A State of Charge Prediction Method, Device, Electronic Device, and Storage Medium

Abstract

The present application provides a method, apparatus, electronic device and storage medium for predicting the state of charge, which is used to improve the problem of low accuracy in predicting the state of charge of a storage battery. The method includes: obtaining a plurality of sensor signals in the storage battery, and performing feature extraction on the plurality of sensor signals to obtain latent space features; extracting signal reconstruction features and latent space attention features from the latent space features, and fusing the latent space attention features and the signal reconstruction features to obtain a reconstructed fusion feature; processing the signal reconstruction features to obtain a reconstructed attention feature, and fusing the reconstructed attention feature and the latent space features to obtain a latent space fusion feature; using a trained neural network model to perform prediction based on multiple features to obtain the state of charge of the storage battery, and the multiple features include any two or more of the latent space features, signal reconstruction features, reconstructed fusion features and latent space fusion features.

Description

Technical Field

[0001] This application relates to the technical fields of batteries, artificial intelligence, and deep learning. Specifically, it relates to a method, device, electronic device, and storage medium for predicting the state of charge. Background Art

[0002] The state of charge of a storage battery refers to the percentage of the actual number of charges (in ampere-hours) present in the energy storage medium during the electrochemical energy storage process to the number of charges (in ampere-hours) contained in the energy storage medium corresponding to the rated energy storage capacity. Specifically, for example, the charge state of a new battery is 100%, and the charge state after the battery is completely discharged is 0%. The state of charge estimation can be used as an important basis for the charge and discharge control and equalization management of lead-acid batteries.

[0003] Currently, the main methods for predicting the state of charge (SOC) of a storage battery include: the ampere-hour method, the open-circuit voltage method, the internal resistance method, the Kalman filter method, etc. However, in the specific practical process, it is found that the battery is easily affected by factors such as the number of charge and discharge cycles, charge and discharge efficiency, electrolyte temperature, battery aging, and self-discharge during operation. Therefore, the current accuracy of predicting the state of charge of the storage battery is relatively low. Summary of the Invention

[0004] The purpose of the embodiments of this application is to provide a method, device, electronic device, and storage medium for predicting the state of charge, which is used to improve the problem of relatively low accuracy in predicting the state of charge of a storage battery.

[0005] The embodiments of this application provide a method for predicting the state of charge, including: obtaining a plurality of sensor signals in the storage battery, and performing feature extraction on the plurality of sensor signals to obtain hidden space features; extracting signal reconstruction features and hidden space attention features from the hidden space features, and fusing the hidden space attention features and the signal reconstruction features to obtain a reconstructed fusion feature; processing the signal reconstruction features to obtain a reconstructed attention feature, and fusing the reconstructed attention feature and the hidden space features to obtain a hidden space fusion feature; using a trained neural network model to perform prediction based on multiple features to obtain the state of charge of the storage battery, and the multiple features include any two or more of the hidden space features, signal reconstruction features, reconstructed fusion features, and hidden space fusion features.

[0006] In the implementation process of the above solution, by extracting multiple features of different manifestations of multiple sensor signals in the storage battery, where the multiple features are any two or more of the latent space feature, signal reconstruction feature, reconstruction fusion feature, and latent space fusion feature, and then using a neural network model to determine the state of charge of the storage battery according to the multiple features, the influence of factors such as the number of charge and discharge cycles, charge and discharge efficiency, electrolyte temperature, battery aging, and self-discharge on the prediction of the charge state is minimized, and the accuracy of predicting the charge state of the storage battery is improved.

[0007] Optionally, in the embodiment of the present application, the neural network model includes: an encoder and a decoder; the feature extraction of multiple sensor signals includes: using the encoder in the neural network model to extract features of multiple sensor signals; extracting the signal reconstruction feature and the latent space attention feature from the latent space feature, including: using the decoder in the neural network model to perform feature reduction on the latent space feature to obtain multiple reduced signals, and extracting the signal reconstruction feature from the multiple reduced signals; extracting the latent space attention feature from the latent space feature.

[0008] In the implementation process of the above solution, by using the encoder in the neural network model to extract features of multiple sensor signals, and using the decoder in the neural network model to perform feature reduction on the latent space feature to obtain multiple reduced signals, and extracting the signal reconstruction feature from the multiple reduced signals, and extracting the latent space attention feature from the latent space feature, the attention mechanism is combined to obtain two or more modal features at each stage, and the charge state of the storage battery is predicted through two or more modal features, effectively improving the accuracy of predicting the charge state of the storage battery, etc.

[0009] Optionally, in the embodiment of the present application, the neural network model further includes: a convolutional long short-term memory network; extracting the signal reconstruction feature from multiple reduced signals includes: using the convolutional long short-term memory network to extract features of multiple reduced signals respectively to obtain multiple manifestation features, and performing fusion reconstruction on the multiple manifestation features to obtain the signal reconstruction feature.

[0010] In the implementation process of the above solution, by using the convolutional long short-term memory network to extract features, not only can the spatio-temporal features of time series data be extracted by the convolutional long short-term memory network, but also the complex time dependence and the spatial dependence of time series data can be learned. The convolutional long short-term memory network can be obtained by replacing all matrix multiplication operations in the long short-term memory network with convolutional operations, so that the local convolutional long short-term memory network can discover more hidden information than the convolutional neural network and the long short-term memory network.

[0011] Optionally, in the embodiments of the present application, the neural network model further includes: a first attention module; extracting hidden space attention features from the hidden space features, including: using the first attention module in the neural network model to process the hidden space features to obtain the hidden space attention features.

[0012] In the implementation process of the above solution, by using the first attention module in the neural network model to process the hidden space features, the attention mechanism is combined to automatically extract high-level semantic features strongly related to the target task, thereby minimizing the influence of factors such as charge and discharge times, charge and discharge efficiency, electrolyte temperature, battery aging, and self-discharge on the prediction of the state of charge, and improving the accuracy of predicting the state of charge of the storage battery.

[0013] Optionally, in the embodiments of the present application, the neural network model further includes: a second attention module; processing the signal reconstruction features to obtain reconstruction attention features, including: using the second attention module in the neural network model to process the signal reconstruction features to obtain the reconstruction attention features.

[0014] In the implementation process of the above solution, by using the second attention module in the neural network model to process the signal reconstruction features, the attention mechanism is combined to automatically extract high-level semantic features strongly related to the target task, thereby minimizing the influence of factors such as charge and discharge times, charge and discharge efficiency, electrolyte temperature, battery aging, and self-discharge on the prediction of the state of charge, and improving the accuracy of predicting the state of charge of the storage battery.

[0015] Optionally, in the embodiments of the present application, using the trained neural network model to make predictions based on multiple features, including: splicing the multiple features to obtain spliced features; using the neural network model to make predictions on the spliced features to obtain the state of charge of the storage battery.

[0016] In the implementation process of the above solution, by splicing the multiple features to obtain spliced features and using the neural network model to make predictions on the spliced features, since the spliced features can reflect high-level semantic features strongly related to the target task, the influence of factors such as charge and discharge times, charge and discharge efficiency, electrolyte temperature, battery aging, and self-discharge on the prediction of the state of charge is minimized, and the accuracy of predicting the state of charge of the storage battery is improved.

[0017] Optionally, in the embodiments of the present application, the neural network model further includes: a fully connected layer; using the neural network model to make predictions on the spliced features, including: using the fully connected layer in the neural network model to make predictions on the spliced features.

[0018] In the implementation process of the above solution, multi-channel dimensional information is effectively extracted and fused through network structures such as fully connected layers in the neural network model, and the high-level semantic features strongly related to the target task are automatically extracted using the correlation of this multi-channel dimensional information, thereby minimizing the influence of factors such as charge and discharge times, charge and discharge efficiency, electrolyte temperature, battery aging, and self-discharge on the prediction of the state of charge, and improving the accuracy of predicting the state of charge of the storage battery.

[0019] The embodiment of the present application also provides a state of charge prediction device, including: a spatial feature acquisition module, configured to acquire a plurality of sensor signals in the storage battery and perform feature extraction on the plurality of sensor signals to obtain hidden space features; a feature extraction and reconstruction module, configured to extract signal reconstruction features and hidden space attention features from the hidden space features, and fuse the hidden space attention features and the signal reconstruction features to obtain a reconstructed fusion feature; a fusion feature acquisition module, configured to process the signal reconstruction features to obtain reconstructed attention features, and fuse the reconstructed attention features and the hidden space features to obtain hidden space fusion features; a state of charge prediction module, configured to use the trained neural network model to perform prediction according to multiple features to obtain the state of charge of the storage battery, and the multiple features include any two or more of the hidden space features, signal reconstruction features, reconstructed fusion features, and hidden space fusion features.

[0020] Optionally, in the embodiment of the present application, the neural network model includes: an encoder and a decoder; the spatial feature acquisition module includes: a spatial feature extraction module, configured to perform feature extraction on the plurality of sensor signals using the encoder in the neural network model; the feature extraction and reconstruction module includes: a spatial feature reduction module, configured to perform feature reduction on the hidden space features using the decoder in the neural network model to obtain a plurality of reduced signals, and extract signal reconstruction features from the plurality of reduced signals; an attention feature extraction module, configured to extract hidden space attention features from the hidden space features.

[0021] Optionally, in the embodiment of the present application, the neural network model further includes: a convolutional long short-term memory network; the spatial feature reduction module includes: a feature fusion and reconstruction module, configured to perform feature extraction on the plurality of reduced signals respectively using the convolutional long short-term memory network to obtain a plurality of manifestation features, and perform fusion and reconstruction on the plurality of manifestation features to obtain signal reconstruction features.

[0022] Optionally, in the embodiment of the present application, the neural network model further includes: a first attention module; the attention feature extraction module includes: a spatial feature processing module, configured to process the hidden space features using the first attention module in the neural network model to obtain hidden space attention features.

[0023] Optionally, in the embodiments of the present application, the neural network model further includes: a second attention module; a fused feature obtaining module, including: a reconstructed feature processing module, configured to process the signal reconstruction feature using the second attention module in the neural network model to obtain a reconstructed attention feature.

[0024] Optionally, in the embodiments of the present application, the charge state prediction module includes: a spliced feature obtaining module, configured to splice multiple features to obtain a spliced feature; a spliced feature prediction module, configured to predict the spliced feature using the neural network model to obtain the charge state of the storage battery.

[0025] Optionally, in the embodiments of the present application, the neural network model further includes: a fully connected layer; the spliced feature prediction module includes: a fully connected layer prediction module, configured to predict the spliced feature using the fully connected layer in the neural network model.

[0026] The embodiments of the present application further provide an electronic device, including: a processor and a memory, where the memory stores machine-readable instructions executable by the processor, and when the machine-readable instructions are executed by the processor, the methods described above are executed.

[0027] The embodiments of the present application further provide a computer-readable storage medium, on which a computer program is stored, and when the computer program is run by a processor, the methods described above are executed. Description of the Drawings

[0028] To more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required to be used in the embodiments of the present application. It should be understood that the following drawings only show some embodiments in the embodiments of the present application, and thus should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.

[0029] Figure 1 A flowchart showing the charge state prediction method provided by the embodiments of the present application;

[0030] Figure 2 A schematic diagram showing the network structure of the neural network model provided by the embodiments of the present application;

[0031] Figure 3 A schematic diagram showing the structure of the charge state prediction device provided by the embodiments of the present application;

[0032] Figure 4 A schematic diagram showing the structure of the electronic device provided by the embodiments of the present application. Detailed Embodiments

[0033] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and illustrated herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed embodiments of the present application, but merely represents the selected embodiments of the present application. All other embodiments obtained by those skilled in the art based on the embodiments of the present application without creative efforts belong to the scope of protection of the embodiments of the present application.

[0034] It can be understood that the "first" and "second" in the embodiments of the present application are used to distinguish similar objects. Those skilled in the art can understand that the words such as "first" and "second" do not limit the quantity and execution order, and the words such as "first" and "second" do not necessarily limit to be different.

[0035] Before introducing the state of charge prediction method provided in the embodiments of the present application, some concepts involved in the embodiments of the present application will be introduced first:

[0036] Variational Auto-Encoder (VAE) refers to a deep generative model and is also an unsupervised learning algorithm; the main function of VAE is to be able to learn a function (model) so that the distribution of the output data can approximate the original data distribution as much as possible. Different from traditional autoencoders that describe the latent space numerically, VAE describes the observation of the latent space in a probabilistic way and shows great application value in data generation.

[0037] Long Short-Term Memory (LSTM) network is a time-recurrent neural network and also a recurrent neural network, suitable for processing and predicting important features with relatively long intervals and delays in time series.

[0038] It should be noted that the state of charge prediction method provided in the embodiments of the present application can be executed by an electronic device, where the electronic device refers to a device terminal or a server with the function of executing a computer program. The device terminal is, for example: a smart phone, a personal computer, a tablet computer, a personal digital assistant or a mobile Internet device, etc. A server refers to a device that provides computing services through a network. The server is, for example: an x86 server and a non-x86 server. The non-x86 server includes: mainframes, minicomputers and UNIX servers.

[0039] The following introduces the application scenarios applicable to the state of charge prediction method. The application scenarios here include but are not limited to: during the use of electric vehicles and photovoltaic energy storage systems, the storage batteries of electric vehicles will be affected by factors such as the number of charge and discharge cycles, charge and discharge efficiency, electrolyte temperature, battery aging, and self-discharge. Therefore, this state of charge prediction method can be used to predict the state of charge of the storage battery of an electric vehicle. Since this state of charge prediction method extracts features at different feature levels from different sensor signals, and combines an attention mechanism to obtain two or more modal features at each stage, and predicts the charge state of the storage battery through two or more modal features, thus effectively improving the accuracy of predicting the charge state of the storage battery, etc.

[0040] Please refer to Figure 1 the schematic flowchart of the state of charge prediction method provided by the embodiment of the present application shown; The embodiment of the present application provides a state of charge prediction method, including:

[0041] Step S110: Obtain multiple sensor signals in the storage battery, and perform feature extraction on the multiple sensor signals to obtain latent space features.

[0042] The multiple sensor signals refer to the sensing signals sent by multiple sensors in the storage battery, including but not limited to: storage battery current signal, storage battery voltage signal, storage battery temperature signal, starting motor current signal, starting motor voltage signal, electrolyte concentration signal, and / or electrolyte liquid level signal, etc.

[0043] Step S120: Extract signal reconstruction features and latent space attention features from the latent space features, and fuse the latent space attention features and the signal reconstruction features to obtain a reconstruction fusion feature.

[0044] Step S130: Process the signal reconstruction features to obtain reconstruction attention features, and fuse the reconstruction attention features and the latent space features to obtain a latent space fusion feature.

[0045] Step S140: Use the trained neural network model to make a prediction based on multiple features to obtain the charge state of the storage battery. The multiple features include any two or more of the latent space features, signal reconstruction features, reconstruction fusion features, and latent space fusion features.

[0046] Please refer to Figure 2Schematic diagram of the network structure of the neural network model provided by the embodiments of the present application; it can be understood that the above neural network model may include: an encoder, a decoder, multiple convolutional long short-term memory networks, a first attention module, a second attention module, a fully connected layer, and other network structures. Here, the encoder and decoder can adopt the encoder and decoder in a variational autoencoder (VAE). The encoder and decoder are connected to each other, and the connection relationships between the other network structures are as Figure 2 shown.

[0047] In the implementation process of the above solution, by extracting multiple features of different forms of multiple sensor signals in the battery, where the multiple features are any two or more of the latent space features, signal reconstruction features, reconstruction fusion features, and latent space fusion features, and then using the neural network model to determine the state of charge of the battery according to the multiple features, the influence of factors such as charge and discharge rate, temperature, aging, etc. on the prediction of the charge state is minimized, and the accuracy of predicting the charge state of the battery is improved.

[0048] As an alternative implementation of step S110, since the above neural network model may include an encoder and a decoder, therefore, the encoder and decoder can be used to extract features from the sensor signals. The implementation of feature extraction may include:

[0049] Step S111: Obtain multiple sensor signals in the battery.

[0050] The implementation of the above step S111 is, for example: since the multiple sensor signals include but are not limited to: battery current signal, battery voltage signal, battery temperature signal, starting motor current signal, starting motor voltage signal, electrolyte concentration signal, and / or electrolyte liquid level signal, etc., therefore, here, taking the acquisition of the electrolyte liquid level signal as an example, use the liquid level sensor of the battery to obtain the liquid height position of the electrolyte, convert the liquid height position into a liquid level signal, and then send the liquid level signal to the electronic device. The electronic device receives the liquid level signal sent by the liquid level sensor, and the same applies to other sensor signals, and multiple sensor signals can be obtained. Among them, the liquid level sensor here (also known as a static pressure liquid level gauge, or liquid level transmitter, or liquid level sensor, or water level sensor) is a pressure sensor for measuring the liquid level.

[0051] Step S112: Use the encoder in the neural network model to extract features from the multiple sensor signals to obtain latent space features.

[0052] For example, the implementation of the above step S112 is as follows: Use the encoder in the neural network model to extract features from multiple sensor signals to obtain latent space features; among them, the encoder here can adopt the encoder in the variational autoencoder (VAE).

[0053] As an alternative implementation of step S120, the implementation of the extraction and fusion of latent space features may include:

[0054] Step S121: Use the decoder in the neural network model to restore the features of the latent space features to obtain multiple restored signals, and extract signal reconstruction features from the multiple restored signals.

[0055] For example, the implementation of the above step S121 is as follows: Use the decoder in the neural network model to restore the features of the latent space features, that is, try to restore the original sensor signals according to the latent space features to obtain multiple restored signals, and extract signal reconstruction features from the multiple restored signals.

[0056] The above neural network model may further include a convolutional long short-term memory network, which can replace all matrix multiplication operations in the long short-term memory network with convolutional operations to obtain a convolutional long short-term memory network. In the specific implementation process, the convolutional long short-term memory network can also be used to extract signal reconstruction features from multiple restored signals. This implementation may include: Use multiple convolutional long short-term memory networks to extract features from multiple restored signals respectively to obtain multiple manifestation features, and perform fusion reconstruction on the multiple manifestation features in terms of dimensions to obtain signal reconstruction features.

[0057] In the above implementation process, using the convolutional long short-term memory network to extract features can not only utilize the spatio-temporal features of time series data extracted by the convolutional long short-term memory network, but also learn complex time dependencies and spatial dependencies of time series data. The convolutional long short-term memory network can be obtained by replacing all matrix multiplication operations in the long short-term memory network with convolutional operations, so that the local convolutional long short-term memory network can discover more hidden information than convolutional neural networks and long short-term memory networks.

[0058] Step S122: Extract latent space attention features from the latent space features.

[0059] For example, the implementation of the above step S122 is as follows: Since the above neural network model may further include: a first attention module, the first attention module can be used to extract latent space attention features from the latent space features. This implementation may include: Use the first attention module in the neural network model to process the latent space features to obtain latent space attention features.

[0060] It can be understood that the calculation process of the attention mechanism included in the above first attention module can be expressed as: Among them, q represents the query feature matrix in the latent space feature, k represents the key feature matrix in the latent space feature, and k T represents the transpose operation of the key feature matrix, v represents the value feature matrix in the latent space feature, is the scale factor, and its specific value is the matrix dimension of q and k. First, the query feature matrix and the key feature matrix are subjected to dot-product attention calculation. To prevent the result of the dot-product attention calculation from being too large, it is necessary to divide by a scale factor (scaling factor); then use softmax to normalize the result into a probability distribution, and finally multiply by the value feature matrix to obtain the calculation result of the attention mechanism.

[0061] Step S123: Fuse the latent space attention feature and the signal reconstruction feature to obtain the reconstructed fusion feature.

[0062] The implementation manner of the above step S123 is, for example: using methods such as mean fusion, weighted fusion, channel fusion, and splicing fusion to fuse the latent space attention feature and the signal reconstruction feature to obtain the reconstructed fusion feature.

[0063] In the implementation process of the above solution, by using the encoder in the neural network model to extract features from multiple sensor signals, and using the decoder in the neural network model to restore the latent space features, multiple restored signals are obtained, and the signal reconstruction features are extracted from the multiple restored signals, and the latent space attention features are extracted from the latent space features, so as to combine the attention mechanism to obtain two or more modal features in each stage, and predict the state of charge of the battery through two or more modal features, effectively improving the accuracy of predicting the state of charge of the battery, etc.

[0064] As an alternative implementation manner of step S130, since the above neural network model may further include: a second attention module, therefore, the implementation manner of further using the second attention module to process the signal reconstruction feature may include:

[0065] Step S131: Use the second attention module in the neural network model to process the signal reconstruction feature to obtain the reconstructed attention feature.

[0066] As an alternative implementation manner of step S140, the implementation manner of using the trained neural network model to make predictions based on multiple features may include:

[0067] Step S141: Concatenate multiple features to obtain a concatenated feature.

[0068] An implementation manner of the above step S141 is, for example: concatenate any two or more features among the latent space feature, the signal reconstruction feature, the reconstruction fusion feature, and the latent space fusion feature in terms of the channel dimension information to obtain a concatenated feature, and this concatenated feature is a feature that can fully represent the state of charge of the storage battery.

[0069] Step S142: Use a neural network model to predict the concatenated feature to obtain the state of charge of the storage battery.

[0070] As an implementation manner of step S142, since the above neural network model may further include: a fully connected layer, therefore, the fully connected layer can also be used to predict the concatenated feature. This implementation manner is specifically, for example: use the fully connected layer in the neural network model to predict the concatenated feature to obtain the state of charge of the storage battery.

[0071] It can be understood that before or after using the above neural network model, the neural network model can also be trained. The training process of the above neural network model is specifically, for example: collect all the sensor signals generated when the storage battery completes one start-up, and use these sensor signals as training data to train the neural network model, and adopt the mean square error loss function as the loss function of the neural network model until the accuracy rate of the neural network no longer increases or the number of iteration times (epoch) is greater than a preset threshold, then the trained neural network model can be obtained. Among them, the above preset threshold can also be set according to specific situations, such as set to 100 or 1000, etc.

[0072] In the above implementation process, through network structures such as the encoder, decoder, multiple convolutional long short-term memory networks, the first attention module, the second attention module, and the fully connected layer in the neural network model, these network structures can effectively extract and fuse multi-channel dimension information, and automatically extract high-level semantic features strongly related to the target task using the correlation of these multi-channel dimension information, thereby minimizing the influence of factors such as the number of charge and discharge cycles, charge and discharge efficiency, electrolyte temperature, battery aging, and self-discharge on the prediction of the state of charge, and improving the accuracy of predicting the state of charge of the storage battery.

[0073] Please refer to Figure 3 the structural schematic diagram of the state of charge prediction device provided by the embodiment of the present application shown; The embodiment of the present application provides a state of charge prediction device 200, including:

[0074] A spatial feature acquisition module 210, configured to acquire a plurality of sensor signals in a storage battery, perform feature extraction on the plurality of sensor signals, and obtain latent space features.

[0075] A feature extraction and reconstruction module 220, configured to extract signal reconstruction features and latent space attention features from the latent space features, and fuse the latent space attention features and the signal reconstruction features to obtain reconstructed fusion features.

[0076] A fusion feature acquisition module 230, configured to process the signal reconstruction features to obtain reconstructed attention features, and fuse the reconstructed attention features and the latent space features to obtain latent space fusion features.

[0077] A state of charge prediction module 240, configured to use a trained neural network model to make a prediction based on multiple features to obtain the state of charge of the storage battery, where the multiple features include any two or more of: latent space features, signal reconstruction features, reconstructed fusion features, and latent space fusion features.

[0078] Optionally, in an embodiment of the present application, the neural network model includes: an encoder and a decoder; the spatial feature acquisition module includes:

[0079] A spatial feature extraction module, configured to perform feature extraction on a plurality of sensor signals by using the encoder in the neural network model.

[0080] The feature extraction and reconstruction module includes:

[0081] A spatial feature reduction module, configured to use the decoder in the neural network model to perform feature reduction on the latent space features to obtain a plurality of reduced signals, and extract signal reconstruction features from the plurality of reduced signals.

[0082] An attention feature extraction module, configured to extract latent space attention features from the latent space features.

[0083] Optionally, in an embodiment of the present application, the neural network model further includes: a convolutional long short-term memory network; the spatial feature reduction module includes:

[0084] A feature fusion and reconstruction module, configured to use the convolutional long short-term memory network to perform feature extraction on the plurality of reduced signals respectively to obtain a plurality of manifestation features, and fuse and reconstruct the plurality of manifestation features to obtain signal reconstruction features.

[0085] Optionally, in an embodiment of the present application, the neural network model further includes: a first attention module; the attention feature extraction module includes:

[0086] A spatial feature processing module, configured to use the first attention module in the neural network model to process the latent space features to obtain latent space attention features.

[0087] Optionally, in the embodiment of the present application, the neural network model further includes: a second attention module; and a fusion feature acquisition module, including:

[0088] The reconstruction feature processing module is used to process the signal reconstruction features using the second attention module in the neural network model to obtain the reconstruction attention features.

[0089] Optionally, in an embodiment of the present application, the charge state prediction module includes:

[0090] The splicing feature acquisition module is used to splice multiple features to obtain splicing features.

[0091] The splicing feature prediction module is used to predict the splicing features using a neural network model to obtain the charge state of the battery.

[0092] Optionally, in the embodiment of the present application, the neural network model further includes: a fully connected layer; a splicing feature prediction module, including:

[0093] The fully connected layer prediction module is used to predict the splicing features using the fully connected layer in the neural network model.

[0094] It should be understood that the device corresponds to the aforementioned state of charge prediction method embodiment and is capable of executing each of the steps involved in the aforementioned method embodiment. The specific functions of the device can be found in the description above, and a detailed description is omitted here to avoid repetition. The device includes at least one software functional module that can be stored in a memory in the form of software or firmware or embedded in the device's operating system (OS).

[0095] See Figure 4 The electronic device 300 provided in the embodiment of the present application includes a processor 310 and a memory 320, wherein the memory 320 stores machine-readable instructions executable by the processor 310, and when the machine-readable instructions are executed by the processor 310, the method described above is performed.

[0096] The embodiment of the present application further provides a computer-readable storage medium 330 , on which a computer program is stored. When the computer program is run by the processor 310 , the above method is executed.

[0097] Among them, the computer-readable storage medium 330 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0098] It should be noted that the embodiments in this specification are all described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. For the same or similar parts among the embodiments, reference can be made to each other. For device embodiments, since they are basically similar to method embodiments, they are described relatively simply. For related parts, reference can be made to the partial description of method embodiments.

[0099] In several embodiments provided by the embodiments of the present application, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device embodiments described above are only illustrative. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architectures, functions and operations of devices, methods and computer program products according to multiple embodiments of the present application. In this regard, each block in the flowchart or block diagram may represent a module, a program segment or a part of code, and the module, program segment or part of code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may also occur in a different order from that marked in the accompanying drawings. For example, two consecutive blocks can actually be executed substantially in parallel, and they can sometimes be executed in the reverse order, which mainly depends on the functions involved.

[0100] In addition, the functional modules in each embodiment of the present application can be integrated together to form an independent part, or each module can exist alone, or two or more modules can be integrated to form an independent part. In addition, in the description of this specification, the descriptions with reference to terms such as "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" etc. mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the embodiments of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0101] In this document, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations.

[0102] The above description is only an optional implementation manner of the embodiments of the present application, but the protection scope of the embodiments of the present application is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the embodiments of the present application, and all should be covered within the protection scope of the embodiments of the present application.

Claims

1. A state of charge prediction method, characterized in that, Including: Obtain a plurality of sensor signals in the storage battery, and perform feature extraction on the plurality of sensor signals to obtain latent space features; Extract a signal reconstruction feature and a latent space attention feature from the latent space features, and fuse the latent space attention feature and the signal reconstruction feature to obtain a reconstruction fusion feature; Process the signal reconstruction feature to obtain a reconstruction attention feature, and fuse the reconstruction attention feature and the latent space feature to obtain a latent space fusion feature; Use the trained neural network model to make a prediction based on multiple features to obtain the state of charge of the storage battery; Wherein, the neural network model includes: a first attention module and a second attention module; the extracting the signal reconstruction feature and the latent space attention feature from the latent space features includes: extracting the signal reconstruction feature from the latent space features; using the first attention module in the neural network model to process the latent space features to obtain the latent space attention feature; the processing the signal reconstruction feature to obtain the reconstruction attention feature includes: using the second attention module in the neural network model to process the signal reconstruction feature to obtain the reconstruction attention feature; The using the trained neural network model to make a prediction based on multiple features includes: concatenating the reconstruction fusion feature and the latent space fusion feature to obtain a concatenated feature; using the neural network model to make a prediction on the concatenated feature to obtain the state of charge of the storage battery.

2. The method according to claim 1, wherein The neural network model includes: an encoder and a decoder; the performing feature extraction on the plurality of sensor signals includes: Using the encoder in the neural network model to perform feature extraction on the plurality of sensor signals; The extracting the signal reconstruction feature and the latent space attention feature from the latent space features includes: Using the decoder in the neural network model to perform feature restoration on the latent space features to obtain a plurality of restored signals, and extracting the signal reconstruction feature from the plurality of restored signals; Extracting the latent space attention feature from the latent space features.

3. The method according to claim 2, characterized in that, The neural network model further includes: a convolutional long short-term memory network; the extracting the signal reconstruction feature from the plurality of restored signals includes: Using the convolutional long short-term memory network to perform feature extraction on the plurality of restored signals respectively to obtain a plurality of manifestation features, and performing fusion reconstruction on the plurality of manifestation features to obtain the signal reconstruction feature.

4. The method according to claim 1, wherein The neural network model further includes: a fully connected layer; the using the neural network model to make a prediction on the concatenated feature includes: Using the fully connected layer in the neural network model to make a prediction on the concatenated feature.

5. A state of charge prediction device, characterized in that Including: A spatial feature obtaining module, configured to obtain a plurality of sensor signals in the storage battery, and perform feature extraction on the plurality of sensor signals to obtain latent space features; A feature extraction and reconstruction module, configured to extract a signal reconstruction feature and a latent space attention feature from the latent space feature, and fuse the latent space attention feature and the signal reconstruction feature to obtain a reconstructed fusion feature; A fusion feature obtaining module, configured to process the signal reconstruction feature to obtain a reconstructed attention feature, and fuse the reconstructed attention feature and the latent space feature to obtain a latent space fusion feature; A charge state prediction module, configured to use a trained neural network model to make a prediction based on multiple features to obtain the charge state of the storage battery; Wherein, the neural network model includes: a first attention module and a second attention module; the extracting the signal reconstruction feature and the latent space attention feature from the latent space feature includes: extracting the signal reconstruction feature from the latent space feature; using the first attention module in the neural network model to process the latent space feature to obtain the latent space attention feature; the processing the signal reconstruction feature to obtain the reconstructed attention feature includes: using the second attention module in the neural network model to process the signal reconstruction feature to obtain the reconstructed attention feature; The using the trained neural network model to make a prediction based on multiple features includes: concatenating the reconstructed fusion feature and the latent space fusion feature to obtain a concatenated feature; using the neural network model to make a prediction on the concatenated feature to obtain the charge state of the storage battery.

6. An electronic device, characterized in that, Comprising: A processor and a memory, the memory stores machine-readable instructions executable by the processor, and when the machine-readable instructions are executed by the processor, the method according to any one of claims 1 to 4 is executed.

7. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium, and when the computer program is run by a processor, the method according to any one of claims 1 to 4 is executed.

Images