Battery health state assessment method, device and equipment based on bidirectional long short-term memory network model, medium and product

Abstract

The invention relates to a battery health state assessment method, device and equipment based on a bidirectional long short-term memory network model, a medium and a product. The method comprises the steps of obtaining operation data of a lithium ion battery in a charge-discharge cycle process, obtaining performance characteristic data of the lithium ion battery according to the operation data, determining target characteristic data from the performance characteristic data, obtaining training set data according to the target characteristic data, and constructing a battery health state evaluation model, in the process of training the battery health state evaluation model through the training set, performing optimization configuration on hyper-parameters of the battery health state evaluation model through an optimization algorithm; according to the optimization algorithm, hyper-parameters serve as search variables, a target hyper-parameter combination is obtained through iterative search so as to obtain an optimized battery health state evaluation model, and battery health state evaluation is carried out through the optimized battery health state evaluation model. By adopting the method, the battery health state evaluation accuracy can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of battery management, in particular to a battery state of health evaluation method and device based on a bidirectional long short-term memory network model, equipment, medium and product. BACKGROUND

[0002] With the gradual development of clean energy driving, power battery technology is gradually becoming an important direction of current and future scientific and technological research and development. Among them, the electric vehicle technology centered on lithium ion batteries is showing an accelerating evolution trend. The driving system of electric vehicles mainly consists of battery packs, motors and electric control systems, etc. These new components are gradually replacing traditional internal combustion engines and gearbox systems.

[0003] At present, the estimation method for the state of health of lithium ion batteries mainly adopts the method based on mechanism modeling, including empirical / semi-empirical model, electrochemical model and equivalent circuit model. The empirical model establishes a capacity attenuation model by analyzing the statistical law of battery performance degradation. However, since different battery models need to establish corresponding mathematical models, the generalization ability is limited, and the accuracy of model construction will affect the accuracy of the final prediction result. SUMMARY

[0004] Therefore, it is necessary to provide a battery state of health evaluation method, device, equipment, medium and product based on a bidirectional long short-term memory network model, which can improve the accuracy of battery state of health evaluation.

[0005] In a first aspect, the present application provides a battery state of health evaluation method based on a bidirectional long short-term memory network model, comprising:

[0006] Obtaining the performance characteristic data of the lithium ion battery according to the operation data of the lithium ion battery in the charging and discharging cycle process;

[0007] Determining the target feature data from the performance characteristic data, and obtaining the training set data according to the target feature data;

[0008] Constructing a battery state of health evaluation model; the battery state of health evaluation model comprises an input layer, a bidirectional long short-term memory network layer, an intermediate layer and an output layer connected in turn;

[0009] In the process of training the battery state of health evaluation model by the training set, the hyperparameters of the battery state of health evaluation model are optimized and configured by an optimization algorithm; the optimization algorithm takes the hyperparameters as search variables, obtains the target hyperparameter combination through iterative search, so as to obtain the optimized battery state of health evaluation model;

[0010] The battery health state is evaluated through the optimized battery health state evaluation model.

[0011] In one of the embodiments, the performance characteristic data includes the constant-voltage rising charging duration, the constant-voltage maintaining duration, the constant-current falling charging duration, the target discharge temperature occurrence duration, and the constant-voltage falling discharging duration; the step of determining the target characteristic data from the performance characteristic data includes:

[0012] The first change trend feature of the performance characteristic data is acquired, and the second change trend feature of the battery health state is acquired;

[0013] The performance characteristic data is filtered according to the similarity between the first change trend feature and the second change trend feature, to obtain intermediate characteristic data;

[0014] The intermediate characteristic data is integrated into the target characteristic data through the principal component analysis algorithm.

[0015] In one of the embodiments, the step of acquiring the second change trend feature of the battery health state includes:

[0016] The full life cycle data of the lithium ion battery is acquired, and the performance degradation index is acquired according to the full life cycle data;

[0017] The battery health state data is acquired based on the performance degradation index, and the second change trend feature is obtained by fitting the battery health state data; the performance degradation index includes at least one of the battery capacity, the battery internal resistance, the battery power, and the battery remaining cycle number; the battery health state data includes a plurality of discrete data points.

[0018] In one of the embodiments, the optimization algorithm is a sparrow search algorithm; the sparrow search algorithm simulates the foraging and anti-predation behavior mechanism of the finder, the follower, and the alarm in the sparrow population, takes the prediction error of the battery health state evaluation model on the training set as the fitness value, iteratively searches and updates the value of the hyperparameter until the training stopping condition is met, and stops to obtain the target hyperparameter combination.

[0019] In one of the embodiments, the bidirectional long short-term memory network layer includes a forward hidden layer and a backward hidden layer; the forward hidden layer is used to extract features in the positive order of the time sequence of the input data; the backward hidden layer is used to extract features in the reverse order of the time sequence of the input data; the hyperparameters include the number of neurons corresponding to the forward hidden layer and the backward hidden layer.

[0020] In one of the embodiments, the method further includes:

[0021] The test set data is acquired according to the target characteristic data; the test set data includes the target characteristic data collected after the collection time of the training set data;

[0022] The test set data is processed by the optimized battery health state evaluation model to obtain a battery health state evaluation value;

[0023] The optimized battery health state evaluation model is verified according to an error between the battery health state evaluation value and a corresponding battery health state historical value.

[0024] In a second aspect, the application further provides a battery health state evaluation device based on a bidirectional long short-term memory network model, comprising:

[0025] A data acquisition module is configured to acquire running data of a lithium ion battery in a charge-discharge cycle process, and acquire performance characteristic data of the lithium ion battery according to the running data.

[0026] A feature screening module is configured to determine target feature data from the performance characteristic data, and acquire training set data according to the target feature data.

[0027] A model construction module is configured to construct a battery health state evaluation model; the battery health state evaluation model comprises an input layer, a bidirectional long short-term memory network layer, an intermediate layer, and an output layer connected in sequence.

[0028] A model training module is configured to optimize and configure hyperparameters of the battery health state evaluation model by an optimization algorithm in a process of training the battery health state evaluation model by the training set; the optimization algorithm takes the hyperparameters as search variables, obtains a target hyperparameter combination by iterative search, and obtains an optimized battery health state evaluation model.

[0029] A state evaluation module is configured to perform battery health state evaluation by the optimized battery health state evaluation model.

[0030] In a third aspect, the application further provides a computer device comprising a memory and a processor; the memory stores a computer program; and the processor implements the method steps of any one of the first aspect when executing the computer program.

[0031] In a fourth aspect, the application further provides a computer readable storage medium having a computer program stored thereon; the computer program is executed by a processor to implement the method steps of any one of the first aspect.

[0032] In a fifth aspect, the application further provides a computer program product comprising a computer program; the computer program is executed by a processor to implement the method steps of any one of the first aspect.

[0033] The battery health state evaluation method, device, equipment, medium and product based on the bidirectional long short-term memory network model, by acquiring the operation data of the lithium ion battery in the charging and discharging cycle process, acquiring the performance characteristic data of the lithium ion battery according to the operation data, determining the target characteristic data from the performance characteristic data, and acquiring the training set data according to the target characteristic data, the battery health state evaluation model is constructed, in the process of training the battery health state evaluation model by the training set, the hyperparameters of the battery health state evaluation model are optimized and configured by the optimization algorithm, and the battery health state evaluation is performed by the optimized battery health state evaluation model, which can improve the model fitting ability, improve the prediction accuracy, and improve the battery health state evaluation accuracy. BRIEF DESCRIPTION OF DRAWINGS

[0034] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related art, the drawings needed to be used in the description of the embodiments of the present application or the related art will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other related drawings can be obtained by those skilled in the art without creative labor.

[0035] Figure 1 An application environment diagram of the battery health state evaluation method based on the bidirectional long short-term memory network model in an embodiment;

[0036] Figure 2 A flowchart of the battery health state evaluation method based on the bidirectional long short-term memory network model in an embodiment;

[0037] Figure 3 A comparison diagram of the model prediction value and the true value of the battery health state in an embodiment;

[0038] Figure 4 A comparison diagram of the model prediction value and the true value of the battery health state in an embodiment;

[0039] Figure 5 A comparison diagram of the model prediction value and the true value of the battery health state in an embodiment;

[0040] Figure 6 A flowchart of the battery health state evaluation method based on the bidirectional long short-term memory network model in another embodiment;

[0041] Figure 7 A structural block diagram of the battery health state evaluation device based on the bidirectional long short-term memory network model in an embodiment;

[0042] Figure 8 An internal structure diagram of a computer device in an embodiment. DETAILED DESCRIPTION

[0043] In order to make the purposes, technical solutions and advantages of the present application clearer, further detailed description will be made to the present application in combination with the drawings and examples. It should be understood that the specific examples described herein are only used to explain the present application, and are not used to limit the present application.

[0044] The battery health state evaluation method based on the bidirectional long short-term memory network model provided by the embodiments of the present application can be applied to the application environment as shown in Figure 1 . Wherein, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data required to be processed by the server 104. The data storage system can be integrated on the server 104, or placed on the cloud or other network servers. Wherein, the terminal 102 is used to obtain the running data of the lithium ion battery in the charging and discharging cycle process, obtain the performance characteristic data of the lithium ion battery according to the running data, determine the target characteristic data from the performance characteristic data, and obtain the training set data according to the target characteristic data, construct the battery health state evaluation model, and optimize the hyperparameters of the battery health state evaluation model through the optimization algorithm in the process of training the battery health state evaluation model through the training set. Wherein, the terminal 102 can be, but is not limited to, various personal computers, notebook computers, smart phones, tablet computers, unmanned aerial vehicles, low-altitude aircraft, Internet of Things devices and portable wearable devices. The Internet of Things device can be a smart speaker, a smart TV, a smart air conditioner, a smart vehicle device, a projection device, etc. The portable wearable device can be a smart watch, a smart bracelet, a head-mounted device, etc. The head-mounted device can be a virtual reality (VR) device, an augmented reality (AR) device, smart glasses, etc. The server 104 can be a stand-alone physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services.

[0045] In an exemplary embodiment, as shown in Figure 2 , a battery health state evaluation method based on a bidirectional long short-term memory network model is provided. Taking the terminal 102 in Figure 1 as an example to illustrate the method, the method includes the following steps 202 to 210. Wherein:

[0046] S202: Obtain the running data of the lithium ion battery in the charging and discharging cycle process, and obtain the performance characteristic data of the lithium ion battery according to the running data.

[0047] Optionally, lithium-ion batteries generate various operating data, such as voltage, current, temperature, time, etc., during charging and discharging, which reflect the real-time state and aging trend of the battery. By extracting performance characteristics from these raw operating data that can reflect the state of health (SOH) of the battery, effective inputs can be provided for subsequent modeling. Among them, the extracted features include: constant voltage rise charging time, constant voltage stage duration, constant current drop charging time, maximum temperature in discharging stage, and constant voltage drop discharging time, etc. These features are closely related to the capacity decay, internal resistance change and other aging phenomena of the battery, and are the basis for building SOH estimation model.

[0048] S204: Determine target feature data from performance feature data, and obtain training set data according to the target feature data.

[0049] Optionally, since not all extracted features are equally important for SOH estimation, in order to improve the efficiency and accuracy of the model, the features with the highest correlation to SOH need to be selected as target features. Grey correlation analysis method is used to evaluate the correlation of each feature with SOH, and the first two principal components are extracted as input variables of the model through principal component analysis method. Then, the data is divided into training set (first 70%) and test set (last 30%) in time sequence, and normalized to unify the dimension and improve the training efficiency of the model.

[0050] S206: Construct a battery state of health evaluation model; the battery state of health evaluation model includes an input layer, a bidirectional long short-term memory network layer, an intermediate layer, and an output layer connected in turn.

[0051] Optionally, the battery state of health evaluation model uses bidirectional long short-term memory network (BiLSTM) as the core structure. BiLSTM can simultaneously learn the forward and reverse dependencies of time series data, fully capture the time sequence features in the battery aging process, and make up for the shortcomings of traditional LSTM which can only learn in one direction. Among them, the model structure includes: an input layer for receiving reduced feature data; a BiLSTM layer for extracting bidirectional time sequence features; an intermediate layer including a fully connected layer and a Dropout layer to prevent overfitting; and an output layer for outputting SOH prediction value.

[0052] For example, the constructed BiLSTM model uses bidirectional long short-term memory network (BiLSTM) as the main structure. The input data first enters the model through the input layer, then enters the BiLSTM layer to extract the bidirectional dependency features in the time sequence, and finally passes through the fully connected layer and the output layer to generate the prediction result of the lithium-ion battery state of health (SOH). The main parameter configuration of the model is shown in Table 1.

[0053] Table 1 Network structure parameter table

[0054]

[0055] S208: In the process of training the battery health state evaluation model through the training set, the hyperparameters of the battery health state evaluation model are optimized and configured through an optimization algorithm; the optimization algorithm takes the hyperparameters as search variables, obtains a target hyperparameter combination through iterative search, and obtains an optimized battery health state evaluation model.

[0056] Optionally, there are multiple hyperparameters (such as the number of neurons, the learning rate, etc.) in the BiLSTM model, and the settings directly affect the performance of the model. The traditional method relies on manual parameter tuning, which is low in efficiency and difficult to find the optimal combination. In order to improve the efficiency of model training, the Sparrow Search Algorithm is used as the optimization algorithm, which simulates the foraging behavior of sparrows to search iteratively in the hyperparameter space to find the hyperparameter combination that minimizes the prediction error of the model. Among them, the Sparrow Search Algorithm balances the search between the global and local through the synergy mechanism of the discoverer-follower-scout, avoids falling into local optimum, and thus realizes efficient and adaptive hyperparameter optimization.

[0057] S210: Perform battery health state evaluation through the optimized battery health state evaluation model.

[0058] In the above battery health state evaluation method based on the BiLSTM model, by obtaining the running data of the lithium ion battery in the charging and discharging cycle process, the performance characteristic data of the lithium ion battery is obtained according to the running data, the target characteristic data is determined from the performance characteristic data, and the training set data is obtained according to the target characteristic data. The battery health state evaluation model is constructed, in the process of training the battery health state evaluation model through the training set, the hyperparameters of the battery health state evaluation model are optimized and configured through an optimization algorithm, and the battery health state evaluation is performed through the optimized battery health state evaluation model, which can improve the model fitting ability, improve the prediction accuracy, and thus improve the battery health state evaluation accuracy.

[0059] In one exemplary embodiment, the performance characteristic data includes the isobaric rising charging duration, the constant voltage duration, the isoflow falling charging duration, the target discharging temperature occurrence duration, and the isobaric falling discharging duration; the step of determining the target characteristic data from the performance characteristic data includes: obtaining a first change trend feature of the performance characteristic data, obtaining a second change trend feature of the battery health state; filtering the performance characteristic data according to the similarity between the first change trend feature and the second change trend feature to obtain intermediate characteristic data; and integrating the intermediate characteristic data into the target characteristic data through a principal component analysis algorithm.

[0060] Optionally, the first change trend feature is a change sequence of the extracted performance feature data with the cycle number of the battery. The second change trend feature refers to a change sequence of the true value of the state of health (SOH) of the battery with the cycle number of the battery. The core of the grey correlation analysis is to judge the representation ability of the feature to the SOH by calculating the similarity (i.e., correlation degree) of the geometric shapes of each feature sequence and the SOH sequence. The higher the correlation degree, the more synchronized and relevant the change trend of the feature to the SOH decay trend. By setting a threshold, the features with high correlation degrees are screened out to form intermediate feature data. This process eliminates the features irrelevant or weakly relevant to the SOH change, and realizes targeted feature preliminary selection.

[0061] Further, on the basis of the screened intermediate feature data, data compression and redundancy elimination are performed. The principal component analysis algorithm combines the intermediate features that may be correlated (such as the time lengths in different charging stages that may be correlated) into several new variables (i.e., principal components) that are independent of each other through linear transformation. These principal components can retain the variance information in the original data set that is most important for SOH prediction to the greatest extent. Finally, the first two principal components are extracted as the input variables of the model. The high-dimensional intermediate feature data is integrated into low-dimensional and irrelevant target feature data, which prepares for the subsequent model input.

[0062] In the embodiment, by acquiring the first change trend feature of the performance feature data, acquiring the second change trend feature of the state of health of the battery, screening the performance feature data according to the similarity between the first change trend feature and the second change trend feature to obtain intermediate feature data, and integrating the intermediate feature data into target feature data through the principal component analysis algorithm, the model prediction accuracy can be improved, the model training efficiency can be enhanced, and therefore the high-precision prediction of the state of health of the battery is realized.

[0063] In one exemplary embodiment, the step of acquiring the second change trend feature of the state of health of the battery includes: acquiring full life cycle data of the lithium ion battery, acquiring a performance degradation index according to the full life cycle data; acquiring battery state of health data based on the performance degradation index, fitting the battery state of health data to obtain the second change trend feature; the performance degradation index includes at least one of battery capacity, battery internal resistance, battery power and battery remaining cycle number; the battery state of health data includes a plurality of discrete data points.

[0064] Optionally, since the SOH is an indirect index that cannot be directly measured, it must be defined and calculated through observable battery performance parameters. It is usually defined based on four measurable or calculable physical quantities:

[0065] (1) Define SOH based on battery capacity:

[0066]

[0067] wherein C now represents the effective capacity value of the battery at present stage, while C0 corresponds to the design rated capacity value. In engineering practice, the capacity parameter is the most common method to evaluate SOH.

[0068] (2) Defining SOH based on the perspective of battery internal resistance:

[0069]

[0070] wherein R EOL represents the internal resistance value when the battery reaches the end-of-life condition, which is usually set as 200% of the initial internal resistance; R C reflects the real-time internal resistance state of the battery, while R NEW represents the reference internal resistance of the battery in brand-new state.

[0071] (3) Defining SOH based on the perspective of battery power:

[0072]

[0073] wherein Q now-max represents the real-time maximum discharge capacity of the measured battery, Q new-max corresponds to the factory-calibrated maximum discharge capacity.

[0074] (4) Defining SOH based on the perspective of battery remaining cycle times:

[0075]

[0076] wherein Cnt remain is used to represent the current available cycle life of the battery, while Cnt total represents the upper limit of its design cycle life.

[0077] Optionally, the lithium-ion battery is tested for full-life cycle charging and discharging cycles in a laboratory or under specific conditions, and voltage, current, etc. are monitored throughout the process, so as to calculate the value of one or more performance degradation indicators corresponding to each cycle point. It should be noted that in actual application, the SOH can be calculated by selecting the corresponding performance degradation indicator value according to the actual parameter acquisition situation or other needs. Since the calculated indicators are a series of discrete SOH data points (each cycle period corresponds to a value) that may have noise. In order to effectively compare the trend similarity with the first change trend feature, it is necessary to convert the discrete SOH points into a continuous and smooth change curve. Curve fitting algorithms (such as polynomial fitting, spline interpolation, moving average, etc.) are used to fit these discrete SOH data points. The fitted curve is the second change trend feature, which reflects the continuous and macroscopic decay law of SOH in the entire battery life cycle, and filters out part of the measurement noise and short-term fluctuations.

[0078] In this embodiment, by obtaining the full-life cycle data of the lithium-ion battery, obtaining the performance degradation indicators according to the full-life cycle data, obtaining the battery health state data based on the performance degradation indicators, fitting the battery health state data, and obtaining the second change trend feature, the accuracy of feature screening can be improved, thereby improving the accuracy of battery health state evaluation.

[0079] In one exemplary embodiment, the optimization algorithm is a sparrow search algorithm; the sparrow search algorithm simulates the foraging and anti-predation behavior mechanism of the discoverer, follower and alarm of the sparrow population, takes the prediction error of the battery health state evaluation model on the training set as the fitness value, iteratively searches and updates the value of the hyperparameter, and stops until the training stopping condition is met, to obtain the target hyperparameter combination.

[0080] Optionally, the hyperparameters to be optimized in the BiLSTM model, such as the number of neurons, the learning rate, the number of iterations, etc., are encoded as the position coordinates of the sparrows in the search space. The prediction error (such as mean square error, MSE) of the SSA-BiLSTM model on the training set is taken as the fitness value, and the smaller the error is, the higher the fitness of the group of hyperparameters (i.e. the position of the sparrow) is, that is, the better the food source is. The iterative process of the algorithm simulates the three types of role division of the sparrow population: the discoverer (explorer) is an individual with high fitness, responsible for finding high-quality food sources (i.e. potential hyperparameter regions) in a wide area, and when the environment is safe, extensive search is carried out; when danger is perceived, the current area is quickly evacuated, guiding the population to escape from local optimization. The follower (utilizer) is used to monitor and follow the discoverer, focusing on promising areas for in-depth development, trying to preempt better positions found by the discoverer, and if the competition fails (low fitness), flying to other areas maintains the diversity of the population. The guard (scout) is a randomly selected part of the individuals (such as 15%), responsible for warning of global danger, when the fitness of an individual is poor (indicating that it is in a marginal dangerous position or the population is trapped in stagnation), the alarm mechanism will drive it to move closer to the global optimal individual or carry out random disturbance, which can effectively help the population jump out of the local optimal trap.

[0081] Optionally, the iteration stopping condition is usually set to reach the maximum number of iterations, or the improvement of fitness tends to be stable. When the iteration stops, the position represented by the sparrow individual with the optimal fitness is the target hyperparameter combination searched, and the BiLSTM model is configured with this combination, that is, the optimized SSA-BiLSTM model is obtained.

[0082] In this embodiment, the sparrow search algorithm is used for hyperparameter optimization, which can improve the model performance and improve the algorithm efficiency, thereby realizing high-precision battery state of health evaluation.

[0083] In one exemplary embodiment, the bidirectional long short-term memory network layer includes a forward hidden layer and a backward hidden layer; the forward hidden layer is used to extract features in the forward order of the time sequence of the input data; the backward hidden layer is used to extract features in the reverse order of the time sequence of the input data; and the hyperparameters include the number of neurons corresponding to the forward hidden layer and the backward hidden layer.

[0084] Optionally, the bidirectional long short-term memory network layer is composed of a forward LSTM (or GRU) hidden layer and a backward LSTM (or GRU) hidden layer in parallel. The forward hidden layer is used to receive and process the sequence data (such as the battery feature sequence after PCA dimensionality reduction) input in time order (from the past to the present). It learns the forward dependency of the sequence, that is, how the past information affects the present. The backward hidden layer is used to receive and process the same sequence data input in time reverse order (from the future to the present, or from the end of the sequence to the front). It learns the reverse dependency of the sequence, that is, how the future context helps to understand the current state. For each time step in the sequence, the model will splice or sum the hidden state of the forward layer at that time and the hidden state of the backward layer at the corresponding time to form a comprehensive feature representation containing both past and future context information, which overcomes the defect of insufficient data information mining of the unidirectional LSTM.

[0085] wherein the number of neurons corresponding to the forward hidden layer and the backward hidden layer is one of the most important structural hyperparameters of the BiLSTM network. It directly determines the capacity (i.e. the complexity of information that can be learned and stored) of each hidden layer. If the number of neurons is too small, it will lead to insufficient network capacity, which cannot fully learn the complex nonlinear time series pattern in the battery aging process, resulting in underfitting and low prediction accuracy. If the number of neurons is too large, it will lead to excessive network capacity, which not only increases the computational overhead, but also is more likely to remember noise and accidental details in the training data, resulting in overfitting, and the model performs well on the training set but has poor generalization ability on the unknown test set. Therefore, through the sparrow search algorithm, an optimal combination of the number of neurons for the forward layer and the backward layer is automatically searched and determined within a predetermined range (such as 5-100), so that the entire BiLSTM model achieves the best balance between exploration ability (capturing complex patterns) and generalization ability (resisting overfitting), thereby minimizing the prediction error (i.e. fitness value) on the training set.

[0086] In an exemplary embodiment, the method further comprises: obtaining test set data according to the target feature data; the test set data comprising the target feature data collected after the collection time of the training set data; processing the test set data by the optimized battery health state evaluation model to obtain a battery health state evaluation value; and verifying the optimized battery health state evaluation model according to the error between the battery health state evaluation value and the corresponding battery health state historical value.

[0087] Optionally, in order to truly evaluate the prediction ability of the model for the future state, samples completely independent of the training data on the timeline are used for testing. The data is divided in chronological order, with the first 70% as the training set and the last 30% as the test set. The test set data is strictly later than the training set in acquisition time, thereby ensuring the time sequence independence and fairness of the test. Then, the target feature data in the test set is input into the optimized SSA-BiLSTM model which has been trained. The model outputs the battery health state evaluation value for each test sample according to the learned feature-SOH mapping relationship. The predicted SOH value output by the model is compared with the corresponding battery health state history value (i.e. the true SOH value calculated by the full life cycle test) of the test set one by one, and the error between the two is calculated, thereby verifying the evaluation effect of the battery health state evaluation model.

[0088] Exemplarily, five key features are extracted from the B5, B6 and B7 lithium-ion battery data sets, the correlation between each feature and the battery SOH is evaluated using the grey correlation analysis method, and the first two principal components are extracted as input variables of the model through principal component analysis. The input data is scaled to 0 to 1 by using the Min-Max normalization method, the dimension is unified, and the model training efficiency is improved. Then the data is divided into training set and test set in chronological order, with the first 70% as the training set and the last 30% as the test set. The SSA algorithm is used to optimize some hyperparameters of the BiLSTM model to improve the fitting ability and generalization performance of the model. Then the SSA-BiLSTM model is trained based on the training set data to obtain the optimal model for SOH prediction. The trained SSA-BiLSTM model is used to predict the SOH of the test set of B5, B6 and B7 batteries respectively. The model output result is compared with the true SOH value, and the prediction error is calculated to verify the prediction accuracy and generalization ability of the model. The optimization of the model uses the Adam gradient descent algorithm as the optimization function, the activation function is ReLU, and the maximum training iteration number is set to 500. The SOH estimation curves of the three types of lithium-ion batteries and their prediction errors are shown in the following Figures 3-5

[0089] Figures 3-5 ​The comparison curves of the real value and the predicted value of the SSA-BiLSTM model of the SOH of three lithium-ion batteries (B5, B6 and B7) are shown respectively. Among them, curve 1 represents the real value of SOH, and curve 2 represents the model prediction result. As can be seen from the figure, for the B5 battery, the prediction error of the SSA-BiLSTM model increases relatively after the capacity regeneration area and the 155th cycle, especially at the 151st cycle, the error reaches the maximum. In addition, at the capacity regeneration point, the model prediction result shows a certain hysteresis. Except for the above period, the prediction error of the remaining period is small, and the overall error is controlled within the acceptable range. For the B6 battery, the model shows high prediction accuracy in the early cycles, especially at the capacity regeneration point, the prediction result is accurate. However, after the 152nd cycle, the prediction error increases and reaches the maximum at the 164th cycle. In the B7 battery, the SSA-BiLSTM model also has a large prediction deviation in the capacity regeneration area, and the maximum error occurs at the 150th cycle. The prediction error of the remaining stages is relatively small, the overall prediction effect is good, and the error is still within the allowable range.

[0090] In this embodiment, by acquiring test set data according to target feature data, processing the test set data through the optimized battery health state evaluation model to obtain a battery health state evaluation value, and verifying the optimized battery health state evaluation model according to the error between the battery health state evaluation value and the corresponding battery health state historical value, the model precision can be further improved, and the accuracy of battery health state evaluation can be ensured.

[0091] In one exemplary embodiment, as shown in Figure 6 a battery health state evaluation method based on a bidirectional long short-term memory network model is provided, which comprises the following steps:

[0092] (1) Performance feature extraction: acquiring the running data of the lithium-ion battery in the charge and discharge cycle process, and acquiring the performance feature data of the lithium-ion battery according to the running data. The performance feature data includes the constant pressure rise charging time, the constant voltage duration, the constant flow drop charging time, the appearance time of the target discharge temperature, and the constant pressure drop discharging time.

[0093] (2) Training set data division: obtain a first change trend feature of performance characteristic data, obtain full life cycle data of the lithium ion battery, and obtain a performance degradation index according to the full life cycle data; obtain battery health state data based on the performance degradation index, fit the battery health state data, and obtain a second change trend feature; the performance degradation index includes at least one of battery capacity, battery internal resistance, battery power, and battery remaining cycle number; the battery health state data includes a plurality of discrete data points; according to the similarity between the first change trend feature and the second change trend feature, the performance characteristic data is screened to obtain intermediate characteristic data; the intermediate characteristic data is integrated into target characteristic data through a principal component analysis algorithm; and the training set data is obtained according to the target characteristic data.

[0094] (3) State evaluation model optimization: a battery health state evaluation model is constructed; the battery health state evaluation model includes an input layer, a bidirectional long short-term memory network layer, an intermediate layer, and an output layer connected in turn. In the process of training the battery health state evaluation model through the training set, the hyperparameters of the battery health state evaluation model are optimized and configured through an optimization algorithm; the optimization algorithm takes the hyperparameters as search variables, obtains a target hyperparameter combination through iterative search, and obtains an optimized battery health state evaluation model. The optimization algorithm is a sparrow search algorithm; the sparrow search algorithm simulates the foraging and anti-predation behavior mechanism of the finder, follower and alarm of the sparrow population, takes the prediction error of the battery health state evaluation model on the training set as the fitness value, iteratively searches and updates the value of the hyperparameters, and stops until the training stop condition is met, to obtain the target hyperparameter combination. The bidirectional long short-term memory network layer includes a forward hidden layer and a backward hidden layer; the forward hidden layer is used to extract features in the positive order of the time sequence of the input data; the backward hidden layer is used to extract features in the reverse order of the time sequence of the input data; and the hyperparameters include the number of neurons corresponding to the forward hidden layer and the backward hidden layer.

[0095] (4) Battery state evaluation: the battery health state evaluation is performed through the optimized battery health state evaluation model.

[0096] (5) Model effect verification: test set data is obtained according to the target characteristic data; the test set data includes target characteristic data collected after the collection time of the training set data; the battery health state evaluation value is obtained by processing the test set data through the optimized battery health state evaluation model; and the optimized battery health state evaluation model is verified according to the error between the battery health state evaluation value and the corresponding battery health state historical value.

[0097] In this embodiment, by acquiring the operation data of the lithium ion battery in the charging and discharging cycle process, the performance characteristic data of the lithium ion battery is acquired according to the operation data, the target characteristic data is determined from the performance characteristic data, and the training set data is acquired according to the target characteristic data. The battery health state evaluation model is constructed. In the process of training the battery health state evaluation model by the training set, the hyperparameters of the battery health state evaluation model are optimized and configured by the optimization algorithm, and the battery health state evaluation is performed by the optimized battery health state evaluation model. The model fitting capability, the prediction accuracy and the battery health state evaluation accuracy can be improved.

[0098] It should be understood that, although each step in the flowchart involved in the above embodiments is displayed in sequence according to the arrow, these steps are not necessarily executed in the order indicated by the arrow. Unless otherwise specified herein, the execution of these steps is not strictly limited in sequence, and these steps can be executed in other orders. Moreover, at least part of the steps in the flowchart involved in the above embodiments can include multiple steps or stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily sequential, but can be alternately or alternately executed with at least part of other steps or steps or stages in other steps. It can be understood that the steps in different embodiments can be freely combined as needed, and various non-contradictory schemes formed by the combination are within the scope of protection of the present application.

[0099] Based on the same inventive concept, the embodiments of the present application also provide a bidirectional long short-term memory network model-based battery health state evaluation device for implementing the above-mentioned battery health state evaluation method based on a bidirectional long short-term memory network model. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme described in the above method, so the specific limitations in one or more bidirectional long short-term memory network model-based battery health state evaluation device embodiments provided below can refer to the limitations of the bidirectional long short-term memory network model-based battery health state evaluation method described above, and will not be repeated here.

[0100] In one exemplary embodiment, as shown in Figure 7 a bidirectional long short-term memory network model-based battery health state evaluation device is provided, comprising: a data acquisition module 10, a feature screening module 20, a model construction module 30, a model training module 40 and a state evaluation module 50, wherein:

[0101] The data acquisition module 10 is configured to acquire operation data of the lithium ion battery in the charging and discharging cycle process, and acquire performance characteristic data of the lithium ion battery according to the operation data.

[0102] The feature screening module 20 is configured to determine target feature data from the performance feature data, and obtain training set data according to the target feature data.

[0103] The model construction module 30 is configured to construct a battery health state evaluation model, which comprises an input layer, a bidirectional long short-term memory network layer, an intermediate layer, and an output layer connected in sequence.

[0104] The model training module 40 is configured to optimize and configure hyperparameters of the battery health state evaluation model through an optimization algorithm in the process of training the battery health state evaluation model through the training set; the optimization algorithm takes the hyperparameters as search variables, obtains a target hyperparameter combination through iterative search, and obtains an optimized battery health state evaluation model.

[0105] The state evaluation module 50 is configured to perform battery health state evaluation through the optimized battery health state evaluation model.

[0106] In an exemplary embodiment, the performance feature data includes an isobaric charging duration, a constant voltage duration, an isoflow discharging duration, an occurrence duration of a target discharging temperature, and an isobaric discharging duration; the feature screening module 20 is further configured to obtain a first change trend feature of the performance feature data, and obtain a second change trend feature of the battery health state; according to the similarity between the first change trend feature and the second change trend feature, the performance feature data is screened to obtain intermediate feature data; and the intermediate feature data is integrated into the target feature data through a principal component analysis algorithm.

[0107] In an exemplary embodiment, the feature screening module 20 is further configured to obtain full life cycle data of the lithium ion battery, obtain a performance degradation index according to the full life cycle data, obtain battery health state data based on the performance degradation index, fit the battery health state data to obtain the second change trend feature; the performance degradation index includes at least one of a battery capacity, a battery internal resistance, a battery power, and a battery remaining cycle number; and the battery health state data includes a plurality of discrete data points.

[0108] In an exemplary embodiment, the optimization algorithm involved in the model training module 40 is a sparrow search algorithm; the sparrow search algorithm simulates the foraging and anti-predation behavior mechanism of the discoverer, follower and alarm of the sparrow population, takes the prediction error of the battery health state evaluation model on the training set as the fitness value, iteratively searches and updates the value of the hyperparameters until the training stop condition is met, and stops to obtain the target hyperparameter combination.

[0109] In an example embodiment, the bidirectional long short-term memory network layer involved by the model training module 40 includes a forward hidden layer and a backward hidden layer; the forward hidden layer is used to extract features in a positive order of the time sequence of the input data; the backward hidden layer is used to extract features in a reverse order of the time sequence of the input data; the hyperparameters include the number of neurons corresponding to the forward hidden layer and the backward hidden layer.

[0110] In an example embodiment, the model training module 40 is further configured to obtain test set data according to the target feature data; the test set data includes the target feature data collected after the collection time of the training set data; the optimized battery health state evaluation model is used to process the test set data to obtain a battery health state evaluation value; and the optimized battery health state evaluation model is verified according to the error between the battery health state evaluation value and the corresponding battery health state historical value.

[0111] The above-mentioned various modules of the battery health state evaluation device based on the bidirectional long short-term memory network model can be realized by software, hardware and combinations thereof, in whole or in part. The above-mentioned various modules can be embedded in or independent of the processor in the computer device in hardware form, or can be stored in the memory in the computer device in software form, so as to be called and executed by the processor to perform the operations corresponding to the above-mentioned various modules.

[0112] In an example embodiment, a computer device is provided, which can be a terminal, and the internal structure diagram thereof can be as shown in Figure 8The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capability. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with external terminals in a wired or wireless manner. The wireless manner can be realized through WIFI, mobile cellular network, near field communication (NFC) or other technologies. The computer program is executed by the processor to realize a battery state of health evaluation method based on a bidirectional long short-term memory network model. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device, or an external keyboard, touchpad or mouse, etc.

[0113] Those skilled in the art can understand that, Figure 8 The skilled in the art can understand that,

[0114] In one exemplary embodiment, a computer device is provided, comprising a memory and a processor, the memory storing a computer program, and the processor implementing the following steps when executing the computer program: obtaining running data of a lithium ion battery in a charge-discharge cycle process, and obtaining performance characteristic data of the lithium ion battery according to the running data; determining target characteristic data from the performance characteristic data, and obtaining training set data according to the target characteristic data; constructing a battery health state evaluation model; the battery health state evaluation model comprises an input layer, a bidirectional long short-term memory network layer, an intermediate layer, and an output layer connected in turn; in the process of training the battery health state evaluation model through the training set, the hyperparameters of the battery health state evaluation model are optimized and configured through an optimization algorithm; the optimization algorithm takes the hyperparameters as search variables, obtains a target hyperparameter combination through iterative search, so as to obtain an optimized battery health state evaluation model; and the battery health state is evaluated through the optimized battery health state evaluation model.

[0115] In one embodiment, the performance characteristic data includes isobaric rising charge duration, constant voltage duration, isoflow falling charge duration, target discharge temperature occurrence duration, and isobaric falling discharge duration; the processor, when executing the computer program, involves determining target characteristic data from the performance characteristic data, including: obtaining a first change trend feature of the performance characteristic data, and obtaining a second change trend feature of the battery health state; according to the similarity between the first change trend feature and the second change trend feature, the performance characteristic data is filtered to obtain intermediate characteristic data; and the intermediate characteristic data is integrated into target characteristic data through a principal component analysis algorithm.

[0116] In one embodiment, the processor, when executing the computer program, involves obtaining a second change trend feature of the battery health state, including: obtaining full life cycle data of the lithium ion battery, and obtaining a performance degradation index according to the full life cycle data; obtaining battery health state data based on the performance degradation index, fitting the battery health state data to obtain the second change trend feature; the performance degradation index includes at least one of battery capacity, battery internal resistance, battery power, and battery remaining cycle number; and the battery health state data includes a plurality of discrete data points.

[0117] In one embodiment, the optimization algorithm involved by the processor when executing the computer program is a sparrow search algorithm; the sparrow search algorithm simulates the foraging and anti-predation behavior mechanism of finders, followers and alarmers in sparrow populations, takes the prediction error of the battery health state evaluation model on the training set as the fitness value, iteratively searches and updates the value of the hyperparameters, and stops until the training stop condition is met, to obtain the target hyperparameter combination.

[0118] In one embodiment, the bidirectional long short-term memory network layer involved when the processor executes the computer program includes a forward hidden layer and a backward hidden layer; the forward hidden layer is used to extract features in a forward order of the time sequence of the input data; the backward hidden layer is used to extract features in a reverse order of the time sequence of the input data; the hyperparameters include the number of neurons corresponding to the forward hidden layer and the backward hidden layer.

[0119] In one embodiment, the processor executing the computer program further implements the following steps: obtaining test set data according to the target feature data; the test set data includes the target feature data collected after the collection time of the training set data; processing the test set data by the optimized battery health state evaluation model to obtain a battery health state evaluation value; verifying the optimized battery health state evaluation model according to the error between the battery health state evaluation value and the corresponding battery health state historical value.

[0120] In one embodiment, a computer readable storage medium is provided, and the computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement the following steps: obtaining running data of a lithium ion battery in a charge-discharge cycle process, and obtaining performance feature data of the lithium ion battery according to the running data; determining target feature data from the performance feature data, and obtaining training set data according to the target feature data; constructing a battery health state evaluation model; the battery health state evaluation model includes an input layer, a bidirectional long short-term memory network layer, an intermediate layer, and an output layer connected in sequence; in the process of training the battery health state evaluation model by the training set, the hyperparameters of the battery health state evaluation model are configured by an optimization algorithm; the optimization algorithm takes the hyperparameters as search variables, and obtains a target hyperparameter combination by iterative search to obtain an optimized battery health state evaluation model; and the battery health state is evaluated by the optimized battery health state evaluation model.

[0121] In one embodiment, the performance feature data includes an isobaric rise charging duration, a constant voltage duration, an isoflow drop charging duration, an appearance duration of a target discharging temperature, and an isobaric drop discharging duration; the computer program executed by the processor involves determining the target feature data from the performance feature data, including: obtaining a first change trend feature of the performance feature data, and obtaining a second change trend feature of the battery health state; filtering the performance feature data according to the similarity between the first change trend feature and the second change trend feature to obtain intermediate feature data; and integrating the intermediate feature data into the target feature data by a principal component analysis algorithm.

[0122] In an embodiment, the second change trend feature of the battery health state is obtained by the computer program when executed by the processor, including: obtaining the full life cycle data of the lithium ion battery, obtaining the performance degradation index according to the full life cycle data; obtaining the battery health state data based on the performance degradation index, fitting the battery health state data to obtain the second change trend feature; the performance degradation index includes at least one of the battery capacity, the battery internal resistance, the battery power and the battery remaining cycle number; the battery health state data includes a plurality of discrete data points.

[0123] In an embodiment, the optimization algorithm is a sparrow search algorithm when the computer program is executed by the processor; the sparrow search algorithm simulates the foraging and anti-predation behavior mechanism of the finder, follower and alarm in the sparrow population, takes the prediction error of the battery health state evaluation model on the training set as the fitness value, iteratively searches and updates the value of the hyperparameter until the training stop condition is met, and stops to obtain the target hyperparameter combination.

[0124] In an embodiment, the bidirectional long short-term memory network layer includes a forward hidden layer and a backward hidden layer; the forward hidden layer is used to extract features in the positive order of the time sequence of the input data; the backward hidden layer is used to extract features in the reverse order of the time sequence of the input data; the hyperparameters include the number of neurons corresponding to the forward hidden layer and the backward hidden layer.

[0125] In an embodiment, the computer program is further executed by the processor to implement the following steps: obtaining the test set data according to the target feature data; the test set data includes the target feature data collected after the collection time of the training set data; processing the test set data by the optimized battery health state evaluation model to obtain the battery health state evaluation value; verifying the optimized battery health state evaluation model according to the error between the battery health state evaluation value and the corresponding battery health state historical value.

[0126] In one embodiment, a computer program product is provided, comprising a computer program which, when executed by a processor, implements the following steps: obtaining running data of a lithium ion battery during a charge-discharge cycle, obtaining performance characteristic data of the lithium ion battery according to the running data; determining target characteristic data from the performance characteristic data, and obtaining training set data according to the target characteristic data; constructing a battery health state evaluation model; the battery health state evaluation model comprises an input layer, a bidirectional long short-term memory network layer, an intermediate layer, and an output layer connected in turn; in the process of training the battery health state evaluation model by the training set, the hyperparameters of the battery health state evaluation model are optimized and configured by an optimization algorithm; the optimization algorithm takes the hyperparameters as search variables, obtains a target hyperparameter combination through iterative search, so as to obtain an optimized battery health state evaluation model; and the battery health state is evaluated by the optimized battery health state evaluation model.

[0127] In one embodiment, the performance characteristic data includes constant-voltage rising charge duration, constant-voltage duration, constant-current falling charge duration, target discharge temperature occurrence duration, and constant-voltage falling discharge duration; the computer program, when executed by the processor, involves determining the target characteristic data from the performance characteristic data, including: obtaining a first change trend feature of the performance characteristic data, obtaining a second change trend feature of the battery health state; according to the similarity between the first change trend feature and the second change trend feature, the performance characteristic data is filtered to obtain intermediate characteristic data; and the intermediate characteristic data is integrated into the target characteristic data by a principal component analysis algorithm.

[0128] In one embodiment, the computer program, when executed by the processor, involves obtaining the second change trend feature of the battery health state, including: obtaining full life cycle data of the lithium ion battery, obtaining a performance degradation index according to the full life cycle data; obtaining battery health state data based on the performance degradation index, fitting the battery health state data to obtain the second change trend feature; the performance degradation index includes at least one of battery capacity, battery internal resistance, battery power and battery remaining cycle number; the battery health state data includes a plurality of discrete data points.

[0129] In one embodiment, the optimization algorithm, when executed by the processor, involves a sparrow search algorithm; the sparrow search algorithm simulates the foraging and anti-predation behavior mechanism of the finder, follower and alarm of the sparrow population, takes the prediction error of the battery health state evaluation model on the training set as the fitness value, iteratively searches and updates the value of the hyperparameters, and stops until the training stop condition is met, to obtain the target hyperparameter combination.

[0130] In one embodiment, the bi-directional long short-term memory network layer involved when the computer program is executed by the processor comprises a forward hidden layer and a backward hidden layer; the forward hidden layer is used to extract features in a forward order of the time sequence of the input data; the backward hidden layer is used to extract features in a reverse order of the time sequence of the input data; the hyperparameters comprise neuron numbers corresponding to the forward hidden layer and the backward hidden layer.

[0131] In one embodiment, the computer program is further executed by the processor to implement the following steps: obtaining test set data according to the target feature data; the test set data comprises the target feature data collected after the collection time of the training set data; processing the test set data by the optimized battery health state evaluation model to obtain a battery health state evaluation value; and verifying the optimized battery health state evaluation model according to the error between the battery health state evaluation value and the corresponding battery health state historical value.

[0132] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when executed, can include the processes of the above-mentioned embodiment methods. Any reference to memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. The non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical storage, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. The volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration but not limitation, the RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The database involved in the embodiments provided in the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a block chain, etc., without being limited thereto. The processor involved in the embodiments provided in the present application can be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., without being limited thereto.

[0133] The technical features of the above embodiments can be combined in any manner. To make the description concise, not all possible combinations of the technical features in the above embodiments are described, but as long as the combinations of the technical features do not exist contradictions, they should be considered as the scope of the present application.

[0134] The above-described embodiments are merely illustrative of several embodiments of the present application, which are described in more detail and in a specific manner, but should not be construed as limiting the scope of the patent of the present application. It should be noted that, for those of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. A battery state of health estimation method based on a bidirectional long short-term memory network model, characterized in that, The method comprises: acquiring operation data of a lithium ion battery in a charge-discharge cycle process, and acquiring performance characteristic data of the lithium ion battery according to the operation data; determining target characteristic data from the performance characteristic data, and acquiring training set data according to the target characteristic data; constructing a battery health state evaluation model; the battery health state evaluation model comprises an input layer, a bidirectional long short-term memory network layer, an intermediate layer, and an output layer connected in sequence; in the process of training the battery health state evaluation model by the training set, the hyperparameters of the battery health state evaluation model are configured by an optimization algorithm; the optimization algorithm takes the hyperparameters as search variables, and obtains a target hyperparameter combination through iterative search to obtain an optimized battery health state evaluation model; evaluating the battery health state by the optimized battery health state evaluation model.

2. The method of claim 1, wherein, The performance characteristic data comprises an isobaric rise charging time, a constant voltage duration, an isoflow drop charging time, an occurrence duration of a target discharging temperature, and an isobaric drop discharging time. The target characteristic data is determined from the performance characteristic data, comprising: acquiring a first change trend feature of the performance characteristic data, and acquiring a second change trend feature of the battery health state; filtering the performance characteristic data according to the similarity between the first change trend feature and the second change trend feature to obtain intermediate characteristic data; integrating the intermediate characteristic data into target characteristic data by a principal component analysis algorithm.

3. The method of claim 2, wherein, The second change trend feature of the battery health state is acquired, comprising: acquiring full life cycle data of the lithium ion battery, and acquiring a performance degradation index according to the full life cycle data; acquiring battery health state data based on the performance degradation index, fitting the battery health state data to obtain a second change trend feature; the performance degradation index comprises at least one of a battery capacity, a battery internal resistance, a battery power, and a battery remaining cycle number; the battery health state data comprises a plurality of discrete data points.

4. The method of claim 1, wherein, The optimization algorithm is a sparrow search algorithm; the sparrow search algorithm simulates the foraging and anti-predation behavior mechanism of discoverers, followers and alarmers in a sparrow population, takes the prediction error of the battery health state evaluation model on the training set as a fitness value, iteratively searches and updates the value of the hyperparameters until a training stop condition is met, and obtains a target hyperparameter combination.

5. The method of claim 1, wherein, The bidirectional long short-term memory network layer comprises a forward hidden layer and a backward hidden layer; the forward hidden layer is used to extract features in a positive order of a time sequence of input data; the backward hidden layer is used to extract features in a reverse order of the time sequence of the input data; the hyperparameters comprise neuron numbers corresponding to the forward hidden layer and the backward hidden layer.

6. The method of claim 1, wherein, The method further comprises: acquiring test set data according to the target characteristic data; the test set data comprises target characteristic data collected after the collection time of the training set data; The test set data is processed by the optimized battery health state evaluation model to obtain a battery health state evaluation value; The optimized battery health state evaluation model is verified according to an error between the battery health state evaluation value and a corresponding battery health state historical value. 7.A battery state of health estimation device based on a bidirectional long short-term memory network model, characterized by The device comprises: The data acquisition module is configured to acquire performance characteristic data of the lithium ion battery according to operation data of the lithium ion battery in a charging and discharging cycle process; The characteristic screening module is configured to determine target characteristic data from the performance characteristic data and acquire training set data according to the target characteristic data; The model construction module is configured to construct a battery health state evaluation model; the battery health state evaluation model comprises an input layer, a bidirectional long short-term memory network layer, an intermediate layer, and an output layer connected in sequence; The model training module is configured to optimize and configure hyperparameters of the battery health state evaluation model by an optimization algorithm in a process of training the battery health state evaluation model by the training set; the optimization algorithm takes the hyperparameters as search variables, obtains a target hyperparameter combination by iterative search, and obtains an optimized battery health state evaluation model. The state evaluation module is configured to perform battery health state evaluation by the optimized battery health state evaluation model.

8. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that, The processor implements the steps of the method of any one of claims 1 to 6 when executing the computer program.

9. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program, when executed by the processor, implements the steps of the method of any one of claims 1 to 6.

10. A computer program product comprising a computer program, characterized in that, The computer program, when executed by the processor, implements the steps of the method of any one of claims 1 to 6.

Images