A battery capacity life prediction method and device based on multi-data fusion

Abstract

The embodiment of the application provides a battery capacity life prediction method and device based on multi-data fusion, computer equipment and a storage medium. The application scheme extracts key features by using a self-adaptive feature extraction algorithm from monitoring data of lithium batteries under a large number of different charging rates, different discharging rates and different environmental temperature conditions, trains and optimizes a bidirectional long short-term recurrent neural network, and finally predicts the life of a lithium ion battery through the bidirectional long short-term recurrent neural network, thereby improving the prediction accuracy.

Description

Technical Field

[0001] This application relates to the field of battery manufacturing technology, and in particular to a method and apparatus for predicting battery capacity and lifespan based on multi-data fusion, as well as a computer device and storage medium. Background Technology

[0002] Lithium-ion batteries possess advantages such as high operating voltage, high energy density, and no memory effect, leading to their widespread application in new energy vehicles, energy storage systems, military equipment, and aerospace. Reliable prediction of lithium-ion battery lifespan is crucial for the safe operation of equipment using lithium-ion batteries. Due to the numerous factors influencing lithium-ion battery lifespan and the complex interactions between these factors, the complexity of the battery aging path is dramatically increased. Therefore, establishing lithium-ion battery capacity and lifespan prediction methods based on failure mechanisms is difficult to implement and suffers from poor accuracy over long periods. While data-driven lithium-ion battery capacity and lifespan prediction methods offer high accuracy and do not require complex internal chemical reaction models of lithium batteries, these methods suffer from drawbacks such as reliance on human experience for data features, limited feature sets, and inability to predict lifespan under changing operating conditions.

[0003] Therefore, there is an urgent need to propose a new method for predicting battery capacity and lifespan in order to solve the above problems. Summary of the Invention

[0004] In order to solve one or more of the technical problems existing in the prior art, this application provides a battery capacity life prediction method and device, computer equipment and storage medium based on multi-data fusion, so as to solve the problems of existing prediction methods relying on human experience for data features, single features, and inability to predict life after changes in operating conditions.

[0005] To achieve the above objectives, the technical solution adopted by this application to solve its technical problem is as follows:

[0006] In a first aspect, this application provides a battery capacity lifetime prediction method based on multi-data fusion, the method comprising:

[0007] The monitoring data of the battery to be predicted under multiple different test conditions is obtained. The monitoring data includes first correlation data and second correlation data. The first correlation data includes at least voltage and current, and the second correlation data includes at least temperature rise, discharge capacitance, test temperature, discharge rate and charge rate.

[0008] The first preset model is used to extract features from the first relevant data and normalize the second relevant data to obtain corresponding feature vectors. The first preset model is then used to fuse the feature vectors to obtain a first fused feature vector.

[0009] The remaining lifespan of the battery to be predicted is obtained using the second preset model based on the first fused feature vector.

[0010] In one specific embodiment, the method includes the training process of the first preset model, comprising:

[0011] Each sample battery was subjected to charge-discharge cycle tests under different test conditions to obtain test data and generate a training dataset. The test data included voltage, current, temperature rise, discharge capacitance, test temperature, discharge rate, and charge rate.

[0012] A first preset model is obtained by training a preset encoder using the voltage and current in the training dataset.

[0013] In a specific embodiment, the preset encoder is trained sequentially using the voltage or current in the training dataset, and the feature vector of the voltage or current obtained from the training is used as input to train the next preset encoder, until all the voltage data or current data in the training dataset have been trained, resulting in multiple encoders and multiple decoders;

[0014] The plurality of encoders and the plurality of decoders are stacked to form the first preset model.

[0015] In one specific embodiment, the first preset model includes a stacked denoising autoencoder model.

[0016] In one specific embodiment, the test data is time series data.

[0017] In one specific embodiment, the method further includes a training process for a second preset model, comprising:

[0018] Normalize the data in the training dataset other than the voltage and the current to obtain the normalization result, and generate a feature vector dataset based on the normalization result and the feature vectors of the voltage and the current.

[0019] A third preset model is obtained by training based on a preset encoder using the feature vector dataset, and the feature vectors in the feature vector dataset are fused using the third preset model to obtain a second fused feature vector.

[0020] The second fused feature vector is used to train a bidirectional long short-time recurrent neural network to obtain the second preset model.

[0021] In one specific embodiment, the third preset model includes a stacked denoising autoencoder model.

[0022] Secondly, this application also provides a battery capacity life prediction device based on multi-data fusion, the device comprising:

[0023] The data acquisition module is used to acquire monitoring data of the battery to be predicted under multiple different test conditions. The monitoring data includes first relevant data and second relevant data. The first relevant data includes at least voltage and current, and the second relevant data includes at least temperature rise, discharge capacitance, test temperature, discharge rate and charge rate.

[0024] The feature extraction module is used to extract features from the first related data using a first preset model and to normalize the second related data to obtain corresponding feature vectors, and to fuse the feature vectors using the first preset model to obtain a first fused feature vector.

[0025] The lifespan prediction module is used to obtain the remaining lifespan of the battery to be predicted based on the first fused feature vector using a second preset model.

[0026] Thirdly, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program that can run on the processor, and when the computer program is executed by the processor, the battery capacity life prediction method based on multi-data fusion is implemented.

[0027] Fourthly, a computer-readable storage medium is also provided, wherein a computer program is stored therein, and when the computer program is executed, the battery capacity and life prediction method based on multi-data fusion is implemented.

[0028] The beneficial effects of the technical solutions provided in this application are:

[0029] This application provides a battery capacity and lifespan prediction method, apparatus, computer device, and storage medium based on multi-data fusion. The method includes acquiring monitoring data of the battery under multiple different test conditions. The monitoring data includes first relevant data and second relevant data. The first relevant data includes at least voltage and current, and the second relevant data includes at least temperature rise, discharge capacitance, test temperature, discharge rate, and charge rate. A first preset model is used to extract features from the first relevant data and normalize the second relevant data to obtain corresponding feature vectors. The first preset model is then used to fuse the feature vectors to obtain a first fused feature vector. Based on the first fused feature vector, a second preset model is used to obtain the remaining lifespan of the battery under the prediction. This application improves the accuracy of lithium-ion battery lifespan prediction by extracting key features from a large amount of monitoring data of lithium batteries under different charge rates, discharge rates, and ambient temperatures during cycling using an adaptive feature extraction algorithm, training and optimizing a Bidirectional Long Short-Time Recurrent Neural Network (Bi-LSTM), and finally predicting the lifespan of lithium-ion batteries using the Bi-LSTM. Attached Figure Description

[0030] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0031] Figure 1 A flowchart illustrating a battery capacity and lifetime prediction method based on multi-data fusion provided in an embodiment of this application;

[0032] Figure 2 A schematic diagram of the overall framework of the battery capacity and lifetime prediction method based on multi-data fusion provided in the embodiments of this application;

[0033] Figure 3 A schematic diagram illustrating the improved SDAE model training process provided in this application embodiment;

[0034] Figure 4 This is a schematic diagram of the Bi-LSTM structure provided in an embodiment of this application;

[0035] Figure 5 This is a schematic diagram of the internal structure of an LSTM cell provided in an embodiment of this application;

[0036] Figure 6 The charge / discharge current of the battery during the 660th cycle under the conditions of 25℃, 0.33C (charging current), 0.88C (discharging current), and 4.3V (cutoff voltage);

[0037] Figure 7 The battery temperature during its 660th discharge cycle is given under the following conditions: 25°C, 0.33C (charging current), 0.88C (discharging current), and 4.3V (cutoff voltage).

[0038] Figure 8 The above are prediction results for some batteries using the battery capacity lifetime prediction method based on multi-data fusion provided in the embodiments of this application.

[0039] Figure 9 A schematic diagram of the structure of a battery capacity and life prediction device based on multi-data fusion provided in an embodiment of this application;

[0040] Figure 10 An example diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0041] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0042] As described in the background section, existing lithium-ion battery capacity and life prediction methods based on failure mechanisms are difficult to implement and have poor accuracy over long periods. Data-driven lithium-ion battery capacity and life prediction methods have high accuracy and do not require the establishment of complex chemical reaction models inside lithium batteries. However, they suffer from problems such as reliance on human experience for the data features, limited features, and inability to predict lifespan after changes in operating conditions.

[0043] To address one or more of the aforementioned problems, this application creatively proposes a novel battery capacity and lifespan prediction method based on multi-data fusion. This method extracts key features from a large amount of monitoring data of lithium batteries under different charging rates, different discharging rates, and different ambient temperatures during cycling. It then trains and optimizes a bidirectional long short-time recurrent neural network (Bi-LSTM) and finally predicts the lifespan of lithium-ion batteries through the Bi-LSTM.

[0044] The solution of this application will now be described in detail with reference to the accompanying drawings and various embodiments.

[0045] Example 1

[0046] To achieve the solution of this application, embodiments of this application provide a battery capacity and lifespan prediction method based on multi-data fusion, referring to... Figure 1 and Figure 2 As shown, the method includes the following steps:

[0047] S100: Acquire monitoring data of the battery to be predicted under multiple different test conditions. The monitoring data includes first relevant data and second relevant data. The first relevant data includes at least voltage and current, and the second relevant data includes at least temperature rise, discharge capacitance, test temperature, discharge rate and charge rate.

[0048] The batteries to be predicted in the embodiments of this application include, but are not limited to, lithium-ion batteries. The following description uses lithium-ion batteries as an example to illustrate the solution of this application.

[0049] Since changes in operating conditions can affect the lifespan of lithium-ion batteries, this application uses multiple different test conditions to test the lithium-ion battery under various test conditions when predicting its remaining lifespan. The data collected during each test process, including voltage, current, temperature rise, discharge capacitance, test temperature, discharge rate, and charge rate, are used to form monitoring data. Among these, voltage and current are classified as first relevant data, while temperature rise, discharge capacitance, test temperature, discharge rate, and charge rate are classified as second relevant data.

[0050] S200: Use a first preset model to extract features from the first related data and normalize the second related data to obtain corresponding feature vectors, and use the first preset model to fuse the feature vectors to obtain a first fused feature vector.

[0051] Preferably, the first preset model in this application embodiment can be a stacked denoising autoencoder model (SDAE), that is, the SDAE model is used to extract features from the first relevant data in the above monitoring data.

[0052] S300: Based on the first fused feature vector, the remaining lifespan of the battery to be predicted is obtained using the second preset model.

[0053] Specifically, in this embodiment, the second preset model can be a bidirectional long short-time recurrent neural network model (Bi-LSTM), and the input of the model is a fused feature vector obtained by fusing the deep features of the above monitoring data, so that the predicted remaining life of the battery to be predicted is more accurate.

[0054] As a preferred implementation, in this embodiment of the application, the method includes the training process of the first preset model, comprising:

[0055] Each sample battery was subjected to charge-discharge cycle tests under different test conditions to obtain test data and generate a training dataset. The test data included voltage, current, temperature rise, discharge capacitance, test temperature, discharge rate, and charge rate.

[0056] A first preset model is obtained by training a preset encoder using the voltage and current in the training dataset.

[0057] Specifically, the sample battery is preferably the same as or of the same type as the battery to be predicted. Taking lithium-ion batteries as an example, lithium-ion batteries from the same batch are subjected to charge-discharge cycle tests under different ambient temperatures and different charge and discharge rates. Data such as battery temperature rise, voltage, and current are recorded during the charge-discharge process. Since the aging path of batteries is complex, to collect as much different aging data as possible, this application preferably sets more than 10 parallel samples under each test condition. After a long period of testing, the battery capacity decays to the retirement index. Based on the aforementioned voltage, current, temperature rise, discharge capacitance, test temperature, discharge rate, and charge rate data, a training dataset for model training is generated as the corresponding test data. Then, the voltage and current in the training dataset are used to train and obtain a first preset model based on a preset encoder. As a preferred implementation, the training dataset can be divided into a first dataset and a second dataset in a 7:3 ratio. The first dataset is used to train and obtain the first preset model based on the preset encoder, and the second dataset is used to validate the trained model.

[0058] Preferably, the preset encoder includes, but is not limited to, an improved stacked denoised auto-encoder (SDAE).

[0059] As a preferred implementation, in this embodiment of the application, the step of using the voltage and current in the training dataset to train and obtain a first preset model based on a preset encoder includes:

[0060] The preset encoder is trained sequentially using the voltage or current in the training dataset, and the feature vector of the voltage or current obtained from the training is used as input to train the next preset encoder, until all the voltage data or current data in the training dataset have been trained, resulting in multiple encoders and multiple decoders.

[0061] The plurality of encoders and the plurality of decoders are stacked to form the first preset model.

[0062] Specifically, as the battery capacity decays during the battery charging cycle, the length of the battery charging and discharging data gradually decreases. To address this issue, this application uses spline interpolation to interpolate the voltage and current data.

[0063] like Figure 3 As shown, the improved stacked denoising autoencoder in this application includes two training processes: training process 1 and training process 2, as detailed below:

[0064] 1) Training process 1

[0065] The process involves sequentially training each stacked denoised autoencoder (DAE) using voltage or current data from the aforementioned training dataset, and then using the trained feature vectors as input to the next DAE.

[0066] Taking the improved SDAE model training process, which extracts voltage data from the above training dataset, as an example, firstly, the voltage data x is assigned according to q D The distribution is modified by adding random noise with probability υ to make it noisy sample data.

[0067] The encoding function f used in the encoding stage of DAE1 θ For sample data containing noise Encode to obtain sample data The deep feature representation h1∈R M Where M is the number of hidden neurons, and the encoding function f θ As shown in the following formula:

[0068] h1 = f θ (x)=s(Wx+b) (1)

[0069] Where s(·) is the Sigmoid activation function, and W and b are the encoding functions f, respectively. θ The weight matrix and bias vector.

[0070] Then, through the decoding function g in the decoding stage of DAE. θ′ The deep feature representation h1 is recovered and reconstructed to obtain the reconstructed voltage data x. The decoding function g in the decoding stage of DAE1 θ′ The formula is shown below:

[0071]

[0072] Where s(·) is the Sigmoid activation function, and W′ and b′ are the decoding functions g, respectively. θ′ The weight matrix and bias vector.

[0073] Finally, by minimizing the voltage data x and the reconstructed voltage data x, The reconstruction error between them is used to obtain the optimal weight parameters {W, W′, b, b′}, which minimizes x and The formula for the reconstruction error is as follows:

[0074]

[0075] Where K is the number of data samples in each training batch, and i is the i-th data sample in each batch.

[0076] After training DAE1, the deep feature representation h1 of voltage data x can be obtained using equation (1). Next, using the deep feature representation h1 of voltage data x as input, DAE2 is trained using equations (1)-(3) to obtain the deep feature representation h2 of voltage data x. Finally, using the deep feature representation h2 of voltage data x as input, DAE3 is trained using equations (1)-(3) to obtain the deep feature representation h3 of voltage data x. This process is repeated until all voltage data in the training dataset has been trained.

[0077] 2) Training process 2

[0078] The encoders and decoders obtained in training process 1 are stacked together to form the encoder and decoder of the SDAE model. First, the voltage data x is fed according to q... D The distribution is modified by adding random noise with probability υ′ to make it noisy sample data.

[0079] The encoding function f used in the encoding phase of SDAE θ For sample data containing noise Encode to obtain sample data Deep feature representation h n ∈R M Where M is the number of hidden neurons and n is the number of encoder layers, the encoding function is shown in the following formula:

[0080]

[0081] in, This is the encoder for the nth DAE.

[0082] Then, the deep feature representation h is processed through the decoding function of the SDAE decoding stage. n Perform recovery and reconstruction to obtain the reconstructed data of voltage data x. The formula for the decoding function in the decoding stage of SDAE is shown below:

[0083]

[0084] in, This is the decoder for the nth DAE.

[0085] Finally, all parameters in the SDAE model are optimized using equation (3). The encoder of the SDAE is then extracted as the final SDAE model for extracting voltage or current data features.

[0086] As a preferred implementation, in this embodiment of the application, the method further includes a training process for a second preset model, comprising:

[0087] Normalize the data in the training dataset other than the voltage and the current to obtain the normalization result, and generate a feature vector dataset based on the normalization result and the feature vectors of the voltage and the current.

[0088] A third preset model is obtained by training based on a preset encoder using the feature vector dataset, and the feature vectors in the feature vector dataset are fused using the third preset model to obtain a second fused feature vector.

[0089] The second fused feature vector is used to train a bidirectional long short-time recurrent neural network to obtain the second preset model.

[0090] Preferably, the temperature rise, discharge capacity, test temperature, charge rate, and discharge rate in the training dataset are normalized using the Min-Max normalization method, and the normalization formula is shown in equation (6):

[0091]

[0092] Where m represents one of the following data: battery temperature rise, discharge capacity, test temperature, charge rate, and discharge rate. min(m) represents the minimum value of the battery temperature rise, discharge capacity, test temperature, charge rate, or discharge rate. The minimum temperature rise can be set to 0 or the battery temperature rise during the first cycle; the minimum discharge capacity can be set to the capacity at which the battery is retired; the minimum test temperature can be set to the minimum of the designed battery operating temperature range; the minimum charge rate can be set to the minimum of the designed battery operating charge rate range; and the minimum discharge rate can be set to the minimum of the designed battery operating discharge rate range. max(m) represents the maximum value of the battery temperature rise, discharge capacity, test temperature, charge rate, or discharge rate. The maximum temperature rise can be set to the temperature at which thermal runaway occurs; the discharge capacity can be set to the capacity during the first cycle; the maximum test temperature can be set to the maximum of the designed battery operating temperature range; the maximum charge rate can be set to the maximum of the designed battery operating charge rate range; and the maximum discharge rate can be set to the minimum of the designed battery operating discharge rate range.

[0093] After obtaining the normalized results of the relevant parameters, the deep feature representations of the voltage and current data extracted using the SDAE model in the previous steps are combined to form a feature vector dataset H.

[0094] Using the feature vector dataset H, another SDAE model is retrained according to the training method of the SDAE model (first preset model) in the aforementioned steps. After several batches of training, the SDAE1 model (third preset model) is obtained and used to fuse feature data. The SDAE1 model can fuse features of data such as voltage, current, temperature rise, discharge capacity, test temperature, charging rate, and discharging rate to obtain a fused feature vector.

[0095] In this application, the second preset model is preferably a bidirectional long short-time recurrent neural network (Bi-LSTM). It can be understood that the test data collected during battery testing is time-series data, and the fused features after feature extraction and fusion are also time-series data. A sliding window of length L is used to reconstruct the data. If the current data's cycle index is d, the fused features with cycle indices within [d, d+L] are used as the input values ​​to the Bi-LSTM, and the capacity with cycle index d+L+1 is used as the data label for the fused features with cycle indices within [d, d+L]. This forms a data sample. Then, the current data's cycle index is increased by d+1 to form a new data sample, until d+L equals the cycle index of the last cycle.

[0096] like Figure 4As shown, Bi-LTSM consists of a forward LSTM layer and a backward LSTM layer. Bi-LSTM can learn bidirectional features of data and can better capture the correlation between fused features and time-related factors. The hidden state output yt of Bi-LSTM at time t is calculated as follows:

[0097]

[0098] In the formula, The output of the hidden state of the backward LSTM layer at time t; The hidden state output of the feedforward LSTM at time t; This refers to the concatenation of vectors. The LSTM cell is a single LSTM cell, mainly consisting of a forget gate, input gate, output gate, and memory unit, and its internal structure is as follows: Figure 5 As shown.

[0099] The input to an LSTM cell is the network's input x at the current time step. t The output h of the LSTM at the previous time step t-1 and the cell state C at the previous time step t-1 The output of the LSTM contains the hidden state h of the LSTM output at the current time. t and the current cell state C t The calculation formula for the forget gate is shown in equation (8). The forget gate determines the cell state C of the previous time step. t-1 It can be utilized in the current moment.

[0100] f t =σ(W f [h t-1 x t ]+b f (8)

[0101] In the formula, σ(·) is the Sigmoid activation function, and W f This is the weight matrix of the forget gate, b f It is the bias vector of the forget gate.

[0102] The calculation formulas for the input gate are shown in equations (9)-(11).

[0103] i t =σ(W i [h t-1 x t ]+b i (9)

[0104]

[0105]

[0106] In the formula, tanh(·) is the hyperbolic tangent function tanh; ⊙ is the Hadamard product; W i b is the weight matrix of the input gate; c W is the bias vector of the input gate; c b is the weight matrix of the memory cells; c This is the bias vector for the memory cell; For memory units.

[0107] The calculation formulas for the output gate are shown in equations (12) and (13).

[0108] O t =σ(W o [h t-1 x t ]+b o (12)

[0109] h t =O t ×tanh(c t (13)

[0110] Optionally, the model parameters can be adjusted by optimizing equation (14) to reduce the prediction error of the Bi-LSTM model and obtain the optimal Bi-LSTM model parameters.

[0111]

[0112] Therefore, the optimal Bi-LSTM model for predicting battery life can be obtained.

[0113] Figure 6 The charge / discharge current of the battery during the 660th cycle under the conditions of 25℃, 0.33C (charging current), 0.88C (discharging current), and 4.3V (cutoff voltage); Figure 7 The battery temperature during its 660th discharge cycle is given by the following parameters: 25℃, 0.33C (charging current), 0.88C (discharging current), and 4.3V (cutoff voltage). Partial testing was conducted under 40℃, 0.33C (charging current), 0.88C (discharging current), and 4.3V (cutoff voltage) conditions to predict the battery's lifespan degradation to 85%. After a period of prediction, the charge / discharge conditions were changed to 40℃, 0.33C (charging current), 2.0C (discharging current), and 4.3V (cutoff voltage). The prediction results are as follows. Figure 8 As shown.

[0114] Example 2

[0115] Corresponding to Embodiment 1 above, this application also provides a battery capacity and lifespan prediction device based on multi-data fusion. In this embodiment, content that is the same as or similar to Embodiment 1 above can be referred to the above description and will not be repeated hereafter. (Refer to...) Figure 9 The device includes:

[0116] The data acquisition module is used to acquire monitoring data of the battery to be predicted under multiple different test conditions. The monitoring data includes first relevant data and second relevant data. The first relevant data includes at least voltage and current, and the second relevant data includes at least temperature rise, discharge capacitance, test temperature, discharge rate and charge rate.

[0117] The feature extraction module is used to extract features from the first related data using a first preset model and to normalize the second related data to obtain corresponding feature vectors, and to fuse the feature vectors using the first preset model to obtain a first fused feature vector.

[0118] The lifespan prediction module is used to obtain the remaining lifespan of the battery to be predicted based on the first fused feature vector using a second preset model.

[0119] In a preferred embodiment of this application, the device further includes a model training module, which is used to perform charge-discharge cycle tests on each sample battery under different test conditions, obtain test data to generate a training dataset, the test data including voltage, current, temperature rise, discharge capacitance, test temperature, discharge rate and charge rate; and use the voltage and current in the training dataset to train and obtain a first preset model based on a preset encoder.

[0120] In a preferred embodiment of this application, the model training module is further configured to sequentially train the preset encoder using the voltage or current in the training dataset, and use the feature vector of the trained voltage or current as input to train the next preset encoder, until all the voltage data or current data in the training dataset have been trained, thereby obtaining multiple encoders and multiple decoders; and stack the multiple encoders and multiple decoders to form the first preset model.

[0121] In a preferred embodiment of this application, the first preset model includes a stacked denoising autoencoder model.

[0122] In a preferred embodiment of this application, the test data is time series data.

[0123] As a preferred implementation, in this embodiment of the application, the model training module is further configured to normalize other data in the training dataset except for the voltage and the current, obtain normalization results, and generate a feature vector dataset based on the normalization results and the feature vectors of the voltage and the current.

[0124] A third preset model is obtained by training based on a preset encoder using the feature vector dataset, and the feature vectors in the feature vector dataset are fused using the third preset model to obtain a second fused feature vector.

[0125] The second fused feature vector is used to train a bidirectional long short-time recurrent neural network to obtain the second preset model.

[0126] As a preferred implementation, in this embodiment of the application, the third preset model includes a stacked denoising autoencoder model.

[0127] Example 3

[0128] Corresponding to Embodiment 1 or 2 above, this application also provides a computer device, including: a processor and a memory, wherein the memory stores a computer program that can run on the processor, and when the computer program is executed by the processor, it executes the battery capacity and life prediction method based on multi-data fusion provided in any of the above embodiments.

[0129] in, Figure 10 An exemplary computer device 1500 is shown, which may specifically include a processor 1510, a video display adapter 1511, a disk drive 1512, an input / output interface 1513, a network interface 1514, and a memory 1520. The processor 1510, video display adapter 1511, disk drive 1512, input / output interface 1513, network interface 1514, and memory 1520 can communicate with each other via a communication bus 1530.

[0130] The processor 1510 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solution provided by the present invention.

[0131] The memory 1520 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 1520 can store the operating system 1521 for controlling the operation of the electronic device, and the basic input / output system (BIOS) for controlling the low-level operations of the electronic device. Additionally, it can store a web browser 1523, a data storage management system 1524, and a device identification information processing system 1525, etc. The aforementioned device identification information processing system 1525 can be the application program that specifically implements the aforementioned steps in this embodiment of the invention. In summary, when implementing the technical solution provided by this invention through software or firmware, the relevant program code is stored in the memory 1520 and is called and executed by the processor 1510.

[0132] Input / output interface 1513 is used to connect input / output modules to realize information input and output. Input / output modules can be configured as components in the device (not shown in the figure) or externally connected to the device to provide corresponding functions. Input devices may include keyboards, mice, touch screens, microphones, various sensors, etc., and output devices may include displays, speakers, vibrators, indicator lights, etc.

[0133] Network interface 1514 is used to connect a communication module (not shown in the figure) to enable communication between this device and other devices. The communication module can communicate via wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, Wi-Fi, Bluetooth, etc.).

[0134] The bus includes a pathway for transmitting information between various components of the device (e.g., processor 1510, video display adapter 1511, disk drive 1512, input / output interface 1513, network interface 1514, and memory 1520).

[0135] In addition, the electronic device can also obtain information on specific claim conditions from the virtual resource object claim condition information database for condition judgment, and so on.

[0136] It should be noted that although the above-described device only shows the processor 1510, video display adapter 1511, disk drive 1512, input / output interface 1513, network interface 1514, memory 1520, bus, etc., in specific implementations, the device may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the above-described device may only include the components necessary for implementing the present invention, and not necessarily all the components shown in the figures.

[0137] Example 4

[0138] Corresponding to Embodiment 1 or 2 above, this application also provides a computer-readable storage medium. In this embodiment, the content that is the same as or similar to Embodiments 1 to 3 above can be referred to the above description and will not be repeated hereafter.

[0139] The computer-readable storage medium has a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0140] The monitoring data of the battery to be predicted under multiple different test conditions is obtained. The monitoring data includes first correlation data and second correlation data. The first correlation data includes at least voltage and current, and the second correlation data includes at least temperature rise, discharge capacitance, test temperature, discharge rate and charge rate.

[0141] The first preset model is used to extract features from the first relevant data and normalize the second relevant data to obtain corresponding feature vectors. The first preset model is then used to fuse the feature vectors to obtain a first fused feature vector.

[0142] The remaining lifespan of the battery to be predicted is obtained using the second preset model based on the first fused feature vector.

[0143] In some implementations of this application, when the computer program is executed by the processor, it can also implement the steps corresponding to the method described in Embodiment 1. Please refer to the detailed description in Embodiment 1, which will not be repeated here.

[0144] As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that the present invention can be implemented by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in various embodiments or some parts of the embodiments of the present invention.

[0145] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, for system or system embodiments, since they are basically similar to method embodiments, the description is relatively simple, and relevant parts can be referred to the descriptions in the method embodiments. The systems and system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0146] The technical solution provided by this invention has been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this invention. Furthermore, those skilled in the art will recognize that, based on the ideas of this invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this invention.

Claims

1. A method for predicting battery capacity and lifespan based on multi-data fusion, characterized in that, The method includes: The monitoring data of the battery to be predicted under multiple different test conditions is obtained. The monitoring data includes first correlation data and second correlation data. The first correlation data includes at least voltage and current, and the second correlation data includes at least temperature rise, discharge capacitance, test temperature, discharge rate and charge rate. The first preset model is used to extract features from the first relevant data and normalize the second relevant data to obtain corresponding feature vectors. The first preset model is then used to fuse the feature vectors to obtain a first fused feature vector. The remaining lifespan of the battery to be predicted is obtained using the second preset model based on the first fused feature vector. The method includes the training process of the first preset model, comprising: Each sample battery was subjected to charge-discharge cycle tests under different test conditions to obtain test data and generate a training dataset. The test data included voltage, current, temperature rise, discharge capacitance, test temperature, discharge rate, and charge rate. A first preset model is obtained by training a preset encoder using the voltage and current in the training dataset. The method also includes a training process for a second preset model, including: Normalize the data in the training dataset other than the voltage and the current to obtain the normalization result, and generate a feature vector dataset based on the normalization result and the feature vectors of the voltage and the current. A third preset model is obtained by training based on a preset encoder using the feature vector dataset, and the feature vectors in the feature vector dataset are fused using the third preset model to obtain a second fused feature vector. The second fused feature vector is used to train a bidirectional long short-time recurrent neural network to obtain the second preset model.

2. The battery capacity and life prediction method based on multi-data fusion according to claim 1, characterized in that, The step of training a first preset model based on a preset encoder using the voltage and current in the training dataset includes: The preset encoder is trained sequentially using the voltage or current in the training dataset, and the feature vector of the voltage or current obtained from the training is used as input to train the next preset encoder, until all the voltage data or current data in the training dataset have been trained, resulting in multiple encoders and multiple decoders. The plurality of encoders and the plurality of decoders are stacked to form the first preset model.

3. The battery capacity and life prediction method based on multi-data fusion according to claim 2, characterized in that, The first preset model includes a stacked denoising autoencoder model.

4. The battery capacity and life prediction method based on multi-data fusion according to claim 1, characterized in that, The test data is time series data.

5. The battery capacity and life prediction method based on multi-data fusion according to claim 1, characterized in that, The third preset model includes a stacked denoising autoencoder model.

6. A battery capacity and life prediction device based on multi-data fusion, characterized in that, The device includes: The data acquisition module is used to acquire monitoring data of the battery to be predicted under multiple different test conditions. The monitoring data includes first relevant data and second relevant data. The first relevant data includes at least voltage and current, and the second relevant data includes at least temperature rise, discharge capacitance, test temperature, discharge rate and charge rate. The feature extraction module is used to extract features from the first related data using a first preset model and to normalize the second related data to obtain corresponding feature vectors, and to fuse the feature vectors using the first preset model to obtain a first fused feature vector. The lifespan prediction module is used to obtain the remaining lifespan of the battery to be predicted based on the first fused feature vector using a second preset model. The training process of the first preset model includes: Each sample battery was subjected to charge-discharge cycle tests under different test conditions to obtain test data and generate a training dataset. The test data included voltage, current, temperature rise, discharge capacitance, test temperature, discharge rate, and charge rate. A first preset model is obtained by training a preset encoder using the voltage and current in the training dataset. The training process of the second preset model includes: Normalize the data in the training dataset other than the voltage and the current to obtain the normalization result, and generate a feature vector dataset based on the normalization result and the feature vectors of the voltage and the current. A third preset model is obtained by training based on a preset encoder using the feature vector dataset, and the feature vectors in the feature vector dataset are fused using the third preset model to obtain a second fused feature vector. The second fused feature vector is used to train a bidirectional long short-time recurrent neural network to obtain the second preset model.

7. A computer device, characterized in that, The method includes a memory and a processor, wherein the memory stores a computer program that can run on the processor, and when the computer program is executed by the processor, it implements the battery capacity life prediction method based on multi-data fusion as described in any one of claims 1 to 5.

8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed, it implements the battery capacity and life prediction method based on multi-data fusion as described in any one of claims 1 to 5.

Images