Intelligent charging and discharging management method and system for lithium battery

Abstract

The invention discloses an intelligent charging and discharging management method and system for a lithium battery, and the method comprises the steps: collecting multi-dimensional parameter data of the lithium battery, and generating a data set with a timestamp; and analyzing the data features based on the battery type classification model, and determining a battery type identifier. And utilizing the recurrent neural network to model the relevance between the capacity attenuation and the health state, and predicting the residual capacity and the health score. If the capacity or health score is lower than the threshold value, extracting the environment temperature and the load demand to generate a temperature-load feature vector; and matching the candidate strategy set from the pre-established index database through the hash table. And distributing weights according to health scores and load demands, sorting strategy efficiency and life influences by adopting a linear regression model, and screening an optimal strategy. And if the strategy calculation complexity exceeds the equipment capability, iteratively optimizing parameters by utilizing a genetic algorithm, simplifying a strategy instruction, generating a charging current, a voltage curve and a discharging rate control instruction, and executing the charging current, the voltage curve and the discharging rate control instruction in real time by a battery management system. The method prolongs the service life of the battery and improves the charge-discharge efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of lithium batteries, and in particular to an intelligent charge and discharge management method and system for lithium batteries. Background Art

[0002] Currently, lithium batteries are a core energy storage technology in the new energy sector. Their performance and lifespan directly impact the efficiency and safety of systems such as electric vehicles and renewable energy storage. Intelligent charging and discharging management requires precise monitoring of battery status and dynamic optimization strategies. However, most existing solutions rely on a single parameter (such as voltage or current), ignoring the dynamic interactions of multiple parameters such as battery type, capacity decay, health status, ambient temperature, and load demand. This results in insufficient policy adaptability in complex scenarios, which can easily lead to reduced efficiency or accelerated lifespan decay. The core challenges are concentrated in three aspects: First, the real-time fusion and dynamic analysis capabilities of multi-dimensional parameters (such as the temperature-capacity relationship in high-temperature and high-load scenarios) are insufficient, making it difficult to support accurate decision-making; second, the lack of a systematic indexing mechanism for massive amounts of data leads to inefficient policy matching; and third, the high computational requirements of complex optimization algorithms conflict with the limited resources (computing power and memory) of embedded devices, making real-time performance difficult to ensure. To this end, it is urgent to build a lightweight index database based on multi-dimensional parameters, and through lightweight optimization methods such as dynamic data fusion modeling, fast hash query and genetic algorithm, achieve efficient strategy matching and low resource consumption execution in complex scenarios, thereby breaking through the bottleneck of coordinated optimization of battery performance and life, and improving the environmental adaptability and reliability of the management system.

[0003] One existing implementation process is as follows: a voltage sensor monitors the terminal voltage of a lithium battery in real time, sampling at a rate of once per second. This generates a timestamped voltage dataset, such as "Timestamp: 2025-04-21 10:00:00, Voltage: 3.7V." The system presets a fixed upper charging voltage limit of 4.2V and a lower discharging voltage limit of 3.0V. The rule base only stores simple action instructions based on voltage thresholds, such as shutting off the charging circuit when the charging voltage reaches 4.2V and forcibly terminating discharge and entering sleep mode when the discharge voltage falls below 3.0V. During execution, the control module directly triggers actions based on the comparison between the real-time voltage and the threshold. For example, if the charging voltage exceeds 4.2V, charging is stopped, and if the discharge voltage falls below 3.0V, the output is turned off. The execution results are reflected through status indicators, such as a green light indicating normal operation and a red light indicating overvoltage or undervoltage protection. However, the system lacks a dynamic parameter adjustment mechanism and does not store historical data for policy optimization. The defects of existing technologies are reflected in the static monitoring mechanism that adopts a single electrochemical characteristic parameter, which does not fully integrate the differences in battery material systems and the dynamic performance evolution during use (such as the drift of the capacity-voltage relationship during the aging process). At the same time, the multi-physics field coupling effects of external working conditions and internal states (such as the time-varying correlation between temperature gradients and current loads) lack systematic representation in strategy formulation, resulting in the difficulty of fixed threshold mechanisms to adapt to complex working conditions. In addition, when dealing with multi-parameter dynamic interaction scenarios, the static strategy library based on prior rules lacks the ability to dynamically reconfigure the strategy generation mechanism driven by real-time parameters, which restricts the balanced optimization of battery system performance and service life. Summary of the Invention

[0004] The present invention provides a method and system for intelligent charge and discharge management of lithium batteries, so as to extend battery life and improve charge and discharge efficiency.

[0005] In a first aspect, in order to solve the above technical problems, the present invention provides a method for intelligent charge and discharge management of a lithium battery, comprising: Acquiring multi-dimensional parameter data including a timestamp, the multi-dimensional parameter data including battery chemical composition, battery structure, ambient temperature, and load demand; Determining a battery type identifier using a pre-established battery type classification model based on the multi-dimensional parameter data; According to the battery type identifier and the multidimensional parameter data, a recurrent neural network model is used to predict the battery remaining capacity and attenuation trend to obtain a battery remaining capacity and health status score; When the remaining capacity of the battery is lower than a preset capacity threshold or the health status score is lower than a preset health threshold, a temperature-load feature is generated using a principal component analysis method according to the ambient temperature and the load demand; According to the temperature-load characteristics, a pre-built index database is used to query a candidate charge and discharge strategy set that matches the battery type identifier, the battery remaining capacity, the health status score, and the temperature-load characteristics through a hash table to obtain a preliminary strategy set; weighting the health status score and the load demand, ranking the efficiency and lifespan impact of the preliminary strategy set using a linear regression model, and determining an optimal charge and discharge strategy; Mapping and comparing the optimal charge and discharge strategy with the benchmark performance parameters of the embedded device, and when the optimal charge and discharge strategy exceeds the benchmark performance parameters of the embedded device, simplifying the strategy parameters using a genetic algorithm and reducing the computational complexity through iterative optimization to obtain a simplified charge and discharge strategy; According to the simplified charge and discharge strategy, control instructions are generated and executed in real time by the battery management system.

[0006] In an optional embodiment, the process of constructing the battery type classification model includes: Acquire historical battery data, remove outliers from the historical battery data using a mean filtering method, and normalize the data using a Z-score normalization method to generate a standardized data set; Performing dimensionality reduction processing on the standardized data set using a principal component analysis algorithm and extracting a battery feature vector, wherein the battery feature vector includes a chemical composition vector and a structural feature vector; The battery feature vector and the preset battery type label are used as training data, and an initial battery type classification model is constructed using a support vector machine algorithm. A radial basis function is selected as the kernel function type, and an initial penalty coefficient and a kernel function parameter range are defined, wherein the battery feature vector serves as the input of the initial battery type classification model, and the battery type label serves as the output of the initial battery type classification model; The training data is divided into training subsets and validation subsets by the multi-fold cross-validation method, and the grid search algorithm is used to traverse the preset candidate penalty coefficient and kernel parameter combinations to obtain the parameter combination; In each parameter combination training stage, a kernel function matrix is ​​calculated based on the training subset, Lagrange multipliers are solved by a sequential minimum optimization algorithm, and a classification hyperplane is constructed; Inputting the verification subset into the initial battery type classification model under the parameter combination, and combining the classification hyperplane, calculating the matching ratio between the predicted category and the true label as the verification accuracy, thereby obtaining the verification accuracy; Selecting the parameter combination with the highest verification accuracy as the optimal model parameters, and performing parameter optimization on the initial battery type classification model; After completing the parameter optimization, the model is retrained using the training data in combination with the optimal model parameters to generate an optimal battery type classification model.

[0007] In an optional embodiment, the method of using a recurrent neural network model to predict the remaining battery capacity and attenuation trend based on the battery type identifier and the multidimensional parameter data to obtain the remaining battery capacity and health status score includes: Based on the battery type identifier and the multidimensional parameter set, after cleaning outliers using the median method, a standardized parameter set is generated through Min-Max normalization; Extracting dynamic features of the standardized parameter set using a recurrent neural network to obtain dynamic features of capacity degradation; Based on the dynamic characteristics of capacity degradation, a recurrent neural network is used to perform time series prediction to obtain the remaining capacity and attenuation trend of the battery; The battery remaining capacity, the decay trend, and the preset capacity weight and decay weight are weighted by using a linear weighting method to obtain a health status score.

[0008] In an optional embodiment, the training process of the recurrent neural network model includes: Acquire historical multidimensional parameter data and corresponding historical battery type identifiers, and perform data cleaning and feature extraction on the historical multidimensional parameter data and the historical battery type identifiers to obtain data features; Inputting the data features into the input layer of the initially constructed recurrent neural network model for training, and obtaining the predicted battery remaining capacity and attenuation trend output by the output layer of the recurrent neural network model; Substituting the predicted battery remaining capacity and attenuation trend and the pre-stored actual battery remaining capacity and attenuation trend empirical values ​​into a loss function to calculate a loss value; Calculate the gradient of the output layer output of the recurrent neural network model according to the loss value, and pass the gradient forward layer by layer through the chain rule to calculate the gradient of the parameters of each layer to obtain gradient data; Update the parameters of each layer of the neural network model according to the gradient data and the preset learning rate; The parameters of each layer are updated repeatedly until the training times of the recurrent neural network model are greater than the preset times, or the loss value data of the recurrent neural network model is less than the preset loss threshold, the training is determined to be completed, and a trained recurrent neural network model is obtained.

[0009] In an optional embodiment, generating the temperature-load characteristics using a principal component analysis method according to the ambient temperature and the load demand includes: Using a linear interpolation algorithm to match the ambient temperature and the load demand to generate temperature-load joint data; According to the temperature-load joint data, a principal component analysis algorithm is used to extract dynamic interaction features of the temperature data and the load curve data to obtain temperature-load features.

[0010] In an optional embodiment, the weighting of the health status score and the load demand, sorting the efficiency and lifespan impact of the preliminary strategy set by a linear regression model, and determining the optimal charge and discharge strategy includes: weighting the health status score and the load demand using an entropy weight method to obtain weight data; Calculating the weighted scores of the preliminary strategy set using a linear weighted summation method based on the weight data to obtain weighted strategy data; Based on the weighted strategy data, a multiple linear regression model is used to analyze the regression coefficients of the efficiency impact and life impact of the preliminary strategy set, and the predicted efficiency and predicted life are calculated; When the predicted efficiency is higher than a preset efficiency threshold and the predicted life is higher than a preset life threshold, the preliminary strategy is included in a ranking list, and a ranking result is obtained by arranging the strategies in the ranking list in descending order according to the regression coefficient; According to the ranking result, the highest-ranked preliminary strategy is extracted from the ranking list and determined as the optimal charging and discharging strategy.

[0011] In an optional embodiment, mapping and comparing the optimal charge and discharge strategy with the benchmark performance parameters of the embedded device, when the optimal charge and discharge strategy exceeds the benchmark performance parameters of the embedded device, simplifying the strategy parameters using a genetic algorithm, and reducing the computational complexity through iterative optimization to obtain a simplified charge and discharge strategy, including: Mapping and comparing the optimal charge and discharge strategy with the benchmark performance parameters of the embedded device, and when the optimal charge and discharge strategy exceeds the benchmark performance parameters of the embedded device, obtaining strategy parameters from the optimal charge and discharge strategy, wherein the strategy parameters include charge rate, discharge depth, and number of cycles; Performing initial population encoding on the strategy parameters using a genetic algorithm to generate a first parameter aggregation; According to the first parameter aggregation, crossover and mutation operations of a genetic algorithm are used to iteratively optimize the first parameter aggregation to generate a second parameter set; The second parameter set is used as a simplified parameter and replaces the strategy parameters in the optimal charge and discharge strategy to generate a simplified charge and discharge strategy.

[0012] In a second aspect, the present invention provides an intelligent charge and discharge management system for lithium batteries, comprising: A data acquisition module for acquiring multi-dimensional parameter data including timestamps, the multi-dimensional parameter data including battery chemical composition, battery structure, ambient temperature and load demand; A battery classification module, configured to determine a battery type identifier based on the multi-dimensional parameter data using a pre-established battery type classification model; A battery status analysis module, configured to predict the battery remaining capacity and decay trend using a recurrent neural network model based on the battery type identifier and the multidimensional parameter data, and obtain a battery remaining capacity and health status score; a temperature-load signature module, configured to generate a temperature-load signature using a principal component analysis method based on the ambient temperature and the load demand when the remaining battery capacity is lower than a preset capacity threshold or the health status score is lower than a preset health threshold; a preliminary strategy generation module, which searches a pre-built index database through a hash table for a candidate charge and discharge strategy set that matches the battery type identifier, the battery remaining capacity, the health status score, and the temperature-load characteristics to obtain a preliminary strategy set; an optimal strategy generation module, configured to assign weights to the health status score and the load demand, sort the efficiency and lifespan impact of the preliminary strategy set using a linear regression model, and determine an optimal charge and discharge strategy; a strategy simplification module, configured to map and compare the optimal charge and discharge strategy with the benchmark performance parameters of the embedded device, and when the optimal charge and discharge strategy exceeds the benchmark performance parameters of the embedded device, simplify the strategy parameters using a genetic algorithm and reduce the computational complexity through iterative optimization to obtain a simplified charge and discharge strategy; The strategy execution module is used to generate control instructions according to the simplified charge and discharge strategy, and execute the control instructions in real time through the battery management system.

[0013] In a third aspect, the present invention also provides an electronic device comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the intelligent charge and discharge management method for a lithium battery as described above is implemented.

[0014] In a fourth aspect, the present invention also provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute any one of the above-mentioned intelligent charging and discharging management methods for lithium batteries.

[0015] Compared with the prior art, the present invention has the following beneficial effects: (1) By collecting multi-dimensional parameters such as ambient temperature and load demand in real time, combining recurrent neural networks to predict the remaining capacity and attenuation trend of the battery, and using principal component analysis to extract temperature-load characteristics, the problem of poor strategy adaptability caused by the existing technology relying on a single parameter is solved, and the comprehensiveness and prediction accuracy of battery status monitoring are significantly improved. (2) Based on the pre-built index database and hash table query mechanism, the candidate charging and discharging strategy set is quickly matched through multi-dimensional feature vectors, which solves the technical problem of low efficiency in massive data retrieval, achieves millisecond-level strategy response in complex scenarios, and ensures the feasibility of real-time dynamic optimization. (3) In view of the computing power limitations of embedded devices, a genetic algorithm is used to iteratively optimize the parameters of the optimal strategy (such as crossover, mutation, and population screening), which reduces the computational complexity while retaining the core effectiveness of the strategy, effectively balances the algorithm accuracy and resource consumption, and ensures the efficient operation of low-power devices. (4) The entropy weight method is used to dynamically assign weights to the health status score and load demand, and the linear regression model is used to quantify the influence coefficient of the strategy on efficiency and life. This solves the problem of strategy rigidity caused by the traditional method relying on fixed weights, realizes the coordinated optimization of charging and discharging efficiency and battery life, and improves the strategy adaptability and comprehensive performance in complex scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 This is a flow chart of the intelligent charge and discharge management method for lithium batteries provided by the first embodiment of the present invention; Figure 2 It is a structural diagram of an intelligent charge and discharge management system for lithium batteries provided by the second embodiment of the present invention. DETAILED DESCRIPTION

[0017] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0018] Reference Figure 1 The first embodiment of the present invention provides a method for intelligent charge and discharge management of a lithium battery, comprising the following steps: S11, acquiring multi-dimensional parameter data including a timestamp, the multi-dimensional parameter data including battery chemical composition, battery structure, ambient temperature, and load demand; S12, determining a battery type identifier using a pre-established battery type classification model based on the multi-dimensional parameter data; S13, predicting the battery remaining capacity and decay trend using a recurrent neural network model based on the battery type identifier and the multidimensional parameter data to obtain a battery remaining capacity and health status score; S14, when the remaining battery capacity is lower than a preset capacity threshold or the health status score is lower than a preset health threshold, generating a temperature-load feature using a principal component analysis method according to the ambient temperature and the load demand; S15, based on the temperature-load characteristics, using a pre-built index database, searching a hash table for a set of candidate charge and discharge strategies that match the battery type identifier, the battery remaining capacity, the health status score, and the temperature-load characteristics, to obtain a preliminary strategy set; S16, assigning weights to the health status scores and the load demands, ranking the efficiency and lifespan impacts of the preliminary strategy set using a linear regression model, and determining an optimal charge and discharge strategy; S17, mapping and comparing the optimal charge and discharge strategy with the benchmark performance parameters of the embedded device. When the optimal charge and discharge strategy exceeds the benchmark performance parameters of the embedded device, simplifying the strategy parameters using a genetic algorithm and reducing the computational complexity through iterative optimization to obtain a simplified charge and discharge strategy. S18: Generate control instructions according to the simplified charge and discharge strategy, and execute the control instructions in real time through the battery management system.

[0019] In step S11 , multi-dimensional parameter data including a timestamp is acquired, where the multi-dimensional parameter data includes battery chemical composition, battery structure, ambient temperature, and load demand.

[0020] Specifically, chemical sensors deployed on the lithium battery pack collect battery chemical composition parameters, including cathode material type, anode material ratio, and electrolyte composition. Structural sensors acquire battery structural parameters, including battery shape, casing material, and number of electrode layers. Temperature sensors collect real-time surface and ambient temperature data. The battery management system's current and voltage monitoring module acquires load demand parameters, including real-time discharge current, operating voltage, and power demand. The data acquisition module synchronously records these parameters at a fixed sampling frequency and assigns a millisecond-accurate timestamp to each sampling point. A timestamp alignment algorithm uses linear interpolation to align asynchronously sampled data from different sensors to the same time base. For example, interpolating the temperature sensor's sampled data every second with the load demand parameter's sampled data every 200 milliseconds creates a timestamp-aligned joint dataset. The data cleaning module uses a sliding window mean filter algorithm with a window size of five data points to remove temperature data that deviates by more than 3 degrees Celsius from the window mean. The cleaned dataset is stored in a structured table in timestamp order. The table fields include acquisition time, positive electrode material, negative electrode material, electrolyte composition, battery shape, housing material, number of electrode layers, ambient temperature, load current, and operating voltage. The standardized multidimensional parameter dataset output from this step provides comprehensive input data for the subsequent battery type classification model, encompassing time dimensions, chemical properties, physical structure, environmental factors, and load conditions.

[0021] In step S12, a battery type identifier is determined based on the multi-dimensional parameter data using a pre-established battery type classification model.

[0022] In a specific embodiment, the process of constructing the battery type classification model includes: Acquire historical battery data, remove outliers from the historical battery data using a mean filtering method, and normalize the data using a Z-score normalization method to generate a standardized data set; Performing dimensionality reduction processing on the standardized data set using a principal component analysis algorithm and extracting a battery feature vector, wherein the battery feature vector includes a chemical composition vector and a structural feature vector; The battery feature vector and the preset battery type label are used as training data, and an initial battery type classification model is constructed using a support vector machine algorithm. A radial basis function is selected as the kernel function type, and an initial penalty coefficient and a kernel function parameter range are defined, wherein the battery feature vector serves as the input of the initial battery type classification model, and the battery type label serves as the output of the initial battery type classification model; The training data is divided into training subsets and validation subsets by the multi-fold cross-validation method, and the grid search algorithm is used to traverse the preset candidate penalty coefficient and kernel parameter combinations to obtain the parameter combination; In each parameter combination training stage, a kernel function matrix is ​​calculated based on the training subset, Lagrange multipliers are solved by a sequential minimum optimization algorithm, and a classification hyperplane is constructed; Inputting the verification subset into the initial battery type classification model under the parameter combination, and combining the classification hyperplane, calculating the matching ratio between the predicted category and the true label as the verification accuracy, thereby obtaining the verification accuracy; Selecting the parameter combination with the highest verification accuracy as the optimal model parameters, and performing parameter optimization on the initial battery type classification model; After completing the parameter optimization, the model is retrained using the training data in combination with the optimal model parameters to generate an optimal battery type classification model.

[0023] Specifically, the construction process of the battery type classification model is implemented as follows: Historical battery data comes from the test database provided by lithium battery manufacturers, which contains structured parameters of the battery's positive electrode material, negative electrode material, electrolyte ratio, shell material, and number of electrode layers. Each data record corresponds to a preset type label for a specific battery model. In the preprocessing stage, a sliding mean filtering algorithm with a window size of 5 is used to calculate the sliding mean of the numerical features of the positive electrode material's lithium cobalt oxide content and the electrolyte solvent concentration, and abnormal data that deviates from the window mean by more than 3 times the standard deviation is eliminated. In the normalization processing stage, the continuous variables of the negative electrode material's graphite ratio and the number of electrode layers are Z-score standardized so that the mean of each feature is 0 and the standard deviation is 1. In the dimensionality reduction stage, the principal component analysis algorithm is used to project the standardized 16-dimensional feature data into a 5-dimensional principal component space, and extract the chemical composition vector containing the main components of the positive electrode material and the electrolyte, and the structural feature vector containing the main components of the electrode structure and the shell material. During the model training phase, the chemical composition vector and the structural feature vector are concatenated into a 10-dimensional feature input vector, which is then combined with three preset type labels: ternary lithium battery, lithium iron phosphate battery, and lithium titanate battery to form a training set. The initial support vector machine model sets the penalty coefficient search range of the radial basis kernel function to 0.1 to 10, and the kernel function parameter search range to 0.01 to 0.5. A 5-fold cross-validation is used to divide the training set into a training subset and a validation subset with a 4:1 ratio, and a grid search is performed on 36 sets of parameter combinations. When training each set of parameters, the support vector is iteratively calculated on the training subset using a sequential minimum optimization algorithm to construct a classification hyperplane separating ternary lithium batteries and lithium iron phosphate batteries, and the type recognition accuracy on the validation subset is calculated. After selecting the parameter combination with the highest validation accuracy (e.g., 92.3%), the model is retrained using the complete training set to ultimately generate the optimal battery type classification model.

[0024] During the model's deployment phase, multidimensional parameter data is input into the trained battery type classification model. The preprocessing module performs the same Z-score normalization on the input data as in the training phase. For example, the ambient temperature of 25°C is normalized to 0.8, and the load current of 50A is normalized to 1.2. The feature extraction module uses the projection matrix of the principal component analysis algorithm to reduce the dimensionality of the normalized 16-dimensional data into a 5-dimensional feature vector. After receiving this 5-dimensional feature vector, the support vector machine classifier calculates the decision value of the distance classification hyperplane using the radial basis kernel function constructed during the training phase. For example, if the chemical feature vector is in the positive direction of the hyperplane and the structural feature vector offset is less than a threshold, the output type identifier is "NCM622 ternary lithium battery." The classification result is mapped to a preset identification code using a type encoder, such as 1 for ternary lithium batteries and 2 for lithium iron phosphate batteries. The identification code is ultimately written to the type register of the battery management system for subsequent modules to access.

[0025] In step S13, a recurrent neural network model is used to predict the battery remaining capacity and attenuation trend based on the battery type identifier and the multidimensional parameter data to obtain the battery remaining capacity and health status score.

[0026] In a specific embodiment, the method of using a recurrent neural network model to predict the remaining capacity and decay trend of the battery based on the battery type identifier and the multidimensional parameter data to obtain the remaining capacity and health status score of the battery includes: Based on the battery type identifier and the multidimensional parameter set, after cleaning outliers using the median method, a standardized parameter set is generated through Min-Max normalization; Extracting dynamic features of the standardized parameter set using a recurrent neural network to obtain dynamic features of capacity degradation; Based on the dynamic characteristics of capacity degradation, a recurrent neural network is used to perform time series prediction to obtain the remaining capacity and attenuation trend of the battery; The battery remaining capacity, the decay trend, and the preset capacity weight and decay weight are weighted by using a linear weighting method to obtain a health status score.

[0027] In a specific embodiment, the training process of the recurrent neural network model includes: Acquire historical multidimensional parameter data and corresponding historical battery type identifiers, and perform data cleaning and feature extraction on the historical multidimensional parameter data and the historical battery type identifiers to obtain data features; Inputting the data features into the input layer of the initially constructed recurrent neural network model for training, and obtaining the predicted battery remaining capacity and attenuation trend output by the output layer of the recurrent neural network model; Substituting the predicted battery remaining capacity and attenuation trend and the pre-stored actual battery remaining capacity and attenuation trend empirical values ​​into a loss function to calculate a loss value; Calculate the gradient of the output layer output of the recurrent neural network model according to the loss value, and pass the gradient forward layer by layer through the chain rule to calculate the gradient of the parameters of each layer to obtain gradient data; Update the parameters of each layer of the neural network model according to the gradient data and the preset learning rate; The parameters of each layer are updated repeatedly until the training times of the recurrent neural network model are greater than the preset times, or the loss value data of the recurrent neural network model is less than the preset loss threshold, the training is determined to be completed, and a trained recurrent neural network model is obtained.

[0028] Specifically, the preprocessing module receives the battery type identifier (e.g., ternary lithium battery type code 1) output from step S12 and the multidimensional parameter data from step S11. The outlier cleaning stage uses the median method to calculate the median of the negative electrode material expansion rate using 10 consecutive data points as a window, removing data points that deviate from the median by more than 20%. The normalization module performs a Min-Max linear transformation on the number of current cycles and voltage fluctuation range after cleaning, mapping the charge rate parameter from the original range of 0.5C-2C to the range of 0 to 1, for example, converting a 1.5C charge rate to 0.75. The recurrent neural network model uses a bidirectional LSTM architecture with a time step of 30. The input layer receives time series data. The dynamic feature extraction layer uses LSTM unit memory cells to store capacity decay patterns from historical charge and discharge cycles. For example, it captures a trend of 1.2% capacity decrease for every 10 cycles over 50 consecutive cycles. The time series prediction layer outputs a capacity forecast for the next five charge and discharge cycles, including the remaining battery capacity and decay trend (e.g., from 82% to 78% of the current capacity), as well as the corresponding decay slope (e.g., a monthly capacity decay rate of 0.8%). The health score generation module inputs the current remaining capacity, predicted battery remaining capacity, a decay trend of 82%, and a decay slope of 0.8% into a weighted calculator, with the capacity weighting factor set to 0.7 and the decay trend weighting factor set to 0.3. The calculation first converts the decay slope to an annual decay of 9.6%. Then, according to the preset scoring formula, the capacity value is multiplied by 0.7 and (100% - annual decay) by 0.3, resulting in a weighted health score of 85.2. When the score falls below the preset threshold of 80, the strategy optimization process is triggered. During the model training phase, historical multi-dimensional parameter data and corresponding historical battery type identifiers are acquired, cleaned, and feature extracted. The processed data features are fed into the input layer of the recurrent neural network model for training, and the output layer generates predicted remaining battery capacity and decay trends. The predicted values ​​and pre-stored empirical values ​​of actual remaining battery capacity and decay trends are substituted into the mean squared error loss function to calculate the current loss value. Based on the loss value, the gradient is backpropagated, and the chain rule is used to calculate the gate weight gradients of each LSTM unit layer by layer from the output layer to the input layer to obtain gradient data. The optimizer updates the parameters of each layer at a preset learning rate of 0.001, specifically adjusting the weight matrices of the forget gate and input gate using the Adam optimization algorithm. Two termination conditions are continuously monitored during the iterative update process: training is terminated immediately when the number of training iterations reaches a preset upper limit of 150 or when the validation set loss value falls below a preset threshold of 0.5%. After parameter updates, the model achieves a stable error level of 0.48% on the validation set. After lightweight compression, it is deployed on embedded devices, achieving real-time prediction capabilities in 15 milliseconds. This training mechanism uses strict iteration control and loss threshold judgment to ensure model accuracy while avoiding overfitting, ensuring the predictive model has engineering practicality. By integrating real-time data and historical decay patterns, the predictive model provides an accurate lifespan assessment basis for different battery types.

[0029] In step S14, when the remaining battery capacity is lower than a preset capacity threshold or the health status score is lower than a preset health threshold, a temperature-load feature is generated using principal component analysis according to the ambient temperature and the load demand.

[0030] In a specific embodiment, generating the temperature-load characteristics using principal component analysis according to the ambient temperature and the load demand includes: Using a linear interpolation algorithm to match the ambient temperature and the load demand to generate temperature-load joint data; According to the temperature-load joint data, a principal component analysis algorithm is used to extract dynamic interaction features of the temperature data and the load curve data to obtain temperature-load features.

[0031] Specifically, when the remaining battery capacity or health score falls below a preset threshold, due to the sampling frequency discrepancy between the temperature sensor and the load monitoring module, a linear interpolation algorithm is used. Based on the timestamp of the load data, the temperature values ​​between two adjacent temperature sampling points are linearly calculated. This generates a temperature estimate that is fully synchronized with the load data timestamp, forming a temperature-load joint data set that includes a unified timestamp, ambient temperature, load current, and operating voltage. In the feature extraction phase, the temperature-load joint data is input into the principal component analysis module, which calculates the covariance matrix of the temperature fluctuation variance and the load current change rate. Feature vectors are extracted to construct an orthogonal projection space. The first principal component captures the positive correlation between the temperature rise phase and the load current increase. The second principal component characterizes the response of the load current fluctuation amplitude within the temperature stability range. The third principal component reflects the relationship between the temperature gradient change rate and the load duration. The final output temperature-load feature vector consists of three principal component scores, corresponding to the synchronization feature code, the stability feature code, and the persistence feature code, providing standardized feature input for the policy index library to characterize the dynamic coupling relationship between the environment and the load. This process converts the original heterogeneous time series data into low-dimensional feature expressions with clear physical meaning through spatiotemporal data fusion and feature dimensionality reduction. It helps the system avoid the risk of overcharging under high-temperature conditions and electrode structure damage caused by high-load impact, significantly extends the battery cycle life, and at the same time improves the energy conversion efficiency of the charging and discharging process, achieving coordinated optimization of battery performance and service life in complex scenarios.

[0032] In step S15, based on the temperature-load characteristics, a pre-built index database is used to query a candidate charge and discharge strategy set that matches the battery type identifier, the battery remaining capacity, the health status score and the temperature-load characteristics through a hash table to obtain a preliminary strategy set.

[0033] Specifically, the matching process for candidate charging and discharging strategy sets uses a hash table mechanism for efficient retrieval. A hash table consists of a fixed number of storage units (hash buckets), each of which stores a set of strategy index pointers corresponding to a specific hash key. A modulo operation involves taking the remainder of the integer value generated by the hash function and the total number of hash buckets. For example, if the hash value is 3587 and the number of buckets is 100, 3587 % 100 = 87, which locates hash bucket number 87. This operation maps hash keys of any length to a fixed range of bucket addresses, ensuring stable query time complexity.

[0034] After the system generates a composite hash key consisting of a battery type identification code (such as a ternary lithium battery type code), a remaining capacity segmentation code (such as a capacity interval code), a health score code, and a temperature-load characteristic code, a hash function converts it into a unique integer value. A modulo operation is used to determine the target hash bucket location. All policy index pointers within that bucket are traversed to quickly retrieve the set of charge and discharge policy parameters that match the current operating conditions (such as charge rate classification values, discharge cutoff voltage thresholds, and cycle interval parameters). Each policy entry is associated with historical execution data, including records of efficiency improvement and capacity fade suppression effects.

[0035] This mechanism transforms the complex parameter space matching problem into a key-value query with constant time complexity through multi-dimensional feature discretization and hash mapping, enabling the system to respond to coordinated changes in battery status and environmental load within milliseconds. Compared to traditional linear traversal of the policy library, hash table indexing speeds up policy matching in high-temperature and high-load scenarios by two orders of magnitude, ensuring that the battery always operates within a safe and efficient range during real-time dynamic optimization, avoiding the risks of overcharging, over-discharging, or thermal runaway caused by policy retrieval delays, thereby extending battery cycle life and improving energy utilization.

[0036] In step S16, weights are assigned to the health status scores and the load demands, and the efficiency and lifespan impacts of the preliminary strategy set are ranked using a linear regression model to determine an optimal charge and discharge strategy.

[0037] In a specific embodiment, the weighting of the health status score and the load demand, sorting the efficiency and lifespan impact of the preliminary strategy set by a linear regression model, and determining the optimal charge and discharge strategy includes: weighting the health status score and the load demand using an entropy weight method to obtain weight data; Calculating the weighted scores of the preliminary strategy set using a linear weighted summation method based on the weight data to obtain weighted strategy data; Based on the weighted strategy data, a multiple linear regression model is used to analyze the regression coefficients of the efficiency impact and life impact of the preliminary strategy set, and the predicted efficiency and predicted life are calculated; When the predicted efficiency is higher than a preset efficiency threshold and the predicted life is higher than a preset life threshold, the preliminary strategy is included in a ranking list, and a ranking result is obtained by arranging the strategies in the ranking list in descending order according to the regression coefficient; According to the ranking result, the highest-ranked preliminary strategy is extracted from the ranking list and determined as the optimal charging and discharging strategy.

[0038] Specifically, the optimal charge and discharge strategy is determined as follows: The input data is the preliminary strategy set (including strategy parameters such as charge rate, depth of discharge, and number of cycles, as well as their historical execution records) output in step S15, the health score generated in step S13, and the current load demand parameters. The weight assignment module uses an entropy weighting method to calculate the information entropy of the health score time series (reflecting the degree of score fluctuation) and the information entropy of the load demand time series (reflecting the severity of load fluctuation). If the health score entropy is lower than the load demand entropy, the health score is given a higher weight (e.g., a weight coefficient of 0.6 and a load demand weight of 0.4). Conversely, if the health score entropy is higher than the load demand entropy, the load demand is given a higher weight (e.g., a weight coefficient of 0.7).

[0039] The weighted calculator analyzes historical execution records in the policy library to calculate each policy's health score relevance value. This value is calculated by multiplying the average rate of change in the health score over the policy's execution cycle (for example, if the health score of a policy increases from 78 to 80 after historical execution, the relevance value is +2 points / cycle) by the linear correlation coefficient of the current health score. It also calculates the load demand satisfaction, which is the degree of match between the policy's allowed peak current parameter and the current load demand curve. For example, if the policy supports a maximum discharge of 50A and the current peak load demand is 40A, (40 / 50) × 100% gives an 80% match. The health score relevance value is multiplied by the health score weight, and then the load demand satisfaction value is added to the corresponding load demand weight to generate the policy's overall score.

[0040] During the regression analysis phase, the strategy parameters, including charge rate and discharge cutoff voltage, are used as independent variables, and the charge and discharge efficiency improvement rate and capacity decay rate from the strategy's historical data are used as dependent variables to construct a multivariate linear regression model. The model calculates the regression coefficients for the effect of charge rate on efficiency (e.g., a coefficient of 0.8 indicates that every 0.1C increase in charge rate increases efficiency by 0.8%) and the regression coefficients for the effect of discharge depth on lifespan (e.g., a coefficient of -1.2 indicates that every 10% increase in discharge depth accelerates capacity decay by 1.2%). The prediction module, based on the regression coefficients of the current strategy parameters, outputs the predicted efficiency (e.g., an estimated efficiency of 89%) and lifespan (e.g., an estimated capacity retention period of 120 cycles) after executing the strategy.

[0041] The strategy filter double-checks the predicted efficiency against a preset efficiency threshold (e.g., a minimum of 85%) and the predicted lifespan against a preset lifespan threshold (e.g., a minimum of 100 cycles). Strategies that pass the threshold are sorted in descending order by the sum of their regression coefficients (e.g., the sum of an efficiency coefficient of 0.8 and a lifespan coefficient of 0.5 is 1.3). The strategy with the highest sum of regression coefficients in the list (e.g., a charge rate of 1C, a depth of discharge of 80%, and a cycle interval of 24 hours) is ultimately selected as the optimal strategy. This strategy strikes the best balance between efficiency improvement and lifespan maintenance, and quantitative metrics ensure objective and verifiable selection.

[0042] In step S17, the optimal charge and discharge strategy is mapped and compared with the benchmark performance parameters of the embedded device. When the optimal charge and discharge strategy exceeds the benchmark performance parameters of the embedded device, the strategy parameters are simplified using a genetic algorithm, and the computational complexity is reduced through iterative optimization to obtain a simplified charge and discharge strategy.

[0043] In a specific embodiment, the optimal charge and discharge strategy is mapped and compared with the benchmark performance parameters of the embedded device. When the optimal charge and discharge strategy exceeds the benchmark performance parameters of the embedded device, a genetic algorithm is used to simplify the strategy parameters, and the computational complexity is reduced through iterative optimization to obtain a simplified charge and discharge strategy, including: Mapping and comparing the optimal charge and discharge strategy with the benchmark performance parameters of the embedded device, and when the optimal charge and discharge strategy exceeds the benchmark performance parameters of the embedded device, obtaining strategy parameters from the optimal charge and discharge strategy, wherein the strategy parameters include charge rate, discharge depth, and number of cycles; Performing initial population encoding on the strategy parameters using a genetic algorithm to generate a first parameter aggregation; According to the first parameter aggregation, crossover and mutation operations of a genetic algorithm are used to iteratively optimize the first parameter aggregation to generate a second parameter set; The second parameter set is used as a simplified parameter and replaces the strategy parameters in the optimal charge and discharge strategy to generate a simplified charge and discharge strategy.

[0044] Specifically, the strategy parameter simplification process is as follows: When the embedded device's computing resource monitoring module detects that the number of floating-point operations required for the optimal charge-discharge strategy exceeds the processor's maximum capability (for example, if the strategy requires a million matrix operations per second while the device only supports 100,000), the genetic algorithm optimization process is triggered. The strategy parameter extraction module extracts three core parameters from the optimal strategy: the charge rate setpoint, the depth of discharge limit, and the charge-discharge cycle interval. These parameters serve as optimization variables for the genetic algorithm.

[0045] The initial population generation phase uses real number encoding, combining the charging rate parameter (e.g., 1.2C) and the discharge depth parameter (e.g., 80%) into chromosome individuals. This randomly generates an initial population containing 50 parameter combinations (e.g., individual 1 is [1.0C, 75%, 30 minutes], individual 2 is [0.8C, 85%, 45 minutes]). The fitness function calculates the strategy simplification potential of each individual. The reduction in computational complexity after parameter simplification is evaluated by analyzing the data accuracy requirements of the charging rate parameter (e.g., simplifying the floating-point value 1.2C to the integer 1C can reduce computational complexity by 32%) and the feasibility of segmented processing of the number of cycles parameter (e.g., changing continuous time control to fixed intervals can reduce computational load by 60%). During the crossover phase, the top 10 fitness individuals are selected as parents. A single-point crossover is performed between the charge rate and discharge depth parameters (for example, a charge rate of 1.0C for parent individual A and a discharge depth of 85% for parent individual B generate a child individual of [1.0C, 85%, 30 minutes]). A mutation phase randomly adjusts the cycle interval parameter with a 5% probability (for example, mutating 30 minutes to 25 minutes or 35 minutes) to ensure population diversity. After 20 iterations, the optimized individual is selected, achieving a 70% reduction in computational complexity and less than a 5% loss in policy effectiveness (for example, the simplified parameter combination of [0.9C, 80%, fixed 1-hour cycle]). The parameter replacement module writes the optimized charge rate, discharge depth, and number of cycles into the original policy instruction set, generating a simplified policy that can be executed on low-computing devices. Instruction compression technology is also used to convert the control curve from a continuous function to a piecewise linear approximation, further reducing computational requirements. This process adapts complex policies to the real-time processing capabilities of embedded devices while preserving the core logic of the policy.

[0046] Through dynamic weight adjustment and quantitative regression analysis, a dynamic balance is established between battery life and charge and discharge efficiency. A weight allocation mechanism based on real-time operating conditions automatically switches the optimization focus—strengthening life protection when health status fluctuates, and prioritizing efficiency when load changes dramatically, achieving multi-objective collaborative optimization. The data-driven regression model accurately quantifies the impact of strategy parameters on performance, combines dual threshold screening to eliminate inefficient strategies, and selects the optimal solution that balances efficiency and life through comprehensive indicator sorting. This fundamentally breaks through the technical bottleneck of traditional methods that make it difficult to coordinate the two, achieving simultaneous improvements in performance and durability.

[0047] In step S18, a control instruction is generated according to the simplified charge and discharge strategy, and the control instruction is executed in real time by the battery management system.

[0048] Specifically, the control command generation and execution process is as follows: The strategy conversion module receives the simplified charge and discharge strategy parameters output in step S17 (e.g., a charge rate of 0.9C, a depth of discharge limit of 80%, and a cycle interval of 60 minutes) and inputs them into the charging curve generator. The charging current curve generator accesses a preset battery type parameter library, matches the corresponding rated capacity (e.g., the rated capacity parameters of a ternary lithium battery) based on the battery type identification code, and converts the charging rate parameters into specific current values ​​(e.g., a 0.9C charge rate for a rated capacity of 2Ah generates a 1.8A constant current charge command). The voltage curve generator uses a piecewise linear approximation algorithm to discretize the complex charging curve into a multi-segment linear control sequence. The algorithm identifies electrochemical characteristics based on battery type and determines segment boundaries by identifying polarization voltage inflection points. In the specific implementation, the charging process is divided into two phases: constant current and constant voltage. The constant current period is divided equally into a steady-state current control segment by the voltage domain, while the constant voltage period is constructed as a current linear decay mode. This method achieves stepped charging voltage regulation through a pre-set sequence of control points, maintaining precise control at the inflection points while simplifying real-time calculations to table lookups, significantly improving control efficiency for low-computing-power devices. This piecewise linearization process effectively balances control accuracy and system response speed while ensuring charging safety.

[0049] The discharge control module analyzes the depth of discharge parameter (e.g., 80% depth of discharge) and, combined with the battery's current remaining capacity (e.g., 78%), calculates the allowable discharge capacity threshold (e.g., total capacity × 80% × 78%). This is then converted into a discharge current control curve (e.g., 2C discharge for the first 50% of capacity, followed by 1C discharge for the next 30%). The command encoder encapsulates the charge current curve, voltage segment parameters, and discharge rate limit into a binary control command packet (containing a command type code, target value, and effective timestamp), which is transmitted via the CAN bus to the battery management system's power control unit. During the real-time execution phase, the current sampling module collects the actual output current every 10 milliseconds and calculates the difference between the actual output current and the command target value (e.g., an actual current of 1.75A when the command requires 1.8A). The PID controller then outputs a PWM duty cycle correction (e.g., a 5% duty cycle increase) to drive the MOSFET switches, ensuring that the charging and discharging process adheres to the policy instructions. This process converts the policy parameters into an executable control sequence and establishes a closed-loop feedback mechanism to ensure precise implementation of the policy while ensuring battery safety.

[0050] Reference Figure 2 The second embodiment of the present invention provides an intelligent charge and discharge management system for lithium batteries, comprising: A data acquisition module for acquiring multi-dimensional parameter data including timestamps, the multi-dimensional parameter data including battery chemical composition, battery structure, ambient temperature and load demand; A battery classification module, configured to determine a battery type identifier based on the multi-dimensional parameter data using a pre-established battery type classification model; A battery status analysis module, configured to predict the battery remaining capacity and decay trend using a recurrent neural network model based on the battery type identifier and the multidimensional parameter data, and obtain a battery remaining capacity and health status score; a temperature-load signature module, configured to generate a temperature-load signature using a principal component analysis method based on the ambient temperature and the load demand when the remaining battery capacity is lower than a preset capacity threshold or the health status score is lower than a preset health threshold; a preliminary strategy generation module, which searches a pre-built index database through a hash table for a candidate charge and discharge strategy set that matches the battery type identifier, the battery remaining capacity, the health status score, and the temperature-load characteristics to obtain a preliminary strategy set; an optimal strategy generation module, configured to assign weights to the health status score and the load demand, sort the efficiency and lifespan impact of the preliminary strategy set using a linear regression model, and determine an optimal charge and discharge strategy; a strategy simplification module, configured to map and compare the optimal charge and discharge strategy with the benchmark performance parameters of the embedded device, and when the optimal charge and discharge strategy exceeds the benchmark performance parameters of the embedded device, simplify the strategy parameters using a genetic algorithm and reduce the computational complexity through iterative optimization to obtain a simplified charge and discharge strategy; The strategy execution module is used to generate control instructions according to the simplified charge and discharge strategy, and execute the control instructions in real time through the battery management system.

[0051] It should be noted that the intelligent charge and discharge management device for a lithium battery provided in an embodiment of the present invention is used to execute all the process steps of the intelligent charge and discharge management method for a lithium battery in the above embodiment. The working principles and beneficial effects of the two correspond one to one, so they will not be repeated here.

[0052] An embodiment of the present invention further provides an electronic device. The electronic device includes: a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a lithium battery intelligent charge and discharge management program. When the processor executes the computer program, the steps in the above-mentioned embodiments of the lithium battery intelligent charge and discharge management method are implemented, such as Figure 1 Alternatively, when the processor executes the computer program, the functions of the modules / units in the above-mentioned device embodiments are realized, such as the intelligent charge and discharge management module of the lithium battery.

[0053] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to implement the present invention. The one or more modules / units may be a series of computer program instruction segments capable of implementing specific functions, and the instruction segments are used to describe the execution process of the computer program in the electronic device.

[0054] The electronic device may be a computing device such as a desktop computer, notebook, PDA, or smart tablet. The electronic device may include, but is not limited to, a processor and memory. Those skilled in the art will appreciate that the aforementioned components are merely examples of electronic devices and do not constitute a limitation of the electronic device. The electronic device may include more or fewer components than those described above, or a combination of certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, and the like.

[0055] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor. The processor is the control center of the electronic device and connects various parts of the entire electronic device using various interfaces and lines.

[0056] The memory can be used to store the computer programs and / or modules. The processor implements the various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory and accessing the data stored in the memory. The memory may primarily include a program storage area and a data storage area. The program storage area may store an operating system and at least one application required for a function (such as a sound playback function or an image playback function); the data storage area may store data generated based on the use of the mobile phone (such as audio data, a phone book, etc.). Furthermore, the memory may include high-speed random access memory and non-volatile memory, such as a hard disk, internal memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state storage device.

[0057] If the module / unit integrated into the electronic device is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention can implement all or part of the process steps in the above-mentioned method embodiments by using a computer program to instruct the relevant hardware. The computer program can be stored in a computer-readable storage medium. When executed by a processor, the computer program can implement the steps of each of the above-mentioned method embodiments. The computer program includes computer program code, which can be in source code form, object code form, executable file, or some intermediate form. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal, and software distribution medium. It should be noted that the content of the computer-readable medium can be appropriately increased or decreased based on the requirements of legislation and patent practice in a jurisdiction. For example, in some jurisdictions, based on legislation and patent practice, computer-readable media does not include electric carrier signals and telecommunication signals.

[0058] It should be noted that the device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed across multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment. In addition, in the drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art can understand and implement the present invention without inventive effort.

[0059] The specific embodiments described above further illustrate the objectives, technical solutions, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for intelligent charge and discharge management of lithium batteries, characterized in that: include: Acquiring multi-dimensional parameter data including a timestamp, the multi-dimensional parameter data including battery chemical composition, battery structure, ambient temperature, and load demand; Determining a battery type identifier using a pre-established battery type classification model based on the multi-dimensional parameter data; According to the battery type identifier and the multidimensional parameter data, a recurrent neural network model is used to predict the battery remaining capacity and attenuation trend to obtain a battery remaining capacity and health status score; When the remaining capacity of the battery is lower than a preset capacity threshold or the health status score is lower than a preset health threshold, a temperature-load feature is generated using a principal component analysis method according to the ambient temperature and the load demand; In a pre-built index database, querying a candidate charge and discharge strategy set that matches the battery type identifier, the battery remaining capacity, the health status score, and the temperature-load characteristics through a hash table to obtain a preliminary strategy set; weighting the health status score and the load demand, ranking the efficiency and lifespan impact of the preliminary strategy set using a linear regression model, and determining an optimal charge and discharge strategy; Mapping and comparing the optimal charge and discharge strategy with the benchmark performance parameters of the embedded device, and when the optimal charge and discharge strategy exceeds the benchmark performance parameters of the embedded device, simplifying the strategy parameters using a genetic algorithm and reducing the computational complexity through iterative optimization to obtain a simplified charge and discharge strategy; According to the simplified charge and discharge strategy, control instructions are generated and executed in real time by the battery management system.

2. The intelligent charge and discharge management method for lithium batteries according to claim 1, characterized in that: The process of constructing the battery type classification model includes: Acquire historical battery data, remove outliers from the historical battery data using a mean filtering method, and normalize the data using a Z-score normalization method to generate a standardized data set; Performing dimensionality reduction processing on the standardized data set using a principal component analysis algorithm and extracting a battery feature vector, wherein the battery feature vector includes a chemical composition vector and a structural feature vector; The battery feature vector and the preset battery type label are used as training data, and an initial battery type classification model is constructed using a support vector machine algorithm. A radial basis function is selected as the kernel function type, and an initial penalty coefficient and a kernel function parameter range are defined, wherein the battery feature vector serves as the input of the initial battery type classification model, and the battery type label serves as the output of the initial battery type classification model; The training data is divided into training subsets and validation subsets by the multi-fold cross-validation method, and the grid search algorithm is used to traverse the preset candidate penalty coefficient and kernel parameter combinations to obtain the parameter combination; In each parameter combination training stage, a kernel function matrix is ​​calculated based on the training subset, Lagrange multipliers are solved by a sequential minimum optimization algorithm, and a classification hyperplane is constructed; Inputting the verification subset into the initial battery type classification model under the parameter combination, and combining the classification hyperplane, calculating the matching ratio between the predicted category and the true label as the verification accuracy, thereby obtaining the verification accuracy; Selecting the parameter combination with the highest verification accuracy as the optimal model parameters, and performing parameter optimization on the initial battery type classification model; After completing the parameter optimization, the model is retrained using the training data in combination with the optimal model parameters to generate an optimal battery type classification model.

3. The intelligent charge and discharge management method of a lithium battery according to claim 1, characterized in that: The method of using a recurrent neural network model to predict the remaining capacity and decay trend of the battery based on the battery type identifier and the multi-dimensional parameter data to obtain the remaining capacity and health status score of the battery includes: Based on the battery type identifier and the multidimensional parameter set, after cleaning outliers using the median method, a standardized parameter set is generated through Min-Max normalization; Extracting dynamic features of the standardized parameter set using a recurrent neural network to obtain dynamic features of capacity degradation; Based on the dynamic characteristics of capacity degradation, a recurrent neural network is used to perform time series prediction to obtain the remaining capacity and attenuation trend of the battery; The battery remaining capacity, the decay trend, and the preset capacity weight and decay weight are weighted by using a linear weighting method to obtain a health status score.

4. The intelligent charge and discharge management method for lithium batteries according to claim 1, characterized in that: The training process of the recurrent neural network model includes: Acquire historical multidimensional parameter data and corresponding historical battery type identifiers, and perform data cleaning and feature extraction on the historical multidimensional parameter data and the historical battery type identifiers to obtain historical capacity degradation dynamic features; Inputting the historical capacity degradation dynamic characteristics into the input layer of the initially constructed recurrent neural network model for training, and obtaining the predicted battery remaining capacity and attenuation trend output by the output layer of the recurrent neural network model; Substituting the predicted battery remaining capacity and decay trend and the pre-stored actual battery remaining capacity and decay trend empirical values ​​into a loss function to calculate a loss value; Calculate the gradient of the output layer output of the recurrent neural network model according to the loss value, and pass the gradient forward layer by layer through the chain rule to calculate the gradient of the parameters of each layer to obtain gradient data; Update the parameters of each layer of the neural network model according to the gradient data and the preset learning rate; The parameters of each layer are updated repeatedly until the training times of the recurrent neural network model are greater than the preset times, or the loss value data of the recurrent neural network model is less than the preset loss threshold, the training is determined to be completed, and a trained recurrent neural network model is obtained.

5. The intelligent charge and discharge management method for lithium batteries according to claim 1, characterized in that: The generating of temperature-load characteristics by principal component analysis according to the ambient temperature and the load requirement includes: Using a linear interpolation algorithm to match the timestamps of the ambient temperature and the load demand to generate temperature-load joint data; According to the temperature-load joint data, a principal component analysis algorithm is used to extract dynamic interaction features of the temperature data and the load curve data to obtain temperature-load features.

6. The intelligent charge and discharge management method for lithium batteries according to claim 1, characterized in that: The weighting of the health status score and the load demand, sorting the efficiency and lifespan impact of the preliminary strategy set by a linear regression model, and determining the optimal charge and discharge strategy includes: weighting the health status score and the load demand using an entropy weight method to obtain weight data; Calculating the weighted scores of the preliminary strategy set using a linear weighted summation method based on the weight data to obtain weighted strategy data; Based on the weighted strategy data, a multiple linear regression model is used to analyze the regression coefficients of the efficiency impact and life impact of the preliminary strategy set, and the predicted efficiency and predicted life are calculated; When the predicted efficiency is higher than a preset efficiency threshold and the predicted life is higher than a preset life threshold, the preliminary strategy is included in a ranking list, and a ranking result is obtained by arranging the strategies in the ranking list in descending order according to the regression coefficient; According to the ranking result, the highest-ranked preliminary strategy is extracted from the ranking list and determined as the optimal charging and discharging strategy.

7. The intelligent charge and discharge management method for lithium batteries according to claim 1, characterized in that: The optimal charge-discharge strategy is mapped and compared with the benchmark performance parameters of the embedded device. When the optimal charge-discharge strategy exceeds the benchmark performance parameters of the embedded device, a genetic algorithm is used to simplify the strategy parameters, and the computational complexity is reduced through iterative optimization to obtain a simplified charge-discharge strategy, including: Decomposing the optimal charge and discharge strategy into quantifiable basic operation units and mapping and comparing them with the benchmark performance parameters of the embedded device; when the optimal charge and discharge strategy exceeds the benchmark performance parameters of the embedded device, obtaining strategy parameters from the optimal charge and discharge strategy, wherein the strategy parameters include charge rate, discharge depth, and number of cycles; Performing initial population encoding on the strategy parameters using a genetic algorithm to generate a first parameter aggregation; According to the first parameter aggregation, crossover and mutation operations of a genetic algorithm are used to iteratively optimize the first parameter aggregation to generate a second parameter set; The second parameter set is used as a simplified parameter and replaces the strategy parameters in the optimal charge and discharge strategy to generate a simplified charge and discharge strategy.

8. An intelligent charge and discharge management system for lithium batteries, characterized in that: include: A data acquisition module for acquiring multi-dimensional parameter data including timestamps, the multi-dimensional parameter data including battery chemical composition, battery structure, ambient temperature and load demand; A battery classification module, configured to determine a battery type identifier based on the multi-dimensional parameter data using a pre-established battery type classification model; A battery status analysis module, configured to predict the battery remaining capacity and decay trend using a recurrent neural network model based on the battery type identifier and the multidimensional parameter data, and obtain a battery remaining capacity and health status score; a temperature-load signature module, configured to generate a temperature-load signature using a principal component analysis method based on the ambient temperature and the load demand when the remaining battery capacity is lower than a preset capacity threshold or the health status score is lower than a preset health threshold; a preliminary strategy generation module, which searches a pre-built index database through a hash table for a candidate charge and discharge strategy set that matches the battery type identifier, the battery remaining capacity, the health status score, and the temperature-load characteristics to obtain a preliminary strategy set; an optimal strategy generation module, configured to assign weights to the health status score and the load demand, sort the efficiency and lifespan impact of the preliminary strategy set using a linear regression model, and determine an optimal charge and discharge strategy; a strategy simplification module, configured to map and compare the optimal charge and discharge strategy with the benchmark performance parameters of the embedded device, and when the optimal charge and discharge strategy exceeds the benchmark performance parameters of the embedded device, simplify the strategy parameters using a genetic algorithm and reduce the computational complexity through iterative optimization to obtain a simplified charge and discharge strategy; The strategy execution module is used to generate control instructions according to the simplified charge and discharge strategy, and execute the control instructions in real time through the battery management system.

9. An electronic device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the method for intelligent charge and discharge management of a lithium battery according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the intelligent charge and discharge management method for a lithium battery according to any one of claims 1 to 7.

Images