Method and device for determining health degree of battery

Abstract

The invention relates to a method and device for determining the health degree of a battery, and the method comprises the steps: obtaining a first feature parameter of a detected battery; wherein the first characteristic parameters comprise battery voltage values at a plurality of sampling moments, current values at the plurality of sampling moments and battery temperature values at the plurality of sampling moments in the charging process of the detected battery; inputting a first characteristic parameter of the detected battery into the trained characteristic generation model, and outputting a second characteristic parameter of the detected battery; wherein the second characteristic parameter of the detected battery is a parameter associated with battery voltage values at a plurality of sampling moments, current values at the plurality of sampling moments and battery temperature values at the plurality of sampling moments; and inputting the second characteristic parameter of the detected battery into the trained health degree prediction model, and outputting the health degree of the detected battery.

Description

Technical Field

[0001] The present disclosure relates to battery detection technology, and more specifically, to a method and device for determining battery health. Background Art

[0002] The explosive growth of electric vehicles has led to the large-scale use of power batteries. Batteries used in automobiles have high requirements for available energy and power. With the use of electric vehicles, power batteries will gradually age, which will lead to problems such as reduced mileage and fail to provide users with the best driving experience. Therefore, if the state of health (SOH) of power batteries can be accurately estimated, it can not only help automakers to effectively manage and improve power batteries, but also provide users with reference opinions on power battery maintenance or replacement, thereby improving user experience.

[0003] Therefore, it is very important to develop a method that can quickly detect the health of batteries. Summary of the invention

[0004] An object of the present invention is to provide a new technical solution for a method of determining battery health.

[0005] According to a first aspect of the present invention, a method for determining battery health is provided, comprising:

[0006] Acquire a first characteristic parameter of the detected battery; wherein the first characteristic parameter includes a battery voltage value at multiple sampling moments, a current value at multiple sampling moments, and a battery temperature value at multiple sampling moments of the detected battery during charging;

[0007] Inputting the first characteristic parameter of the detected battery into the trained feature generation model, and outputting the second characteristic parameter of the detected battery; wherein the second characteristic parameter of the detected battery is a parameter associated with the battery voltage values ​​at the multiple sampling moments, the current values ​​at the multiple sampling moments, and the battery temperature values ​​at the multiple sampling moments;

[0008] The second characteristic parameter of the detected battery is input into the trained health prediction model, and the health of the detected battery is output.

[0009] Optionally, before inputting the first characteristic parameter of the detected battery into the trained feature generation model and outputting the second characteristic parameter of the detected battery, the method further includes:

[0010] Screening the first characteristic parameter of the detected battery to obtain an abnormal value in the first characteristic parameter of the detected battery;

[0011] Performing a replacement process on the abnormal value in the first characteristic parameter of the detected battery to obtain a processed first characteristic parameter of the detected battery;

[0012] The step of inputting the first characteristic parameter of the detected battery into the trained feature generation model and outputting the second characteristic parameter of the detected battery includes:

[0013] The processed first characteristic parameter of the detected battery is input into the trained characteristic generation model, and the second characteristic parameter of the detected battery is output.

[0014] Optionally, before inputting the first characteristic parameter of the detected battery into the trained feature generation model and outputting the second characteristic parameter of the detected battery, the method further includes:

[0015] Performing Fourier transform processing on each parameter of the first characteristic parameter of the detected battery to obtain a frequency domain signal characteristic value corresponding to each parameter;

[0016] Performing inverse Fourier transform processing on the frequency domain signal characteristic values ​​corresponding to the parameters to obtain a time domain signal characteristic value associated with the first characteristic parameter of the detected battery, and using the time domain signal characteristic value as the third characteristic parameter of the detected battery;

[0017] The step of inputting the first characteristic parameter of the detected battery into the trained feature generation model and outputting the second characteristic parameter of the detected battery includes:

[0018] The first characteristic parameter of the detected battery and the third characteristic parameter of the detected battery are input into the trained feature generation model, and the second characteristic parameter of the detected battery is output.

[0019] Optionally, before inputting the first characteristic parameter of the detected battery into the trained feature generation model and outputting the second characteristic parameter of the detected battery, the method further includes:

[0020] Extracting statistical characteristic values ​​based on the first characteristic parameter of the detected battery; wherein the statistical characteristic values ​​include at least one of a maximum value, a minimum value, a mean value, a standard deviation, a variance, a median value, and a peak area value;

[0021] Merging the statistical characteristic value with the first characteristic parameter of the detected battery and the third characteristic parameter of the detected battery respectively to obtain the merged first characteristic parameter and the merged third characteristic parameter of the detected battery;

[0022] The first characteristic parameter of the detected battery and the third characteristic parameter of the detected battery are input into the trained feature generation model, and the second characteristic parameter of the detected battery is output, including:

[0023] The combined first characteristic parameter and the combined third characteristic parameter of the detected battery are input into the trained feature generation model, and the second characteristic parameter of the detected battery is output.

[0024] Optionally, extracting a statistical characteristic value based on the first characteristic parameter of the detected battery includes:

[0025] Generate a voltage value-time relationship curve, a current value-time relationship curve, and a temperature value-time relationship curve based on the battery voltage values ​​at the multiple sampling moments, the current values ​​at the multiple sampling moments, and the battery temperature values ​​at the multiple sampling moments, respectively;

[0026] Respectively performing smoothing and denoising processing on the relationship curve between the voltage value and time, the relationship curve between the current value and time, and the relationship curve between the temperature value and time to obtain a smoothed relationship curve between the voltage value and time, a smoothed relationship curve between the current value and time, and a smoothed relationship curve between the temperature value and time;

[0027] Corresponding statistical characteristic values ​​are extracted from the smoothed voltage value and time relationship curve, the smoothed current value and time relationship curve, and the smoothed temperature value and time relationship curve, respectively.

[0028] Optionally, before inputting the first characteristic parameter of the detected battery into the trained feature generation model and outputting the second characteristic parameter of the detected battery, the method further includes:

[0029] Using the first characteristic parameter of the sample battery in the training set as the first input data, inputting it into the feature generation model to be trained, and obtaining the first output data;

[0030] Obtaining a first loss function value according to the first input data and the first output data;

[0031] Adjust the parameters of the feature generation model to be trained according to the first loss function value until the first loss function value is less than a first preset threshold, stop training the feature generation model to be trained, and obtain the trained feature generation model.

[0032] Optionally, before inputting the first characteristic parameter of the detected battery into the trained feature generation model and outputting the second characteristic parameter of the detected battery, the method further includes:

[0033] Performing Fourier transform processing on each parameter in the first characteristic parameter of the sample battery in the training set to obtain the frequency domain signal characteristic value corresponding to each parameter;

[0034] Performing inverse Fourier transform processing on the frequency domain signal characteristic values ​​corresponding to the parameters to obtain the time domain signal characteristic value associated with the first characteristic parameter of the sample battery in the training set, and using the time domain signal characteristic value as the third characteristic parameter of the sample battery in the training set;

[0035] The first characteristic parameter and the third characteristic parameter of the sample battery in the training set are used as the second input data, and input into the feature generation model to be trained to obtain the second output data;

[0036] Obtaining a second loss function value according to the second input data and the second output data;

[0037] Adjust the parameters of the feature generation model to be trained according to the second loss function value until the second loss function value is less than the first preset threshold, stop training the feature generation model to be trained, and obtain the trained feature generation model.

[0038] Optionally, before the first characteristic parameter and the third characteristic parameter of the sample battery in the training set are used as the second input data and input into the feature generation model to be trained to obtain the second output data, the method further includes:

[0039] Extracting statistical characteristic values ​​based on the first characteristic parameters of the sample batteries in the training set; wherein the statistical characteristic values ​​include at least one of the maximum value, the minimum value, the mean value, the standard deviation, the variance, the median value, and the peak area value;

[0040] Merging the statistical characteristic value with the first characteristic parameter of the sample battery in the training set and the third characteristic parameter of the sample battery in the training set respectively to obtain the merged first characteristic parameter and the merged third characteristic parameter of the sample battery in the training set;

[0041] The method of using the first characteristic parameter and the third characteristic parameter of the sample battery in the training set as the second input data and inputting the feature generation model to be trained to obtain the second output data includes:

[0042] The combined first characteristic parameter and the combined third characteristic parameter of the sample battery in the training set are used as second input data and input into the feature generation model to be trained to obtain second output data.

[0043] Optionally, before inputting the second characteristic parameter of the detected battery into the trained health prediction model and outputting the health of the detected battery, the method further includes:

[0044] Inputting the second characteristic parameter of the sample battery in the training set into the health prediction model to be trained, and outputting the predicted value of the health of the sample battery;

[0045] Obtaining a third loss function value according to an actual value of the health of the sample battery in the training set and a predicted value of the health of the sample battery;

[0046] Adjusting the parameters of the feature generation model to be trained according to the third loss function value until the third loss function value is less than a second preset threshold, stopping the training of the health prediction model to be trained, and obtaining a trained health prediction model;

[0047] Using the second characteristic parameter of the sample battery in the test set to verify whether the trained health prediction model meets the requirements;

[0048] When the trained health prediction model meets the requirements, the second characteristic parameter of the detected battery is input into the trained health prediction model, and the health of the detected battery is output.

[0049] According to a second aspect of the present invention, there is provided a device for determining battery health, comprising:

[0050] An acquisition module, used to acquire a first characteristic parameter of a detected battery; wherein the first characteristic parameter includes a battery voltage value at multiple sampling moments, a current value at multiple sampling moments, and a battery temperature value at multiple sampling moments of the detected battery during charging;

[0051] a first output module, configured to input the first characteristic parameter of the detected battery into the trained characteristic generation model, and output a second characteristic parameter of the detected battery; wherein the second characteristic parameter of the detected battery is a parameter associated with the battery voltage values ​​at the multiple sampling moments, the current values ​​at the multiple sampling moments, and the battery temperature values ​​at the multiple sampling moments;

[0052] The second output module is used to input the second characteristic parameter of the detected battery into the trained health prediction model to output the health of the detected battery.

[0053] According to a third aspect of the present invention, a device for determining a battery health is provided, comprising a memory and a processor, wherein the memory stores a computer program, and the computer program is used to control the processor to operate to execute a method for determining a battery health according to any one of the first aspect of the present invention.

[0054] The method for determining the battery health provided by the present invention, the second characteristic parameter of the detected battery output by the trained feature generation model can capture the potential connection between the battery voltage value, current value and battery temperature value in the first characteristic parameter while retaining the original information of the first characteristic parameter of the detected battery, and then input the second characteristic parameter of the detected battery into the trained health prediction model to obtain the health of the detected battery, thereby improving the prediction accuracy of the battery health.

[0055] Features and advantages of the embodiments of the present specification will become apparent from the following detailed description of exemplary embodiments of the present specification with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate embodiments of the specification and, together with the description, serve to explain the principles of the embodiments of the specification.

[0057] Figure 1 is a processing flow chart of a method for determining battery health according to an embodiment of the present invention.

[0058] Figure 2 A schematic diagram of a training process flow of a feature generation model to be trained according to an embodiment of the present invention is shown.

[0059] Figure 3 A schematic diagram of a training process flow of a feature generation model to be trained according to an embodiment of the present invention is shown.

[0060] Figure 4 A schematic diagram of the construction of the VAE model is shown.

[0061] Figure 5 A schematic diagram of a training process flow of a health prediction model to be trained according to an embodiment of the present invention is shown.

[0062] Figure 6 A schematic diagram of the construction of the GRU model is shown.

[0063] Figure 7 A schematic diagram of the structure of each GRU unit of the GRU model is shown.

[0064] Figure 8 It is a principle block diagram of a device for determining battery health according to an embodiment of the present invention.

[0065] Fig. 9 4 is a schematic diagram of the hardware structure of a device for determining battery health according to an embodiment of the present invention. DETAILED DESCRIPTION

[0066] Various exemplary embodiments of the present specification will now be described in detail with reference to the accompanying drawings.

[0067] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the embodiments of the present specification and its application or uses.

[0068] It should be noted that like reference numerals and letters refer to similar items in the following figures, and therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0069] <Method Example>

[0070] In this embodiment, a method for determining battery health is provided. Figure 1 As shown, the method for determining the battery health of this embodiment may include the following S110 to S130.

[0071] S110, obtaining a first characteristic parameter of the detected battery; wherein the first characteristic parameter includes a battery voltage value, a current value and a battery temperature value at multiple sampling moments during the charging process of the detected battery.

[0072] The battery under test may be a battery pack including multiple single cells, or a single cell.

[0073] When the detected battery is a battery pack, the first characteristic parameter includes the battery pack voltage value at multiple sampling moments during the charging process, the current value at multiple sampling moments, and the battery pack temperature value at multiple sampling moments.

[0074] When the detected battery is a single cell, the first characteristic parameter includes a single cell voltage value at multiple sampling moments during the charging process of the single cell, a current value at multiple sampling moments, and a single cell temperature value at multiple sampling moments.

[0075] The detected battery may be a battery installed in an electric vehicle. The first characteristic parameter of the detected battery may be obtained from data reported by a battery management system (BMS) of the electric vehicle.

[0076] S120, input the first characteristic parameter of the detected battery into the trained feature generation model, and output the second characteristic parameter of the detected battery; wherein the second characteristic parameter of the detected battery is a parameter associated with the battery voltage values ​​at multiple sampling moments, the current values ​​at multiple sampling moments, and the battery temperature values ​​at multiple sampling moments.

[0077] The second characteristic parameter of the detected battery is a characteristic parameter reconstructed by using the trained feature generation model, which reflects the potential connection between the battery voltage values ​​at multiple sampling moments, the current values ​​at multiple sampling moments and the battery temperature values ​​at multiple sampling moments of the detected battery.

[0078] S130, inputting the second characteristic parameter of the detected battery into the trained health prediction model, and outputting the health of the detected battery.

[0079] The health prediction model reflects the corresponding relationship between the second characteristic parameter of the detected battery and the health of the detected battery.

[0080] The method for determining the battery health provided in an embodiment of the present invention is a method for determining the battery health. The second characteristic parameter of the detected battery output by the trained feature generation model can capture the potential connection between the battery voltage value, current value and battery temperature value in the first characteristic parameter while retaining the original information of the first characteristic parameter of the detected battery. The second characteristic parameter of the detected battery is then input into the trained health prediction model to obtain the health of the detected battery, thereby improving the prediction accuracy of the battery health.

[0081] In one embodiment, before S120, the method for determining the battery health further includes S140 to S150.

[0082] S140, screening the first characteristic parameters of the detected battery to obtain abnormal values ​​in the first characteristic parameters of the detected battery.

[0083] S150, performing replacement processing on the abnormal value in the first characteristic parameter of the detected battery to obtain the processed first characteristic parameter of the detected battery.

[0084] The abnormal value in the first characteristic parameter of the detected battery is a missing value or a value out of range. The abnormal value in the first characteristic parameter of the detected battery can be replaced by an adjacent value or an average value to obtain the processed first characteristic parameter of the detected battery.

[0085] In this embodiment, the processed first characteristic parameter of the detected battery is input into the trained characteristic generation model, and the second characteristic parameter of the detected battery is output.

[0086] In one embodiment, before S120, the method for determining the battery health further includes S160-S170.

[0087] S160, performing Fourier transform processing on each parameter in the first characteristic parameter of the detected battery to obtain a frequency domain signal characteristic value corresponding to each parameter.

[0088] S170, performing inverse Fourier transform processing on the frequency domain signal characteristic values ​​corresponding to each parameter to obtain a time domain signal characteristic value associated with the first characteristic parameter of the detected battery, and using the time domain signal characteristic value as the third characteristic parameter of the detected battery.

[0089] The input data of the Fourier transform processing may be the first characteristic parameter of the detected battery obtained directly, or may be the first characteristic parameter of the detected battery processed according to S140 to S150.

[0090] Based on the Fourier transform method, the Fast Fourier Transform (FFT) is used to map each parameter in the first characteristic parameter of the detected battery into a frequency domain signal characteristic value x(k).

[0091]

[0092] Wherein, N is the number of sampling moments, x(n) is the battery voltage value, current value or battery temperature value at a specific sampling moment in the first characteristic parameter, W is the frequency, n is the ranking number of the specific sampling moment in all sampling moments, and j is the rotation factor.

[0093] Then, the frequency domain signal characteristic value is mapped to the time domain signal characteristic value X(k) associated with the first characteristic parameter of the detected battery by using the Inverse Fast Fourier Transform (IFFT).

[0094]

[0095] Wherein, N is the number of sampling moments, x(k) is the frequency domain signal eigenvalue, W is the frequency, n is the ranking number of the sampling moment corresponding to the frequency domain signal eigenvalue among all sampling moments, and j is the rotation factor.

[0096] Specifically, using calculation formula (1), the battery voltage values ​​of the detected battery at multiple sampling moments during the charging process are respectively subjected to FFT to obtain the frequency domain signal characteristic values ​​corresponding to the battery voltage values ​​at each sampling moment. Then, using calculation formula (2), the frequency domain signal characteristic values ​​corresponding to the battery voltage values ​​at each sampling moment are respectively subjected to IFFT to obtain the corresponding time domain signal characteristic values. The obtained time domain signal characteristic values ​​can reflect the overall trend of the battery voltage values ​​at multiple sampling moments.

[0097] Using calculation formula (1), the current of the detected battery at multiple sampling moments during the charging process is subjected to FFT respectively, and the frequency domain signal characteristic value corresponding to the current value at each sampling moment is obtained. Then, using calculation formula (2), the frequency domain signal characteristic value corresponding to the current value at each sampling moment is subjected to IFFT respectively, and the corresponding time domain signal characteristic value is obtained. The obtained time domain signal characteristic value can reflect the overall trend of the current value at multiple sampling moments.

[0098] Using calculation formula (1), the battery temperature values ​​of the detected battery at multiple sampling moments during the charging process are respectively subjected to FFT to obtain the frequency domain signal characteristic values ​​corresponding to the battery temperature values ​​at each sampling moment. Then, using calculation formula (2), the frequency domain signal characteristic values ​​corresponding to the battery temperature values ​​at each sampling moment are respectively subjected to IFFT to obtain the corresponding time domain signal characteristic values. The obtained time domain signal characteristic values ​​can reflect the overall trend of the battery temperature values ​​at multiple sampling moments.

[0099] In this embodiment, after obtaining the second characteristic parameter of the detected battery, the first characteristic parameter of the detected battery and the third characteristic parameter of the detected battery are input into the trained characteristic generation model, and the second characteristic parameter of the detected battery is output. The second characteristic parameter of the detected battery is a characteristic parameter reconstructed by using the trained characteristic generation model, reflecting the potential connection between the battery voltage value at multiple sampling moments, the current value at multiple sampling moments, the battery temperature value at multiple sampling moments, the time domain signal characteristic value corresponding to the battery voltage value at each sampling moment, the time domain signal characteristic value corresponding to the current value at each sampling moment, and the time domain signal characteristic value corresponding to the battery temperature value at each sampling moment.

[0100] In one embodiment, before S120, the method for determining the battery health further includes S180-S190.

[0101] S180, extracting statistical characteristic values ​​based on the first characteristic parameter of the detected battery. The statistical characteristic values ​​include at least one of maximum value, minimum value, mean value, standard deviation, variance, median, and peak area value.

[0102] S190, merging the statistical characteristic value with the first characteristic parameter of the detected battery and the third characteristic parameter of the detected battery respectively to obtain a merged first characteristic parameter and a merged third characteristic parameter.

[0103] The statistical characteristic value may be extracted based on the first characteristic parameter of the detected battery directly obtained, or may be extracted based on the first characteristic parameter of the detected battery processed according to S140 to S150.

[0104] Based on the first characteristic parameter of the detected battery, extracting the statistical characteristic value specifically includes the following S181 to S183.

[0105] S181, based on the battery voltage values ​​at multiple sampling moments, the current values ​​at multiple sampling moments, and the battery temperature values ​​at multiple sampling moments, respectively generate a voltage value and time relationship curve, a current value and time relationship curve, and a temperature value and time relationship curve.

[0106] S182, respectively perform smoothing and denoising processing on the relationship curve between the voltage value and the time, the relationship curve between the current value and the time, and the relationship curve between the temperature value and the time to obtain a smoothed relationship curve between the voltage value and the time, a smoothed relationship curve between the current value and the time, and a smoothed relationship curve between the temperature value and the time.

[0107] The above curve can be processed by any of the following smoothing and denoising algorithms: sliding average algorithm, Savitzky-Golay convolution smoothing algorithm.

[0108] The sliding average algorithm can be implemented using the convolution function conv() or the filter function.

[0109] The Savitzky-Golay convolution smoothing algorithm first sets the width of the filter window to N = 2m + 1, and each measurement point is x = (-m, -m + 1, ..., 0, 1, m-1, m). The data in the window are fitted using a K-1 degree polynomial. Based on the K-1 degree equation, a system of N equations consisting of K-variable linear equations is formed, and finally the smoothed output value is solved using the least squares method.

[0110] S183, extracting corresponding statistical characteristic values ​​from the smoothed voltage value and time relationship curve, the smoothed current value and time relationship curve, and the smoothed temperature value and time relationship curve, respectively.

[0111] The statistical characteristic value extracted from the smoothed voltage value and time relationship curve is recorded as the first statistical characteristic value. The statistical characteristic value extracted from the smoothed current value and time relationship curve is recorded as the second statistical characteristic value. The statistical characteristic value extracted from the smoothed temperature value and time relationship curve is recorded as the third statistical characteristic value.

[0112] The first statistical characteristic value is combined with the battery voltage value at multiple sampling moments of the detected battery during the charging process to obtain the first combined data. The second statistical characteristic value is combined with the current value at multiple sampling moments of the detected battery during the charging process to obtain the second combined data. The third statistical characteristic value is combined with the current value at multiple sampling moments of the detected battery during the charging process to obtain the third combined data. The first combined data, the second combined data and the third combined data constitute the combined first characteristic parameter.

[0113] The first statistical characteristic value is combined with the time domain signal characteristic value corresponding to the battery voltage value at each sampling moment to obtain the fourth combined data. The second statistical characteristic value is combined with the time domain signal characteristic value corresponding to the current value at each sampling moment to obtain the fifth combined data. The third statistical characteristic value is combined with the time domain signal characteristic value corresponding to the battery temperature value at each sampling moment to obtain the sixth combined data. The fourth combined data, the fifth combined data and the sixth combined data constitute the combined third characteristic parameter.

[0114] In this embodiment, the combined first characteristic parameter is input into the trained characteristic generation model, and the second characteristic parameter of the detected battery is output, or the combined first characteristic parameter and the combined third characteristic parameter are input into the trained characteristic generation model, and the second characteristic parameter of the detected battery is output. The second characteristic parameter of the detected battery is a characteristic parameter reconstructed by using the trained characteristic generation model, reflecting the potential connection between the above six combined data.

[0115] It should be noted that the input data of the trained feature generation model is first dedimensionalized and then converted into data that obeys the standard normal distribution before being input into the model.

[0116] In one embodiment, the feature generation model to be trained can be trained according to the following S201 to S203. Figure 2 .

[0117] S201, taking the first characteristic parameter of the sample battery in the training set as the first input data, inputting it into the feature generation model to be trained, and obtaining the first output data.

[0118] The feature generation model to be trained can be any of the following: AE (Auto-Encoder) model, SAE (Stacked Denoising Auto-Encoders) model, DAE (Denoising Auto-Encoders) model, VAE (Variational Autoencoder) model and Transformer model.

[0119] S202: Obtain a first loss function value according to the first input data and the first output data.

[0120] S203, adjusting the parameters of the feature generation model to be trained according to the first loss function value until the first loss function value is less than a first preset threshold, stopping the training of the feature generation model to be trained, and obtaining the trained feature generation model.

[0121] In another embodiment, the feature generation model to be trained can be trained according to the following S301 to S305. Figure 3 .

[0122] S301, performing Fourier transform processing on each parameter in the first characteristic parameter of the sample battery in the training set to obtain the frequency domain signal characteristic value corresponding to each parameter.

[0123] S302, perform inverse Fourier transform processing on the frequency domain signal characteristic values ​​corresponding to each parameter to obtain the time domain signal characteristic values ​​associated with the first characteristic parameters of the sample batteries in the training set, and use the time domain signal characteristic values ​​as the third characteristic parameters of the sample batteries in the training set.

[0124] S303, taking the first characteristic parameter of the sample battery in the training set and the third characteristic parameter of the sample battery in the training set as second input data, inputting into the feature generation model to be trained, and obtaining second output data.

[0125] S304: Obtain a second loss function value according to the second input data and the second output data.

[0126] S305, adjusting the parameters of the feature generation model to be trained according to the second loss function value until the second loss function value is less than the first preset threshold, stopping the training of the feature generation model to be trained, and obtaining the trained feature generation model.

[0127] In another embodiment, before executing the above S303, a statistical characteristic value is extracted based on the first characteristic parameter of the sample battery in the training set. The statistical characteristic value includes at least one of the maximum value, the minimum value, the mean value, the standard deviation, the variance, the median, and the peak area value. The statistical characteristic value is respectively merged with the first characteristic parameter of the sample battery in the training set and the third characteristic parameter of the sample battery in the training set to obtain the merged first characteristic parameter of the sample battery in the training set and the merged third characteristic parameter of the sample battery in the training set. Then, the merged first characteristic parameter of the sample battery in the training set and the merged third characteristic parameter of the sample battery in the training set are used as the second input data to execute S303.

[0128] In another embodiment, before executing the above S301, the first characteristic parameters of the sample batteries in the training set are screened to obtain abnormal values ​​in the first characteristic parameters of the sample batteries in the training set. Then, the abnormal values ​​in the first characteristic parameters of the sample batteries in the training set are replaced to obtain the first characteristic parameters of the processed sample batteries. S301 is executed based on the first characteristic parameters of the processed sample batteries.

[0129] The replacement processing method of the outlier values ​​in the first characteristic parameter of the sample battery in the training set can refer to the replacement processing method of the outlier values ​​in the first characteristic parameter of the detected battery.

[0130] Taking the VAE model as an example, the VAE model is trained using the training set.

[0131] The data in the training set is preprocessed before being input into the VAE model. Preprocessing includes de-dimensionalizing the data in the training set and converting it into data that obeys the standard normal distribution. The preprocessed data in the training set is used as the input data of the VAE model to be trained.

[0132] Figure 4 The schematic diagram of the construction of the VAE model is shown. The principle of the VAE model is as follows: the data X in the training set is input into the VAE model, mapped to the distribution of the latent space through the encoder, resampled and input into the decoder to calculate the reconstruction error, back-propagated according to the reconstruction error, and the parameters of the VAE model are updated using the optimization algorithm, and finally the reconstructed latent feature data X' is output.

[0133] Specifically, the preprocessed data in the training set is input into the VAE model to generate the features of the latent space. Among them, the Encoder of the VAE model is composed of 3 layers of fully connected layers and relu activation function, and the Decoder is composed of 3 layers of fully connected layers, relu activation function and sigmoid activation function.

[0134] The Encoder transforms the feature information carried by the input data into a Gaussian-like distribution, also known as an implicit distribution. The Encoder outputs the mean matrix and standard deviation matrix of the Gaussian-like distribution.

[0135] The Encoder does not directly generate a Gaussian distribution, but randomly generates a mean and standard deviation. In order to make the actual output distribution close to the target distribution, KL divergence is used as the loss function of the Encoder.

[0136]

[0137] Among them, K is the number of groups of mean and standard deviation generated by a sample, σ j is the standard deviation of the jth group, μ j is the expected value of the jth group.

[0138] Based on each group of mean and standard deviation, a Gaussian distribution is constructed in the latent space, which is the target distribution.

[0139] Resampling is performed based on the target distribution, randomly extracting a high-dimensional implicit representation matrix Z, and inputting the high-dimensional implicit representation matrix Z into the Decoder.

[0140] The Decoder decodes based on the implicit representation matrix Z and outputs data with higher feature dimensions than the original input data as the output data of the VAE model. MSE (Mean Squred Error) can be used as the loss function of the Decoder.

[0141]

[0142] Among them, m is the number of input data, that is, the number of samples, x i is the i-th input data, is the ith output data.

[0143] The loss function of the VAE model is as follows:

[0144]

[0145] Based on the loss function of the VAE model, the VAE model is back-propagated, and then the parameters of the VAE model are adjusted using an optimization algorithm until the loss function value of the VAE model meets the requirements. The optimization algorithm may be an ADAM (adaptive moment estimation) optimization algorithm.

[0146] In one embodiment, the health prediction model to be trained can be trained and tested according to the following S501 to S505. Figure 5 .

[0147] S501, input the second characteristic parameter of the sample battery in the training set into the health prediction model to be trained, and output the predicted value of the health of the sample battery. The second characteristic parameter of the sample battery in the training set is the output data of the trained feature generation model.

[0148] The health prediction model can be any one of the following: a GRU (Gated Recurrent Unit) model, a network model of a bidirectional LSTM (Long Short-Term Memory) layer, or a BP (Back Propagation) neural network.

[0149] S502, obtaining a third loss function value according to the actual value of the health of the sample batteries in the training set and the predicted value of the health of the sample batteries.

[0150] S503, adjusting the parameters of the feature generation model to be trained according to the third loss function value until the third loss function value is less than the second preset threshold, stopping the training of the health prediction model to be trained, and obtaining the trained health prediction model.

[0151] S504, using the second characteristic parameters of the sample battery in the test set to verify whether the trained health prediction model meets the requirements. The second characteristic parameters of the sample battery in the test set are output data of the trained feature generation model.

[0152] Specifically, verify whether the accuracy and generalization of the trained health prediction model meet the requirements.

[0153] S505, when the trained health prediction model meets the requirements, the second characteristic parameter of the detected battery is input into the trained health prediction model, and the health of the detected battery is output.

[0154] If the trained health prediction model does not meet the requirements, the parameters of the health prediction model to be trained are adjusted, and the training is continued until the trained health prediction model meets the requirements.

[0155] Taking the GRU model as an example, the training set is used to train and test the GRU model.

[0156] See also Figure 6 The GRU model includes an input layer (Input Layer), two GRU hidden layers (Hidden-layer1, Hidden-layer2), a discard layer ( Figure 6 ), a fully connected layer (FC Layer), an output layer (Output Layer), and set model parameters. Among them, the dropout layer is set to a dropout probability of 0.6.

[0157] Based on the GRU hidden layer, each GRU unit contains a reset gate, an update gate, a candidate hidden state, an input (Input X(t)) and an output (H(t)), refer to Figure 7 .

[0158] Based on the reset gate, the calculation method is:

[0159] r t =Sigmoid(W r ·[H t-1 , x t ])-Calculation formula (6),

[0160] Among them, W r is the weight of the gate, h t-1 is the hidden state at the previous moment, x t is the input at the current moment, and Sigmoid is the gating function.

[0161] Based on the update gate, the calculation method is:

[0162] z t =Sigmoid(W z ·[H t-1 , x t ])-Calculation formula (7),

[0163] Among them, W z is the weight of the gate, h t-1 is the hidden state at the previous moment, x t is the input at the current moment, and Sigmoid is the gating function.

[0164] Based on the candidate hidden state, the calculation method is:

[0165] H' t =Tanh(W xh x t +(r t ⊙H t-1 )W hh )-Calculation formula (8),

[0166] Among them, W xh X(t) and H' t The weight between hh H' t With H t-1 The weight between t-1 is the hidden state at the previous moment, x t is the input at the current moment, ⊙ is the logical operation symbol, and Tanh is the activation function.

[0167] The output H(t) is calculated as:

[0168] H t =(1-z t )*H t-1 +z t *H' t -Calculate formula (9).

[0169] The training set is input into the GRU model for training, the loss function is used to perform back propagation on the GRU model, and then the optimization algorithm is used to adjust the parameters of the GRU model until the loss function value of the GRU model meets the requirements. The optimization algorithm may be the NADAM optimization algorithm. The NAdam optimization algorithm is a combination of the Nesterov accelerated gradient (NAG) algorithm and the ADAM algorithm.

[0170] The loss function of the GRU model can be any of the following: mean square error MSE, mean absolute percentage error MAPE, goodness of fit R 2 .

[0171] MSE can predict the accuracy of the model. The smaller the MSE value, the more accurate the fitting degree. The calculation method is:

[0172]

[0173] Where m is the number of predicted values, is the predicted value, y i is the actual value.

[0174] MAPE can predict the accuracy of the model. The smaller the MAPE value, the more accurate the fitting degree. The calculation method is:

[0175]

[0176] Where m is the number of predicted values, is the predicted value, y i is the actual value.

[0177] R 2 The closer the value is to 1, the better the fit is. The calculation method is:

[0178]

[0179] Where m is the number of predicted values, is the predicted value, y i is the actual value, is the average of all measured values.

[0180] Use the test set to evaluate the trained GRU model. Determine whether the accuracy and generalization of the trained GRU model meet the requirements. If the accuracy and generalization of the trained GRU model do not meet the requirements, adjust the parameters of the GRU model to be trained and continue training until the accuracy and generalization of the trained GRU model meet the requirements.

[0181] <Device Example>

[0182] An embodiment of the present invention provides a device for determining battery health, such as Figure 8 The battery health determination device 800 includes an acquisition module 810 , a first output module 820 and a second output module 830 .

[0183] The acquisition module 810 is used to acquire the first characteristic parameter of the detected battery; wherein the first characteristic parameter includes the battery voltage value, the current value and the battery temperature value at multiple sampling moments during the charging process of the detected battery.

[0184] The first output module 820 is used to input the first characteristic parameter of the detected battery into the trained feature generation model, and output the second characteristic parameter of the detected battery; wherein the second characteristic parameter of the detected battery is a parameter associated with the battery voltage values ​​at multiple sampling moments, the current values ​​at multiple sampling moments, and the battery temperature values ​​at multiple sampling moments.

[0185] The second output module 830 is used to input the second characteristic parameter of the detected battery into the trained health prediction model, and output the health of the detected battery.

[0186] The health prediction model reflects the corresponding relationship between the second characteristic parameter of the detected battery and the health of the detected battery.

[0187] In one embodiment of the present invention, the device for determining the battery health further includes a screening module and a replacement processing module.

[0188] The screening module is used to screen the first characteristic parameter of the detected battery to obtain abnormal values ​​in the first characteristic parameter of the detected battery.

[0189] The replacement processing module is used to perform replacement processing on the abnormal value in the first characteristic parameter of the detected battery to obtain the processed first characteristic parameter of the detected battery.

[0190] The first output module 820 is further used to input the processed first characteristic parameter of the detected battery into the trained characteristic generation model, and output the second characteristic parameter of the detected battery.

[0191] In one embodiment of the present invention, the device for determining the battery health further includes a Fourier transform processing module and an inverse Fourier transform processing module.

[0192] The Fourier transform processing module is used to perform Fourier transform processing on each parameter in the first characteristic parameter of the detected battery to obtain the frequency domain signal characteristic value corresponding to each parameter.

[0193] The Fourier transform processing module is used to perform inverse Fourier transform processing on the frequency domain signal characteristic values ​​corresponding to each parameter, obtain the time domain signal characteristic value associated with the first characteristic parameter of the detected battery, and use the time domain signal characteristic value as the third characteristic parameter of the detected battery.

[0194] The first output module 820 is further used to input the first characteristic parameter of the detected battery and the third characteristic parameter of the detected battery into the trained feature generation model, and output the second characteristic parameter of the detected battery.

[0195] In one embodiment of the present invention, the device for determining the battery health further includes an extraction module and a merging module.

[0196] The extraction module is used to extract statistical characteristic values ​​based on the first characteristic parameter of the detected battery; wherein the statistical characteristic values ​​include at least one of maximum value, minimum value, mean value, standard deviation, variance, median, and peak area value.

[0197] The merging module is used to merge the statistical characteristic value with the first characteristic parameter of the detected battery and the third characteristic parameter of the detected battery respectively, to obtain the merged first characteristic parameter and the merged third characteristic parameter of the detected battery.

[0198] The first output module 820 is further used to input the combined first characteristic parameter and the combined third characteristic parameter of the detected battery into the trained feature generation model, and output the second characteristic parameter of the detected battery.

[0199] In one embodiment of the present invention, the extraction module is also used to generate a voltage value and time relationship curve, a current value and time relationship curve, and a temperature value and time relationship curve based on the battery voltage values ​​at multiple sampling moments, the current values ​​at multiple sampling moments, and the battery temperature values ​​at multiple sampling moments; respectively perform smoothing and denoising processing on the voltage value and time relationship curve, the current value and time relationship curve, and the temperature value and time relationship curve to obtain a smoothed voltage value and time relationship curve, a smoothed current value and time relationship curve, and a smoothed temperature value and time relationship curve; respectively extract corresponding statistical feature values ​​from the smoothed voltage value and time relationship curve, the smoothed current value and time relationship curve, and the smoothed temperature value and time relationship curve.

[0200] In one embodiment of the present invention, the device for determining the battery health further includes a first input module, a first loss function value determination module and a first adjustment module.

[0201] The first input module is used to input the first characteristic parameter of the sample battery in the training set as the first input data into the feature generation model to be trained to obtain the first output data.

[0202] The first loss function value determination module is used to obtain a first loss function value according to the first input data and the first output data.

[0203] The first adjustment module is used to adjust the parameters of the feature generation model to be trained according to the first loss function value until the first loss function value is less than a first preset threshold, stop training the feature generation model to be trained, and obtain the trained feature generation model.

[0204] In one embodiment of the present invention, the device for determining the battery health further includes a second loss function value determination module.

[0205] The first input module is also used to use the merged first feature parameters and the merged third feature parameters of the sample batteries in the training set as second input data, input them into the feature generation model to be trained, and obtain second output data; wherein, the method for determining the merged first feature parameters and the merged third feature parameters of the sample batteries in the training set is the same as the method for determining the merged first feature parameters and the merged third feature parameters of the battery being tested.

[0206] The second loss function value determination module is also used to obtain a second loss function value based on the second input data and the second output data.

[0207] The first adjustment module is also used to adjust the parameters of the feature generation model to be trained according to the second loss function value until the second loss function value is less than the first preset threshold, stop training the feature generation model to be trained, and obtain the trained feature generation model.

[0208] In one embodiment of the present invention, the device for determining the battery health further includes a second input module, a third loss function value determination module, a second adjustment module and a second verification module.

[0209] The second input module is used to input the second characteristic parameter of the sample battery in the training set into the health prediction model to be trained, and output the predicted value of the health of the sample battery.

[0210] The third loss function value determination module is used to obtain a third loss function value according to the actual value of the health of the sample battery in the training set and the predicted value of the health of the sample battery.

[0211] The second adjustment module is used to adjust the parameters of the feature generation model to be trained according to the third loss function value until the third loss function value is less than the second preset threshold, stop the training of the health prediction model to be trained, and obtain the trained health prediction model.

[0212] An embodiment of the present invention provides a device for determining battery health, such as Fig. 9 As shown. The battery health determination device 900 includes a memory 920 and a processor 910. The memory 920 stores a computer program, and the computer program is used to control the processor 910 to operate to execute the battery health determination method in any of the above embodiments.

[0213] In one embodiment, the battery health determination device is arranged on the server side. The detected battery is a battery installed in an electric vehicle. The electric vehicle can be connected to the server for communication.

[0214] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. For the electric vehicle embodiment, its related parts can be referred to the partial description of the method embodiment.

[0215] The above is a description of a specific embodiment of the present specification. Other embodiments are within the scope of the appended claims. In some cases, the actions recorded in the claims may be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0216] The embodiments of the present specification may be systems, methods and / or computer program products. The computer program product may include a computer-readable storage medium carrying computer instructions for causing a processor to implement various aspects of the embodiments of the present specification.

[0217] A computer-readable storage medium may be a tangible device that can hold and store computer instructions used by a computer instruction execution device. A computer-readable storage medium may be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. More specific examples of computer-readable storage media (a non-exhaustive list) include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disk read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, such as a punch card or a raised structure in a groove on which computer instructions are stored, and any suitable combination of the foregoing. As used herein, a computer-readable storage medium is not to be interpreted as a transient signal per se, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagating through a waveguide or other transmission medium (e.g., a light pulse through a fiber optic cable), or an electrical signal transmitted through a wire.

[0218] The computer instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, optical fiber transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer instructions from the network and forwards the computer instructions for storage in the computer-readable storage medium in each computing / processing device.

[0219] The flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the systems, methods and computer program products according to the multiple embodiments of this specification. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of a computer instruction, and a part of a module, a program segment or a computer instruction contains one or more executable computer instructions for realizing the specified logical function. 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 the flowchart, and the combination of the boxes in the block diagram and / or the flowchart can be implemented by a dedicated hardware-based system that performs the specified function or action, or can be implemented by a combination of dedicated hardware and computer instructions. It is well known to those skilled in the art that it is equivalent to implement it by hardware, implement it by software, and implement it by combining software and hardware.

[0220] The embodiments of the present specification have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and variations are obvious to those of ordinary skill in the art without departing from the scope of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A method for determining battery health, characterized in that: include: Acquire a first characteristic parameter of the detected battery; wherein the first characteristic parameter includes a battery voltage value at multiple sampling moments, a current value at multiple sampling moments, and a battery temperature value at multiple sampling moments of the detected battery during charging; Inputting the first characteristic parameter of the detected battery into the trained feature generation model, and outputting the second characteristic parameter of the detected battery; wherein the second characteristic parameter of the detected battery is a parameter associated with the battery voltage values ​​at the multiple sampling moments, the current values ​​at the multiple sampling moments, and the battery temperature values ​​at the multiple sampling moments; The second characteristic parameter of the detected battery is input into the trained health prediction model, and the health of the detected battery is output.

2. The method according to claim 1, characterized in that Before inputting the first characteristic parameter of the detected battery into the trained feature generation model and outputting the second characteristic parameter of the detected battery, the method further includes: Screening the first characteristic parameter of the detected battery to obtain an abnormal value in the first characteristic parameter of the detected battery; Performing a replacement process on the abnormal value in the first characteristic parameter of the detected battery to obtain a processed first characteristic parameter of the detected battery; The step of inputting the first characteristic parameter of the detected battery into the trained feature generation model and outputting the second characteristic parameter of the detected battery includes: The processed first characteristic parameter of the detected battery is input into the trained characteristic generation model, and the second characteristic parameter of the detected battery is output.

3. The method according to claim 1, characterized in that Before inputting the first characteristic parameter of the detected battery into the trained feature generation model and outputting the second characteristic parameter of the detected battery, the method further includes: Performing Fourier transform processing on each parameter of the first characteristic parameter of the detected battery to obtain a frequency domain signal characteristic value corresponding to each parameter; Performing inverse Fourier transform processing on the frequency domain signal characteristic values ​​corresponding to the parameters to obtain a time domain signal characteristic value associated with the first characteristic parameter of the detected battery, and using the time domain signal characteristic value as the third characteristic parameter of the detected battery; The step of inputting the first characteristic parameter of the detected battery into the trained feature generation model and outputting the second characteristic parameter of the detected battery includes: The first characteristic parameter of the detected battery and the third characteristic parameter of the detected battery are input into the trained feature generation model, and the second characteristic parameter of the detected battery is output.

4. The method according to claim 3, characterized in that Before inputting the first characteristic parameter of the detected battery into the trained feature generation model and outputting the second characteristic parameter of the detected battery, the method further includes: Extracting statistical characteristic values ​​based on the first characteristic parameter of the detected battery; wherein the statistical characteristic values ​​include at least one of a maximum value, a minimum value, a mean value, a standard deviation, a variance, a median value, and a peak area value; Merging the statistical characteristic value with the first characteristic parameter of the detected battery and the third characteristic parameter of the detected battery respectively to obtain the merged first characteristic parameter and the merged third characteristic parameter of the detected battery; The first characteristic parameter of the detected battery and the third characteristic parameter of the detected battery are input into the trained feature generation model, and the second characteristic parameter of the detected battery is output, including: The combined first characteristic parameter and the combined third characteristic parameter of the detected battery are input into the trained feature generation model, and the second characteristic parameter of the detected battery is output.

5. The method according to claim 4, characterized in that The step of extracting a statistical characteristic value based on the first characteristic parameter of the detected battery includes: Generate a voltage value-time relationship curve, a current value-time relationship curve, and a temperature value-time relationship curve based on the battery voltage values ​​at the multiple sampling moments, the current values ​​at the multiple sampling moments, and the battery temperature values ​​at the multiple sampling moments, respectively; Respectively performing smoothing and denoising processing on the relationship curve between the voltage value and time, the relationship curve between the current value and time, and the relationship curve between the temperature value and time to obtain a smoothed relationship curve between the voltage value and time, a smoothed relationship curve between the current value and time, and a smoothed relationship curve between the temperature value and time; Corresponding statistical characteristic values ​​are extracted from the smoothed voltage value and time relationship curve, the smoothed current value and time relationship curve, and the smoothed temperature value and time relationship curve, respectively.

6. The method according to claim 1, characterized in that Before inputting the first characteristic parameter of the detected battery into the trained feature generation model and outputting the second characteristic parameter of the detected battery, the method further includes: Using the first characteristic parameter of the sample battery in the training set as the first input data, inputting it into the feature generation model to be trained, and obtaining the first output data; Obtaining a first loss function value according to the first input data and the first output data; Adjust the parameters of the feature generation model to be trained according to the first loss function value until the first loss function value is less than a first preset threshold, stop training the feature generation model to be trained, and obtain the trained feature generation model.

7. The method according to claim 1, characterized in that Before inputting the first characteristic parameter of the detected battery into the trained feature generation model and outputting the second characteristic parameter of the detected battery, the method further includes: Performing Fourier transform processing on each parameter in the first characteristic parameter of the sample battery in the training set to obtain the frequency domain signal characteristic value corresponding to each parameter; Performing inverse Fourier transform processing on the frequency domain signal characteristic values ​​corresponding to the parameters to obtain the time domain signal characteristic value associated with the first characteristic parameter of the sample battery in the training set, and using the time domain signal characteristic value as the third characteristic parameter of the sample battery in the training set; The first characteristic parameter and the third characteristic parameter of the sample battery in the training set are used as the second input data, and input into the feature generation model to be trained to obtain the second output data; Obtaining a second loss function value according to the second input data and the second output data; Adjust the parameters of the feature generation model to be trained according to the second loss function value until the second loss function value is less than the first preset threshold, stop training the feature generation model to be trained, and obtain the trained feature generation model.

8. The method according to claim 7, characterized in that Before the first characteristic parameter and the third characteristic parameter of the sample battery in the training set are used as the second input data and input into the feature generation model to be trained to obtain the second output data, the method further includes: Extracting statistical characteristic values ​​based on the first characteristic parameters of the sample batteries in the training set; wherein the statistical characteristic values ​​include at least one of the maximum value, the minimum value, the mean value, the standard deviation, the variance, the median value, and the peak area value; Merging the statistical characteristic value with the first characteristic parameter of the sample battery in the training set and the third characteristic parameter of the sample battery in the training set respectively to obtain the merged first characteristic parameter and the merged third characteristic parameter of the sample battery in the training set; The method of using the first characteristic parameter and the third characteristic parameter of the sample battery in the training set as the second input data and inputting the feature generation model to be trained to obtain the second output data includes: The combined first characteristic parameter and the combined third characteristic parameter of the sample battery in the training set are used as second input data and input into the feature generation model to be trained to obtain second output data.

9. The method according to any one of claims 6 to 8, characterized in that: Before inputting the second characteristic parameter of the detected battery into the trained health prediction model and outputting the health of the detected battery, the method further includes: Inputting the second characteristic parameter of the sample battery in the training set into the health prediction model to be trained, and outputting the predicted value of the health of the sample battery; Obtaining a third loss function value according to an actual value of the health of the sample battery in the training set and a predicted value of the health of the sample battery; Adjusting the parameters of the feature generation model to be trained according to the third loss function value until the third loss function value is less than a second preset threshold, stopping the training of the health prediction model to be trained, and obtaining a trained health prediction model; Using the second characteristic parameter of the sample battery in the test set to verify whether the trained health prediction model meets the requirements; When the trained health prediction model meets the requirements, the second characteristic parameter of the detected battery is input into the trained health prediction model, and the health of the detected battery is output.

10. A device for determining battery health, characterized in that: The device comprises: An acquisition module, used to acquire a first characteristic parameter of a detected battery; wherein the first characteristic parameter includes a battery voltage value at multiple sampling moments, a current value at multiple sampling moments, and a battery temperature value at multiple sampling moments of the detected battery during charging; a first output module, configured to input the first characteristic parameter of the detected battery into the trained characteristic generation model, and output a second characteristic parameter of the detected battery; wherein the second characteristic parameter of the detected battery is a parameter associated with the battery voltage values ​​at the multiple sampling moments, the current values ​​at the multiple sampling moments, and the battery temperature values ​​at the multiple sampling moments; The second output module is used to input the second characteristic parameter of the detected battery into the trained health prediction model to output the health of the detected battery.

11. A device for determining battery health, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program, and the computer program is used to control the processor to operate so as to execute the method according to any one of claims 1 to 9.

Images