, the impact of battery health state degradation leading to inaccurate rated capacity drift and thus affecting battery state of charge (SOC) changes is addressed by coupling SOC and SOC modeling. A multi-innovative adaptive unscented Kalman filter algorithm is proposed, extending the state vector to include SOC and DC internal resistance in the online-updated state vector. Furthermore, by introducing constant-current charging time into the observation vector, real-time correction of SOC and DC internal resistance is achieved, reducing estimation errors in SOC estimation and further improving estimation accuracy, thus realizing online joint estimation of SOC and SOC. Additionally, a three-channel innovation approach with adaptive covariance matching effectively reduces estimation errors caused by sensor zero drift, ensuring the stability and accuracy of the joint estimation.

[0069] Through experiments, such as Figure 6 As shown, in a 200 kW / 430 kWh integrated platform, with two charge-discharge cycles per day, this solution was continuously operated for 180 days (≈1200 cycles), and the operating results are as follows: Cycles 0–200: SOH decreased from 100% to 94.2%, with a joint estimation error of 0.3% using the double exponential model and GPR. 600th cycle: SOH = 88.1%, SOC full-cycle error 1.5%; 1200th cycle: SOH=80.4%, triggering an early warning; RUL predicted value 248 cycles, actual value 251 cycles, error 3 cycles; It can be seen that the maximum error of SOC estimation has decreased from 8.1% of the traditional coulomb counting + EKF method to 1.8%, and the maximum error of SOH has decreased from 4.7% of the offline internal resistance method to 1.3%, both of which are better than the industry threshold of 5%.

[0070] This solution can complete capacity degradation calibration using daily "two charge and two discharge" operation data, eliminating the need to arrange 0.33C offline full charge and discharge. It shortens the cycle life test cycle from the traditional 60-day offline life test to "zero extra time", which can be completed using daily operation, greatly saving test costs and obtaining the required life assessment report in advance.

[0071] Furthermore, this scheme triggers power reduction and warning on the 1080th cycle, while the offline method can only confirm it on the 1200th cycle. This allows for an early warning of SOH < 80% 120 cycles in advance, avoiding revenue loss for energy storage power stations due to sudden capacity drops.

[0072] This solution requires only 1.2 kB of RAM and 14 kB of Flash memory per cycle, with a CPU utilization of 4.3%. The required computing power is less than 5 DMIPS. It can be embedded into an existing BMU (Cortex-M4 120 MHz) without hardware upgrades. Compared with cloud AI solutions, it also saves power, which is beneficial for energy conservation and emission reduction.

[0073] In addition, this solution reduces the overcharge / over-discharge probability from 0.9% to 0.05% through SOC-SOH joint constraint with a precision of 2% level. According to the empirical formula for NAS batteries, every 1% reduction in overcharge / over-discharge can extend the cycle life by 6%. This method increases the cycle life of sodium-ion battery clusters from 4,000 cycles to approximately 4,500 cycles under the same operating conditions, which translates to an 8% reduction in the levelized cost of electricity (LCOE) over the entire life cycle, and can significantly extend the service life of the system.

[0074] In another embodiment, this method is adapted to a 48 V 50 Ah electric two-wheel battery pack, with a cycle condition of 1 C charging and 2 C discharging. Due to the increased rate of change, Revised to 520 cycles. =0.007; after running 500 cycles, the maximum error of SOC is 2.1% and the maximum error of SOH is 1.6%, which still meet the industry standard requirements.

[0075] In summary, this method achieves quantifiable benefits across five dimensions: accuracy, cost, time, energy consumption, and lifespan. It addresses the pain points of state estimation and lifespan testing in large-scale sodium-ion battery energy storage applications, demonstrating significant economic and social benefits. Through a closed-loop process of "integrated operating conditions → dual timescale model → DMI-AUKF collaboration → GPR rapid correction → lifespan end determination," it achieves online integration of cycle life testing and SOC-SOH collaborative estimation in an integrated sodium-ion battery energy storage and charging system. By organically combining DMI-AUKF and GPR, it enables collaborative updating and rapid correction of "second-level SOC to cycle-level SOH," improving estimation accuracy and response speed. Cycle life testing, state estimation, and lifespan warning are completed simultaneously without additional downtime testing, achieving "evaluation while operating," significantly reducing testing costs and time, and substantially improving estimation accuracy while shortening the testing cycle.

[0076] The above describes the method for co-estimating the cycle life and SOC-SOH of sodium-ion batteries provided in this application. Based on the same inventive concept, this application also provides a corresponding system for co-estimating the cycle life and SOC-SOH of sodium-ion batteries, such as... Figure 7 As shown.

[0077] Figure 7 This is a schematic diagram of the structure of the sodium-ion battery cycle life and SOC-SOH co-estimation system provided in the embodiments of this application, specifically including: Module 701 is used to construct a dual-time-scale state-space model based on the battery's state of charge and state of health, and to observe the battery's terminal voltage. Definition module 702 is used to define the state vector of multi-state joint observation and the observation vector based on three channels; wherein, the multi-state includes battery state of charge, battery health state, DC internal resistance, first polarization voltage and second polarization voltage, and the three channels include the terminal voltage, cumulative ampere-hours and constant current charging time under set conditions; The observation module 703 is used to perform three-channel observation of the observation vector based on the multi-innovation adaptive unscented Kalman filter algorithm, and to update the state vector under dual time scales by adaptively adjusting the covariance through the three-channel innovation, thereby obtaining a joint estimate of the battery state of charge and battery health state.

[0078] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0079] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0080] The systems and methods provided in this application are one-to-one correspondences. Therefore, the system also has similar beneficial technical effects as its corresponding method. Since the beneficial technical effects of the method have been described in detail above, the beneficial technical effects of the system will not be repeated here.

[0081] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using dedicated hardware combined with computer instructions. The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0082] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0083] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for co-estimating the cycle life and SOC-SOH of a sodium-ion battery, characterized in that, include: Based on the battery's state of charge and state of health, a dual-timescale state-space model is constructed to observe the battery terminal voltage, including: through joint observation of variables. The battery terminal voltage, coupled with the battery's state of charge and state of health, is observed; among them, Indicates terminal voltage. The open-circuit voltage that represents the coupling between the battery's state of charge and its state of health. Represents current. Indicates DC internal resistance. and Let represent the first polarization voltage and the second polarization voltage, respectively; wherein, the dual-timescale state-space model includes a slow-timescale model and a fast-timescale model; the slow-timescale model describes the degradation of the battery's state of health through the cycle life, degradation coefficient, and temperature of the double-exponential terms; the fast-timescale model describes the changes in the battery's state of charge through the battery's state of health, rated capacity, coulombic efficiency, current, and process noise; the slow-timescale model is expressed as... ,in, This indicates the battery health status in the k-th cycle. and Indicates the attenuation coefficient. and Indicates cycle life, where T represents temperature. The fast timescale model is expressed as: ,in, This represents the state of charge of the battery at time t. Indicates Coulomb efficiency. Represents current. Indicates the rated capacity. Indicates process noise; Define a state vector for multi-state joint observation and an observation vector based on three channels; wherein, the multi-state includes battery state of charge, battery health state, DC internal resistance, first polarization voltage and second polarization voltage, and the three channels include the terminal voltage, cumulative ampere-hours and constant current charging time under set conditions; Based on a multi-innovation adaptive unscented Kalman filter algorithm, the observation vector is observed through three channels, and the covariance is adaptively adjusted through three-channel innovation to update the state vector under dual time scales, obtaining a joint estimate of the battery state of charge (SOC) and battery health state. This includes: initializing the state vector; using a preset scaling factor, generating Sigma points through unscented transformation (UT) to update the battery SOC and polarization voltage, obtaining a preliminary state prediction; performing three-channel observation prediction and calculating the predicted observation values; calculating a three-channel innovation sequence based on the actual observation values ​​and the predicted observation values; adaptively adjusting the observation noise covariance based on the three-channel innovation sequence and correcting and updating the preliminary state prediction values ​​to obtain a corrected state prediction value, which serves as the estimate of the battery SOC; and determining the estimate of the battery health state based on the estimated battery SOC and the dual time scale state-space model. The three-channel observation prediction includes: through... Terminal voltage observations were conducted; among them, The open-circuit voltage that represents the coupling between the battery's state of charge and its state of health. Represents current. Indicates DC internal resistance. and These represent the first polarization voltage and the second polarization voltage, respectively; through Hourly observations were performed, and the results were discretized to obtain... ;in, This represents the state of charge of the battery at time t. Indicates the rated capacity. This indicates the battery health status at time t; under the set condition of charging from 20% SOC to a terminal voltage of 3.6V during the constant current charging phase, through... Constant current charging time was observed, and simplified to 0.5C constant current. ,in, Indicates the constant current charging time. Indicates constant current charging current. This represents the change in polarization voltage. Represents the polarization coefficient. This indicates the DC internal resistance.

2. The method for co-estimating the cycle life and SOC-SOH of a sodium-ion battery according to claim 1, characterized in that, The adaptive adjustment of the observation noise covariance based on the three-channel information sequence includes: Based on the three-channel information sequence and chi-square test, determine whether there is zero drift of the current sensor according to the values ​​of the ampere-hour information and the terminal voltage information; If so, adaptively adjust the observation noise covariance to modify the weights of the three channels.

3. The method for co-estimating the cycle life and SOC-SOH of a sodium-ion battery according to claim 1, characterized in that, Before obtaining the joint estimate of the battery state of charge and battery state of health, the method further includes: Extract health factors related to battery health status; A Gaussian process regression (GPR) model was trained, using RBF and white noise kernel functions, and online hyperparameter optimization was performed to obtain the predicted battery health status. When the preset correction conditions are met, the battery health state estimate obtained by the multi-innovation adaptive unscented Kalman filter algorithm is replaced with the observation results of the GPR model to achieve battery health state correction.

4. The method for co-estimating the cycle life and SOC-SOH of a sodium-ion battery according to claim 1, characterized in that, After obtaining the joint estimate of the battery state of charge and battery state of health, the method further includes: Based on the collaborative estimation value, when it is determined that the battery meets the preset lifespan conditions, the charge / discharge rate is reduced and a lifespan termination warning is issued. The preset lifespan conditions include at least one of the following: the battery health status drops to 80%, and the single-cell voltage variance is greater than 150mV.

Images