Energy storage battery health state evaluation method and device in random charging scene, equipment and storage medium

Abstract

The invention discloses an energy storage battery health state evaluation method and device in a random charging scene, equipment and a storage medium, and the method comprises the steps: obtaining operation data from the start to the end of battery charging, and the operation data comprises voltage and charging capacity data; preprocessing the operation data; according to the preprocessed data, converting the preprocessed data into a voltage-capacity sequence of an incomplete charging process; inputting the voltage-capacity sequence of the incomplete charging process into a pre-constructed sequence reconstruction model, and reconstructing a corresponding voltage-capacity sequence of the complete charging process; and inputting the reconstructed voltage-capacity sequence of the complete charging process into a pre-constructed health state estimation model based on a layered multi-head attention mechanism and a bidirectional gating circulation unit to obtain a health state value of the battery. The method improves the practicability and flexibility of battery SOH evaluation, is suitable for various complex working conditions, effectively prolongs the service life of the battery, and optimizes the performance of a battery management system.

Description

Technical Field

[0001] The present invention relates to the field of battery management technology, and in particular to a method, device and storage medium for evaluating the health status of an energy storage battery in a random charging scenario. Background Art

[0002] Battery energy storage technology has become a key driver in many areas, including electric vehicles, mobile devices, and power systems. This technology has played an important role in improving energy efficiency and reducing carbon emissions. By evaluating the health status of the battery, we can understand the degree of battery degradation in a timely manner, so as to maintain the battery system more efficiently and effectively, while improving the stability and safety of the battery system.

[0003] Unlike the battery discharge process, the battery charging process usually follows a specific charging protocol, so the charging data is more stable and contains a lot of information related to the battery health status. Therefore, the current mainstream SOH estimation methods are based on charging process data. However, most of the existing methods for SOH estimation based on charging process data require a complete charging process, that is, charging from 0% state of charge to 100% state of charge, which rarely occurs in actual situations.

[0004] To address this problem, the present invention proposes a new method for estimating the health status of a battery under random charging conditions. This method can estimate the SOH value when the battery undergoes only a short charging process, and has high practicality and broad application prospects. Summary of the invention

[0005] In order to solve the problems existing in the prior art, the present invention provides a method, device and storage medium for evaluating the health status of an energy storage battery in a random charging scenario, which solves the problems mentioned in the above background technology.

[0006] To achieve the above object, the present invention provides the following technical solution: a method for evaluating the health status of an energy storage battery in a random charging scenario, comprising the following steps: S1. Acquire the operating data of the battery from the beginning to the end of charging, wherein the operating data includes voltage and charging capacity data; S2. Preprocess the operation data to obtain preprocessed data; S3, converting the preprocessed data into a voltage-capacity sequence of an incomplete charging process; S4, inputting the voltage-capacity sequence of the incomplete charging process into a pre-built sequence reconstruction model to reconstruct the corresponding voltage-capacity sequence of the complete charging process; S5. Input the reconstructed voltage-capacity sequence of the complete charging process into a pre-built health state estimation model based on a hierarchical multi-head attention mechanism and a bidirectional gated recurrent unit to obtain the health state value of the battery.

[0007] Preferably, in step S2, the preprocessing includes data normalization, outlier removal and capacity interpolation processing, which specifically includes the following steps: S21, data normalization: Assume that the upper cut-off voltage of the battery during operation is ; The lower cut-off voltage is ; The rated capacity of the battery is , the normalized calculation formula is as follows: ; In the formula, represents the voltage value to be normalized, Represents the normalized voltage value, represents the capacity value to be normalized, shows the normalized capacity value; S22, outlier removal: Delete outliers in the charging data, specifically including treating data points exceeding ±3 times the standard deviation of the mean as outliers and deleting them. The specific screening formula is as follows: ; In the formula, is the number of collected data points, It is data points, is the mean of the collected data, is the standard deviation of the collected data; Determine the range of outliers, i.e. the upper limit , lower limit ; If the data point satisfy or , then the data point It is considered as an outlier and removed; S23, capacity interpolation processing: interpolating the voltage-capacity sequence, including: Set the starting voltage , Termination voltage and voltage increment , through the voltage range Determine the sequence length ; Then, the capacity value corresponding to the new sequence voltage value is calculated by the following interpolation formula: ; in, It's location The known voltage point and capacity value on the left; It's location The known voltage point and capacity value on the right; is the position of the midpoint in the new sequence, and ; After interpolation processing, a voltage-capacity sequence with a fixed length is obtained.

[0008] Preferably, in step S3, the conversion into a voltage-capacity sequence of an incomplete charging process specifically includes the following: Assume the length of the voltage-capacity sequence is , randomly select the voltage starting point , Voltage termination point , and the shortest subsequence ratio ,satisfy ; According to the preprocessed data, the voltage-capacity sequence of each complete cycle is divided into subsequences with different starting and ending points and different lengths. The relative positions of the divided subsequences are retained, and the capacity value of each position of the subsequence is subtracted The corresponding capacity value is then set to 0 for the voltage-capacity value not selected in the sequence. The process formula is as follows: ; in Indicates the capacity-voltage curve Point voltage The corresponding capacity value; The extracted sequence is converted into a voltage-capacity sequence, that is, all the capacity values ​​in the extracted sequence are subtracted from the capacity value corresponding to the starting voltage. The conversion process is as follows: ; in is a randomly selected starting voltage The corresponding capacity value; The voltage-capacity sequence of the treated incomplete charging process is as follows: ; in, It is the voltage-capacity sequence of the incomplete charging process; It is the voltage-capacity data at a certain moment.

[0009] Preferably, in step S4, the voltage-capacity sequence of the incomplete charging process is input into a pre-constructed sequence reconstruction model to reconstruct the voltage-capacity sequence of the corresponding complete charging process, and the formula is expressed as follows: ; in, To reconstruct the voltage-capacity sequence of the complete charging process, Reconstruct the model for the sequence; The sequence reconstruction model consists of an encoder and a decoder. The encoder is used to encode the features of the input incomplete sequence to obtain the reconstructed features; the decoder is used to decode the encoded reconstructed features to reconstruct the complete sequence corresponding to the incomplete sequence; the encoder is composed of a number of FC fully connected layers activated by ReLu functions, and the decoder is composed of a number of FC fully connected layers activated by LeakyReLU functions.

[0010] Preferably, in step S4, the reconstruction process of the sequence reconstruction model is as follows: Assume the encoder has Layer, The weight matrix of the layer for , the bias vector is , the encoder's layer-by-layer output operation rules are as follows: ; in, Indicates Layer encoder calculation results; the voltage-capacity data of the incomplete charging process is used as input data ,Right now ; Then the last layer output of the encoder is the hidden feature : ; in is the dimension of the latent feature space; Represents the dimension of the last layer of the encoder; Assume the decoder has Layer, The weight matrix of the layer is , the bias vector is , the layer-by-layer output operation rules of the decoder are as follows: ; in, Indicates The decoder is used to decode the last layer output of the encoder, that is, ;in, The calculation formula of the activation function is: ; in is the coefficient of the negative slope; The output of the last layer of decoder is the reconstruction result of the model As shown below: ; in, is the complete sequence length, and d is the feature dimension.

[0011] Preferably, in step S5, the health state estimation model based on the hierarchical multi-head attention mechanism and the bidirectional gated recurrent unit specifically includes: Feature embedding layer, used to map discrete voltage and capacity inputs into continuous vector space to capture potential correlation features; Short-term attention layer, in which short-term dependencies are captured through several multi-head attention mechanism modules; Long-term attention layer, in which long-term dependencies are captured through a multi-head attention mechanism module; The long- and short-term attention concatenation layer is used to integrate the different time scale features of the multi-head attention layer to form a comprehensive feature representation; The BiGRU layer further enhances the ability to capture time series data and extract battery timing features; The fully connected layer is used to map the features extracted by BiGRU into the estimated value of the battery's health status.

[0012] Preferably, in step S5, the reconstructed voltage-capacity sequence of the complete charging process is input into a pre-built health state estimation model based on a hierarchical multi-head attention mechanism and a bidirectional gated recurrent unit to obtain a health state value of the battery, specifically including the following: Divide the voltage-capacity data of the complete charging process into several parts evenly ,in Indicates part; then in the feature embedding layer, several Embedding modules are used to convert each part into an input embedding vector, and position encoding is added to the input embedding to utilize the order of the input data; by adding the input embedding and the position encoding, the input data is obtained, that is: ; in, Indicates The processing result of each Embedding module is: is a linear map, Represents the sine-cosine encoding process; Then, each feature extracted by the Embedding module is input into the short-term attention layer, which captures short-term dependencies through several multi-head attention mechanism modules: ; in, Represents the multi-head attention mechanism operation; Indicates The operation results of the multi-head attention mechanism; Concatenate the outputs of all multi-head attention modules and combine them through a linear transformation: ; in, is a learnable weight matrix used to project the concatenated vector back to the original dimension; Represent query, key, and value vectors respectively; Integrate the last vector of each result of the multi-head attention mechanism operation in the short-term attention layer to obtain the input for the long-term attention layer ,in express The last vector of ; it is input into the long-term attention layer, where a multi-head attention mechanism module is used to capture long-term dependencies: ; in, represents the feature result extracted by the long-term attention layer, and then the calculation results of the short-term attention layer and the long-term attention layer are spliced ​​through the long-term and short-term attention splicing layer to obtain a feature vector containing long-term and short-term dependencies, that is, ; Finally, by taking the feature vector containing the long-term and short-term dependencies LSD The data is input into the BiGRU layer to extract the battery timing features, and then passes through the fully connected layer to map out the final battery health status SOH value. The formula is as follows: ; in, Represents BiGRU operation; Represents a linear mapping, whose activation function is the RELU function; Represents the BiGRU layer output; Represents the linear mapping result.

[0013] On the other hand, to achieve the above purpose, the present invention also provides the following technical solution: a device for evaluating the health status of an energy storage battery in a random charging scenario, comprising the following modules: A data acquisition module, which acquires the operating data of the battery from the beginning to the end of charging, wherein the operating data includes voltage and charging capacity data; A data preprocessing module preprocesses the running data to obtain preprocessed data; The conversion module converts the preprocessed data into a voltage-capacity sequence of an incomplete charging process; A sequence reconstruction module, inputting the voltage-capacity sequence of the incomplete charging process into a pre-built sequence reconstruction model, and reconstructing the corresponding voltage-capacity sequence of the complete charging process; The health state estimation module inputs the reconstructed voltage-capacity sequence of the complete charging process into a pre-built health state estimation model based on a hierarchical multi-head attention mechanism and a bidirectional gated recurrent unit to obtain the health state value of the battery.

[0014] On the other hand, to achieve the above-mentioned purpose, the present invention also provides the following technical solution: an electronic device, the electronic device comprising: a processor; and a storage device for storing one or more programs; When the one or more programs are executed by the processor, the processor executes the method for evaluating the health status of the energy storage battery in a random charging scenario.

[0015] On the other hand, to achieve the above-mentioned purpose, the present invention also provides the following technical solution: a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method for evaluating the health status of a storage battery in a random charging scenario is implemented.

[0016] The beneficial effect of the present invention is that the present invention can complete the health status assessment when the battery undergoes only a short period of charging, without the need for a complete charging process. This advantage makes it more flexible in practical applications, especially suitable for scenarios where the battery does not always undergo a complete charge and discharge, significantly improving the flexibility and efficiency of health status estimation. In addition, the method can help users timely and effectively assess the degradation of the battery, thereby improving the maintenance efficiency of the battery system, extending the battery life, and enhancing the stability and safety of the system operation. This technology is particularly suitable for applications in electric vehicles, mobile devices, power systems and other fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a schematic diagram of the steps of a method for evaluating the health status of an energy storage battery in a random charging scenario in an embodiment of the present invention; Figure 2 This is a schematic diagram of a sequence reconstruction model in an embodiment of the present invention; Figure 3 A schematic diagram of a health status estimation model in an embodiment of the present invention; Figure 4 It is a schematic diagram of a module of a device for evaluating the health status of an energy storage battery in a random charging scenario in an embodiment of the present invention; Figure 5 This is a schematic diagram of the structure of an electronic device in an embodiment of the present invention; In the figure, 110 - data acquisition module; 120 - data preprocessing module; 130 - conversion module; 140 - sequence reconstruction module; 150 - health status estimation module; 210 - processor; 220 - storage. DETAILED DESCRIPTION

[0018] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] See also Figure 1 The present invention provides a technical solution: a method for evaluating the health status of an energy storage battery in a random charging scenario, comprising the following steps: S1. Obtaining operation data of the battery from the beginning to the end of charging, wherein the operation data includes voltage and charging capacity data.

[0020] Specifically, the battery operation data can be used to indicate the health status of the battery. The calculation formula of the SOH label during training is as follows: ; In the formula, Cap is the charge and discharge capacity, for The current value at the moment, is the charge and discharge capacity of a certain cycle, is the charge and discharge capacity value of the first cycle.

[0021] S2. Preprocess the operating data to obtain preprocessed data.

[0022] Specifically, abnormal data is the charging data containing invalid data deleted, normalization processing can speed up the training of the model, and capacity interpolation can ensure the consistency of the length of the processed sequence, which is convenient for subsequent model calculation.

[0023] The preprocessing includes data normalization, outlier removal and capacity interpolation processing, which specifically includes the following steps: S21. Data normalization: In order to avoid the adverse effects of dimensional differences in voltage and capacity on model training and speed up training, the operating data is processed by normalization to scale the voltage and capacity to the range of 0~1.

[0024] Assume that the upper cut-off voltage of the battery is ; The lower cut-off voltage is ; The rated capacity of the battery is , the normalized calculation formula is as follows: ; In the formula, represents the voltage value to be normalized, Represents the normalized voltage value, represents the capacity value to be normalized, shows the normalized capacity value; S22, outlier removal: Delete outliers in the charging data. Outliers in the charging data will interfere with the training of the model and make it difficult to train the model, so this part of the data is deleted.

[0025] Specifically, data points exceeding the mean ± 3 times the standard deviation are considered as outliers and deleted. The specific screening formula is as follows: ; In the formula, is the number of collected data points, It is data points, is the mean of the collected data, is the standard deviation of the collected data; Determine the range of outliers, i.e. the upper limit , lower limit ; If the data point satisfy or , then the data point It is considered as an outlier and is removed; Outliers can affect the uniform dimension and cause the model to misunderstand the data, resulting in deviations. The above-mentioned operation of removing outliers can prevent this from happening.

[0026] S23, Capacity interpolation processing: As the battery ages, the charging time will continue to shorten, resulting in unequal lengths of voltage-capacity sequences in each cycle. In this case, it is impossible to use the advantages of GPU parallel computing for batch training, which will lead to a long training time when the amount of data is large. Therefore, the interpolation method can unify the data length and speed up model training.

[0027] The voltage-capacity sequence is interpolated, and a suitable interpolation voltage range and voltage increment are set. After the interpolation, a voltage-capacity sequence with a fixed length can be obtained, including: Set the starting voltage , Termination voltage and voltage increment , through the voltage range Determine the sequence length ; Then, the capacity value corresponding to the new sequence voltage value is calculated by the following interpolation formula: ; in, It's location The known voltage point and capacity value on the left; It's location The known voltage point and capacity value on the right; is the position of the midpoint in the new sequence, and ; After interpolation processing, a voltage-capacity sequence with a fixed length is obtained.

[0028] S3. Convert the preprocessed data into a voltage-capacity sequence of an incomplete charging process, and obtain input data for reconstructing a voltage-capacity sequence of a complete charging process.

[0029] The voltage-capacity sequence converted into an incomplete charging process specifically includes the following: Assume the length of the voltage-capacity sequence is , randomly select the voltage starting point , Voltage termination point , and the shortest subsequence ratio ,satisfy and , , indicating that the length of the selected subsequence must be between the length of the original sequence to 100%; according to the preprocessed data, the voltage-capacity sequence of each complete cycle is divided into subsequences with different starting and ending points and different lengths, the relative positions of the divided subsequences are retained, and the capacity value of each position of the subsequence is subtracted The corresponding capacity value is then set to 0 for the voltage-capacity value not selected in the sequence. The process formula is as follows: ; in Indicates the capacity-voltage curve Point voltage The corresponding capacity value; The extracted sequence is converted into a voltage-capacity sequence, that is, all the capacity values ​​in the extracted sequence are subtracted from the capacity value corresponding to the starting voltage. The conversion process is as follows: ; in is a randomly selected starting voltage The corresponding capacity value; The voltage-capacity sequence of the treated incomplete charging process is as follows: ; in, It is the voltage-capacity sequence of the incomplete charging process; It is the voltage-capacity data at a certain moment.

[0030] S4. Input the voltage-capacity sequence of the incomplete charging process into a pre-built sequence reconstruction model to reconstruct the corresponding voltage-capacity sequence of the complete charging process.

[0031] The voltage-capacity sequence of the incomplete charging process is input into the pre-built sequence reconstruction model to reconstruct the corresponding voltage-capacity sequence of the complete charging process. The formula is as follows: ; in, To reconstruct the voltage-capacity sequence of the complete charging process, Reconstruct the model for the sequence; like Figure 2 As shown in the figure, the sequence reconstruction model consists of an encoder and a decoder. The encoder is used to encode the features of the input incomplete sequence to obtain the reconstructed features; the decoder is used to decode the encoded reconstructed features to reconstruct the complete sequence corresponding to the incomplete sequence; the encoder is composed of a number of FC fully connected layers activated by ReLu functions, and the decoder is composed of a number of FC fully connected layers activated by LeakyReLU functions.

[0032] The reconstruction process of the sequence reconstruction model is as follows: Assume the encoder has Layer, The weight matrix of the layer for , the bias vector is , the encoder's layer-by-layer output operation rules are as follows: ; in, Indicates Layer encoder calculation results; the voltage-capacity data of the incomplete charging process is used as input data ,Right now ; The calculation formula of the activation function is: , that is, when hour, ,when hour, . Then the last layer output of the encoder is the hidden feature : ; in is the dimension of the latent feature space; Represents the dimension of the last layer of the encoder; Assume the decoder has Layer, The weight matrix of the layer is , the bias vector is , the layer-by-layer output operation rules of the decoder are as follows: ; in, Indicates The decoder is used to decode the last layer output of the encoder, that is, ;in, The calculation formula of the activation function is: ; in is the coefficient for negative slopes, usually a small positive number; The output of the last layer of decoder is the reconstruction result of the model As shown below: ; in, is the complete sequence length, and d is the feature dimension.

[0033] The goal of the sequence reconstruction model is to minimize the input incomplete sequence and reconstruction sequence Therefore, when training the model, the loss function uses the mean square error (MSE), and its formula is: ; in is the length of the sequence; is the actual sequence data; The sequence data calculated for the model.

[0034] The encoder performs feature encoding on the input incomplete sequence to obtain reconstructed features, and these features are decoded by the decoder to finally reconstruct a complete sequence that is as close to the original sequence as possible.

[0035] Training of sequence reconstruction models Steps S1 to S3 are performed on the battery data set, and then the data is divided into a training set and a test set and the model is trained. The training set is used to optimize the model, and the test set is used to evaluate the model.

[0036] The model training process includes: defining the learning rate of the sequence reconstruction model; defining the loss function of the sequence reconstruction model; defining the optimization algorithm of the sequence reconstruction model; performing multiple iterations, calculating the loss value between the output of the sequence reconstruction model and the true value in each iteration, and optimizing the model parameters through back propagation of the optimization algorithm.

[0037] Use the leave-one-out method to divide the test set and the training set. The leave-one-out method is: each data set is used as a test set, and the remaining data sets are divided as training sets. For the model training process, the loss function is defined as MSE loss, the learning rate is 0.0001, and the optimization algorithm is set to Adam. Perform multiple iterations, and each iteration calculates the loss value between the output of the sequence reconstruction model and the true value, and optimizes the model parameters through back propagation of the optimization algorithm. After multiple trainings, the reconstruction curve of the model can continuously approach the true curve.

[0038] After the sequence reconstruction model is trained, the model can be used to reconstruct the voltage-capacity sequence of the complete charging process. The sequence reconstruction model of the embodiment of the present invention only needs one model to reconstruct the voltage-capacity sequence with any length and any start and end voltage as the voltage-capacity sequence of the complete charging process, and can maintain high reconstruction accuracy even in a short charging time.

[0039] S5. Input the reconstructed voltage-capacity sequence of the complete charging process into a pre-built health state estimation model based on a hierarchical multi-head attention mechanism and a bidirectional gated recurrent unit to obtain the health state value of the battery.

[0040] like Figure 3 As shown in the figure, the health state estimation model based on the hierarchical multi-head attention mechanism and the bidirectional gated recurrent unit specifically includes: Feature embedding layer, used to map discrete voltage and capacity inputs into a continuous vector space to capture potential correlation features; Short-term attention layer, in which short-term dependencies are captured through several multi-head attention mechanism modules (capturing short-term time dependencies); Long-term attention layer, in which long-term dependencies are captured through a multi-head attention mechanism module (focusing on long-term dependencies); The long- and short-term attention concatenation layer is used to integrate the different time scale features of the multi-head attention layer to form a comprehensive feature representation; The BiGRU layer further enhances the ability to capture time series data and extract battery timing features; The fully connected layer is used to map the features extracted by BiGRU into the estimated value of the battery's health status.

[0041] The training method of the health state estimation model based on the hierarchical multi-head attention mechanism and bidirectional gated recurrent unit is consistent with the training method of the sequence reconstruction model.

[0042] Furthermore, the reconstructed voltage-capacity sequence of the complete charging process is input into a pre-built health state estimation model based on a hierarchical multi-head attention mechanism and a bidirectional gated recurrent unit to obtain the health state value of the battery, which includes the following: Divide the voltage-capacity data of the complete charging process into several parts evenly ,in Indicates Then, in the feature embedding layer, several Embedding modules are used to convert each part into an input embedding vector, and position encoding is added to the input embedding to utilize the order of the input data; by adding the input embedding and the position encoding, the input data is obtained, that is, ; in, Indicates The processing result of each Embedding module is: is a linear map, Represents the sine-cosine encoding process; Then, each feature extracted by the Embedding module is input into the short-term attention layer, which captures short-term dependencies through several multi-head attention mechanism modules: ; in, Represents the multi-head attention mechanism operation; Indicates The operation results of the multi-head attention mechanism; Furthermore, the operation process of the multi-head attention mechanism is as follows: First, let’s look at how a single attention mechanism is calculated. First, we calculate the query, key, and value vectors: For each input element , generating the corresponding query, key, and value vectors. This is done through three learnable weight matrices To achieve: ; in They are query matrix, key matrix, and value matrix vector respectively.

[0043] The attention score is then calculated: Next, the correlation between each query and all keys is calculated. This correlation is usually measured using scaled dot product attention, which is formulated as: ; in, is the dimension of the key vector, which is used to scale the dot product result to make it more stable. The softmax function is used to normalize the scores so that they become weights and are combined with the value vector Perform weighted summation to get the output.

[0044] In multi-head attention operations, each attention head has independent query, key, and value matrix weights: ; in Indicates attention heads, a total of Attention heads, each head independently calculates a self-attention.

[0045] Compute the attention output of each head separately.

[0046] Each attention head uses the scaled dot product attention formula described above to calculate its output: ; Merge multiple outputs.

[0047] Concatenate the outputs of all multi-head attention modules and combine them through a linear transformation: ; in, It is a learnable weight matrix used to project the concatenated vector back to the original dimension; the final output combines the attention information of multiple heads and has richer expressive power.

[0048] Integrate the last vector of each result of the multi-head attention mechanism operation in the short-term attention layer to obtain the input for the long-term attention layer ,in express The last vector of ; it is input into the long-term attention layer, where a multi-head attention mechanism module is used to capture long-term dependencies: ; in, represents the feature result extracted by the long-term attention layer, and then the calculation results of the short-term attention layer and the long-term attention layer are spliced ​​through the long-term and short-term attention splicing layer to obtain a feature vector containing long-term and short-term dependencies, that is, ; Finally, by taking the feature vector containing the long-term and short-term dependencies LSDThe data is input into the BiGRU layer to extract the battery timing features, and then passes through the fully connected layer to map out the final battery health status SOH value. The formula is as follows: ; in, Represents BiGRU operation; Represents a linear mapping, whose activation function is the RELU function; Represents the BiGRU layer output; Represents the linear mapping result.

[0049] Furthermore, the long-term and short-term dependency vectors are input into the BiGRU layer to further extract the battery timing features: ; in, Represents BiGRU operation, the rules are as follows: Forward GRU calculation.

[0050] Update Gate: ; Reset the gate: ; Candidate status: ; Current status: ; Inverse GRU calculation.

[0051] Update Gate: ; Reset the gate: ; Candidate status: ; Current status: ; The forward and reverse GRU operation results are concatenated to obtain the BiGRU layer feature processing results: ; Finally, the features extracted by BiGRU are mapped to the battery SOH value through the fully connected layer, that is: ; in, Represents a linear mapping, whose activation function is the RELU function; Represents the BiGRU layer output; Represents the linear mapping result.

[0052] Based on the same inventive concept as the above method embodiment, the present application embodiment also provides a device for evaluating the health status of an energy storage battery in a random charging scenario, which can implement the functions provided by the above method embodiment, such as Figure 4 As shown, the device includes the following modules: The data acquisition module 110 acquires the operating data of the battery from the start to the end of charging, wherein the operating data includes voltage and charging capacity data; The data preprocessing module 120 preprocesses the operation data to obtain preprocessed data; The conversion module 130 converts the preprocessed data into a voltage-capacity sequence of an incomplete charging process; The sequence reconstruction module 140 inputs the voltage-capacity sequence of the incomplete charging process into a pre-built sequence reconstruction model to reconstruct the corresponding voltage-capacity sequence of the complete charging process; The health state estimation module 150 inputs the reconstructed voltage-capacity sequence of the complete charging process into a pre-built health state estimation model based on a hierarchical multi-head attention mechanism and a bidirectional gated recurrent unit to obtain the health state value of the battery.

[0053] Based on the same inventive concept as the above method embodiment, the present application embodiment also provides an electronic device, such as Figure 5 As shown, the device includes: a processor 210; and a storage 220 for storing one or more programs; When the one or more programs are executed by the processor 210, the processor executes the method for evaluating the health status of the energy storage battery in a random charging scenario.

[0054] The method for evaluating the health status of energy storage batteries in the random charging scenario specifically includes the following: Obtaining operating data from the start to the end of battery charging, wherein the operating data includes voltage and charging capacity data; Preprocessing the running data to obtain preprocessed data; According to the preprocessed data, the data is converted into a voltage-capacity sequence of an incomplete charging process; Input the voltage-capacity sequence of the incomplete charging process into a pre-built sequence reconstruction model to reconstruct the corresponding voltage-capacity sequence of the complete charging process; The reconstructed voltage-capacity sequence of the complete charging process is input into a pre-built health state estimation model based on a hierarchical multi-head attention mechanism and a bidirectional gated recurrent unit to obtain the health state value of the battery.

[0055] Based on the same inventive concept as the above method embodiment, the embodiment of the present application also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by the processor 210, the method for evaluating the health status of the energy storage battery in the random charging scenario is implemented.

[0056] The method for evaluating the health status of energy storage batteries in the random charging scenario specifically includes the following: Obtaining operating data from the start to the end of battery charging, wherein the operating data includes voltage and charging capacity data; Preprocessing the running data to obtain preprocessed data; According to the preprocessed data, the data is converted into a voltage-capacity sequence of an incomplete charging process; Input the voltage-capacity sequence of the incomplete charging process into a pre-built sequence reconstruction model to reconstruct the corresponding voltage-capacity sequence of the complete charging process; The reconstructed voltage-capacity sequence of the complete charging process is input into a pre-built health state estimation model based on a hierarchical multi-head attention mechanism and a bidirectional gated recurrent unit to obtain the health state value of the battery.

[0057] In several embodiments provided in the embodiments of the present invention, it should be understood that the disclosed apparatus and method can also be implemented in other ways. The apparatus and method embodiments described above are merely schematic. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the apparatus, method and computer program product according to multiple embodiments of the present invention. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of a code, and the module, program segment or a part of the code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0058] In addition, the functional modules in the various embodiments of the present invention may be integrated together to form an independent part, or each module may exist independently, or two or more modules may be integrated to form an independent part.

[0059] If the function is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions to enable a computer device (which can be a personal computer, an electronic device, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program code. It should be noted that in this article, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements that are not explicitly listed, or also includes elements inherent to such a process, method, article or device. Without more constraints, an element defined by the phrase "comprising a..." does not exclude the existence of other identical elements in the process, method, article or apparatus comprising the element.

[0060] The terms used in the embodiments of the present invention are only for the purpose of describing specific embodiments, and are not intended to limit the present invention. The singular forms "a", "said" and "the" used in the embodiments of the present invention and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings.

[0061] It should be understood that the term "and / or" used in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0062] The word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining" or "in response to detecting", depending on the context. Similarly, the phrases "if it is determined" or "if (stated condition or event) is detected" may be interpreted as "when it is determined" or "in response to determining" or "when detecting (stated condition or event)" or "in response to detecting (stated condition or event)", depending on the context.

[0063] The "first\second" mentioned in the embodiments is only to distinguish similar objects, and does not represent a specific order for the objects. It is understandable that the "first\second" can be interchanged with the specific order or sequence where permitted. It should be understood that the objects distinguished by "first\second" can be interchanged where appropriate, so that the embodiments described herein can be implemented in an order other than those illustrated or described herein.

[0064] Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for evaluating the health status of an energy storage battery in a random charging scenario, characterized in that: The steps include: S1. Acquire the operating data of the battery from the beginning to the end of charging, wherein the operating data includes voltage and charging capacity data; S2. Preprocess the operation data to obtain preprocessed data; S3, converting the preprocessed data into a voltage-capacity sequence of an incomplete charging process; S4, inputting the voltage-capacity sequence of the incomplete charging process into a pre-built sequence reconstruction model to reconstruct the corresponding voltage-capacity sequence of the complete charging process; S5. Input the reconstructed voltage-capacity sequence of the complete charging process into a pre-built health state estimation model based on a hierarchical multi-head attention mechanism and a bidirectional gated recurrent unit to obtain the health state value of the battery.

2. The method for evaluating the health status of an energy storage battery in a random charging scenario according to claim 1, characterized in that: In step S2, the preprocessing includes data normalization, outlier removal and capacity interpolation processing, which specifically includes the following steps: S21, data normalization: Assume that the upper cut-off voltage of the battery during operation is ; The lower cut-off voltage is ; The rated capacity of the battery is , the normalized calculation formula is as follows: ； In the formula, represents the voltage value to be normalized, Represents the normalized voltage value, represents the capacity value to be normalized, shows the normalized capacity value; S22, outlier removal: Delete outliers in the charging data, specifically including treating data points exceeding ±3 times the standard deviation of the mean as outliers and deleting them. The specific screening formula is as follows: ； In the formula, is the number of collected data points, It is data points, is the mean of the collected data, is the standard deviation of the collected data; Determine the range of outliers, i.e. the upper limit , lower limit ; If the data point satisfy or , then the data point It is considered as an outlier and is removed; S23, capacity interpolation processing: interpolating the voltage-capacity sequence, including: Set the starting voltage , Termination voltage and voltage increment , through the voltage range Determine the sequence length ; Then, the capacity value corresponding to the new sequence voltage value is calculated by the following interpolation formula: ； in, It's location The known voltage point and capacity value on the left; It's location The known voltage point and capacity value on the right; is the position of the midpoint in the new sequence, and ; After interpolation processing, a voltage-capacity sequence with a fixed length is obtained.

3. The method for evaluating the health status of an energy storage battery in a random charging scenario according to claim 1, characterized in that: In step S3, the conversion into a voltage-capacity sequence of an incomplete charging process specifically includes the following: Assume the length of the voltage-capacity sequence is , randomly select the voltage starting point , Voltage termination point , and the shortest subsequence ratio ,satisfy ; According to the preprocessed data, the voltage-capacity sequence of each complete cycle is divided into subsequences with different starting and ending points and different lengths. The relative positions of the divided subsequences are retained, and the capacity value of each position of the subsequence is subtracted The corresponding capacity value is then set to 0 for the voltage-capacity value not selected in the sequence. The process formula is as follows: ； in Indicates the capacity-voltage curve Point voltage The corresponding capacity value; The extracted sequence is converted into a voltage-capacity sequence, that is, all the capacity values ​​in the extracted sequence are subtracted from the capacity value corresponding to the starting voltage. The conversion process is as follows: ； in is a randomly selected starting voltage The corresponding capacity value; The voltage-capacity sequence of the treated incomplete charging process is as follows: ； in, It is the voltage-capacity sequence of the incomplete charging process; It is the voltage-capacity data at a certain moment.

4. The method for evaluating the health status of an energy storage battery in a random charging scenario according to claim 1, characterized in that: In step S4, the voltage-capacity sequence of the incomplete charging process is input into a pre-built sequence reconstruction model to reconstruct the voltage-capacity sequence of the corresponding complete charging process, which is expressed as follows: ； in, To reconstruct the voltage-capacity sequence of the complete charging process, Reconstruct the model for the sequence; The sequence reconstruction model consists of an encoder and a decoder. The encoder is used to encode the features of the input incomplete sequence to obtain the reconstructed features; the decoder is used to decode the encoded reconstructed features to reconstruct the complete sequence corresponding to the incomplete sequence; the encoder is composed of a number of FC fully connected layers activated by ReLu functions, and the decoder is composed of a number of FC fully connected layers activated by LeakyReLU functions.

5. The method for evaluating the health status of an energy storage battery in a random charging scenario according to claim 1, characterized in that: In step S4, the reconstruction process of the sequence reconstruction model is as follows: Assume the encoder has Layer, The weight matrix of the layer for , the bias vector is , the encoder's layer-by-layer output operation rules are as follows: ； in, Indicates Layer encoder calculation results; the voltage-capacity data of the incomplete charging process is used as input data ,Right now ; Then the last layer output of the encoder is the hidden feature : ； in is the dimension of the latent feature space; Represents the dimension of the last layer of the encoder; Assume the decoder has Layer, The weight matrix of the layer is , the bias vector is , the layer-by-layer output operation rules of the decoder are as follows: ； in, Indicates The decoder is used to decode the last layer output of the encoder, that is, ;in, The calculation formula of the activation function is: ； in is the coefficient of the negative slope; The output of the last layer of decoder is the reconstruction result of the model As shown below: ； in, is the complete sequence length, and d is the feature dimension.

6. The method for evaluating the health status of an energy storage battery in a random charging scenario according to claim 1, characterized in that: In step S5, the health state estimation model based on the hierarchical multi-head attention mechanism and the bidirectional gated recurrent unit specifically includes: Feature embedding layer, used to map discrete voltage and capacity inputs into continuous vector space to capture potential correlation features; Short-term attention layer, in which short-term dependencies are captured through several multi-head attention mechanism modules; Long-term attention layer, in which long-term dependencies are captured through a multi-head attention mechanism module; The long- and short-term attention concatenation layer is used to integrate the different time scale features of the multi-head attention layer to form a comprehensive feature representation; The BiGRU layer further enhances the ability to capture time series data and extract battery timing features. The fully connected layer is used to map the features extracted by BiGRU into the estimated value of the battery's health status.

7. The method for evaluating the health status of an energy storage battery in a random charging scenario according to claim 1, characterized in that: In step S5, the reconstructed voltage-capacity sequence of the complete charging process is input into a pre-built health state estimation model based on a hierarchical multi-head attention mechanism and a bidirectional gated recurrent unit to obtain a health state value of the battery, specifically including the following: Divide the voltage-capacity data of the complete charging process into several parts evenly ,in Indicates part; then in the feature embedding layer, several Embedding modules are used to convert each part into an input embedding vector, and position encoding is added to the input embedding to utilize the order of the input data; by adding the input embedding and the position encoding, the input data is obtained, that is: ； in, Indicates The processing result of each Embedding module is: is a linear map, Represents the sine-cosine encoding process; Then, each feature extracted by the Embedding module is input into the short-term attention layer, which captures short-term dependencies through several multi-head attention mechanism modules: ； in, Represents the multi-head attention mechanism operation; Indicates The operation results of the multi-head attention mechanism; Concatenate the outputs of all multi-head attention modules and combine them through a linear transformation: ； in, is a learnable weight matrix used to project the concatenated vector back to the original dimension; Represent query, key, and value vectors respectively; Integrate the last vector of each result of the multi-head attention mechanism operation in the short-term attention layer to obtain the input for the long-term attention layer ,in express The last vector of ; it is input into the long-term attention layer, where a multi-head attention mechanism module is used to capture long-term dependencies: ； in, represents the feature result extracted by the long-term attention layer, and then the calculation results of the short-term attention layer and the long-term attention layer are spliced ​​through the long-term and short-term attention splicing layer to obtain a feature vector containing long-term and short-term dependencies, that is, ； Finally, by taking the feature vector containing the long-term and short-term dependencies LSD The data is input into the BiGRU layer to extract the battery timing features, and then passes through the fully connected layer to map out the final battery health status SOH value. The formula is as follows: ； in, Represents BiGRU operation; Represents a linear mapping, whose activation function is the RELU function; Represents the BiGRU layer output; Represents the linear mapping result.

8. A device for evaluating the health status of an energy storage battery in a random charging scenario according to any one of claims 1 to 7, characterized in that: Includes the following modules: A data acquisition module (110) acquires operating data of the battery from the start to the end of charging, wherein the operating data includes voltage and charging capacity data; A data preprocessing module (120) preprocesses the operating data to obtain preprocessed data; A conversion module (130) converts the pre-processed data into a voltage-capacity sequence of an incomplete charging process; A sequence reconstruction module (140) inputs the voltage-capacity sequence of the incomplete charging process into a pre-built sequence reconstruction model to reconstruct the corresponding voltage-capacity sequence of the complete charging process; The health state estimation module (150) inputs the reconstructed voltage-capacity sequence of the complete charging process into a pre-built health state estimation model based on a hierarchical multi-head attention mechanism and a bidirectional gated recurrent unit to obtain a health state value of the battery.

9. An electronic device, characterized in that: The electronic device comprises: a processor (210); and a storage (220) for storing one or more programs; When the one or more programs are executed by the processor (210), the processor executes the method for evaluating the health status of an energy storage battery in a random charging scenario as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed by the processor (210), the method for evaluating the health status of an energy storage battery in a random charging scenario according to any one of claims 1 to 7 is implemented.

Images