An adaptive multi-mode battery activation method and device based on multi-data fusion

Abstract

The application relates to a self-adaptive multi-mode storage battery activation method and device based on multi-data fusion, and relates to the technical field of battery optimization.The method comprises the following steps: correcting and updating multi-dimensional parameters of collected storage batteries, introducing dynamic noise covariance and Kalman gain for optimized preprocessing, obtaining a characteristic vector, and inputting the dynamic threshold classification model.In the classification model, a nonlinear mapping is used to make a decision and classify the sulfuration state of the storage battery, the characteristic vector is used to analyze activation control parameters according to the classified sulfuration state, and pulse activation parameters are obtained.Finally, a genetic algorithm is used, combined with fitness function improvement and evolution operation improvement, to obtain an optimal pulse parameter combination through iterative circulation, so that the storage battery is activated and optimized, intelligent battery state recognition, dynamic switching of the activation mode and self-adaptive optimization of the pulse parameters are realized, and the problems of low repair efficiency and poor compatibility existing in the existing battery activation technology are solved.

Description

Technical Field

[0001] This application relates to the field of battery optimization technology, and in particular to an adaptive multi-mode battery activation method and apparatus based on multi-data fusion. Background Technology

[0002] Currently, in industries such as industry, transportation, and communications, batteries serve as a core component of backup power, and their performance directly impacts system reliability. Prolonged static or float charging conditions easily lead to plate sulfation, increased internal resistance, and capacity decay. Existing battery activation technologies mainly suffer from the following problems: Limited single-mode operation: Existing devices mostly use fixed high-frequency or low-frequency pulses, unable to dynamically adjust according to the degree of battery sulfation, resulting in poor repair effects on deep sulfation. Insufficient adaptive capability: Relying on manually set parameters, lacking real-time identification of battery status and automatic mode switching, resulting in low compatibility. The contradiction between portability and efficiency: Large devices, while comprehensive in function, are bulky, while small devices suffer from low activation efficiency due to power limitations.

[0003] It is evident that existing battery activation technologies suffer from low repair efficiency and poor compatibility. Summary of the Invention

[0004] This application provides an adaptive multi-mode battery activation method and system based on multi-data fusion. By introducing five core algorithms—adaptive filtering, decision tree, state machine, genetic algorithm, and protection logic—a closed loop of "acquisition-processing-decision-optimization-protection" is formed, realizing intelligent identification of battery status, dynamic switching of activation mode, and adaptive optimization of pulse parameters. This solves the problems of low repair efficiency and poor compatibility in existing battery activation technologies.

[0005] In a first aspect, this application provides an adaptive multi-mode battery activation method based on multi-data fusion, comprising:

[0006] A multidimensional feature vector is constructed based on the multidimensional parameters of the battery collected in real time, and the multidimensional feature vector is updated according to the correction factor. The multidimensional parameters are the real-time status data of the battery.

[0007] By estimating the covariance and combining it with Kalman gain, the multidimensional feature vector is optimized to obtain a target feature vector containing the filtered and optimized parameter sequence.

[0008] Based on the dynamic threshold classification model, a nonlinear mapping function is used to perform decision tree generalization and multi-level sulfation state classification on the target feature vector to determine the sulfation state information of the battery.

[0009] Based on the vulcanization state information, the activation control parameters are analyzed using the target feature vector to obtain the pulse activation parameters and charging control parameters;

[0010] A genetic algorithm is used to improve the fitness function and evolutionary operation based on the pulse activation parameters, and the optimal combination of pulse parameters is obtained through crossover iteration.

[0011] Based on the optimal pulse parameter combination, the battery activation is optimized in conjunction with the charging control parameters.

[0012] Optionally, a multidimensional feature vector is constructed based on the multidimensional parameters of the battery acquired in real time, and the multidimensional feature vector is updated according to a correction factor, including:

[0013] Based on the multidimensional parameters of the battery acquired in real time, a multidimensional feature vector is constructed. ;

[0014] Based on the analysis of the internal electrochemical processes of the battery, temperature and frequency correction terms were used as correction factors. For internal resistance to communication Make corrections and update the multidimensional feature vectors;

[0015] in, Terminal voltage, This is the charging current. For temperature, This refers to the settling time. This refers to the battery's nominal capacity. This is a temperature correction factor. This is the difference between the actual temperature and the reference temperature. This is the frequency correction factor. This is the difference between the actual injection frequency and the standard frequency. This is the capacity change calculated based on ampere-hour integration. These are factors related to health status.

[0016] Optionally, the multidimensional feature vector is optimized by covariance estimation and Kalman gain to obtain a target feature vector containing the filtered and optimized parameter sequence, including:

[0017] Construct a time window based on the collected historical time-series data;

[0018] Define prediction error covariance And using the multidimensional feature vector, time window, and data weighting factor For input, according to Obtain the prediction error covariance;

[0019] Obtaining measurement difference covariance ;

[0020] Kalman gain Introducing prediction error covariance Measurement difference covariance and data credibility factor ,according to The multidimensional feature vector is optimized to obtain the target feature vector;

[0021] in, The size of the time window. This represents the data weighting factor that decays over time. and All of these are preset parameters.

[0022] Optionally, based on a dynamic threshold classification model, a nonlinear mapping function is used to perform decision tree generalization and multi-level sulfation state classification on the target feature vector to determine the sulfation state information of the battery, including:

[0023] Using the target feature vector as input, the changes in battery state are analyzed through a preset dynamic threshold classification model, and a nonlinear mapping function is employed based on... Analysis of internal resistance threshold and voltage threshold ;

[0024] Construct a decision tree and use in-depth evaluation metrics. Combined with attribute correlation factors and according to The contribution of the target feature vector to the classification is quantified to obtain the threshold priority;

[0025] with internal resistance threshold Voltage threshold Based on threshold priority and target feature vector, the sulfation state is classified using a preset multi-level sulfation state classification index to obtain the sulfation state information of the battery.

[0026] in, , , , All coefficients are obtained by fitting experimental data. The internal resistance threshold reference value at the reference temperature. This is the battery's nominal voltage. Information gain ratio The Gini index, Used to measure the correlation between attributes in a target feature vector.

[0027] Optionally, based on the vulcanization state information, the activation control parameters are analyzed using the target feature vector to obtain pulse activation parameters and charging control parameters, including:

[0028] For the vulcanization state information, using the target feature vector as input, according to Analysis of high-frequency pulse duty cycle and according to Analysis of low-frequency pulse frequency ;

[0029] Based on high-frequency pulse duty cycle and low frequency pulse frequency Determine the pulse activation parameters;

[0030] A battery health state factor is introduced into the charging current of the target feature vector. ,according to Update the charging current to obtain constant current charging parameters. Furthermore, based on the target characteristic parameters, temperature compensation and self-discharge rate are analyzed, according to... Analysis of float charge voltage ;

[0031] The charging control parameters are determined based on the constant current charging parameters and the float charge voltage.

[0032] in, Based on duty cycle, Based on the base frequency, It is a multi-parameter index based on the analysis of each parameter in the target feature vector.

[0033] Optionally, a genetic algorithm is used to improve the fitness function and evolutionary operations based on the pulse activation parameters, and the optimal combination of pulse parameters is obtained through crossover iteration, including:

[0034] Gene mapping and encoding are performed based on the pulse activation parameters to construct a real number vector. Using the real number vector as a benchmark and combined with preset weight coefficients, a fitness function is used for analysis to obtain the fitness used to evaluate the merits of the pulse parameters.

[0035] Based on the real number vector and the fitness, the dynamic differences in fitness are analyzed, evolutionary operations are performed to improve the system, and the optimal combination of pulse parameters is obtained through cross-iteration loops.

[0036] Optionally, gene mapping and encoding are performed based on the pulse activation parameters to construct a real-number vector. Using this real-number vector as a benchmark, and combined with preset weight coefficients, a fitness function is employed for analysis to obtain the fitness used to evaluate the merits of the pulse parameters, including:

[0037] Gene coding mapping is performed based on the pulse activation parameters, according to the constructed real number vector. ;

[0038] Based on real number vectors Analysis of the rate of decrease in internal resistance Battery capacity recovery rate , gas evolution volume Combined with weighting coefficients and pulse energy utilization rate ,according to Analysis of fitness ;

[0039] in, , , , All are weighting coefficients.

[0040] Optionally, based on the real-valued vector and the fitness, the dynamic differences in fitness are analyzed, evolutionary improvements are performed, and the optimal combination of pulse parameters is obtained through cross-iteration loops, including:

[0041] The real-valued vectors are used as chromosomes. The fitness differences between parent chromosomes are analyzed in conjunction with the fitness analysis, and an adaptive crossover probability is introduced based on the crossover operation. Dynamically adjust for fitness differences;

[0042] The average fitness and current chromosome fitness of the population are analyzed based on the chromosomes and fitness values, and an adaptive mutation probability is introduced based on the mutation operation. The average fitness and chromosome fitness are dynamically adjusted.

[0043] Based on adaptive crossover probability and adaptive mutation probability The optimal combination of pulse parameters is obtained through cross-iteration loops;

[0044] in, The initial crossover probability, For fitness standard deviation, The initial mutation probability, The average fitness of the population.

[0045] Optionally, after obtaining the optimal combination of pulse parameters through cross-iteration loops, the process also includes:

[0046] Based on the target feature vector, a battery aging factor is introduced to analyze the degree of battery aging and obtain an adjusted safety coefficient.

[0047] Based on the preset power adjustment strategy and exponential decay strategy, the temperature parameters in the target feature vector are adjusted to obtain temperature segmentation control information.

[0048] The battery is monitored for safety based on the safety factor and the temperature segmentation control information, and when the battery malfunctions, abnormal information is fed back to adjust the activation control parameters.

[0049] Secondly, this application provides an adaptive multi-mode battery activation device based on multi-data fusion, comprising:

[0050] A multidimensional data acquisition module is used to construct a multidimensional feature vector based on the multidimensional parameters of the battery acquired in real time, and update the multidimensional feature vector according to a correction factor. The multidimensional parameters are the real-time status data of the battery.

[0051] The data preprocessing module is used to optimize the multidimensional feature vector by means of covariance estimation and Kalman gain, so as to obtain a target feature vector containing the parameter sequence after filtering optimization.

[0052] The battery status identification module is used to determine the sulfation status information of the battery by performing decision tree generalization and multi-level sulfation status classification on the target feature vector based on a dynamic threshold classification model and a nonlinear mapping function.

[0053] The activation mode control module is used to analyze the activation control parameters based on the vulcanization state information using the target feature vector to obtain pulse activation parameters and charging control parameters.

[0054] The pulse parameter optimization module is used to improve the fitness function and evolutionary operation based on the pulse activation parameters using a genetic algorithm, and to obtain the optimal combination of pulse parameters through cross-iteration loops.

[0055] The activation optimization module is used to optimize battery activation based on the optimal pulse parameter combination and the charging control parameters.

[0056] In summary, this application first corrects and updates the multidimensional parameters of the collected battery by introducing dynamic noise covariance and Kalman gain for optimization preprocessing to obtain an optimized feature vector, which is then input into the dynamic threshold classification model. Next, in the classification model, a nonlinear mapping is applied to the input feature vector to classify the sulfation state of the battery. For each classified sulfation state, the activation control parameters are analyzed using the feature vector to obtain pulse activation parameters. Finally, a genetic algorithm, combined with improvements to the fitness function and evolutionary operations, iteratively obtains the optimal combination of pulse parameters to optimize battery activation. This achieves intelligent battery state recognition, dynamic switching of activation modes, and adaptive optimization of pulse parameters, solving the problems of low repair efficiency and poor compatibility in existing battery activation technologies. Attached Figure Description

[0057] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0058] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0059] Figure 1 A flowchart illustrating an adaptive multi-mode battery activation method based on multi-data fusion provided in this application embodiment;

[0060] Figure 2 This is a schematic flowchart of an adaptive multi-mode battery activation method based on multi-data fusion provided in an optional embodiment of this application;

[0061] Figure 3 A software layer architecture diagram is provided as an example for this application;

[0062] Figure 4 A flowchart of a battery state recognition decision tree is provided as an example of this application;

[0063] Figure 5 This is a block diagram of an adaptive multi-mode battery activation device based on multi-data fusion provided in an embodiment of this application. Detailed Implementation

[0064] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0065] To facilitate understanding of the embodiments of this application, further explanations and descriptions will be provided below in conjunction with the accompanying drawings and specific embodiments. These embodiments do not constitute a limitation on the embodiments of this application.

[0066] Figure 1 This is a schematic flowchart illustrating an adaptive multi-mode battery activation method based on multi-data fusion, provided as an embodiment of this application. Figure 1 As shown in the embodiments of this application, the adaptive multi-mode battery activation method based on multi-data fusion may specifically include the following steps:

[0067] Step 110: Construct a multidimensional feature vector based on the multidimensional parameters of the battery collected in real time, and update the multidimensional feature vector according to the correction factor.

[0068] The multidimensional parameters are the real-time status data of the battery.

[0069] In this embodiment, the multidimensional parameters mainly include real-time monitored battery status data, including but not limited to: voltage (such as terminal voltage), current (such as charging current, discharging current, etc.), resting time, temperature (such as charging temperature, discharging temperature, etc.), battery capacity (such as nominal battery capacity), internal resistance (such as AC internal resistance), capacitance change, and health status factor. Among these, parameters such as voltage, current, and temperature can be acquired through relevant sensors; for example, current can be detected using a Hall effect sensor. AC internal resistance, capacity change rate, and health status factor can be acquired through corresponding measurement / calculation methods. Parameters such as resting time and battery capacity can be acquired through preset or timing parameters; this embodiment does not impose any limitations on these parameters.

[0070] In this specific implementation, the multidimensional feature vector is defined using the collected multidimensional parameters to describe the battery state, and a correction factor is introduced to correct the parameters in order to improve the accuracy of the data.

[0071] Step 120: The multidimensional feature vector is optimized by covariance estimation and Kalman gain to obtain a target feature vector containing the filtered and optimized parameter sequence.

[0072] In its implementation, to better adapt to real-time data changes and analyze data volatility, this embodiment uses a multidimensional feature vector as input. First, covariance estimation is used to determine the corresponding covariance value, thereby quantifying the deviation between the predicted and actual values ​​of the multidimensional parameters. The error is then adjusted in real-time, updating the parameters of the multidimensional feature vector. Next, using the covariance value as input, Kalman gain is used to dynamically balance the predicted and measured values, improving data reliability. The multidimensional parameters are then filtered and optimized, forming a parameter sequence to obtain the target feature vector.

[0073] In practical implementation, the covariance estimation introduced in this embodiment mainly uses dynamic noise covariance estimation and measurement difference covariance estimation. Specifically, dynamic noise covariance estimation quantifies the inaccuracy of predicted values ​​(i.e., multidimensional parameters) to obtain the dynamic noise covariance estimate, such as the deviation between the predicted and actual values ​​of each parameter in the multidimensional parameters, reducing the interference of data noise. For example, during battery activation, when there are sudden temperature changes, the latest temperature data has a greater impact on noise estimation. Measurement difference covariance estimation describes the real-time characteristics of noise in the measured values ​​to obtain the measurement difference covariance estimate, so as to more accurately adapt to real-time changes in data.

[0074] Step 130: Based on the dynamic threshold classification model, a nonlinear mapping function is used to perform decision tree generalization and multi-level sulfation state classification on the target feature vector to determine the sulfation state information of the battery.

[0075] The vulcanization status information includes the vulcanization status grading results, which mainly include, but are not limited to, deep vulcanization, light vulcanization, and normal vulcanization.

[0076] This embodiment uses a dynamic threshold classification model to process the parameters of the feature vector and distinguish the sulfation state of the battery. Specifically, firstly, a nonlinear mapping function is used to nonlinearly map the internal resistance and voltage parameters in the multidimensional parameters, and the complex changes in battery state are considered to capture the complex relationship between battery state and temperature and resting time, thus obtaining the internal resistance threshold and voltage threshold. Then, using the established multi-level sulfation state classification, the corresponding multidimensional parameters are analyzed based on each threshold to identify the sulfation state of the battery.

[0077] When identifying the sulfation state, this embodiment introduces a decision tree to construct a hierarchical state logic, matching differentiated activation strategies for batteries with different sulfation levels, and mapping the various thresholds and multidimensional parameters analyzed to the battery sulfation state, providing a basis for subsequent activation control.

[0078] For example, based on internal resistance threshold and voltage threshold, decision trees can capture the complex relationship between battery state and environmental parameters. For instance, when the internal resistance is greater than the internal resistance threshold and the terminal voltage is greater than the voltage threshold, the decision tree can directly determine that the sulfation state is deep sulfation, and subsequently trigger a high-frequency + low-frequency composite pulse activation mode.

[0079] Therefore, this embodiment realizes the construction of a multi-level sulfation state classification system, and then uses a decision tree to match differentiated activation strategies for batteries with different sulfation degrees, transforming multi-dimensional parameters into control instructions for activation strategies. Its anti-correlation design and dynamic threshold mechanism are key technologies to solve the complex and variable nature of battery states, directly improving the engineering applicability and repair efficiency of the activation algorithm.

[0080] Step 140: Based on the vulcanization state information, the activation control parameters are analyzed using the target feature vector to obtain pulse activation parameters and charging control parameters.

[0081] In this embodiment, based on different vulcanization states, the activation control mode is analyzed using multi-dimensional parameters, including a pulse activation mode and a charging mode (including a constant current charging mode and a constant voltage maintenance mode), to obtain pulse activation parameters and charging control parameters. The pulse activation parameters are used to control the pulse activation mode, while the charging control parameters are mainly used to control the constant current charging mode and the constant voltage maintenance mode.

[0082] Specifically, when analyzing the pulse activation mode, the analysis mainly focuses on high-frequency and low-frequency pulses, fully considering the influence of multiple parameters such as internal resistance, capacitance, and temperature. The activation effect is improved by comprehensively adjusting the pulse parameters. When analyzing the constant current charging mode, the analysis mainly considers charging current and battery health status. When analyzing the constant voltage maintenance mode, the analysis mainly considers parameters such as voltage, temperature, and battery self-discharge rate.

[0083] Therefore, this embodiment realizes a dynamically switching composite pulse activation strategy, enabling the analysis of the "high-frequency crushing + low-frequency penetration" composite pulse mode, and the relevant parameters of pulse activation are dynamically adjusted according to internal resistance, temperature, and capacity change rate.

[0084] Step 150: Using a genetic algorithm, the fitness function and evolutionary operation are improved based on the pulse activation parameters, and the optimal combination of pulse parameters is obtained through cross-iteration loop.

[0085] In this implementation, a genetic algorithm is used to encode the pulse activation parameters as "chromosomes," with each parameter representing a "gene." A fitness function is introduced to analyze the pulse activation parameters and determine the corresponding fitness value. Then, the genetic algorithm optimizes the fitness function through improved adaptive crossover / mutation operations. Crossover and mutation operations are performed on the chromosomes to dynamically adjust the fitness, retaining superior genes and increasing the mutation probability to drive the process towards a better solution. Ultimately, this achieves global optimization of the parameter combination, i.e., searching for the optimal combination of pulse parameters to optimize the battery activation effect.

[0086] Step 160: Optimize battery activation based on the optimal pulse parameter combination and the charging control parameters.

[0087] In this specific implementation, the optimal combination of pulse parameters is used to perform composite pulse activation, and constant current charging or constant voltage maintenance is adjusted by charging control parameters.

[0088] As can be seen, this embodiment first corrects and updates the multidimensional parameters of the collected battery, introducing dynamic noise covariance and Kalman gain for optimization preprocessing to obtain an optimized feature vector, which is then input into the dynamic threshold classification model. Next, in the classification model, a nonlinear mapping is applied to the input feature vector to classify the sulfation state of the battery. For the classified sulfation state, the activation control parameters are analyzed using the feature vector to obtain pulse activation parameters. Finally, a genetic algorithm, combined with fitness function improvement and evolutionary operation improvement, is used to iteratively obtain the optimal pulse parameter combination. Based on the optimal pulse parameter combination and charging control parameters, battery activation optimization is performed. This embodiment achieves efficient, fully automated battery activation through a closed-loop process of data acquisition → preprocessing → state identification → activation execution → parameter optimization, possessing strong compatibility and solving the problems of low repair efficiency and poor compatibility in existing battery activation technologies.

[0089] Reference Figure 2 This illustration shows a flowchart of an adaptive multi-mode battery activation method based on multi-data fusion, provided in an optional embodiment of this application. The method specifically includes the following steps:

[0090] Step 210: Construct a multidimensional feature vector based on the multidimensional parameters of the battery collected in real time, and update the multidimensional feature vector according to the correction factor.

[0091] The multidimensional parameters are the real-time status data of the battery.

[0092] In this embodiment, the construction of multidimensional feature vectors is the core foundation for achieving accurate battery state analysis. By fusing multidimensional parameters and introducing a correction mechanism, it provides reliable data support for subsequent state identification and activation strategies.

[0093] To address the limitations of existing battery activation technologies that rely on single parameters (such as voltage and current), this embodiment improves battery activation compatibility by integrating electrochemical parameters (internal resistance), environmental parameters (temperature), and time parameters (resting time) to comprehensively characterize the battery state, achieving multi-dimensional data fusion in the form of feature vectors. Furthermore, to address the issue that measurements of internal resistance are easily affected by temperature and frequency, this embodiment introduces a correction mechanism to improve accuracy. Dynamic correction terms are used to calibrate the data and avoid misjudgments of the battery state due to environmental fluctuations.

[0094] Furthermore, in this embodiment, the multidimensional feature vector also lays the foundation for intelligent decision-making. Specifically, the feature vector serves as the input basis for subsequent Kalman filter preprocessing, dynamic threshold classification, and genetic algorithm optimization, and its accuracy directly affects the effectiveness of activation mode switching and pulse parameter adjustment.

[0095] Optionally, a multidimensional feature vector is constructed based on the multidimensional parameters of the battery acquired in real time, and the multidimensional feature vector is updated according to a correction factor. Specifically, this may include: constructing a multidimensional feature vector based on the multidimensional parameters of the battery acquired in real time. Based on the analysis of the internal electrochemical processes of the battery, temperature and frequency correction terms were used as correction factors. For internal resistance to communication Make corrections and update the multidimensional feature vector; where, Terminal voltage, This is the charging current. For temperature, This refers to the settling time. This refers to the battery's nominal capacity. Self-discharge rate This is a temperature correction factor. This is the difference between the actual temperature and the reference temperature. This is the frequency correction factor. This is the difference between the actual injection frequency and the standard frequency. This is the capacity change calculated based on ampere-hour integration. These are factors related to health status.

[0096] Specifically, the multidimensional feature vector defined in this embodiment It covers nine key parameters, describing battery status from different dimensions (including conventional physical parameters, dynamic characteristic parameters, and comprehensive health indicators). Among them, through... , , , , These parameters directly reflect the basic operating state of the battery; AC internal resistance Capacity change rate Self-discharge rate These parameters can be used to characterize the internal electrochemical changes and capacity decay trends of a battery; health state factor By performing multi-parameter normalization, the overall health of the battery is quantified.

[0097] The following describes the acquisition and / or function of each parameter:

[0098] Terminal voltage Used to reflect the real-time energy state of the battery, and to preliminarily determine whether the battery is in a fully charged, discharging, or over-discharged state; charging current. The magnitude and direction of the current (charging or discharging) can be detected using a Hall sensor, which is crucial for understanding the battery's operating state and nominal capacity. It is used to dynamically adjust the charging current.

[0099] internal resistance of communication The internal resistance can be measured using the AC injection method. Considering the complex electrochemical processes inside the battery, the basic internal resistance can be obtained by applying a small AC signal and calculating the rate of change of voltage / current. Introduce a temperature correction term and frequency correction term Corrections are made to eliminate the interference of environmental factors on internal resistance measurements. Capacity change (or capacity change rate) It can be calculated based on ampere-hour integration, reflecting changes in battery capacity over time (such as dynamic degradation or recovery). For example, The calculation formula can be: , The charge / discharge efficiency coefficient. This represents the self-discharge rate.

[0100] Health status factors The normalized index (0-1) that integrates internal resistance, voltage, and capacitance can be obtained using the formula: Calculations show that the degree of deviation from internal resistance is converted into a health indicator between 0 and 1 using an exponential function; the higher the internal resistance, the better. The smaller the value, the more directly it reflects the degree of sulfation or aging of the battery. Among them, Sensitivity coefficient This is a reference value for normal internal resistance.

[0101] Step 220: The multidimensional feature vector is optimized by covariance estimation and Kalman gain to obtain a target feature vector containing the filtered and optimized parameter sequence.

[0102] In practical implementation, dynamic noise covariance estimation and adaptive Kalman gain adjustment are used to denoise and optimize the reliability of battery state data, providing reliable data support for subsequent processing. Specifically, dynamic noise covariance estimation provides real-time noise parameters for Kalman gain calculation, while the adaptive factor of the Kalman gain can in turn optimize the gain's adaptability to noise, forming a closed loop of "noise assessment - gain adjustment - data filtering" to ensure the measurement accuracy of multi-dimensional feature vectors (such as internal resistance and voltage).

[0103] As can be seen, this embodiment can effectively solve problems such as sensor noise and environmental interference during battery activation by preprocessing multi-dimensional parameters. For example, when the temperature changes drastically, the internal resistance misjudgment caused by temperature drift can be avoided by dynamically adjusting the weight factor and confidence factor.

[0104] In an optional embodiment, the above-mentioned optimization of the multidimensional feature vector by covariance estimation combined with Kalman gain yields a target feature vector containing a filtered and optimized parameter sequence. Specifically, this may include: constructing a time window based on collected historical time-series data; defining the prediction error covariance. And using the multidimensional feature vector, time window, and data weighting factor For input, according to Obtain the prediction error covariance; obtain the measurement difference covariance. ; In Kalman gain Introducing prediction error covariance Measurement difference covariance and data credibility factor ,according to The multidimensional feature vector is optimized to obtain the target feature vector; wherein, The size of the time window. This represents the data weighting factor that decays over time.

[0105] In this embodiment, the optimized preprocessing of the real-time acquired multidimensional feature vectors mainly combines historically acquired multidimensional parameters to form historical time-series data. Then, a time window is used. Limit the scope of historical data used in the calculation to balance computational efficiency and data timeliness.

[0106] This embodiment transforms and represents each parameter in the multidimensional parameters to obtain the first... Predicted value at time and measured values Then, dynamic noise covariance estimation is used, first using the prediction error covariance. ,exist The update formula introduces a data weighting factor that decays over time. Its calculation formula is ,in, and All parameters are adjustable and can be adjusted as needed. Measurement difference covariance For updates, please refer to The update is achieved by dynamically adjusting the weighting factors. This allows for more accurate adaptation to real-time changes in data.

[0107] In this embodiment, the covariance calculation uses a time decay weighting mechanism, meaning that recent data has a larger weighting factor (e.g., when...). The closer to the current moment hour, The closer to 1), the lower the weight of longer-term data, reflecting the logic that "recent data is more reliable" and adapting to dynamic changes in battery status. Noise suppression and dynamic adaptation are achieved by calculating the error covariance through weighted averaging, which suppresses sudden noise (such as transient interference) and adjusts the time window. , , It can adapt to data fluctuation characteristics under different operating conditions (such as drastic changes during fast charging or slow drift during static conditions). The synchronous update of measurement difference covariance and error covariance adopt the same logic update. The two together describe the uncertainty of the prediction model and measurement data, and provide dynamic noise parameters for Kalman filtering.

[0108] In this embodiment, by fusing credibility factors Combined with error covariance and measurement covariance Improved Kalman gain The formula. Where, This is calculated dynamically based on the data's fluctuation range and historical stability. For example, when data fluctuations are large or historical stability is poor, Decreasing the value reduces the weight of the current measurement in the gain calculation, preventing noise from dominating the filtering result; conversely, when the data is stable, Approaching 1, fully utilize the accuracy of the measured value.

[0109] As can be seen, this scheme addresses the problem of traditional Kalman filtering assuming fixed noise characteristics by... The dynamic adjustment solves the problem of non-stationary measurement noise caused by sudden changes in operating conditions (such as pulse switching) during battery activation, and improves the robustness of state estimation.

[0110] Step 230: Based on the dynamic threshold classification model, a nonlinear mapping function is used to perform decision tree generalization and multi-level sulfation state classification on the target feature vector to determine the sulfation state information of the battery.

[0111] The main shortcomings of existing battery activation technologies include: ① Manual mode switching: "Normal / Fast" mode must be manually selected, failing to achieve intelligent decision-making based on real-time data; ② Coarse parameter adjustment: Pulse frequency and duty cycle are fixed, without dynamic optimization based on internal resistance, potentially leading to plate damage or incomplete repair; ③ Simple energy management: Relying on external power supply, lacking built-in energy storage units and low-power design, failing to meet the needs of portable scenarios. Specifically, these shortcomings mainly lead to the following problems: A. Deficiencies in pulse activation technology: While breaking down sulfides through electrical pulses, the pulse parameters are fixed, failing to form a composite optimization strategy; B. Deficiencies in battery state detection, such as simple threshold judgment based on voltage and current, without integrating multi-dimensional parameters such as internal resistance and temperature; C. Deficiencies in charging control algorithms: Using constant current / constant voltage segmented control, lacking coordinated optimization with the activation mode. Therefore, existing battery activation technologies suffer from low repair efficiency, poor compatibility, and insufficient portability.

[0112] To address the aforementioned technical problems, this embodiment optimizes the battery activation algorithm, referring to... Figure 3 and Figure 4 As shown, where, Figure 3 The software layer architecture diagram shows that battery activation includes five core algorithms: adaptive filtering, decision tree, state machine, genetic algorithm, and protection logic, forming a closed loop of "acquisition-processing-decision-optimization-protection". Figure 4 Flowchart of the decision tree for battery status identification.

[0113] like Figure 3 As shown, this embodiment achieves accurate classification of the three stages of battery sulfation through a dynamic threshold classification model, multi-level sulfation state division, and an improved decision tree algorithm. Specifically, the dynamic threshold classification model integrates decision tree attribute selection and the establishment of multi-level sulfation state classification. The target feature vector is used as the input to the classification model. The classification model uses a decision tree algorithm to analyze the correlation between various attributes (i.e., multi-dimensional parameters) of the target feature vector, avoiding the selection of highly correlated attributes and improving the generalization ability of the decision tree. The decision tree can use non-linear decision-making to classify the battery state into deep sulfation, light sulfation, and normal state based on the input data. Thus, the classification of sulfation degree is achieved based on the dynamic threshold decision tree.

[0114] In an optional embodiment, the above-mentioned determination of the sulfation state information of the battery by using a nonlinear mapping function to perform decision tree generalization and multi-level sulfation state classification on the target feature vector based on a dynamic threshold classification model includes: taking the target feature vector as input, analyzing the changes in battery state through a preset dynamic threshold classification model, and using a nonlinear mapping function according to... Analysis of internal resistance threshold and voltage threshold ; Construct a decision tree and use in-depth evaluation indicators Combined with attribute correlation factors and according to The contribution of the target feature vector to classification is quantified to obtain the threshold priority; the internal resistance threshold is used as the basis for this. Voltage threshold Based on threshold priority and target feature vector, the sulfation state is classified using a preset multi-level sulfation state classification index to obtain the sulfation state information of the battery; wherein, All coefficients are obtained by fitting experimental data. The internal resistance threshold reference value at the reference temperature. This is the battery's nominal voltage. Information gain ratio The Gini index, Used to measure the correlation between attributes in a target feature vector.

[0115] In the specific implementation, refer to Figure 3 , Figure 4 The dynamic threshold classification model fully considers the complex changes in battery state and employs a nonlinear mapping function to analyze the internal resistance threshold and voltage threshold. Specifically, the nonlinear mapping function includes functions for calculating the internal resistance threshold and the voltage threshold. When calculating the internal resistance threshold, the internal resistance parameter in the feature vector is used as the input to the internal resistance threshold calculation function, and the temperature coefficient is used as the input. (This temperature coefficient can reflect the nonlinear trend of the internal resistance threshold changing with temperature, including the quadratic term capturing the accelerated rate of change of internal resistance at high temperatures.) Temperature sensitivity correction is performed, and the internal resistance threshold is calculated. The internal resistance threshold value at the reference temperature is used as the reference value; when calculating the voltage threshold, the nominal battery voltage is used as the input, and the correlation coefficient of the resting time is used. Dynamically adjust the settling time and calculate the voltage threshold. .

[0116] in, The coefficients were all obtained by fitting a large amount of experimental data, and the square root term It characterizes the rapid changes in voltage during the initial stage of recovery. Utilizing nonlinear functions allows for a better capture of the complex relationship between battery state and temperature, resting time.

[0117] This embodiment can establish the following multi-level vulcanization state classification:

[0118] Deep vulcanization: meets the requirements ,and ,and ,in, These are all corresponding threshold coefficients. This is normal internal resistance.

[0119] Mild vulcanization: meets the requirements ,and ,and .

[0120] Normal vulcanization: meets the requirements ,and ,and .

[0121] In this embodiment, the decision tree attribute selection adopts a comprehensive evaluation index. At that time, attribute correlation factor was introduced. .in, Used to measure attributes The higher the correlation with other attributes, the better. The larger the value, the better. By introducing this factor, we can avoid selecting highly correlated attributes, determine the threshold priority, and improve the generalization ability of the decision tree.

[0122] For example, internal resistance With health factors There is a strong correlation; if selection is based solely on IGR or Gini, duplicate selections may occur, leading to overfitting of the decision tree. (Introduction) After that, highly correlated attributes Will because Increased but suppressed attributes are prioritized, while independent attributes (such as temperature and resting time) are preferred. Improved generalization ability: By reducing the influence of correlated attributes, decision trees are more stable in classifying new samples (such as different battery models), avoiding misclassification caused by accidental correlations in the training data.

[0123] In practical implementation, refer to Figure 3 Decision trees achieve multi-level sulfur classification through multi-level node splitting. For example, at the root node: determine the AC internal resistance. Does it exceed the correction threshold? If the threshold is exceeded, proceed to the "Deep Sulfation Candidate Branch"; otherwise, proceed to the "Mild / Normal Branch". Intermediate Node: In the "Mild / Normal Branch", further determine whether the terminal voltage is below the correction threshold. Combined with the threshold of health factor H (e.g. The system further subdivides mild sulfidation into normal conditions. At the leaf node, the final sulfidation classification result is output, and different activation modes are triggered accordingly.

[0124] In summary, the battery state recognition decision tree in this embodiment has a three-level classification: the battery state is divided into deep sulfation, light sulfation, and normal state by using dynamic thresholds of internal resistance, voltage, and health status.

[0125] Nonlinear decision-making in battery state recognition decision trees: introducing a temperature-corrected internal resistance threshold (e.g.) (and health status factors) to improve classification accuracy.

[0126] Step 240: Based on the vulcanization state information, the activation control parameters are analyzed using the target feature vector to obtain pulse activation parameters and charging control parameters.

[0127] Reference Figure 4 As shown, this embodiment determines the activation mode by analyzing activation control parameters for different vulcanization states. The activation modes include pulse activation mode, constant current charging mode, and constant voltage maintenance mode.

[0128] The pulse activation mode mainly adopts a composite strategy of high-frequency fragmentation + low-frequency penetration, the constant current charging mode mainly drives dynamic current control based on health status, and the constant voltage maintenance mode mainly compensates through temperature and self-discharge rate.

[0129] This embodiment breaks through the limitations of traditional fixed parameters and achieves: ① Multi-parameter collaborative optimization: Unlike traditional single-parameter adjustment (such as relying solely on voltage), this module achieves precise mapping of "sulfation degree - pulse parameter" through the linkage of multiple dimensions such as internal resistance, temperature, and self-discharge rate; ② Adaptive mode switching: Combined with the output of the battery status recognition module (such as high-frequency + low-frequency composite pulse triggered by deep sulfation), manual intervention is avoided, improving repair efficiency; ③ Balance between safety and efficiency: The constant current mode dynamically limits power according to the health status, and the constant voltage mode introduces temperature compensation to prevent overcharging / overheating, balancing activation effect and equipment safety.

[0130] In an optional embodiment, this embodiment uses the target feature vector to analyze activation control parameters based on the vulcanization state information to obtain pulse activation parameters and charging control parameters. Specifically, it may include: using the target feature vector as input, based on the vulcanization state information, and... Analysis of high-frequency pulse duty cycle and according to Analysis of low-frequency pulse frequency Based on high-frequency pulse duty cycle and low frequency pulse frequency Determine the pulse activation parameters; introduce a battery health state factor into the charging current of the target feature vector. ,according to Update the charging current to obtain constant current charging parameters. Furthermore, based on the target characteristic parameters, temperature compensation and self-discharge rate are analyzed, according to... Analysis of float charge voltage The charging control parameters are determined based on the constant current charging parameters and the float charge voltage; among which, It is a multi-parameter index based on the analysis of each parameter in the target feature vector.

[0131] In this embodiment, the sulfation state is related to the activation mode of the battery, such as... Figure 4 As shown, the vulcanization state is determined through a decision tree, and based on this state, a specific activation mode can be further determined, such as pulse activation mode, constant current charging mode, and constant voltage maintenance mode. In the pulse activation mode, activation control is mainly based on calculated pulse activation parameters, including the calculated high-frequency pulse duty cycle. and low frequency pulse frequency In constant current charging mode, the main parameters are based on the constant current charging parameters. Constant current charging control is performed. In constant voltage maintenance mode, it is mainly based on the float charge voltage. Perform constant pressure maintenance control. The following explains each parameter:

[0132] High-frequency pulse duty cycle based on base duty cycle Internal resistance adjustment coefficient , as well as The calculated inputs employ multi-parameter fusion logic to achieve dynamic adjustment, meaning that in addition to considering internal resistance, the effects of capacity change rate and temperature are also taken into account. For example, internal resistance deviation... The larger the value, the higher the degree of sulfidation, requiring a larger high-frequency pulse duty cycle to enhance the energy output for breaking down the sulfides. Although the above formula does not directly reflect the rate of temperature and volume change, in actual calculations... Temperature correction item approved Calibration, and capacity change rate It is negatively correlated with changes in internal resistance, indirectly affecting the regulation logic.

[0133] The low-frequency pulse frequency takes into account the battery's self-discharge rate and internal resistance change rate, and is mainly based on the fundamental frequency. Self-discharge rate and rate of change of internal resistance Calculations show that adjusting pulse parameters comprehensively across multiple parameters can improve the activation effect. For example, when the self-discharge rate is high, the low-frequency pulse frequency is reduced to decrease energy loss; when the internal resistance rises rapidly, the frequency is increased to promote electrolyte penetration and enhance the repair effect.

[0134] Constant current charging parameters (i.e., constant current charging current) Battery health status factor As a benchmark, and Taking into account multiple parameters such as battery internal resistance, voltage, and capacity, therefore A higher value indicates a better battery health condition, and the charging current can be appropriately increased.

[0135] float voltage The calculations, in addition to incorporating temperature compensation, also consider the battery's self-discharge rate. Float charge voltage. In the calculation formula, This indicates a temperature compensation term, which reduces the float charge voltage when the temperature rises to prevent thermal runaway. The self-discharge rate compensation coefficient (SDR) increases the float charge voltage more to offset the capacity loss caused by self-discharge. Application scenarios include long-term float charging scenarios (such as backup power supplies), where dynamic voltage compensation maintains battery capacity and extends battery life.

[0136] Step 250: Gene mapping and encoding are performed based on the pulse activation parameters to construct a real number vector. Using the real number vector as a benchmark and combined with preset weight coefficients, a fitness function is used for analysis to obtain the fitness used to evaluate the merits of the pulse parameters.

[0137] Step 260: Based on the real number vector and the fitness, analyze the dynamic differences in fitness, perform evolutionary operations to improve the results, and obtain the optimal combination of pulse parameters through cross-iteration loops.

[0138] Steps 250-260 are described uniformly as follows:

[0139] In this embodiment, the optimization of pulse activation parameters is mainly achieved using a genetic algorithm. The pulse optimization parameters are encoded and represented as "chromosomes," which are then converted into real-number vectors. Each parameter in the real-number vector corresponds to a parameter in the pulse activation parameters, and can be understood as a "gene." Each "gene" in the real-number vector chromosome (such as high-frequency pulse frequency, duty cycle, etc.) is essentially a parameter for generating the activation electrical signal, directly determining the characteristics of the pulse waveform applied to the battery. For example, the high-frequency pulse duty cycle determines the duration of the high-voltage pulse per unit time, directly affecting the efficiency of sulfide breakup on the plates. The low-frequency pulse frequency controls the periodic interval of the pulse, affecting the penetration depth of the electrolyte into the plates. The conduction time determines the energy release intensity of a single pulse, directly related to the gas evolution rate (safety) and energy utilization rate.

[0140] Changes in genes (pulse parameters) affect the battery state through electrochemical processes, which in turn become input parameters for the fitness function. Therefore, this embodiment analyzes key parameters affecting the battery state based on real number vectors, including the rate of decrease in internal resistance, the rate of capacity recovery, and the amount of gas evolution. Using these as inputs, combined with weighting coefficients and pulse energy utilization, the fitness function is improved to evaluate the merits of the parameter combination. The multidimensional optimization objective is converted into a fitness index to obtain a fitness value, which serves as the evaluation basis for the evolutionary operation of the genetic algorithm.

[0141] Then, crossover and mutation operations are performed based on real-valued vectors and fitness values. Specifically, this embodiment improves the crossover and mutation operations. The improvement to the crossover operation aims to generate new solutions through gene recombination, mainly considering the fitness of the parent chromosome. By exchanging some genes of the parent chromosome to generate the offspring chromosome, gene exchange is promoted, and new parameter combinations are explored. The improvement to the mutation operation aims to explore the solution space through random perturbation, mainly analyzing the adaptive mutation probability and randomly adjusting certain genes (parameters) of the chromosome to avoid the algorithm getting trapped in local optima and enhance global search capabilities. The crossover and mutation operations are executed iteratively to explore the optimal combination of impulse parameters.

[0142] In this embodiment, the evolved chromosome is decoded into pulse parameters and fed back to the activation mode control module to drive the pulse activation process. At the same time, the new data generated after the new parameters are applied to the battery (such as changes in internal resistance and capacity recovery) are input into the fitness function again, forming a closed-loop optimization process of "evaluation-evolution-execution-re-evaluation".

[0143] In an optional embodiment, the above-mentioned gene mapping encoding based on the pulse activation parameters, constructing a real-number vector, and using the real-number vector as a benchmark, combined with preset weight coefficients, and analyzing using a fitness function to obtain the fitness used to evaluate the merits of the pulse parameters, includes: performing gene encoding mapping based on the pulse activation parameters, and constructing a real-number vector... Based on real number vectors Analysis of the rate of decrease in internal resistance Battery capacity recovery rate , gas evolution volume Combined with weighting coefficients and pulse energy utilization rate ,according to Analysis of fitness ;in, All are weighting coefficients.

[0144] Optionally, the above-mentioned analysis of the dynamic differences in fitness based on the real-valued vector and the fitness, the improvement of evolutionary operations, and the acquisition of the optimal pulse parameter combination through crossover iteration can include: using the real-valued vector as a chromosome, analyzing the fitness differences of parent chromosomes in combination with the fitness, and introducing an adaptive crossover probability based on the crossover operation. The fitness differences are dynamically adjusted; the average fitness and current chromosome fitness of the population are analyzed based on the chromosomes and the fitness values, and an adaptive mutation probability is introduced based on the mutation operation. The average fitness and chromosome fitness are dynamically adjusted based on the adaptive crossover probability. and adaptive mutation probability The optimal combination of pulse parameters is obtained through a cross-iteration loop; where... The initial crossover probability, For fitness standard deviation, The initial mutation probability, The average fitness of the population.

[0145] A unified description is provided for the fitness function analysis and evolutionary operation improvements described above:

[0146] In this embodiment, a "chromosome" is an encoded representation of a potential solution to a problem, simulating a carrier of biological genetic information. In the above context, a chromosome is a set of codes for battery pulse activation parameters, such as pulse parameters (e.g., high-frequency pulse frequency). High-frequency pulse conduction time High-frequency pulse duty cycle Low-frequency pulse frequency Low-frequency pulse conduction time Low-frequency pulse duty cycle (etc.) is encoded into a real number vector Such a vector constitutes a "chromosome," with each parameter in the vector corresponding to a "gene" on the chromosome. Genetic algorithms use evolutionary operations such as crossover and mutation on these chromosomes to search for the optimal combination of pulse parameters, thereby optimizing the battery activation effect. For example, the crossover operation generates offspring chromosomes by exchanging some genes (parameters) of two parent chromosomes, while the mutation operation randomly adjusts certain genes (parameters) of the chromosomes to explore a better solution space.

[0147] This embodiment performs evolutionary operations such as crossover and mutation on chromosomes, mainly by combining fitness values ​​and corresponding parameters.

[0148] The fitness value analysis is based on the fitness function calculation. The logic is that the internal resistance reduction rate (positive correlation) and capacity recovery rate (positive correlation) increase fitness, gas evolution rate (negative correlation) decreases fitness, energy utilization rate (positive correlation) balances energy consumption, and weight coefficients regulate the priority of each indicator.

[0149] Interpretation of fitness analysis:

[0150] Based on real vector analysis, the key parameters affecting fitness values ​​include the rate of decrease in internal resistance. Battery capacity recovery rate , gas evolution volume and pulse energy utilization rate etc. Among them, the rate of decrease in internal resistance It can be based on the intensity of the high-frequency pulse (by...) , Analysis revealed that, for example, the greater the intensity of the high-frequency pulse, the more thoroughly the sulfide is broken down, and the lower the internal resistance. The more pronounced the decrease, the better. (This refers to the frequency of low-frequency pulses.) Linked to the rate of change of internal resistance (see the formula in the activation mode control module), it optimizes the permeation effect and accelerates the recovery of internal resistance.

[0151] Gas evolution volume Primarily based on pulse conduction time (including and Analysis revealed that excessively long pulse conduction time or excessively high duty cycle may lead to battery overcharging and increased gas evolution. Increased saturation leads to decreased security. Genetic algorithms utilize the fitness function... The penalty for gas evolution volume forces the pulse parameters (such as...) Adjust to a safe range.

[0152] Capacity recovery rate The combination of pulse parameters (such as the synergy of high-frequency breakage and low-frequency penetration) directly affects the recovery degree of the active material on the electrode, thereby increasing the battery capacity. Capacity change rate High frequency duty cycle The adjustment formulas work together to dynamically optimize recovery efficiency.

[0153] Energy utilization rate Optimizing pulse frequency and duty cycle can reduce ineffective energy loss (such as heat loss) and improve energy efficiency. In the fitness function This encourages the generation of pulse parameter combinations with high energy utilization efficiency (such as reasonable...). and (Proportioning).

[0154] In this embodiment, chromosomes are the objects of the genetic algorithm; changes in their genes directly affect the calculation of fitness values, which in turn influence chromosome evolution. In the fitness value calculation, the weight coefficients are optimized. It balances activation effect, safety, capacity recovery and energy utilization efficiency.

[0155] Explanation of the improvement to crossover operations:

[0156] The crossover operation dynamically adjusts the adaptive crossover probability based on the fitness differences of the parent chromosomes. Its inputs include: parental chromosome fitness ( and ), initial crossover probability (Cross-cross probability baseline) and fitness standard deviation (Used to measure the dispersion of population fitness), it explores new parameter combinations through crossover operations and outputs adaptive crossover probabilities. For example, the crossover probability can be dynamically adjusted based on the difference in parent fitness: if the difference is large, the crossover probability is reduced. To preserve superior genes; if the difference is small, increase This promotes gene exchange, that is, it promotes the generation of new solutions. The offspring chromosomes after crossover will participate in fitness assessment as new parameter combinations.

[0157] Explanation of the improved mutation operation:

[0158] Mutation operation introduces adaptive mutation probability This allows for dynamic adjustments based on the average fitness of the population and the fitness of the current chromosome, preventing the algorithm from getting trapped in local optima and enhancing its global search capabilities. Specifically, it first adjusts based on fitness... Analyze population to assess fitness (Measures the overall optimization level of the population), combined with fitness and initial mutation probability Analyze adaptive mutation probability This is used to control the probability of random mutations in chromosomal genes. For example, when the chromosome fitness is below the population average, it is increased... To encourage it to evolve towards a better solution; conversely, to reduce This is to preserve high-quality genes. The mutated chromosome will generate new parameter combinations and continue to participate in fitness assessment.

[0159] Therefore, this embodiment realizes parameter connection and closed-loop process between modules. In the parameter input stage: the multi-dimensional data acquisition module provides real-time data such as battery internal resistance and capacity as the basic input for fitness function calculation; the initial pulse parameters are provided by the activation mode control module and encoded as chromosomes. In the fitness evaluation stage: the fitness function combines real-time data with pulse parameters to calculate the fitness value of each chromosome. In the evolutionary operation stage: the crossover and mutation probabilities are adjusted according to the fitness values, and gene recombination and mutation are performed on the chromosomes to generate new parameter combinations. Finally, in the parameter output stage: the evolved chromosomes are decoded into pulse parameters and fed back to the activation mode control module to drive the pulse activation process; simultaneously, new data generated after the new parameters act on the battery (such as changes in internal resistance and capacity recovery) are again input into the fitness function, forming a closed-loop optimization process of "evaluation-evolution-execution-re-evaluation".

[0160] As can be seen, in this embodiment, the genetic algorithm searches for the optimal combination of pulse parameters by performing evolutionary operations such as crossover and mutation on chromosomes to optimize the activation effect of the battery. For example, the crossover operation generates offspring chromosomes by exchanging some genes (parameters) of two parent chromosomes, while the mutation operation randomly adjusts some genes (parameters) of the chromosomes to explore a better solution space.

[0161] Furthermore, this embodiment can also achieve battery safety protection, namely voltage / current protection, through battery activation.

[0162] Optionally, after obtaining the optimal pulse parameter combination through cross-iterative loops, this embodiment may further include: using the target feature vector as a benchmark, introducing a battery aging factor, analyzing the battery aging degree, and obtaining an adjusted safety factor; based on a preset power adjustment strategy and an exponential decay strategy, adjusting the temperature parameters in the target feature vector to obtain temperature segmentation control information; performing safety monitoring on the battery based on the safety factor and the temperature segmentation control information, and feeding back abnormal information to adjust the activation control parameters when the battery malfunctions.

[0163] In its implementation, battery protection includes current / voltage protection and temperature-segmented control. Current / voltage protection is primarily adjusted based on an aging factor. Temperature-segmented control is mainly implemented using power adjustment and exponential degradation strategies. This will be explained in detail below:

[0164] In current / voltage protection, a battery aging factor is introduced. Adjust the safety factor:

[0165] when ,or When the protection is triggered, the following conditions apply: ;

[0166] in, As the initial safety factor, This is an aging adjustment factor. As the battery ages, the safety threshold is lowered to enhance protection.

[0167] In temperature-segmented control, different power adjustment strategies are introduced for different temperature segments. In addition to linearly reducing power, an exponential decay strategy can also be used.

[0168] when hour, ;

[0169] in, For initial power, It is the power decay coefficient based on temperature. Through exponential decay, the power can be reduced more quickly, thus protecting the battery.

[0170] Therefore, this embodiment achieves a closed loop of "collection-processing-decision-optimization-protection" through the core algorithm.

[0171] Step 270: Optimize battery activation based on the optimal pulse parameter combination and the charging control parameters.

[0172] In summary, firstly, based on a multi-dimensional feature vector construction method combined with a dynamic threshold decision tree classification algorithm, multi-dimensional parameters such as voltage, current, internal resistance, and temperature are integrated. Adaptive Kalman filtering and nonlinear decision trees are used to achieve accurate classification of sulfation levels, realizing intelligent state recognition through multi-data fusion. Secondly, in the battery activation mode, a composite pulse waveform generation method and a multi-parameter linkage pulse parameter adjustment algorithm are employed to construct a dynamically switching composite pulse activation strategy, realizing a "high-frequency breakage + low-frequency penetration" composite pulse mode. Parameters are dynamically adjusted based on internal resistance, temperature, and capacity change rate. Thirdly, in the activation parameter optimization, a genetic algorithm-based parameter optimization technique is introduced. A chromosome encoding method is used to encode pulse parameters as "chromosomes," and a fitness function combining the internal resistance decrease rate and gas evolution rate, along with an adaptive evolutionary operation algorithm, is integrated. Adaptive crossover / mutation operations optimize the fitness function, achieving global optimization of parameter combinations. Finally, a hardware-software collaborative safety protection mechanism is constructed. Battery aging factors and temperature exponential decay strategies are introduced into battery safety protection, along with a safety threshold adjustment method based on aging level and a temperature-triggered power exponential decay algorithm, achieving multi-level protection of voltage, current, and temperature. As can be seen, this application optimizes the activation of batteries, enabling intelligent identification of battery status, dynamic switching of activation modes, and adaptive optimization of pulse parameters. It is portable and solves the problems of low repair efficiency and poor compatibility in existing battery activation technologies.

[0173] It should be noted that, for the sake of simplicity, the method embodiments are described as a series of actions. However, those skilled in the art should know that the embodiments of this application are not limited to the described order of actions, because according to the embodiments of this application, some steps may be performed in other orders or simultaneously.

[0174] In practical implementation, this embodiment can achieve battery activation protection through multi-module collaboration, including data interaction protocols between modules and a feedback-based dynamic adjustment mechanism.

[0175] The main modules of machine dynamic interaction include:

[0176] ① The multi-dimensional data acquisition module, mainly working in conjunction with the data preprocessing module, acquires feature vectors containing multiple parameters and outputs these real-time feature vectors to the data preprocessing module. The data preprocessing module performs noise suppression through a Kalman filter preprocessing module, and then adaptively adjusts the Kalman gain. The Kalman gain depends on the data reliability factor, which is dynamically calculated based on the data fluctuation range and historical stability. Then, a temperature correction mechanism is incorporated: the calculation of AC internal resistance introduces a temperature correction term, and the correction coefficient is obtained by fitting historical data from the preprocessing module, achieving accurate measurement under temperature variations.

[0177] ② The battery state identification module classifies the degree of sulfation based on a dynamic threshold decision tree. It works in conjunction with the activation mode control module, primarily implementing dynamic threshold decision-making and automatic mode switching. Dynamic threshold decision-making: Based on preprocessed data, the battery state identification module uses a nonlinear threshold function (e.g., to determine the degree of sulfation) and incorporates correlation factors through the attribute selection of the decision tree to avoid redundant parameters affecting the judgment. Automatic mode switching: The state identification result directly drives the activation mode control module. For example, a deeply sulfated battery triggers a "high-frequency + low-frequency" composite pulse mode, which uses a formula to calculate and adjust the high and low frequency duty cycles. The high-frequency duty cycle calculation integrates multiple parameters such as internal resistance and capacity change rate.

[0178] ③ The activation mode control module performs composite pulse activation, constant current charging, or constant voltage maintenance based on the state classification results. The genetic algorithm iterative optimization module feeds back the activation effect (such as the rate of decrease in internal resistance) to the fitness function and generates new parameter combinations through crossover and mutation operations, forming a closed loop of "execution-evaluation-optimization".

[0179] ④ The pulse parameter optimization module uses a genetic algorithm to encode pulse parameters into real number vector chromosomes. It optimizes the fitness function through adaptive crossover and mutation operations, forming a closed loop of "execution-evaluation-optimization".

[0180] ⑤ Safety protection module with embedded safety constraints: The safety protection module monitors parameters such as voltage and current in real time. When the threshold is exceeded, it immediately triggers hardware cutoff and feeds back the abnormal information to the activation mode control module to adjust the strategy; the composite pulse activation mode adopts a method of dynamically adjusting the high-frequency pulse duty cycle according to the internal resistance.

[0181] like Figure 5 As shown in the figure, this application embodiment also provides an adaptive multi-mode battery activation device 500 based on multi-data fusion, including:

[0182] The multidimensional data acquisition module 510 is used to construct a multidimensional feature vector based on the multidimensional parameters of the battery acquired in real time, and update the multidimensional feature vector according to the correction factor, wherein the multidimensional parameters are the real-time status data of the battery.

[0183] The data preprocessing module 520 is used to optimize the multidimensional feature vector by means of covariance estimation and Kalman gain to obtain a target feature vector containing the parameter sequence after filtering optimization.

[0184] The battery state identification module 530 is used to determine the sulfation state information of the battery by performing decision tree generalization and multi-level sulfation state classification on the target feature vector based on a dynamic threshold classification model and a nonlinear mapping function.

[0185] The activation mode control module 540 is used to analyze the activation control parameters based on the vulcanization state information using the target feature vector to obtain pulse activation parameters and charging control parameters.

[0186] The pulse parameter optimization module 550 is used to improve the fitness function and evolutionary operation based on the pulse activation parameters using a genetic algorithm, and to obtain the optimal pulse parameter combination through cross-iteration loop.

[0187] The activation optimization module 560 is used to optimize the activation of the battery based on the optimal pulse parameter combination and the charging control parameters.

[0188] Optional, the pulse parameter optimization module includes:

[0189] The fitness analysis submodule is used to perform gene mapping encoding based on the pulse activation parameters, construct a real number vector, and use the real number vector as a benchmark, combined with preset weight coefficients, to perform analysis using a fitness function to obtain the fitness used to evaluate the merits of the pulse parameters.

[0190] The crossover and mutation operation processing submodule is used to analyze the dynamic differences in fitness based on the real number vector and the fitness, perform evolutionary improvement operations, and obtain the optimal combination of pulse parameters through crossover iteration loop.

[0191] Optionally, the adaptive multi-mode battery activation device based on multi-data fusion also includes:

[0192] The safety protection module is used to analyze the battery aging degree by introducing a battery aging factor based on the target feature vector, and obtain an adjusted safety factor; based on the preset power adjustment strategy and exponential decay strategy, it adjusts the temperature based on the temperature parameter in the target feature vector to obtain temperature segment control information; based on the safety factor and the temperature segment control information, it monitors the battery for safety, and when the battery malfunctions, it feeds back abnormal information to adjust the activation control parameters.

[0193] It should be noted that the adaptive multi-mode battery activation device based on multi-data fusion provided in the embodiments of this application can execute the adaptive multi-mode battery activation method based on multi-data fusion provided in any embodiment of this application, and has the corresponding functions and beneficial effects of the method.

[0194] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0195] The above description is merely a specific embodiment of this application, enabling those skilled in the art to understand or implement this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. An adaptive multi-mode battery activation method based on multi-data fusion, characterized in that, include: A multidimensional feature vector is constructed based on the multidimensional parameters of the battery collected in real time, and the multidimensional feature vector is updated according to the correction factor. The multidimensional parameters are the real-time status data of the battery. By estimating the covariance and combining it with Kalman gain, the multidimensional feature vector is optimized to obtain a target feature vector containing the filtered and optimized parameter sequence. Based on the dynamic threshold classification model, a nonlinear mapping function is used to perform decision tree generalization and multi-level sulfation state classification on the target feature vector to determine the sulfation state information of the battery. Based on the vulcanization state information, the activation control parameters are analyzed using the target feature vector to obtain the pulse activation parameters and charging control parameters; A genetic algorithm is used to improve the fitness function and evolutionary operation based on the pulse activation parameters, and the optimal combination of pulse parameters is obtained through crossover iteration. Based on the optimal pulse parameter combination, and in conjunction with the charging control parameters, the battery activation is optimized. The process of constructing a multidimensional feature vector based on real-time acquired multidimensional parameters of the battery, and updating the multidimensional feature vector according to a correction factor, includes: constructing a multidimensional feature vector based on real-time acquired multidimensional parameters of the battery. Based on the analysis of the internal electrochemical processes of the battery, temperature and frequency correction terms were used as correction factors. For internal resistance to communication Make corrections and update the multidimensional feature vectors; Terminal voltage, This is the charging current. For temperature, This refers to the settling time. This refers to the battery's nominal capacity. Self-discharge rate This is a temperature correction factor. This is the difference between the actual temperature and the reference temperature. This is the frequency correction factor. This is the difference between the actual injection frequency and the standard frequency. This is the capacity change calculated based on ampere-hour integration. These are factors related to health status.

2. The method according to claim 1, characterized in that, By estimating the covariance and combining it with Kalman gain to optimize the multidimensional feature vector, a target feature vector containing the filtered and optimized parameter sequence is obtained, including: Construct a time window based on the collected historical time-series data; Define prediction error covariance And using the multidimensional feature vector, time window, and data weighting factor For input, according to Obtain the prediction error covariance; Obtaining measurement difference covariance ; Kalman gain Introducing prediction error covariance Measurement difference covariance and data credibility factor ,according to The multidimensional feature vector is optimized to obtain the target feature vector; in, The size of the time window. This represents the data weighting factor that decays over time. and All of these are preset parameters.

3. The method according to claim 1, characterized in that, Based on a dynamic threshold classification model, a nonlinear mapping function is used to perform decision tree generalization and multi-level sulfation state classification on the target feature vector to determine the sulfation state information of the battery, including: Using the target feature vector as input, the changes in battery state are analyzed through a preset dynamic threshold classification model, and a nonlinear mapping function is employed based on... Analysis of internal resistance threshold and voltage threshold ; Construct a decision tree and use in-depth evaluation metrics. Combined with attribute correlation factors and according to The contribution of the target feature vector to the classification is quantified to obtain the threshold priority; with internal resistance threshold Voltage threshold Based on threshold priority and target feature vector, the sulfation state is classified using a preset multi-level sulfation state classification index to obtain the sulfation state information of the battery. in, , , , All coefficients are obtained by fitting experimental data. The internal resistance threshold reference value at the reference temperature. This is the battery's nominal voltage. Information gain ratio The Gini index, Used to measure the correlation between attributes in a target feature vector.

4. The method according to claim 1, characterized in that, Based on the vulcanization state information, the activation control parameters are analyzed using the target feature vector to obtain pulse activation parameters and charging control parameters, including: For the vulcanization state information, using the target feature vector as input, according to Analysis of high-frequency pulse duty cycle and according to Analysis of low-frequency pulse frequency ; Based on high-frequency pulse duty cycle and low frequency pulse frequency Determine the pulse activation parameters; A battery health state factor is introduced into the charging current of the target feature vector. ,according to Update the charging current to obtain constant current charging parameters. Furthermore, based on the target feature vector, temperature compensation and self-discharge rate are analyzed, according to... Analysis of float charge voltage ; The charging control parameters are determined based on the constant current charging parameters and the float charge voltage. in, Based on duty cycle, Based on the base frequency, It is a multi-parameter index based on the analysis of each parameter in the target feature vector.

5. The method according to claim 1, characterized in that, A genetic algorithm is used to improve the fitness function and evolutionary operations based on the pulse activation parameters, and the optimal combination of pulse parameters is obtained through crossover iteration, including: Gene mapping and encoding are performed based on the pulse activation parameters to construct a real number vector. Using the real number vector as a benchmark and combined with preset weight coefficients, a fitness function is used for analysis to obtain the fitness used to evaluate the merits of the pulse parameters. Based on the real number vector and the fitness, the dynamic differences in fitness are analyzed, evolutionary operations are performed to improve the system, and the optimal combination of pulse parameters is obtained through cross-iteration loops.

6. The method according to claim 5, characterized in that, Gene mapping and encoding are performed based on the pulse activation parameters to construct real-valued vectors. Using these real-valued vectors as a benchmark, and combined with preset weighting coefficients, a fitness function is employed for analysis to obtain the fitness used to evaluate the merits of the pulse parameters, including: Gene coding mapping is performed based on the pulse activation parameters, according to the constructed real number vector. ; Based on real number vectors Analysis of the rate of decrease in internal resistance Battery capacity recovery rate , gas evolution volume Combined with weighting coefficients and pulse energy utilization rate ,according to Analysis of fitness ; in, , , , All are weighting coefficients.

7. The method according to claim 5, characterized in that, Based on the real-valued vector and the fitness, the dynamic differences in fitness are analyzed, evolutionary operations are performed to improve the results, and the optimal combination of pulse parameters is obtained through cross-iteration loops, including: The real-valued vectors are used as chromosomes. The fitness differences between parent chromosomes are analyzed in conjunction with the fitness analysis, and an adaptive crossover probability is introduced based on the crossover operation. Dynamically adjust for fitness differences; The average fitness and current chromosome fitness of the population are analyzed based on the chromosomes and fitness values, and an adaptive mutation probability is introduced based on the mutation operation. The average fitness and chromosome fitness are dynamically adjusted. Based on adaptive crossover probability and adaptive mutation probability The optimal combination of pulse parameters is obtained through cross-iteration loops; in, The initial crossover probability, For fitness standard deviation, The initial mutation probability, The average fitness of the population.

8. The method according to claim 1, characterized in that, After obtaining the optimal pulse parameter combination through cross-iteration loops, the following steps are also included: Based on the target feature vector, a battery aging factor is introduced to analyze the degree of battery aging and obtain an adjusted safety coefficient. Based on the preset power adjustment strategy and exponential decay strategy, the temperature parameters in the target feature vector are adjusted to obtain temperature segmentation control information. The battery is monitored for safety based on the safety factor and the temperature segmentation control information, and when the battery malfunctions, abnormal information is fed back to adjust the activation control parameters.

9. An adaptive multi-mode battery activation device based on multi-data fusion, characterized in that, include: A multidimensional data acquisition module is used to construct a multidimensional feature vector based on the multidimensional parameters of the battery acquired in real time, and update the multidimensional feature vector according to a correction factor. The multidimensional parameters are the real-time status data of the battery. The data preprocessing module is used to optimize the multidimensional feature vector by means of covariance estimation and Kalman gain, so as to obtain a target feature vector containing the parameter sequence after filtering optimization. The battery status identification module is used to determine the sulfation status information of the battery by performing decision tree generalization and multi-level sulfation status classification on the target feature vector based on a dynamic threshold classification model and a nonlinear mapping function. The activation mode control module is used to analyze the activation control parameters based on the vulcanization state information using the target feature vector to obtain pulse activation parameters and charging control parameters. The pulse parameter optimization module is used to improve the fitness function and evolutionary operation based on the pulse activation parameters using a genetic algorithm, and to obtain the optimal combination of pulse parameters through cross-iteration loops. The activation optimization module is used to optimize battery activation based on the optimal pulse parameter combination and the charging control parameters. The process of constructing a multidimensional feature vector based on real-time acquired multidimensional parameters of the battery, and updating the multidimensional feature vector according to a correction factor, includes: constructing a multidimensional feature vector based on real-time acquired multidimensional parameters of the battery. Based on the analysis of the internal electrochemical processes of the battery, temperature and frequency correction terms were used as correction factors. For internal resistance to communication Make corrections and update the multidimensional feature vectors; Terminal voltage, This is the charging current. For temperature, This refers to the settling time. This refers to the battery's nominal capacity. Self-discharge rate This is a temperature correction factor. This is the difference between the actual temperature and the reference temperature. This is the frequency correction factor. This is the difference between the actual injection frequency and the standard frequency. This is the capacity change calculated based on ampere-hour integration. These are factors related to health status.

Images