Hydrogen storage alloy, hydrogen storage alloy electrode, and sodium-ion battery

Abstract

The application discloses a water-based sodium ion battery equalization control method, equipment and device, relates to the water-based sodium ion battery control technical field, and comprises the following steps: collecting the operation data of each single battery in the battery pack, constructing a single battery state prediction model, and predicting the SOC change trend of the single battery in the future; according to the SOC change trend, the priority equalization object of the single battery is obtained, and the equalization scheduling of the priority equalization object is corrected based on the SOH change trend; according to the corrected priority, the energy of the priority equalization object is transferred to the energy storage element; based on the fuzzy game method, the energy in the energy storage element is released to the low-priority single battery; the application combines the SOC prediction, the SOH priority correction, the energy transfer and release method combining the fuzzy game and the reinforcement learning, and solves the problem that the single SOC of the battery pack is not balanced and the service life decline is difficult to consider.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of aqueous sodium-ion battery control, and more particularly, relates to an aqueous sodium-ion battery equalization control method, device and apparatus. BACKGROUND

[0002] Aqueous sodium-ion batteries (ASIB) have become an important candidate for new energy storage systems due to their abundant materials, low cost, high safety, and excellent environmental performance. However, in practical applications, aqueous sodium-ion batteries have the problems of uneven single battery SOC (state of charge) and large differences in state of health (SOH), which directly affect the overall performance and service life of the battery pack. As the number of cycles increases, the SOC difference between single batteries gradually increases, and some single batteries may prematurely degrade or exhibit overcharging / overdischarging phenomena, resulting in accelerated battery pack capacity degradation, increased safety hazards, and decreased energy storage efficiency.

[0003] Currently, battery equalization technology mainly includes passive equalization and active equalization. Passive equalization adjusts SOC through the method of dissipating excess energy by resistance, which is simple to control, but has low energy utilization rate and cannot effectively prolong battery life; active equalization redistributes electrical energy between single batteries through energy transfer or energy storage, which has high energy utilization rate, but existing technologies generally use fixed thresholds or control strategies based on current SOC, lack predictive consideration of future SOC trends and single battery health, and equalization actions are prone to be excessive or insufficient, making it difficult to balance efficiency while considering battery life and safety constraints. SUMMARY

[0004] To overcome the above-mentioned defects of the prior art, embodiments of the present application provide an aqueous sodium-ion battery equalization control method, device and apparatus, which solves the problem of balancing the SOC of single batteries in a battery pack and the difficulty of considering battery life by combining SOC prediction, SOH priority correction, and energy transfer and release methods combining fuzzy game and reinforcement learning.

[0005] To achieve the above-mentioned purposes, the present application provides the following technical solutions:

[0006] In a first aspect, the application provides a water-based sodium-ion battery equalization control method, which comprises: collecting the operation data of each single battery in the battery pack, constructing a single battery state prediction model, predicting the SOC change trend of the single battery in the future period of time; according to the SOC change trend, obtaining the priority equalization object of the single battery, and correcting the equalization scheduling priority of the priority equalization object based on the SOH change trend, the priority correction is to construct an SOH prediction model through a multi-layer LSTM network, predict the SOH change trend, and dynamically correct the equalization scheduling priority based on the SOH deviation and the SOC deviation; according to the corrected priority, the energy of the priority equalization object is transferred to the energy storage element; based on the fuzzy game method, the energy in the energy storage element is discharged to the low-priority single battery.

[0007] In one embodiment, a single battery state prediction model is constructed to predict the SOC change trend of the single battery in the future period of time, specifically: the operation data is preprocessed, the key feature parameters are extracted according to the processed data, the key feature parameters include voltage feature, charge and discharge capacity cumulative value feature, instantaneous temperature change feature and internal resistance change feature; a single battery state prediction model is constructed based on the key feature parameters; based on the single battery state prediction model, the SOC change trend of each single battery in the future period of time is predicted in combination with the real-time operation data.

[0008] In one embodiment, according to the SOC change trend, the priority equalization object of the single battery is obtained, specifically: based on the SOC change trend, the SOC difference value between any two single batteries in the battery pack is obtained; based on the reinforcement learning algorithm, the optimal dynamic equalization threshold is generated according to the operation state of the battery pack; the SOC difference value is compared with the optimal dynamic equalization threshold, and the potential unbalanced battery is identified; the SOC deviation of the potential unbalanced battery is calculated, and the priority equalization object list is generated based on the deviation.

[0009] In one embodiment, the priority of the equalization scheduling of the priority equalization object is corrected based on the SOH change trend, specifically: an SOH prediction model is constructed based on a multi-layer LSTM network to obtain the future SOH change trend of each single battery in the priority equalization object list; according to the SOH change trend, the deviation of the SOH of each single battery is determined; based on the deviation of the SOH, the original priority of each single battery in the priority equalization object list is dynamically corrected to generate a comprehensive priority; the priority equalization object list is updated and sorted according to the comprehensive priority.

[0010] In one of the embodiments, the energy of the priority equalization object is transferred to the energy storage element, specifically: selecting an energy transfer object according to the updated priority equalization object list; connecting the energy transfer object and the energy storage element through the control of the bidirectional switch module to form an energy transfer loop; controlling the switch module through the intelligent control strategy to transfer the energy of the energy transfer object to the energy storage element through the energy transfer loop; monitoring the energy state of the energy storage element, and adjusting the energy transfer process according to the error between it and the preset target value until the energy transfer is completed.

[0011] In one of the embodiments, the energy of the energy transfer object is transferred to the energy storage element through the intelligent control strategy, specifically: defining the battery pack state as a state vector and defining the energy transfer adjustment parameter as an action to construct an energy transfer reward function; training the state vector and the action to output the optimal energy transfer adjustment parameter; establishing a dynamic mathematical model of the energy transfer system, the dynamic mathematical model including operation constraints; performing MPC optimization on the optimal energy transfer adjustment parameter based on the dynamic mathematical model to generate a final control instruction that meets the operation constraints; and performing the energy transfer operation according to the final control instruction.

[0012] In one of the embodiments, the energy in the energy storage element is discharged to the low-priority single battery based on the fuzzy game method, specifically: determining a target single battery according to the corrected priority equalization object list; modeling the discharging process of the energy storage element as a multi-objective optimization problem and using fuzzy logic to express it; using the Bayesian optimization method to adaptively optimize the membership function of the fuzzy logic to obtain an optimized fuzzy inference model; generating a candidate solution of the discharging control parameter based on the optimized fuzzy inference model and real-time data of the system; optimizing the candidate solution to output a final discharging control scheme; and controlling the energy storage element to release energy to the target single battery according to the final discharging control scheme.

[0013] In one of the embodiments, the membership function of the fuzzy logic is adaptively optimized using the Bayesian optimization method, specifically: constructing a membership optimization objective function, the membership optimization objective function including the SOC difference variation after equalization, the energy transfer efficiency, and the temperature rise amplitude; setting an initial search space of the to-be-optimized parameters of the membership function; based on the Bayesian optimization framework, iteratively searching for the optimal membership function parameters through Gaussian process modeling and function guidance; and when the convergence condition is met, outputting the membership function parameters that optimize the objective function.

[0014] In a second aspect, the application provides a device for water-based sodium ion battery equalization control method, the device comprising: an SOC prediction module for collecting operation data of each single battery in the battery pack, constructing a single battery state prediction model, and predicting the SOC change trend of the single battery in the future; an equalization object selection module for obtaining a priority equalization object of the single battery according to the SOC change trend, and correcting the equalization scheduling priority of the priority equalization object based on the SOH change trend, wherein the priority correction is performed by constructing an SOH prediction model through a multi-layer LSTM network, predicting the SOH change trend, and dynamically correcting the equalization scheduling priority based on the SOH deviation and the SOC deviation; an energy transfer module for transferring the energy of the priority equalization object to an energy storage element according to the corrected priority; and an energy release module for releasing the energy in the energy storage element to a low-priority single battery based on a fuzzy game method.

[0015] In a third aspect, the application provides an electronic device, comprising:

[0016] a memory for storing a computer program;

[0017] a processor for executing the computer program to implement the steps of the water-based sodium ion battery equalization control method provided in the foregoing aspects

[0018] As can be seen from the above technical solutions, the embodiments of the application have the following advantages:

[0019] Through the multi-level, closed-loop, and intelligent equalization control strategy, accurate management of the single batteries in the battery pack is achieved. First, by collecting high-precision voltage, current, temperature, internal resistance, and historical cycle data, and constructing an SOC prediction model, the future SOC change of the single battery is accurately predicted, so that potential unbalanced batteries can be identified in advance. Second, the dynamic priority correction of the priority equalization object is performed in combination with the SOH change trend, so that healthy batteries can undertake more equalization tasks and degraded batteries can reduce the load, thereby balancing the battery life and equalization efficiency. Third, the high-priority single battery energy is efficiently and safely transferred to the energy storage element by using the reinforcement learning strategy and MPC optimization control bidirectional switch module, and then the energy in the energy storage element is accurately distributed to the low-priority single battery by using the fuzzy game method combined with the Bayesian optimization adjustment membership function, thereby achieving closed-loop equalization. The overall scheme has the advantages of predictability, intelligence, adaptability, and high efficiency and safety, and significantly improves the overall equalization effect of the battery pack, prolongs the life, and reduces the energy consumption, and is more accurate and reliable than the traditional fixed threshold or single SOC equalization method. BRIEF DESCRIPTION OF DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings described below only constitute a part of the embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of the provided drawings.

[0021] Figure 1 The water-based sodium ion battery equalization control method flowchart provided in the embodiments of the present application.

[0022] Figure 2 The water-based sodium ion battery equalization control device structure diagram provided in the embodiments of the present application.

[0023] Figure 3 The electronic device structure diagram provided in the embodiments of the present application. DETAILED DESCRIPTION

[0024] The technical solutions in the embodiments of the present application will be described clearly and completely in the following with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments only constitute a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0025] Referring to Figure 1 The water-based sodium ion battery equalization control method flowchart provided in the embodiments of the present application includes the following steps:

[0026] S1, collecting the operation data of each single battery in the battery pack, constructing a single battery state prediction model, and predicting the SOC (state of charge) change trend of the single battery in the future period of time.

[0027] In the embodiments, the operation data of each single battery in the battery pack is collected, a single battery state prediction model is constructed, and the SOC change trend of the single battery in the future period of time is predicted, specifically:

[0028] The operation data of each single battery in the battery pack is collected in real time, and the operation data includes voltage, current, temperature, internal resistance and historical cycle data;

[0029] Among them, high-precision sampling sensors are used, and the historical cycle data refers to the long-term operation record generated by the single battery in all past charging and discharging cycles, including the charging and discharging capacity each time, the cycle number, the voltage, current and temperature curve in each charging and discharging process.

[0030] The operation data is preprocessed, and the preprocessing includes noise elimination, outlier rejection and normalization processing;

[0031] According to the processed data, key characteristic parameters for predicting the SOC are extracted, including voltage characteristics, charge-discharge capacity cumulative value characteristics, instantaneous temperature change characteristics, and internal resistance change characteristics;

[0032] Among them, the key characteristic parameters for predicting the SOC extracted according to the processed data are specifically:

[0033] According to the preprocessed single battery voltage data, an instantaneous voltage change rate is calculated;

[0034] The specific calculation formula of the instantaneous voltage change rate is as follows:

[0035]

[0036] In the formula, is the instantaneous voltage change rate, is the current sampling time single battery voltage, is the voltage at the last sampling time, is the sampling time interval.

[0037] The second-order change rate of the voltage is calculated based on the finite difference method through the instantaneous voltage change rate, which is used to capture the nonlinear fluctuation characteristics in the charge-discharge process;

[0038] The specific calculation formula of the second-order change rate of the voltage is as follows:

[0039]

[0040] In the formula, is the second-order change rate of the voltage, is the voltage at the next sampling time.

[0041] The instantaneous voltage change rate and the second-order change rate are taken as voltage characteristics;

[0042] According to the charge-discharge single battery current of each cycle and the charge-discharge cycle duration, the charge-discharge cumulative capacity is calculated, and the different cycle charge-discharge cumulative capacities are normalized to form a charge-discharge cumulative capacity change trend;

[0043] The charge-discharge cumulative capacity and its change trend are taken as capacity characteristics;

[0044] The specific calculation formula of the charge-discharge cumulative capacity is as follows:

[0045]

[0046] In the formula, is the charge-discharge cumulative capacity, is the single battery current, This refers to the duration of the charge / discharge cycle.

[0047] Trend analysis was performed on the temperature curve of each individual cell to calculate the instantaneous temperature change rate, which was used as the instantaneous temperature change characteristic.

[0048] The instantaneous temperature change rate is calculated using the following formula:

[0049]

[0050] In the formula, The instantaneous rate of temperature change The temperature at the current sampling time. The temperature at the previous sampling time.

[0051] For the internal resistance data of a single cell, calculate the instantaneous rate of change of internal resistance as a characteristic of internal resistance change;

[0052] The instantaneous rate of change of internal resistance is calculated using the following formula:

[0053]

[0054] In the formula, The instantaneous rate of change of internal resistance, The internal resistance of the single cell at the current sampling time. This is the internal resistance at the previous sampling time.

[0055] A single-cell state prediction model is constructed based on key feature parameters. The key feature parameters are input into the model for training to obtain the ability to predict the change of single-cell state of charge (SOC) over time or cycle number.

[0056] Based on the single-cell state prediction model, and combined with real-time operating data, the SOC change trend (SOC curve) of each single cell is predicted over a period of time, and the prediction results are output. The prediction results include the SOC high and low points reached by each cell in the future charge and discharge stages, the predicted SOC value, and the SOC change rate.

[0057] Furthermore, a single-cell state prediction model is constructed based on key feature parameters, specifically as follows:

[0058] Correlation analysis is performed on key feature parameters to screen out higher-order features. The screening method is to check whether the Pearson correlation coefficient between the key feature parameters and the SOC of a single cell exceeds a preset screening threshold. If it exceeds the threshold, the feature is retained; otherwise, it is discarded.

[0059] Construct a multidimensional feature matrix based on higher-order features;

[0060] An ensemble learning model based on time series regression is used to capture the nonlinear relationship between the feature matrix and SOC changes, thereby constructing a single-cell state prediction model.

[0061] The training objective of the model is determined to minimize the SOC prediction error, and the mean squared error is used as the loss function.

[0062] The multidimensional feature matrix is ​​divided into training set, validation set and test set to ensure that the data is evenly distributed and free from temporal overlap interference.

[0063] The training set is input into the single-cell state prediction model, and the model parameters are optimized by cross-validation. The model parameters include maximum depth, learning rate, and subsample ratio.

[0064] During training, the model parameters are dynamically adjusted based on the validation set error to avoid overfitting, improve the model's generalization ability, and obtain a trained single-cell state prediction model.

[0065] S2, based on the SOC change trend, obtain the priority balancing objects of the individual cells, and adjust the priority balancing scheduling of the priority balancing objects based on the SOH (state of health) change trend.

[0066] In this embodiment, based on the SOC change trend, the priority balancing target of a single battery cell is obtained, specifically as follows:

[0067] Prediction time intervals are obtained based on the future SOC change trends of each individual cell. The SOC curve within the range is calculated, and the SOC difference between any two monomers is calculated.

[0068] The specific formula for calculating the SOC difference is as follows:

[0069]

[0070] In the formula, Let $\frac{i}{j}$ be the average SOC difference between the $i$-th and $j$-th individual cells in the battery pack during the predicted time interval. For the predicted SOC of the i-th single cell at time t, For the predicted SOC of the j-th cell at time t, To predict the start time, To predict the termination time.

[0071] The operating state of the battery pack is defined as a state vector, which includes the current SOC of each individual cell, the temperature of each individual cell, the internal resistance of each individual cell, and the historical balancing actions and balancing effects.

[0072] Actions are defined as dynamic equilibrium thresholds, and a reward function is constructed to measure the impact of each action on the equilibrium effect. The reward function includes the change in SOC difference, equilibrium energy consumption, and battery degradation loss.

[0073] The reward function is calculated using the following formula:

[0074]

[0075] In the formula, For the reward function value, This refers to the change in SOC difference after equalization, that is, the reduction in the SOC difference between individual cells after the battery pack has undergone equalization control. For energy consumption, For battery degradation losses, , , These are the weighting coefficients.

[0076] By using reinforcement learning algorithms to train policies based on states, actions, and reward functions, the dynamic equilibrium threshold selected under different operating states can maximize the cumulative reward.

[0077] Output the optimal dynamic equilibrium threshold based on the trained strategy;

[0078] The SOC difference is compared with the dynamic equilibrium threshold. If the SOC difference is greater than the dynamic equilibrium threshold, it is determined that there is a potential imbalance in the individual cell in the future.

[0079] For individual cells identified as potentially unbalanced, the deviation of their SOC from the average SOC of the battery pack is obtained.

[0080] The specific formula for calculating the SOC deviation is as follows:

[0081]

[0082] In the formula, This represents the SOC deviation. The predicted SOC value for cell i. This represents the predicted average SOC value of the battery pack.

[0083] Sort the objects in descending order based on their deviation to obtain priority balancing objects. The objects with the greater deviation are determined as priority balancing objects, and a priority balancing object list is formed.

[0084] It should be noted that the method of selecting priority equalization targets for individual cells by combining reinforcement learning algorithms with SOC change trends and deviations has the following main advantages: it can dynamically and adaptively optimize the equalization threshold based on the overall operating state of the battery pack (SOC, temperature, internal resistance, and historical equalization effects). This allows the equalization action to effectively reduce the SOC differences between individual cells while also taking into account energy consumption and battery life. As a result, it can identify potentially unbalanced cells in advance and achieve predictive, intelligent, and forward-looking priority equalization control. Compared with traditional methods that use fixed thresholds or are simply based on the current SOC, this method is more accurate, efficient, and has less impact on battery degradation.

[0085] Furthermore, based on the SOH change trend, priority adjustments are made to the balance scheduling of priority balance objects, specifically as follows:

[0086] The input feature sequence for training LSTM is constructed based on the historical cycle data of the battery pack.

[0087] Construct a multi-layer LSTM network to learn the long-term dependency characteristics of battery SOH evolution;

[0088] The network is configured with an input layer, a hidden layer (containing several LSTM units), and an output layer. The output is the predicted SOH sequence. The network is iteratively optimized using the input feature sequence to minimize the SOH loss function, thus obtaining the trained SOH prediction model.

[0089] The specific calculation formula for the SOH loss function is as follows:

[0090]

[0091] In the formula, The training loss function value measures the error between the SOH predicted by the LSTM model and the true SOH, where N is the total number of training samples. Let SOH be the predicted SOH value of the i-th individual cell at a certain time point. The actual SOH value of the i-th single cell is usually obtained through experimental measurement or historical data. is the aging rate weighting coefficient for the i-th sample.

[0092] The aging rate weighting coefficient can be determined based on the historical capacity degradation rate. The specific calculation formula for the aging rate weighting coefficient is as follows:

[0093]

[0094] In the formula, This is an empirical coefficient. This represents the SOH decay rate.

[0095] Based on the current running data of the priority balancing objects, input the trained SOH prediction model and output the SOH change trend of each priority balancing object over a period of time in the future;

[0096] The deviation of the predicted SOH is obtained based on the trend of SOH change;

[0097] The specific formula for calculating the deviation of the predicted SOH is as follows:

[0098]

[0099] In the formula, To predict the deviation of SOH, The average SOH of the battery pack.

[0100] Priority correction rules are constructed for the priority balancer list based on the deviation of the predicted SOH.

[0101] If the deviation of the SOH prediction of a single cell is lower than the preset health threshold, its equalization priority is reduced to avoid applying excessive equalization load to cells with severe degradation.

[0102] If the deviation of the SOH prediction of a single cell is higher than the preset health threshold, its balancing priority is increased, so that healthy cells can undertake more balancing tasks.

[0103] Among them, the preset health threshold can be determined through empirical values ​​and statistical analysis;

[0104] The SOH correction function is generated based on the priority correction rule, and the overall priority is obtained by weighted calculation in combination with the SOC deviation.

[0105] The specific calculation formula for the SOH correction function is as follows:

[0106]

[0107] The specific calculation formula for the overall priority is as follows:

[0108]

[0109] In the formula, For overall priority, The weighting coefficient for SOC deviation. The weighting coefficients for the SOH correction function, the SOC deviation, and the SOH correction function can all be set through engineering experience.

[0110] The priority balance object list is revised based on the overall priority, and then sorted in descending order according to the overall priority to obtain the revised priority balance object list.

[0111] It should be noted that prioritizing the equalization scheduling of individual cells by combining the SOC change trend and the SOH change trend has the following advantages: It can not only identify potentially unbalanced individual cells in advance based on the prediction of SOC changes in the future period and achieve predictive equalization control, but also dynamically adjust the equalization order by prioritizing the change trend of the individual cell's state of health (SOH). This allows healthy cells to take on more equalization tasks and degraded cells to reduce their load, thereby taking into account SOC equalization efficiency, battery life and energy consumption. This achieves intelligent, forward-looking, efficient and lifespan-friendly battery pack equalization management, which is more accurate and reliable than traditional methods based solely on the current SOC or a fixed threshold.

[0112] S3, based on the corrected priority, transfers the energy of the priority balancing object to the energy storage element.

[0113] In this embodiment, according to the modified priority, the energy of the priority balancing object is transferred to the energy storage element, specifically as follows:

[0114] Based on the revised priority balancing object list, select the individual cell with the highest SOC value and overall priority that is higher than the average SOC value of the battery pack as the energy transfer object.

[0115] Initialize the energy storage element and place it in an energy receiving state. The energy storage element also includes a bidirectional switch module that controls the energy storage element. The bidirectional switch module is kept in an initially open state to prevent energy loss without control. The capacity of the energy storage element should meet the maximum energy transferable by a single cell.

[0116] A bidirectional switching module connects high-priority individual cells to energy storage elements, forming an energy transfer loop;

[0117] Based on a bidirectional switching control combining reinforcement learning strategy and model predictive control, the energy of high-priority individual cells is transferred to the energy storage element.

[0118] The energy transfer is judged to determine whether the expected energy transfer has been achieved by acquiring the energy of the energy storage element in real time and comparing it with the preset target energy transfer energy. The target energy transfer energy is obtained from the energy that the high-priority single battery cell needs to transfer.

[0119] If the error is within the allowable error range, the energy transfer is complete. If it is not within the allowable error range, the energy transfer steps are repeated until the error is within the allowable error range, at which point the transfer stops.

[0120] Furthermore, based on bidirectional switching control combining reinforcement learning strategies and model predictive control, the energy of high-priority individual cells is transferred to the energy storage element, specifically as follows:

[0121] The battery pack state is defined as a state vector, which includes the predicted SOC value, predicted SOH value, voltage, temperature, internal resistance of individual cells, as well as historical energy transfer data and historical equalization actions.

[0122] The action is defined as an energy transfer regulation parameter, which includes the transfer current and the transfer duration.

[0123] Construct an energy transfer reward function, which includes the temperature rise, energy transfer efficiency, and equilibrium response time;

[0124] The energy transfer reward function is calculated using the following formula:

[0125]

[0126] In the formula, The value of the energy transfer reward function. This represents the temperature rise of a single battery cell during energy transfer. For energy transfer efficiency, To balance the response time, i.e. the time required from detecting that the SOC has reached the threshold to completing the energy transfer, , , These are the weights.

[0127] Training is based on reinforcement learning to maximize cumulative reward and output the optimal action strategy, i.e., the optimal energy transfer adjustment parameters, under different states;

[0128] A dynamic mathematical model of energy storage elements and individual batteries is established, the dynamic mathematical model including voltage constraints, current constraints and power constraints;

[0129] The optimal energy transfer adjustment parameters of the reinforcement learning output are optimized by MPC to ensure that the constraints of the dynamic mathematical model are met.

[0130] Specifically, the voltage constraint is:

[0131]

[0132]

[0133]

[0134] In the formula, Let be the instantaneous terminal voltage of the i-th individual cell in the battery pack. This is the minimum allowable voltage for a single cell. This represents the maximum permissible voltage of a single cell. For energy transfer current, For the maximum allowable transfer current, The current energy of the energy storage element. This represents the maximum permissible energy of the energy storage element.

[0135] It should be noted that voltage constraint protects individual battery cells, ensuring they do not exceed safe operating ranges during balancing; current constraint protects energy storage components and switching devices, preventing overcurrent damage; and energy storage constraint prevents overcharging or energy overflow of energy storage components, ensuring safe temporary energy storage. MPC (Model Predictive Control) is responsible for rapidly closing-loop adjusting the transfer current and time to prevent overcharging, over-discharging, and overload of energy storage components. MPC's rapid closed-loop adjustment of transfer current and time can be understood as predicting the energy transfer effect within a few sampling steps based on the current battery voltage, temperature, internal resistance, and energy storage component state. During the prediction process, if a certain current is expected to cause battery over-discharge or energy storage component overcharging, the MPC will automatically adjust... Within a safe range, while ensuring energy transfer as fully as possible, achieving both high efficiency and safety; if the prediction indicates that the battery will overheat or reach the safe voltage limit at a certain time, the MPC will shorten the transfer time in advance to avoid overcharging or excessive temperature rise, ensuring intelligent control.

[0136] Based on the reinforcement learning strategy and MPC optimization results, the final control command is generated to control the bidirectional switching module, which transfers the energy of high-priority individual battery cells to the energy storage element.

[0137] It should be noted that by controlling the bidirectional switching module through reinforcement learning strategies and MPC optimization results, fast, accurate, and adaptive energy transfer can be achieved while ensuring the safety constraints of the battery and energy storage components. The reinforcement learning strategy optimizes the process based on a reward function that considers the temperature rise, energy transfer efficiency, and equalization response time. This ensures that each transfer action can quickly complete equalization while reducing temperature rise, improving transfer efficiency, and shortening response time. MPC performs real-time closed-loop optimization of the actions to prevent overcharging, over-discharging, and overload of energy storage components. At the same time, it dynamically adjusts the transfer current and duration to achieve efficient, safe, and intelligent equalization control.

[0138] S4, based on the fuzzy game theory method, releases the energy in the energy storage element to the low-priority single cell.

[0139] In this embodiment, the energy in the energy storage element is released to the low-priority single battery cell based on the fuzzy game theory method, specifically as follows:

[0140] Based on the revised priority balancing object list, select the single cell with the lowest priority as the target;

[0141] The energy release process of energy storage elements is modeled as a multi-objective optimization problem, namely a fuzzy inference model. The multi-objective optimization problem includes low-priority single-unit SOC improvement demand, energy storage element lifetime maintenance, and system thermal safety constraints. The multi-objective optimization problem is converted into fuzzy linguistic variables, namely high / medium / low demand, high / medium / low lifetime risk, and high / medium / low temperature rise risk.

[0142] A fuzzy rule base is constructed based on expert experience. The rule form is: if the input conditions are met, the output is the energy release intensity level. The input conditions include SOC deviation, energy storage element life risk and temperature rise risk. The output is high, medium and low energy release intensity.

[0143] The membership function is adaptively tuned using the Bayesian optimization method to output the optimal membership function and obtain the optimized fuzzy inference model.

[0144] The first data of the energy storage element collected in real time is fuzzified through an optimized fuzzy inference model to output a fuzzy set of energy release intensities. Candidate solutions for energy release current and duration are obtained through a defuzzification method. The first data of the energy storage element includes the energy storage element voltage, current, temperature and SOC deviation.

[0145] A fast game layer is constructed, which includes a fast game utility function. Candidate solutions are input into the fast game layer, and the fast equilibrium candidate solution set is output by iterative calculation using an approximate game method combined with the fast game utility function.

[0146] The specific calculation formula for the fast game utility function is as follows:

[0147]

[0148] The specific calculation formula for the iterative calculation is as follows:

[0149]

[0150] In the formula, Let be the utility value of the i-th individual cell in the fast game layer, used to measure the quality of the current candidate solution. The weighting coefficient for the change in SOC difference. The weighting coefficient for the temperature rise. This is a candidate solution for the i-th cell in the m-th iteration.

[0151] A slow game layer is constructed, which includes a slow game utility function. The fast equilibrium candidate solution set is input into the slow game layer for correction, and the comprehensive energy release current and duration scheme is output.

[0152] The energy storage element is controlled to release energy to the lowest priority individual cells based on a scheme that combines the energy release current and duration.

[0153] The specific formula for calculating the slow game utility function is as follows:

[0154]

[0155] The specific calculation formula for the integrated energy release current and duration scheme is as follows:

[0156]

[0157] In the formula, For the slow game utility function value, This represents the state of harmonic oxygen (SOH) degradation of a single cell. The weighting coefficients of the SOC difference change in the slow game layer. The weighting coefficients for the temperature increase in the slow game layer. The weighting coefficient for SOH degradation. To integrate the energy release current and duration scheme, To quickly balance the candidate solution set, For correction factor, This represents the average slow game utility value.

[0158] It should be noted that the advantage of using the fuzzy game theory method to transfer energy from energy storage elements to low-priority individual cells is that it can comprehensively consider multiple objective constraints such as individual cell SOC deviation, energy storage element lifespan, and system temperature rise. Through fuzzification, complex continuous indicators are transformed into easily decision-making hierarchical inputs, and Bayesian optimization is used to adaptively tune the membership function to accurately generate candidate energy release strategies. Subsequently, by combining fast and slow game layers to iteratively optimize and prioritize candidate solutions, a comprehensive energy release current and duration scheme is obtained. Thus, while ensuring equalization efficiency, thermal safety, and battery life, intelligent energy compensation for low-priority individual cells is achieved, forming an adaptive closed-loop control, which significantly improves the overall equalization effect and long-term reliability of the battery pack.

[0159] Furthermore, a Bayesian optimization method is used to adaptively tune the membership function, outputting the optimal membership function, specifically as follows:

[0160] Based on the battery equalization control requirements, a membership degree optimization objective function is constructed. The membership degree optimization objective function includes the change in SOC difference after equalization, energy transfer efficiency, and temperature rise. By weighted summation, each performance index is unified into a single optimization objective function.

[0161] For the initial membership function (triangular, trapezoidal, or Gaussian function), a search space for the parameters to be optimized is defined, wherein the parameters to be optimized include the center position, width, and slope of the function;

[0162] Fuzzy inference is performed using the initial membership function to calculate the objective function value and form an initial sample set.

[0163] Based on the Bayesian optimization framework, the optimal membership function parameters are iteratively searched through Gaussian process modeling and acquisition function guidance.

[0164] After the convergence condition is met, the membership function parameters corresponding to the optimal objective function value are output, thus obtaining the optimal membership function.

[0165] The Bayesian optimization framework is specifically as follows:

[0166] Based on the initial sample set, a Gaussian process model is trained to approximately describe the mapping relationship between the membership function parameters and the objective function value, and outputs the predicted mean and uncertainty variance.

[0167] Using the Gaussian process model, a data acquisition function is constructed to balance the exploration of unknown parameter regions and the utilization of known good solutions; by maximizing the value of the data acquisition function, the next set of candidate parameters for the membership function is determined.

[0168] In fuzzy inference, candidate parameters are verified and the sample set is updated. This involves updating the candidate parameters and performing fuzzy inference, executing the battery balancing process, calculating the objective function value based on the running results, and adding the parameters and the objective function value as new samples to the sample set.

[0169] Based on the updated sample set, the Gaussian process model is retrained until a preset convergence condition is met. The convergence condition includes the objective function value changing less than a preset change threshold or the number of iterations reaching an upper limit.

[0170] It's worth noting that the advantage of adaptively tuning the membership function using Bayesian optimization is that it can adaptively reflect the battery pack's operating characteristics under different equalization scenarios. It considers key indicators such as SOC variation, energy transfer efficiency, and temperature rise uniformly, and continuously iterates and adjusts parameters through Bayesian optimization to achieve the optimal mapping, thus providing more accurate and reliable control output in fuzzy inference. Compared to traditional fixed membership functions, this method significantly improves the intelligence, efficiency, and safety of equalization control, generating highly adaptable and optimized energy release strategies under different operating conditions, thereby improving the overall battery pack lifespan and equalization effect.

[0171] Reference Figure 2As shown in the schematic diagram, the aqueous sodium-ion battery equalization control device provided by the present invention includes a SOC prediction module, an equalization object selection module, an energy transfer module, and an energy release module, with connections between the modules:

[0172] The SOC prediction module is used to collect the operating data of each individual cell in the battery pack, build a single cell state prediction model, and predict the SOC change trend of the single cell over a period of time.

[0173] The equalization object selection module is used to obtain the priority equalization object of the single cell according to the SOC change trend, and to perform priority correction on the equalization scheduling of the priority equalization object based on the SOH change trend. The priority correction is to construct an SOH prediction model through a multi-layer LSTM network to predict the SOH change trend of the battery, and to dynamically correct the equalization scheduling priority based on the SOH deviation and SOC deviation.

[0174] The energy transfer module is used to transfer the energy of the priority balancing object to the energy storage element according to the corrected priority.

[0175] The energy release module releases energy from the energy storage element to low-priority individual cells based on a fuzzy game theory method.

[0176] Furthermore, embodiments of this application also disclose an electronic device, Figure 3 This is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content of the diagram should not be construed as limiting the scope of this application.

[0177] Figure 3 This is a schematic diagram of the structure of an electronic device 20 provided in an embodiment of this application. Specifically, the electronic device 20 may include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 stores a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the aqueous sodium-ion battery equalization control method disclosed in any of the foregoing embodiments. Alternatively, the electronic device 20 in this embodiment may specifically be an electronic computer.

[0178] In this embodiment, the power supply 23 is used to provide operating voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and external devices, and the communication protocol it follows can be any communication protocol applicable to the technical solution of this application, and is not specifically limited here; the input / output interface 25 is used to acquire external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs, and is not specifically limited here.

[0179] The processor 21 may include one or more processing cores, such as a quad-core processor or an octa-core processor. The processor 21 may be implemented using at least one hardware form selected from DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). The processor 21 may also include a main processor and a coprocessor. The main processor, also known as a CPU (Central Processing Unit), is used to process data in the wake-up state; the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, the processor 21 may integrate a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the screen. In some embodiments, the processor 21 may also include an AI (Artificial Intelligence) processor, which is used to handle computational operations related to machine learning.

[0180] In addition, the memory 22, as a carrier for resource storage, can be a read-only memory, random access memory, disk or optical disk, etc. The resources stored thereon can include operating system 221, computer program 222, etc., and the storage method can be temporary storage or permanent storage.

[0181] The operating system 221 manages and controls the various hardware devices and computer programs 222 on the electronic device 20 to enable the processor 21 to perform calculations and processing on the massive amounts of data 223 in the memory 22. It can be Windows Server, Netware, Unix, Linux, etc. The computer program 222, in addition to including a computer program capable of performing the aqueous sodium-ion battery equalization control method executed by the electronic device 20 as disclosed in any of the foregoing embodiments, may further include computer programs capable of performing other specific tasks. The data 223 may include data received by the electronic device from external devices, as well as data collected by its own input / output interface 25.

[0182] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0183] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0184] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0185] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0186] 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.

[0187] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for equalization control of aqueous sodium-ion batteries, characterized in that, Includes the following steps: Collect operational data from each individual cell in the battery pack, construct a state prediction model for each individual cell, and predict the SOC change trend of each individual cell over a future period of time. Based on the SOC change trend, the priority balancing objects of individual cells are obtained, and the priority of the balancing scheduling of the priority balancing objects is corrected based on the SOH change trend. The priority correction is achieved by constructing an SOH prediction model through a multi-layer LSTM network to predict the SOH change trend of the battery, and dynamically correcting the balancing scheduling priority based on the SOH deviation and SOC deviation. The step of prioritizing the balancing targets of individual cells based on the SOC change trend is as follows: The SOC difference between any two individual cells in the battery pack is obtained based on the SOC change trend. Based on reinforcement learning algorithms, the optimal dynamic equilibrium threshold is generated according to the battery pack's operating state. The SOC difference value is compared with the optimal dynamic equilibrium threshold to identify potentially unbalanced cells; Calculate the SOC deviation of potentially unbalanced batteries and generate a priority balancing object list based on the deviation; The priority adjustment of the equalization scheduling of priority equalization objects based on the SOH change trend is specifically as follows: A SOH prediction model is constructed based on a multi-layer LSTM network to obtain the future SOH change trend of each cell in the priority balancing object list. Based on the SOH change trend, determine the deviation of SOH for each individual cell. Based on the deviation of SOH, the original priority of each individual cell in the priority balancing object list is dynamically corrected, and a comprehensive priority is generated by combining the SOC deviation. The priority balance object list is updated and sorted according to the overall priority. Based on the revised priority, the energy of the priority balancing object is transferred to the energy storage element; The energy in the energy storage element is released to the low-priority single cell based on the fuzzy game theory method.

2. The equalization control method for an aqueous sodium-ion battery according to claim 1, characterized in that, The construction of a single-cell state prediction model to predict the SOC change trend of a single cell over a future period of time is as follows: Preprocess the running data and extract key feature parameters based on the processed data; A single-cell state prediction model is constructed based on key feature parameters; Based on the single-cell state prediction model, combined with real-time operating data, the SOC change trend of each single cell is predicted over a future period.

3. The equalization control method for an aqueous sodium-ion battery according to claim 1, characterized in that, The transfer of energy from the priority balancing object to the energy storage element specifically involves: Select the energy transfer object based on the updated priority balancing object list; By controlling the bidirectional switch module, the energy transfer object is connected to the energy storage element to form an energy transfer loop; The intelligent control strategy controls the switching module to transfer the energy from the energy-exporting object to the energy storage element through the energy transfer loop; Monitor the energy state of the energy storage element and adjust the energy transfer process according to the error between it and the preset target value until the energy transfer is completed.

4. The equalization control method for an aqueous sodium-ion battery according to claim 3, characterized in that, The intelligent control strategy controls the switching module to transfer the energy from the energy-exporting object to the energy storage element through an energy transfer loop, specifically as follows: The battery pack state is defined as a state vector, and the energy transfer adjustment parameters are defined as actions. Construct an energy transfer reward function; The policy is trained using the state vector and actions, and the optimal energy transfer adjustment parameters are output. A dynamic mathematical model of the energy transfer system is established, and the dynamic mathematical model includes operational constraints. Based on the dynamic mathematical model, the optimal energy transfer regulation parameters are optimized by MPC to generate the final control command that meets the operating constraints. The energy transfer operation is executed according to the final control command.

5. The equalization control method for an aqueous sodium-ion battery according to claim 1, characterized in that, The method based on fuzzy game theory to release energy from the energy storage element to low-priority individual cells specifically involves: The target single cell is determined based on the revised priority balancing object list; The energy release process of energy storage elements is modeled as a multi-objective optimization problem and expressed using fuzzy logic. The membership function of fuzzy logic is adaptively tuned using Bayesian optimization methods to obtain an optimized fuzzy inference model. Based on the optimized fuzzy inference model and real-time system data, candidate solutions for energy release control parameters are generated. Optimize the candidate solutions and output the final energy release control scheme; The final energy release control scheme controls the energy storage element to release energy to the target single cell.

6. The equalization control method for an aqueous sodium-ion battery according to claim 5, characterized in that, The adaptive tuning of the membership function of fuzzy logic using the Bayesian optimization method specifically involves: Construct a membership degree optimization objective function, which includes the change in SOC difference after equilibrium, energy transfer efficiency, and temperature rise. Define the search space for the parameters to be optimized in the initial membership function; Based on the Bayesian optimization framework, the optimal membership function parameters are iteratively searched through Gaussian process modeling and acquisition function guidance. When the convergence condition is met, output the membership function parameters that make the objective function optimal.

7. An apparatus using the equalization control method for an aqueous sodium-ion battery as described in any one of claims 1-6, characterized in that, include: The SOC prediction module is used to collect the operating data of each individual cell in the battery pack, build a single cell state prediction model, and predict the SOC change trend of the single cell over a period of time. The equalization object selection module is used to obtain the priority equalization object of the single cell according to the SOC change trend, and to perform priority correction on the equalization scheduling of the priority equalization object based on the SOH change trend. The priority correction is to construct an SOH prediction model through a multi-layer LSTM network to predict the SOH change trend of the battery, and to dynamically correct the equalization scheduling priority based on the SOH deviation and SOC deviation. The energy transfer module is used to transfer the energy of the priority balancing object to the energy storage element according to the corrected priority. The energy release module releases energy from the energy storage element to low-priority individual cells based on a fuzzy game theory method.

8. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program to implement the steps of the equalization control method for aqueous sodium-ion batteries as described in any one of claims 1 to 6.

Images