A method and system for estimating the life expectancy of a hybrid energy storage system

Abstract

The application discloses a kind of expected life estimation method and system for hybrid energy storage system, and hybrid energy storage system is composed of water series hydrogen ion battery and lithium iron phosphate battery.Method includes: defining single cell basic parameters and operating boundary conditions, based on two battery aging paths to build battery aging submodel;Through hybrid operating condition acceleration test, operating data is collected, difference temperature correction coefficient is calculated, and interactive influence factor library is quantified;Using BiLSTM neural network to complete full life cycle data and optimize submodel;Integrate multi-dimensional correction factor to calculate single cell corrected life, determine system expected life by combining preset judging rule, and realize model iterative optimization by comparing actual operating data.System includes parameter definition and model construction, data acquisition and processing, data completion and model optimization, life estimation and iterative optimization module.The application improves the accuracy and adaptability of expected life estimation, and supports the large-scale application of hybrid energy storage system.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and more specifically, to a method and system for estimating the expected lifetime of a hybrid energy storage system. Background Technology

[0002] With the rapid development of new energy storage technologies, hybrid energy storage systems have been widely used in various scenarios such as laboratories, industrial plants, and commercial buildings due to their ability to integrate the performance advantages of different types of batteries. Among them, hybrid energy storage systems composed of aqueous batteries and lithium iron phosphate batteries achieve synergistic optimization of safety, capacity, and cost by leveraging the high safety and wide temperature range adaptability of aqueous batteries and the high energy density and mature cycle performance of lithium iron phosphate batteries. This effectively compensates for the shortcomings of single-battery energy storage systems, such as the risk of thermal runaway or insufficient energy density, making it the preferred solution for scenarios with high safety requirements.

[0003] In the design, operation, maintenance, and lifecycle cost management of hybrid energy storage systems, expected lifespan calculation is a core technical aspect, as its results directly determine the initial configuration selection, operation and maintenance strategy formulation, and cost optimization scheme. The industry generally uses a specific percentage of battery capacity degradation to its initial level and a system charge / discharge efficiency drop to a specific threshold as the end-of-life standard. Traditional lifespan calculation methods mainly focus on the characteristics of individual batteries: for lithium iron phosphate batteries, the impact of factors such as temperature on their core aging path is quantified, and cycle life is predicted using empirical models; for aqueous batteries, experimental testing is used to obtain the correlation between key operating parameters and the aging process, thereby estimating lifespan.

[0004] However, when the traditional single-cell lifespan calculation method is directly applied to an aqueous battery-lithium iron phosphate hybrid energy storage system, there are many technical problems that are difficult to solve, as follows:

[0005] The coupling interference caused by differences in aging mechanisms has not been considered: the core aging paths of lithium iron phosphate (LFP) batteries and aqueous batteries are fundamentally different. The aging of LFP batteries is mainly related to the growth of specific interface films and the loss of active materials, and is more sensitive to temperature and charge / discharge rates; while the aging of aqueous batteries is concentrated on electrode material loss and electrolyte performance degradation, and is more sensitive to depth of discharge and cycle frequency. In hybrid systems, the two types of batteries work together through power distribution strategies. Aqueous batteries typically undertake peak power regulation tasks, while LFP batteries undertake long-term energy storage tasks. Their collaborative operation will interfere with each other's aging processes—for example, the power sharing of aqueous batteries can reduce overload conditions in LFP batteries, thereby delaying their aging; while high-frequency peak discharge may accelerate the aging of aqueous batteries. Existing technologies have not established an interaction model for the aging of the two types of batteries, and simply apply a single battery life model or perform superposition calculations, resulting in significant deviations between the calculated results and the actual lifespan.

[0006] The differential temperature response has not been precisely adapted: lithium iron phosphate batteries are extremely sensitive to temperature changes, and increased temperature significantly accelerates their aging; aqueous batteries, on the other hand, have better temperature adaptability and can operate stably over a wider temperature range. Although the hybrid system is equipped with a temperature control device, the existing calculation method only uses the overall system ambient temperature for uniform correction, without considering the local temperature differences and temperature sensitivity differences between the two types of batteries. This leads to a mismatch between the temperature correction logic and actual operating conditions, further amplifying the calculation error.

[0007] Insufficient parameter adaptability under dynamic operating conditions: The lifespan calculation of a single battery largely relies on fixed operating parameters, while the operating conditions of hybrid energy storage systems are dynamic: Under low-load scenarios, lithium iron phosphate batteries cycle at low rates and medium depths of discharge; under high-load scenarios, aqueous batteries respond instantaneously at high rates and shallow depths of discharge. Existing technologies lack experimental data to support these dynamic operating conditions, making it impossible to quantify the impact of dynamic parameters on the aging of the two types of batteries. This results in models that are difficult to adapt to actual operating scenarios and insufficient calculation accuracy.

[0008] The criteria for determining system-level lifespan are vague: Existing technologies mostly use the lifespan end-of-life criteria for single batteries, without clearly defining a unified determination rule for hybrid energy storage systems. When there are significant differences in the cycle life of the two types of batteries, it is difficult to determine whether to use the battery with the shorter lifespan as the determination benchmark or to consider the performance compensation effect of the longer-life battery. This results in a lack of uniformity and reliability in the determination results, failing to provide effective guidance for system design and operation and maintenance.

[0009] The aforementioned issues result in significant errors in the calculation of the expected lifespan of existing hybrid energy storage systems, making it difficult to meet the needs of engineering applications. If the calculated result is too long, it will lead to premature system failure, causing safety hazards or additional operation and maintenance costs; if the calculated result is too short, it will result in over-design, increasing initial investment costs. Therefore, there is an urgent need for a method to calculate the expected lifespan of hybrid energy storage systems that can take into account the aging coupling effects of the two types of batteries, the impact of temperature differences, dynamic operating condition adaptability, and clear judgment rules, in order to improve calculation accuracy and support the large-scale application of hybrid energy storage systems. Summary of the Invention

[0010] To address the aforementioned technical problems, this invention provides a method for estimating the expected lifetime of a hybrid energy storage system, comprising:

[0011] Define the basic parameters and operating boundary conditions of a single battery in a hybrid energy storage system, which consists of an aqueous hydrogen-ion battery and a lithium iron phosphate battery.

[0012] Accelerate the collection of operational data through mixed operating conditions; calculate the differential temperature correction coefficient; and quantify and form a library of interactive influencing factors.

[0013] Complete the full lifecycle data and optimize the battery aging sub-model;

[0014] Calculate the corrected lifetime of a single cell by integrating multi-dimensional correction factors;

[0015] The expected lifespan of the system is determined by combining preset judgment rules; the model is iteratively optimized by comparing actual operating data.

[0016] Furthermore, the basic parameters of the single battery include: the interface film growth characteristics parameters and active material loss coefficient of the lithium iron phosphate battery; the electrode dissolution rate coefficient and electrolyte performance degradation coefficient of the aqueous hydrogen ion battery; and the operating boundary conditions include the average number of charge-discharge cycles per day, the power distribution ratio range, the ambient temperature range, and the system charge-discharge efficiency threshold.

[0017] Furthermore, the mixed-condition accelerated test includes: designing mixed-conditions with multiple sets of variables, including the power ratio of the aqueous hydrogen-ion battery, the system ambient temperature, and the average number of daily cycles; the test cycle of each condition covers the early to mid-stages of accelerated battery aging; and simultaneously collecting the actual operating temperature, charge / discharge rate, depth of discharge, and capacity decay rate of the two types of batteries, while also collecting system-level charge / discharge efficiency data.

[0018] Furthermore, the calculation of the differential temperature correction coefficient includes: obtaining the deviation between the actual operating temperature and the reference temperature of the two types of batteries; and calculating the targeted differential temperature correction coefficient based on the temperature sensitivity difference between the two types of batteries.

[0019] Furthermore, the quantitative formation of the interaction factor library includes: acquiring capacity decay rate data of two types of batteries under different power allocation ratios; establishing a correspondence between power allocation ratio and interaction factors based on the capacity decay rate data; constructing a mapping model characterizing the correlation between power allocation ratio and interaction factors through linear fitting or nonlinear fitting algorithms; and extracting interaction factors corresponding to different operating condition combinations and incorporating them into the interaction factor library.

[0020] Furthermore, the completion of the full life cycle data and optimization of the sub-model include: the full life cycle data includes the capacity retention rate, internal resistance change value, charge and discharge efficiency decay curve, electrode material wear degree, and electrolyte performance parameter evolution data of the two types of batteries under different cycle numbers; a BiLSTM neural network is used as a deep learning model, and the input features include battery cycle number, actual operating temperature, charge and discharge rate, and depth of discharge; the full life cycle data of the two types of batteries is completed through this model, and the error between the completed data and the measured data is controlled within a preset threshold; the completed full life cycle data is substituted into the battery aging sub-model to correct the model parameters.

[0021] Furthermore, the preset judgment rules include: using lithium iron phosphate batteries as the key unit for lifespan determination, and using the capacity decay to 80% of the initial capacity as the basic judgment standard; if the corrected lifespan of the aqueous hydrogen ion battery is longer than that of the lithium iron phosphate battery, then the discharge depth of the aqueous hydrogen ion battery is adjusted, the capacity output data of the adjusted aqueous hydrogen ion battery is obtained, and it is determined whether the data can make up for the capacity decay gap of the lithium iron phosphate battery; when the capacity of the lithium iron phosphate battery decays to 80% of the initial capacity, or the system charge and discharge efficiency drops to 85%, the system is determined to have reached the end of its lifespan.

[0022] Furthermore, the model iterative optimization includes: comparing the actual operating data of the hybrid energy storage system with the expected lifetime estimation results in real time and analyzing the sources of deviation; updating the parameters of the interaction influence factor library based on the latest actual operating data after each preset number of cycles; and simultaneously correcting the coefficients of the battery aging sub-model.

[0023] A lifetime estimation system for hybrid energy storage systems, implementing the aforementioned lifetime estimation method for hybrid energy storage systems, includes:

[0024] The parameter definition and model building module defines the basic parameters and operating boundary conditions of a single cell and builds a cell aging sub-model.

[0025] The data acquisition and processing module conducts accelerated testing under mixed operating conditions, collects operational data, calculates differential temperature correction coefficients, and quantifies and forms a library of interactive influencing factors.

[0026] The data completion and model optimization module completes the full lifecycle data and optimizes the battery aging sub-model.

[0027] The lifespan estimation module integrates multi-dimensional correction factors such as differential temperature correction coefficient and interaction influence factor to calculate the corrected lifespan of a single cell and determine the expected lifespan of the system by combining preset judgment rules.

[0028] The iterative optimization module compares the actual operating data of the hybrid energy storage system with the estimated expected lifespan, analyzes the sources of deviation, and updates the interactive influencing factor library and the parameters of the battery aging sub-model.

[0029] The positive and progressive effects of this invention are as follows:

[0030] This invention precisely addresses the technical challenges of aging coupling between two types of batteries in hybrid energy storage systems, temperature response differences, dynamic operating condition adaptation, and ambiguous lifespan determination by constructing a battery aging sub-model, differentiating temperature corrections, quantifying interactive influencing factors, and completing full lifespan data. This significantly reduces the error in expected lifespan estimation and avoids the risk of over-design or premature failure. Its iterative optimization mechanism can continuously adapt to actual operating scenarios, providing reliable support for the design and selection of hybrid energy storage systems, the formulation of operation and maintenance strategies, and the control of full lifespan costs, significantly improving the safety and economy of engineering applications. Attached Figure Description

[0031] Figure 1 This is a flowchart of the steps of the present invention. Detailed Implementation

[0032] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0033] Reference Figure 1 A method for estimating the expected lifespan of a hybrid energy storage system includes: defining the basic parameters and operating boundary conditions of a single cell in the hybrid energy storage system, wherein the hybrid energy storage system consists of an aqueous hydrogen-ion battery and a lithium iron phosphate battery; collecting operating data through accelerated testing under mixed operating conditions; calculating a differential temperature correction coefficient; quantifying and forming a library of interactive influencing factors; supplementing the full life cycle data and optimizing the cell aging sub-model; integrating multi-dimensional correction factors to calculate the corrected lifespan of a single cell; determining the expected lifespan of the system by combining preset judgment rules; and achieving iterative optimization of the model through comparison with actual operating data.

[0034] Furthermore, the basic parameters of the single battery include: the interface film growth characteristics and active material loss coefficient of the lithium iron phosphate battery; the electrode dissolution rate coefficient and electrolyte performance degradation coefficient of the aqueous hydrogen ion battery; and the operating boundary conditions include the average number of charge-discharge cycles per day, the power allocation ratio range, the ambient temperature range, and the system charge-discharge efficiency threshold. In one example, the interface film growth characteristics of the lithium iron phosphate battery can be set to 0.002~0.005 / cycle, and the active material loss coefficient to 0.001~0.003 / cycle; the electrode dissolution rate coefficient of the aqueous hydrogen ion battery is 0.0008~0.002 / cycle, and the electrolyte performance degradation coefficient is 0.0015~0.004 / cycle. In the operating boundary conditions, the average number of charge-discharge cycles per day is 1~3 times, the power allocation ratio range is 10%~30% for the aqueous hydrogen ion battery, the ambient temperature range is 10℃~35℃, and the system charge-discharge efficiency threshold is set to 85%.

[0035] As further explanation, the aqueous hydrogen-ion battery adopts a stacked structure of "positive electrode-separator-negative electrode". The positive electrode uses a porous carbon-supported platinum-based catalyst, and the negative electrode uses a titanium-based hydrogen ion storage material. The electrolyte is a 0.5~2 mol / L sulfuric acid aqueous solution (excluding flammable organic solvents). The separator is a perfluorosulfonic acid proton exchange membrane, which allows for rapid hydrogen ion migration and blocks cross-contamination of electrode materials. The battery casing is made of flame-retardant ABS material and has a built-in pressure relief valve and electrolyte circulation channel to ensure safe operation. The lithium iron phosphate battery adopts a wound structure. The positive electrode is a mixture of lithium iron phosphate active material, conductive carbon black, and PVDF binder coated on an aluminum foil current collector. The negative electrode is a graphite material coated on a copper foil current collector. The electrolyte is a mixture of lithium hexafluorophosphate and carbonate solvents. The separator is a polypropylene / polyethylene composite porous membrane. The battery module is equipped with an aluminum alloy casing, a built-in temperature sensor interface, and heat dissipation fins, making it suitable for conventional laboratory installation environments.

[0036] The hybrid energy storage system is deployed in a dedicated laboratory energy storage cabinet, employing a "layered" and "partitioned" layout: the upper layer houses lithium iron phosphate battery modules, each containing 16 individual cells connected in series, with 4 parallel groups and a total capacity of 100Ah. A 15cm heat dissipation channel is reserved between the modules, and their backs are attached to the energy storage cabinet's heat dissipation panel. The lower layer houses aqueous hydrogen ion battery modules, each containing 8 individual cells connected in series, with 2 parallel groups and a total power of 50kW. Electrolyte circulation pipelines and cooling water tanks are located below the modules, controlling temperature through water cooling. The main control unit and power distributor are installed in the middle of the energy storage cabinet, connected to both types of battery modules via copper busbars for energy transfer and command interaction. Temperature acquisition points are located on the surface of the middle individual cells of the lithium iron phosphate battery modules and at the electrolyte outlet of the aqueous hydrogen ion battery modules, ensuring that the collected data reflects the actual operating temperature of the batteries.

[0037] During system operation, the main control unit dynamically allocates power according to the laboratory's power load: In low-load scenarios (such as only the constant temperature incubator and chromatograph operating, with power ≤10kW), the power distributor disconnects the aqueous hydrogen-ion battery circuit, and the lithium iron phosphate battery module supplies power independently, maintaining stable output through constant current discharge mode, with the charge / discharge rate controlled at 0.3~0.8C to match its long-term energy storage characteristics; In high-load scenarios (such as the start-up of large synthesizers and X-ray diffractometers, with power 10~50kW), the main control unit triggers the aqueous hydrogen-ion battery to intervene, supplementing the power gap through instantaneous high-rate discharge (2~5C). At this time, the lithium iron phosphate battery still maintains basic load power supply to avoid accelerated aging due to overload; The operating status, voltage, current, and temperature data of both batteries are collected in real time by the main control unit and transmitted to the data processing module for subsequent lifetime estimation. This arrangement and functional allocation method ensures the performance of both battery characteristics at the structural level, providing a stable hardware foundation for expected lifetime estimation.

[0038] Furthermore, the accelerated testing under mixed operating conditions includes: designing mixed operating conditions with multiple combinations of variables, including the power percentage of the aqueous hydrogen-ion battery, system ambient temperature, and average daily cycle count; the test cycle for each operating condition covers the early to mid-stages of accelerated battery aging; simultaneously collecting the actual operating temperature, charge / discharge rate, depth of discharge, and capacity decay rate of both types of batteries, as well as system-level charge / discharge efficiency data. In one example, the variable combinations for the mixed operating conditions can be designed as three groups: the first group has an aqueous hydrogen-ion battery power percentage of 10%, an ambient temperature of 20°C, and an average daily cycle count of 1; the second group has a power percentage of 20%, an ambient temperature of 25°C, and an average daily cycle count of 2; and the third group has a power percentage of 30%, an ambient temperature of 30°C, and an average daily cycle count of 2. The test cycle for each operating condition is set to 200 cycles. When collecting data, a high-precision charge and discharge test system (model: CT-4008-5V100A, charge and discharge accuracy ±0.01%FS) is used for testing. The charge and discharge rate accuracy is controlled within ±0.05C, the discharge depth recording accuracy is ±1%, the capacity decay rate is statistically analyzed every 20 cycles, the actual operating temperature is recorded every 10 seconds using a PT100 temperature sensor (measurement accuracy ±0.1℃), and the system-level charge and discharge efficiency is recorded once per cycle. The collected raw data needs to be preprocessed, and noise is removed by using a moving average filter (window size of 5). Outliers (data exceeding 3 times the standard deviation) are replaced by linear interpolation to ensure data reliability. By designing mixed operating conditions with multiple power allocation ratios and multiple temperatures, and simultaneously collecting aging-related data of two types of batteries, the problem of "coupling interference caused by differences in aging mechanisms not being considered" is directly solved. Different power ratios correspond to different collaborative working modes of the two batteries. The collected capacity decay rate data can directly reflect the intervention of the operating state of one battery on the aging of the other battery, providing a practical basis for the subsequent quantification of interaction factors and breaking the limitation of traditional single-battery testing being unable to capture coupling effects.

[0039] Furthermore, the calculation of the differential temperature correction coefficient includes: obtaining the deviation between the actual operating temperature and the reference temperature of the two types of batteries; and calculating the targeted differential temperature correction coefficient based on the temperature sensitivity difference between the two types of batteries. In one example, the reference temperature is set to 25℃ (according to the standard test temperature specified in GB / T31484-2015 "Lithium-ion Battery Cycle Life Test Method"). The actual operating temperature of the lithium iron phosphate battery is obtained as 28℃ through the temperature acquisition device, with a deviation of 3℃; the actual operating temperature of the aqueous hydrogen-ion battery is 26℃, with a deviation of 1℃. The temperature sensitivity coefficients of the two types of batteries were determined through a single-factor variable experiment: With parameters such as power distribution ratio (20%), charge / discharge rate (0.5C), and depth of discharge (50%) kept constant, the two types of batteries were placed in a constant temperature environmental chamber (model: HWS-250, temperature control accuracy ±0.5℃) at 15℃, 20℃, 25℃, 30℃, and 35℃ for 300 cycles of testing. The capacity decay rate λ at different temperatures was calculated, and the relationship between the capacity decay rate and temperature was obtained using linear fitting: λ = α × T + β (where α is the temperature sensitivity coefficient and β is a constant term). The final fitting yielded a temperature sensitivity coefficient of 0.12 / ℃ for lithium iron phosphate batteries and 0.08 / ℃ for aqueous hydrogen ion batteries. The specific mathematical formula for calculating the differential temperature correction coefficient is KT=1+α×(Tact-Tref), where Tact is the actual operating temperature of the battery, Tref is the reference temperature, and α is the temperature sensitivity coefficient of the corresponding battery. Substituting the example data, the differential temperature correction coefficient for lithium iron phosphate batteries is calculated to be 1+0.12×3=1.36, and for aqueous hydrogen ion batteries, it is 1+0.08×1=1.08. By calculating correction coefficients separately for the temperature sensitivity differences of the two types of batteries, the problem of "inaccurate correction of temperature differential response" is precisely solved. This approach abandons the crude method of traditional uniform temperature correction. By combining the actual operating temperature deviation with the sensitivity coefficient, and clarifying the measurement method of the sensitivity coefficient and the calculation logic of the correction coefficient, the quantification of the impact of temperature on the aging of each type of battery is more accurate, avoiding lifespan calculation errors caused by local temperature differences.

[0040] Furthermore, the quantitative formation of the interaction factor library includes: acquiring capacity decay rate data of two types of batteries under different power allocation ratios; establishing a correspondence between power allocation ratio and interaction factors based on the capacity decay rate data; constructing a mapping model characterizing the correlation between power allocation ratio and interaction factors through linear fitting or nonlinear fitting algorithms; and extracting interaction factors corresponding to different operating condition combinations and incorporating them into the interaction factor library. In one example, when the power allocation ratio is 10%, the capacity decay rate of the lithium iron phosphate battery is 0.002 / cycle, and that of the aqueous hydrogen ion battery is 0.0018 / cycle; when the power allocation ratio is 20%, the decay rate of the lithium iron phosphate battery is 0.0017 / cycle, and that of the aqueous hydrogen ion battery is 0.0021 / cycle; when the power allocation ratio is 30%, the decay rate of the lithium iron phosphate battery is 0.0015 / cycle, and that of the aqueous hydrogen ion battery is 0.0023 / cycle. Additionally, two other sets of measured data for power allocation ratios (15% and 25%) are provided: when the power allocation ratio is 15%, the decay rate of the lithium iron phosphate battery is 0.0019 / cycle, and that of the aqueous hydrogen ion battery is 0.00195 / cycle; when the power allocation ratio is 25%, the decay rate of the lithium iron phosphate battery is 0.0016 / cycle, and that of the aqueous hydrogen ion battery is 0.00225 / cycle. The fitting algorithm uses the least squares method for linear fitting. The specific formula is as follows: For sample points (Pi, Ki) (i = 1, 2, ..., n), the fitting coefficients are a = (n × Σ(Pi × Ki) - ΣPi × ΣKi) / (n × ΣPi² - (ΣPi)²), b = (ΣKi - a × ΣPi) / n, where Ki = λi / λ0 (λ0 is the capacity decay rate of a single battery operating independently; for lithium iron phosphate batteries, λ0 = 0.0025 / cycle; for aqueous hydrogen ion batteries, λ0 = 0.0015 / cycle). Substituting the above 5 sets of data, the relationship between the interaction factor and the power distribution ratio of lithium iron phosphate batteries is calculated as K = The relationship between the power allocation ratio (1.0 + 0.005P, where P is the power distribution ratio) and the aqueous hydrogen-ion battery is K = 1.0 - 0.003P. The interaction factors corresponding to the above five operating conditions are extracted as 1.05, 1.075, 1.10, 1.125, 1.15 (lithium iron phosphate) and 0.97, 0.955, 0.94, 0.925, 0.91 (aqueous hydrogen-ion), and included in the factor library. If subsequent measured data exhibit a nonlinear relationship, the Gauss-Newton method can be used for nonlinear fitting, and the mapping model is adjusted to K = a × P² + b × P + c. The fitting coefficients are solved using the lsqcurvefit function in MATLAB. By quantifying the correlation between the power distribution ratio and the interaction factors, a dedicated factor library is constructed, clarifying the interaction law of aging for the two types of batteries. This makes the lifespan calculation no longer a simple superposition of a single battery model, but a precise deduction based on the actual coupling effect.

[0041] Furthermore, the completion of the full life cycle data and optimization of the sub-model include: the full life cycle data includes capacity retention rate, internal resistance change value, charge / discharge efficiency decay curve, electrode material wear degree, and electrolyte performance parameter evolution data of the two types of batteries at different cycle counts; a BiLSTM neural network is used as a deep learning model, and the input features include battery cycle count, actual operating temperature, charge / discharge rate, and depth of discharge; the full life cycle data of the two types of batteries is completed using this model, and the error between the completed data and the measured data is controlled within a preset threshold; the completed full life cycle data is substituted into the battery aging sub-model to correct the model parameters. In one example, the full life cycle data covers 0 to 6000 cycles, of which the measured data is 0 to 200 cycles, and the completed data is 201 to 6000 cycles. The specific architecture of the BiLSTM neural network is as follows: the input layer has a dimension of 4 (corresponding to 4 input features), followed by an embedding layer that maps the input features to a 128-dimensional vector, with ReLU as the activation function; the output of the embedding layer is connected to the first BiLSTM layer (64 neurons), with sigmoid activation functions for the forget gate, input gate, and output gate, tanh activation function for the cell state, and the return sequence set to True; after the first BiLSTM layer, a Dropout layer (dropout probability of 0.2) is connected, followed by a second BiLSTM layer (32 neurons), with the same activation function configuration as the first BiLSTM layer, and the return sequence set to False; after the second BiLSTM layer, a second Dropout layer (dropout probability of 0.2) is connected, followed by a fully connected layer (16 neurons, ReLU activation function), and finally, the output layer (1 neuron, sigmoid activation function) outputs the capacity retention data. The model training parameters were set as follows: the optimizer used was the Adam optimizer, the initial learning rate was 0.001, and a learning rate decay strategy was adopted (decaying to 0.9 of the original value every 100 rounds); the loss function was mean squared error (MSE=(1 / m)×Σ(ypred-ytrue)², where m is the number of samples, ypred is the predicted value, and ytrue is the measured value); the number of training iterations was 500 rounds, and the batch size was 32; L2 regularization (weight decay coefficient 0.001) combined with Dropout layer was used to suppress overfitting.Data preprocessing employs Min-Max normalization, with the formula xnorm=(x-xmin) / (xmax-xmin), where xmin and xmax are the minimum and maximum values ​​of each feature (based on statistical data from actual measurements, such as the number of iterations xmin=0, xmax=200, and the actual working temperature xmin=10℃, xmax=35℃). The actual measurements are divided into training, validation, and test sets in a 7:2:1 ratio. The training set is used for updating model parameters, the validation set is used to monitor overfitting (training is stopped if the validation set loss increases for 20 consecutive iterations), and the test set is used to evaluate model accuracy. The error between the completed data and the actual measurements is required to be controlled within 5% (test set MSE≤0.0025). After substituting the completed full lifecycle data into the sub-model of individual batteries, the interface film growth characteristic parameter of lithium iron phosphate batteries was corrected to 0.0035 / cycle, and the electrode dissolution rate coefficient of aqueous hydrogen ion batteries was corrected to 0.0012 / cycle. The correction process used the least squares method to ensure that the goodness of fit between the model predictions and the completed data was R² ≥ 0.95. By using a deep learning model to complete the full lifecycle data under dynamic operating conditions, the problem of poor model adaptability caused by missing parameters under dynamic operating conditions was effectively solved. The measured data can only cover short-term static operating conditions. The complete data chain after completion can cover dynamic scenarios with alternating low and high loads in the laboratory. Furthermore, the model architecture, training parameters, and data processing flow were clearly defined, enabling the sub-model of individual batteries to accurately adapt to actual operating conditions and improve the versatility and accuracy of lifespan calculation.

[0042] Furthermore, the preset judgment rules include: using the lithium iron phosphate battery as the key unit for lifespan determination, and using its capacity decay to 80% of the initial capacity as the basic judgment criterion; if the corrected lifespan of the aqueous hydrogen ion battery is longer than that of the lithium iron phosphate battery, then the discharge depth of the aqueous hydrogen ion battery is adjusted, and the capacity output data of the adjusted aqueous hydrogen ion battery is obtained to determine whether the data can compensate for the capacity decay gap of the lithium iron phosphate battery; when the capacity of the lithium iron phosphate battery decays to 80% of the initial capacity, or the system charge / discharge efficiency drops to 85%, the system is determined to have reached the end of its lifespan. In one example, the initial capacity of the lithium iron phosphate battery is 100Ah, and the basic judgment criterion is triggered when it decays to 80Ah.If the corrected life of the lithium iron phosphate battery is 3500 cycles and the corrected life of the aqueous hydrogen ion battery is 4200 cycles, and the discharge depth of the aqueous hydrogen ion battery is adjusted from 40% to 55%, its capacity output data is increased from 40 Ah to 55 Ah. After calculation, it can compensate for the 10 Ah capacity attenuation gap of the lithium iron phosphate battery, and the system life can be corrected according to 3675 cycles. The specific algorithm logic of this adjustment process is as follows: First, calculate the capacity attenuation gap of the lithium iron phosphate battery ΔCFe = CFeinit×(1 - 80%) = 0.2×CFeinit (CFeinit is the initial capacity of the lithium iron phosphate battery, which is 100 Ah in the example, so ΔCFe = 20 Ah); Second, calculate the upper limit of the compensable capacity of the aqueous hydrogen ion battery ΔCWatermax = CWaterinit×(DODmax - DODcurrent), where CWaterinit is the initial capacity of the aqueous hydrogen ion battery (80 Ah in the example), DODmax is the maximum allowable discharge depth of the aqueous hydrogen ion battery (set to 80% to avoid accelerated aging caused by over-discharge), and DODcurrent is the current discharge depth (40% in the example), so ΔCWatermax = 80×(80% - 40%) = 32 Ah; Third, judge the feasibility of compensating the capacity gap: Since ΔCFe (20 Ah) ≤ ΔCWatermax (32 Ah), the adjusted discharge depth DODnew = DODcurrent + ΔCFe / CWaterinit = 40% + 20 / 80 = 65%. In the example, adjusting to 55% is a reasonable fine-tuning after comprehensively considering the battery aging rate and system stability. If ΔCFe > ΔCWatermax, then DODnew = DODmax, and the uncompensated capacity gap ΔCunfilled = ΔCFe - ΔCWatermax. The system life is corrected according to "the life of the lithium iron phosphate battery × (1 - ΔCunfilled / ΔCFe)"; Fourth, verify the stability after adjustment: Calculate the cycle life attenuation rate of the aqueous hydrogen ion battery after adjustment ΔLWater = (DODnew - DODcurrent)×0.005×LWater (0.005 is the correlation coefficient between the discharge depth and life attenuation fitted based on experimental data). In the example, ΔLWater = (55% - 40%)×0.005×4200 = 31.5 cycles, and after adjustment, LWater' = 4200 - 31.5 = 4168.5 cycles ≥ 3500 cycles, confirming that the adjustment is effective; If LWater' < LFe after adjustment, recalculate DODnew = DODcurrent + (LWater - LFe) / (0.005×LWater). When the capacity of the lithium iron phosphate battery drops to 80 Ah, or the charge-discharge efficiency of the system drops to 84%, directly determine that the system life ends.

[0043] Furthermore, the model iterative optimization includes: comparing the actual operating data of the hybrid energy storage system with the expected lifetime estimate in real time to analyze the sources of deviation; updating the parameters of the interaction influencing factor library based on the latest actual operating data after each preset number of cycles; and simultaneously correcting the coefficients of the battery aging sub-model. In one example, the preset number of cycles is 100. After the system actually runs for 100 cycles, the actual capacity retention rate of the lithium iron phosphate battery is 98.5%, while the expected estimate is 97.8%. The source of deviation is the temperature correction coefficient deviation caused by ambient temperature fluctuations. Based on the actual data, the factor corresponding to a power allocation ratio of 20% in the interaction factor library was updated to 1.12, and the active material loss coefficient of the lithium iron phosphate battery aging sub-model was simultaneously corrected to 0.0028 / cycle. During the iteration process, the deviation analysis adopted the deviation rate ε = |actual value - estimated value| / actual value. If ε > 5%, it is necessary to focus on analyzing the sources such as temperature fluctuations, power allocation strategy adjustments, and individual battery differences. Every 500 cycles, the newly added actual running data needs to be added to the training set of the BiLSTM model, and the model is retrained to optimize the network weights and biases, ensuring that the data completion accuracy continues to meet the standard. The update of the interaction factor library requires recalculating the actual capacity decay rate λact under different power allocation ratios, updating the coefficients a and b of the mapping model according to the aforementioned fitting algorithm, and then updating the factor library values. The coefficient correction of the battery aging sub-model adopts the gradient descent method to make the deviation rate between the model prediction value and the actual value ≤ 5%. Through the closed-loop comparison between the actual running data and the estimation results, the model parameters are continuously optimized, further reducing the cumulative error caused by various problems, so that the lifetime estimation accuracy continues to improve with the operation of the system, adapting to the needs of long-term stable operation in the laboratory.

[0044] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for estimating the expected lifetime of a hybrid energy storage system, characterized in that, include: Define the basic parameters and operating boundary conditions of a single battery in a hybrid energy storage system, which consists of an aqueous hydrogen-ion battery and a lithium iron phosphate battery. Accelerate the collection of operational data through mixed operating conditions; Calculate the differential temperature correction coefficient; quantify and form a library of interactive influencing factors; Complete the full lifecycle data and optimize the battery aging sub-model; Calculate the corrected lifetime of a single cell by integrating multi-dimensional correction factors; The expected lifespan of the system is determined by combining preset judgment rules; Iterative optimization of the model is achieved through comparison with actual operational data; The quantitative formation of the interaction factor library includes: acquiring capacity decay rate data of two types of batteries under different power allocation ratios; establishing a correspondence between power allocation ratio and interaction factors based on the capacity decay rate data; constructing a mapping model characterizing the relationship between power allocation ratio and interaction factors through linear fitting or nonlinear fitting algorithms; and extracting interaction factors corresponding to different operating condition combinations and incorporating them into the interaction factor library. The completion of the full life cycle data and optimization of the sub-model include: the full life cycle data includes the capacity retention rate, internal resistance change value, charge and discharge efficiency decay curve, electrode material wear degree, and electrolyte performance parameter evolution data of the two types of batteries under different cycle numbers; a BiLSTM neural network is used as a deep learning model, and the input features include battery cycle number, actual operating temperature, charge and discharge rate, and depth of discharge; the full life cycle data of the two types of batteries is completed through this model, and the error between the completed data and the measured data is controlled within a preset threshold; the completed full life cycle data is substituted into the battery aging sub-model to correct the model parameters.

2. The method for estimating the expected lifetime of a hybrid energy storage system according to claim 1, characterized in that, The basic parameters of the single battery include: the interface film growth characteristics and active material loss coefficient of lithium iron phosphate batteries; the electrode dissolution rate coefficient and electrolyte performance degradation coefficient of aqueous hydrogen ion batteries; and the operating boundary conditions include the average number of charge-discharge cycles per day, the power distribution ratio range, the ambient temperature range, and the system charge-discharge efficiency threshold.

3. The method for estimating the expected lifetime of a hybrid energy storage system according to claim 1, characterized in that, The mixed-condition accelerated test includes: designing mixed-conditions with multiple sets of variables, including the power ratio of the aqueous hydrogen-ion battery, the system ambient temperature, and the average number of daily cycles; the test cycle of each condition covers the early to mid-stages of battery accelerated aging; and simultaneously collecting the actual operating temperature, charge / discharge rate, depth of discharge, and capacity decay rate of the two types of batteries, as well as system-level charge / discharge efficiency data.

4. The method for estimating the expected lifetime of a hybrid energy storage system according to claim 1, characterized in that, The calculation of the differential temperature correction coefficient includes: obtaining the deviation between the actual operating temperature and the reference temperature of the two types of batteries; and calculating the targeted differential temperature correction coefficient based on the temperature sensitivity difference between the two types of batteries.

5. The method for estimating the expected lifetime of a hybrid energy storage system according to claim 1, characterized in that, The preset judgment rules include: using lithium iron phosphate batteries as the key unit for lifespan determination, and using the capacity decay to 80% of the initial capacity as the basic judgment standard; if the corrected lifespan of the aqueous hydrogen ion battery is longer than that of the lithium iron phosphate battery, then the discharge depth of the aqueous hydrogen ion battery is adjusted, the capacity output data of the adjusted aqueous hydrogen ion battery is obtained, and it is determined whether the data can make up for the capacity decay gap of the lithium iron phosphate battery; when the capacity of the lithium iron phosphate battery decays to 80% of the initial capacity, or the system charge and discharge efficiency drops to 85%, the system is determined to have reached the end of its lifespan.

6. The method for estimating the expected lifetime of a hybrid energy storage system according to claim 1, characterized in that, The model iterative optimization includes: comparing the actual operating data of the hybrid energy storage system with the expected lifetime estimation results in real time and analyzing the sources of deviation; updating the parameters of the interaction influence factor library based on the latest actual operating data after each preset number of cycles; and simultaneously correcting the coefficients of the battery aging sub-model.

7. A system for estimating the expected lifetime of a hybrid energy storage system, characterized in that, Implementing the expected lifetime estimation method for a hybrid energy storage system as described in any one of claims 1-6, comprising: The parameter definition and model building module defines the basic parameters and operating boundary conditions of a single cell and builds a cell aging sub-model. The data acquisition and processing module conducts accelerated testing under mixed operating conditions, collects operational data, calculates differential temperature correction coefficients, and quantifies and forms a library of interactive influencing factors. The data completion and model optimization module completes the full lifecycle data and optimizes the battery aging sub-model. The lifespan estimation module integrates differential temperature correction coefficients and multi-dimensional correction factors of interactive influence factors to calculate the corrected lifespan of a single cell and determine the expected lifespan of the system by combining preset judgment rules. The iterative optimization module compares the actual operating data of the hybrid energy storage system with the estimated expected lifespan, analyzes the sources of deviation, and updates the interactive influencing factor library and the parameters of the battery aging sub-model.

Images