Electric quantity determination method and device of vehicle-mounted battery system, equipment, medium and product

Abstract

The invention provides an electric quantity determination method and device for a vehicle-mounted battery system, equipment, a medium and a product, which are used for improving the accuracy of predicted residual electric quantity in the vehicle-mounted battery system. The method comprises the following steps: acquiring historical data; wherein the historical data comprises electric quantity data, and the electric quantity data comprises first residual electric quantity corresponding to historical time; processing the electric quantity data through a first convolutional network in a first target model to obtain a first feature matrix, and processing the electric quantity data through a gated loop network in the first target model to obtain a second feature matrix; fusing the first feature matrix and the second feature matrix, and based on a third feature matrix obtained by fusion, determining a second residual electric quantity of the target time; wherein the target time is after the historical time.

Description

Technical Field

[0001] The present application relates to the field of vehicle technology, and in particular to a method, device, equipment, medium, and product for determining the power level of an on-board battery system. Background Art

[0002] In new energy vehicles, the battery pack, as the primary energy storage component and power source, plays a crucial role in vehicle safety and driving experience. The remaining charge, which reflects the battery's endurance, is a key parameter describing its status. During driving, remaining charge prediction can be used to match the vehicle with an appropriate power mode or plan a charging schedule. However, the accuracy of current remaining charge prediction methods still needs to be improved. Summary of the Invention

[0003] Based on this, it is necessary to provide a method, device, equipment, medium and product for determining the power level of an on-board battery system to address the above technical problems, so as to improve the accuracy of the predicted remaining power level in the on-board battery system.

[0004] In a first aspect, an embodiment of the present application provides a method for determining the power level of an on-board battery system, comprising:

[0005] Acquire historical data; wherein the historical data includes power data, and the power data includes a first remaining power corresponding to a historical time;

[0006] Processing the power data through a first convolutional network in a first target model to obtain a first feature matrix, and processing the power data through a gated recurrent network in the first target model to obtain a second feature matrix;

[0007] The first characteristic matrix and the second characteristic matrix are fused, and based on a third characteristic matrix obtained by the fusion, a second remaining power at a target time is determined; wherein the target time is after the historical time.

[0008] In some embodiments, determining the second remaining power at the target time based on the fused third characteristic matrix includes:

[0009] Performing feature mapping processing on the third feature matrix to obtain a fourth feature matrix;

[0010] The fourth feature matrix is ​​input into a fully connected layer to obtain the second remaining power.

[0011] In some embodiments, the first convolutional network includes a Ghost module, a first convolutional layer, and a global pooling layer;

[0012] The first convolutional network in the first target model processes the power data to obtain a first feature matrix, including:

[0013] Processing the power data through the Ghost module to obtain a fifth characteristic matrix;

[0014] Processing the fifth characteristic matrix through the first convolutional layer to obtain a sixth characteristic matrix;

[0015] The sixth feature matrix is ​​processed by the global pooling layer to obtain the first feature matrix.

[0016] In some embodiments, processing the power data by the Ghost module to obtain a fifth characteristic matrix includes:

[0017] Processing the power data through the second convolutional layer in the Ghost module and compressing it to obtain an intrinsic feature matrix;

[0018] Performing linear transformation on the inherent characteristic matrix to obtain a Ghost characteristic matrix;

[0019] The intrinsic characteristic matrix and the Ghost characteristic matrix are concatenated to obtain the fifth characteristic matrix.

[0020] In some embodiments, the historical data further includes driving behavior data of the vehicle, the driving behavior data corresponding to the historical time;

[0021] After obtaining the historical data, the method further includes:

[0022] Based on a target graph convolutional network, a weight matrix corresponding to the driving behavior data is generated; wherein the weight matrix is ​​used to perform feature calibration processing on the third feature matrix to obtain the second remaining power.

[0023] In some embodiments, generating a weight matrix corresponding to the driving behavior data based on a target graph convolutional network includes:

[0024] Converting the driving behavior data into an adjacency matrix and a node feature matrix respectively;

[0025] The adjacency matrix and the node feature matrix are processed by the target graph convolutional network to generate the weight matrix.

[0026] In some embodiments, fusing the first characteristic matrix and the second characteristic matrix and determining the second remaining power at the target time based on a third characteristic matrix obtained by the fusion includes:

[0027] concatenating the first characteristic matrix and the second characteristic matrix to obtain the third characteristic matrix;

[0028] Performing a dot product process on the third characteristic matrix and the weight matrix to obtain a seventh characteristic matrix;

[0029] The second remaining power is obtained based on the seventh characteristic matrix.

[0030] In a second aspect, an embodiment of the present application provides a device for determining the power level of an on-board battery system, comprising:

[0031] A history module, configured to obtain historical data; wherein the historical data includes power data, and the power data includes a first remaining power corresponding to a historical time;

[0032] a feature module, configured to process the power data through a first convolutional network in a first target model to obtain a first feature matrix, and to process the power data through a gated recurrent network in the first target model to obtain a second feature matrix;

[0033] The power module is configured to fuse the first characteristic matrix and the second characteristic matrix, and determine a second remaining power at a target time based on a third characteristic matrix obtained by the fusion; wherein the target time is after the historical time.

[0034] In some embodiments, the power module is specifically configured to perform feature mapping processing on the third feature matrix to obtain a fourth feature matrix; and input the fourth feature matrix into a fully connected layer to obtain the second remaining power.

[0035] In some embodiments, the first convolutional network includes a Ghost module, a first convolutional layer and a global pooling layer; the feature module is specifically used to process the power data through the Ghost module to obtain a fifth feature matrix; process the fifth feature matrix through the first convolutional layer to obtain a sixth feature matrix; and process the sixth feature matrix through the global pooling layer to obtain the first feature matrix.

[0036] In some embodiments, the feature module is specifically used to process the power data through the second convolutional layer in the Ghost module and compress it to obtain an inherent feature matrix; perform linear transformation on the inherent feature matrix to obtain a Ghost feature matrix; and splice the inherent feature matrix and the Ghost feature matrix to obtain the fifth feature matrix.

[0037] In some embodiments, the historical data also includes driving behavior data of the vehicle, and the driving behavior data corresponds to the historical time; the device also includes a weight module for generating a weight matrix corresponding to the driving behavior data based on a target graph convolutional network; wherein the weight matrix is ​​used to perform feature calibration processing on the third feature matrix to obtain the second remaining power.

[0038] In some embodiments, the weight module is specifically used to convert the driving behavior data into an adjacency matrix and a node feature matrix respectively; and process the adjacency matrix and the node feature matrix through the target graph convolutional network to generate the weight matrix.

[0039] In some embodiments, the power module is further used to concatenate the first characteristic matrix and the second characteristic matrix to obtain the third characteristic matrix; perform dot multiplication on the third characteristic matrix and the weight matrix to obtain a seventh characteristic matrix; and obtain the second remaining power based on the seventh characteristic matrix.

[0040] In a third aspect, an embodiment of the present application provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and for running on the processor, characterized in that when the processor executes the computer program, the method described in the first aspect and any embodiment is implemented.

[0041] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method described in the first aspect and any embodiment is implemented.

[0042] In a fifth aspect, an embodiment of the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the method described in the first aspect and any embodiment.

[0043] In the method for determining the power level of an on-board battery system provided in an embodiment of the present application, the power level data in the historical data is processed separately by a first convolutional network and a gated recurrent network. The first convolutional network can work together to efficiently focus on local features in the power level data, exhibiting characteristics of better robust performance. The gated recurrent network can alleviate the vanishing gradient and capture the characteristics of long-term dependent information, effectively improving the accuracy of the predicted second remaining power level, thereby improving driving safety performance and user experience.

[0044] Other features and advantages of the present application will be described in the following description and, in part, will become apparent from the description or may be learned through practice of the present application. The objectives and other advantages of the present application may be achieved and obtained through the structures particularly pointed out in the written description, claims, and drawings. It should be understood that the above general description and the detailed description that follow are merely exemplary and explanatory and do not limit the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0046] Figure 1 A diagram illustrating an application environment of a method for determining the power level of an on-board battery system provided in an embodiment of the present application;

[0047] Figure 2 A flowchart of a method for determining the power level of an on-board battery system provided in an embodiment of the present application;

[0048] Figure 3 A schematic diagram of the structure of the first convolutional network provided in an embodiment of the present application;

[0049] Figure 4 A schematic diagram of the structure of the second target network provided in an embodiment of the present application;

[0050] Figure 5 Schematic diagram of information flow in a gated recurrent network when the gated recurrent network provided in an embodiment of the present application is a Bi-LSTM;

[0051] Figure 6 A schematic diagram of the structure of a device for determining the power level of an on-board battery system according to one embodiment;

[0052] Figure 7 FIG. 1 is a diagram showing the internal structure of an electronic device in one embodiment. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0054] It should be noted that the diagrams provided in the present embodiment are only schematic illustrations of the basic concept of the present application. The diagrams only show the components related to the present application rather than the number, shape and size of the components when actually implemented. The type, quantity and ratio of each component can be changed at will during actual implementation, and the component layout pattern may also be more complicated. The structures, ratios, sizes, etc. illustrated in the drawings of this specification are only used to match the content disclosed in the specification for people familiar with this technology to understand and read. They are not used to limit the restrictive conditions that can be implemented in this application. Therefore, they have no technical significance. Any modification of the structure, change of the proportional relationship or adjustment of the size should still fall within the scope of the technical content disclosed in this application without affecting the effect and purpose that can be achieved by this application. At the same time, the terms such as "upper", "lower", "left", "right", "middle" and "one" quoted in this specification are only for the convenience of description and are not used to limit the scope of the implementation of this application. The change or adjustment of their relative relationship should also be considered as the scope of the implementation of this application without substantial change in the technical content.

[0055] References herein to "embodiments" mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various places herein does not necessarily refer to the same embodiment, nor does it constitute an independent or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0056] As used herein, unless the context clearly indicates otherwise, the terms "a," "an," "an," and / or "the" are not intended to refer to the singular but may include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements specifically identified, and these steps and elements do not constitute an exclusive list. A method or apparatus may also include additional steps or elements.

[0057] The definition of inclusion herein, such as the terms “having”, “may have”, “include” or “may include” as used herein, indicates the existence of the corresponding functions, operations, elements, etc. herein, and does not limit the existence of one or more other functions, operations, elements, etc. In addition, it should be understood that the terms “including” or “having” as used herein indicate the existence of the features, numbers, steps, operations, elements, components or their combination described in the specification, and do not exclude the existence or addition of one or more other features, numbers, steps, operations, elements, components or their combination.

[0058] In the embodiments of the present application, prefixes such as "first" and "second" are used only to distinguish different description objects and have no limiting effect on the position, order, priority, quantity or content of the described objects. In the embodiments of the present application, the use of prefixes such as ordinal numbers to distinguish description objects does not constitute a restriction on the described objects. For the statement of the described + object, please refer to the description in the context of the claims or embodiments, and no unnecessary restrictions should be constituted due to the use of such prefixes. In addition, in the description of this embodiment, unless otherwise specified, the meaning of "plurality" is two or more.

[0059] The method for determining the amount of power of a vehicle battery system provided in this application can be applied to Figure 1 In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The terminal 102 can query the historical data on the vehicle's power consumption, so as to realize the prediction of power and obtain the second remaining power by performing feature calculation based on the historical data through the deep learning model deployed in the terminal 102 and including the first target model. The server 104 can maintain the deep learning model (for example, the first target model) in the terminal through the network; for example, the server 104 can send an optimization instruction to the terminal 102 at intervals of corresponding time lengths according to preset rules, so that the terminal 102 can optimize the deep learning model running on the terminal 102 according to the optimization instruction and the actual remaining power corresponding to the target time of the predicted second remaining power, thereby further improving the generalization of the deep learning model and improving the accuracy of the predicted second remaining power.

[0060] The terminal 102 may be, but is not limited to, any terminal capable of acquiring historical data of the vehicle in real time, such as a vehicle-mounted terminal or a driving computer, and the server 104 may be implemented as an independent server or a server cluster consisting of multiple servers.

[0061] In the embodiment of the present application, the vehicle-mounted battery system may be a power battery pack including one or more battery modules.

[0062] In one embodiment, Figure 2 As shown, a method for determining the power of a vehicle battery system is provided, and the method is applied to Figure 1 The following steps are used as an example to illustrate the terminal in the figure:

[0063] Step 201: Obtain historical data.

[0064] The historical data includes power data, and the power data includes a first remaining power corresponding to a historical time.

[0065] Specifically, the historical time may include a single moment, or multiple moments with time intervals. For example, in a 24-hour format, the historical times include 16:00, 16:20, 16:40, etc.

[0066] In one embodiment, if the historical time includes multiple moments, among the vehicle states corresponding to any two adjacent moments, the vehicle state corresponding to at least one moment is a running state.

[0067] The first remaining capacity is a first SOC (State of Charge), which is obtained by querying a BMS (Battery Management System) at a corresponding historical time.

[0068] The aforementioned power data may be power data of the vehicle battery system. For example, when the vehicle battery system is a power battery pack, the aforementioned power data is power data of the power battery pack.

[0069] The power data may include a first remaining power corresponding to a historical time, where the first remaining power is the remaining power of the power battery pack at the historical time.

[0070] In step 202 , the power data is processed by the first convolutional network in the first target model to obtain a first feature matrix, and the power data is processed by the gated recurrent network in the first target model to obtain a second feature matrix.

[0071] Specifically, the power data can be input into a first target model, which is a pre-trained deep learning model.

[0072] The gated recurrent network may include multiple cell states (memory cells) and be configured based on a gating mechanism. The cell states can be used to store and update characteristic information in the power data. The gating mechanism can control the flow of the aforementioned characteristic information based on a forget gate, an input gate, and an output gate to capture the long-term dependency of the first remaining power in the power data.

[0073] In one embodiment, the gated recurrent network may be selected from: LSTM (Long Short-Term Memory) or Bi-LSTM (Bidirectional Long Short-Term Memory).

[0074] Preferably, the above-mentioned gated recurrent network is a Bi-LSTM.

[0075] Furthermore, in an embodiment of the present application, the first convolutional network in the first target model can be connected in parallel with the gated recurrent network. Thus, after the power data in the historical data is input into the first target model, the first convolutional network and the gated recurrent network can respectively process the power data to obtain the aforementioned first and second feature matrices.

[0076] The implementation method of connecting the first convolutional network and the gated recurrent network in parallel to process the power data separately can effectively improve the prediction accuracy of the second remaining power obtained in step 203; it effectively avoids the situation where the local features in the first convolutional network are over-smoothed due to the unidirectional flow of features when the first convolutional network and the gated recurrent network are connected in series, and the gated recurrent network cannot supplement the lost details, resulting in the defect of over-smoothing of the aforementioned local features being further amplified, thereby resulting in insufficient prediction accuracy of the second remaining power in step 203.

[0077] Step 203 : The first characteristic matrix and the second characteristic matrix are integrated, and a second remaining power at the target time is determined based on a third characteristic matrix obtained by the integration.

[0078] Specifically, the above target time is after the historical time.

[0079] The target time can be a plurality of consecutive moments. The target time can also be a designated moment separated from the latest moment in the aforementioned plurality of historical times by a preset time length.

[0080] In one embodiment, the second remaining power can be obtained in the following manner: first, feature mapping is performed on the third feature matrix to obtain a fourth feature matrix. Then, the fourth feature matrix is ​​input into a fully connected layer to obtain the second remaining power.

[0081] In the aforementioned method for determining the charge level of an onboard battery system, the first convolutional network in the first target model captures the local features of the charge data, thereby extracting a first feature matrix. Simultaneously, the gated recurrent network in the first target model captures the long-term temporal dependency features of the first remaining charge in the charge data, thereby obtaining a second feature matrix corresponding to the global features. This fusion of the first and second feature matrices achieves synergistic effects between local and global features, improving the accuracy of the predicted second remaining charge.

[0082] It is understandable that the target time is after the execution time of the method for determining the power level of the vehicle battery system described in steps 210 to 203, thereby achieving the second remaining power prediction. In terms of time sequence, the historical time is before the execution time, and the target time is after the execution time.

[0083] In one embodiment, the first convolutional network includes a first convolutional layer and a global pooling layer; wherein the output segment of the first convolutional layer is connected to the input of the global pooling layer. By inputting the power data into the first convolutional layer, a first feature matrix can be obtained from the output of the global pooling layer.

[0084] To further improve the accuracy of the second remaining power, in one embodiment, a Ghost Module (ghost module) can be embedded in the first convolutional network and placed in front of the first convolutional network to reduce the computational load of the first convolutional network. At the same time, the first convolutional layer can be placed in the back to ensure that the first convolutional network has sufficient expression capabilities and efficiently outputs a first feature matrix that can accurately reflect the feature information in the power data. The following is a detailed description:

[0085] Please refer to Figure 3 After the power data is input into the first convolutional network, it can be processed by the Ghost module to obtain the fifth feature matrix. Then, the output of the Ghost module is connected to the input of the first convolutional layer, and the output of the first convolutional layer is connected to the input of the global pooling layer.

[0086] The global pooling layer may be a global maximum pooling layer or a global average pooling layer. Thus, the fifth feature matrix may be processed by the first convolutional layer to obtain a sixth feature matrix. Finally, the sixth feature matrix may be processed by the global average pooling layer to obtain the first feature matrix.

[0087] Preferably, in the above-mentioned first convolutional network, the global pooling layer connected to the first convolutional layer is a global average pooling layer to adapt to the continuity and dynamics of the first remaining power corresponding to the historical time in the power number, and reduce overfitting, thereby improving the robustness of the first convolutional network.

[0088] Further, the following is an explanation of how the Ghost module processes power data to output the fifth characteristic matrix. Please continue to refer to Figure 3 :

[0089] The Ghost module includes a second convolutional layer, which is used to extract features from the power data. Therefore, when the power data is input into the first convolutional network, it can first be processed and compressed by the second convolutional layer to obtain an intrinsic feature matrix of a specified dimension. Then, the intrinsic feature matrix is ​​linearly transformed to obtain the Ghost feature matrix. Finally, the intrinsic feature matrix and the Ghost feature matrix can be spliced ​​in a preset order to obtain a fifth feature matrix.

[0090] To further improve the accuracy of the second remaining battery level, in one embodiment, the historical data also includes vehicle driving behavior data, which is processed together with the battery level data by a corresponding deep learning model. The driving behavior data corresponds to historical time. Therefore, the driving behavior data corresponds to the battery level data.

[0091] The driving behavior data may include driving modes and pedal information corresponding to historical time and driving modes. The driving behavior data in the historical data can be processed in the following manner: First, a weight matrix corresponding to the driving behavior data can be generated based on a target graph convolutional network. This weight matrix is ​​used to perform feature calibration on the third feature matrix obtained by the aforementioned fusion feature to obtain the second remaining battery capacity. The target graph convolutional network is a pre-trained graph convolutional network.

[0092] In one embodiment, the driving mode in the driving behavior data may include a first mode, a second mode, a third mode, etc. The pedal information may include at least one of an accelerator pedal change rate, an average accelerator pedal change rate, a maximum accelerator pedal change rate, an average accelerator pedal stroke, a standard deviation of accelerator pedal stroke, and a preset mileage energy consumption. The preset mileage energy consumption may be, for example, energy consumption per 100 kilometers (unit: kilowatt-hour).

[0093] The following table 1 illustrates the corresponding relationship between driving mode, historical time, and pedal information in driving behavior data.

[0094] Table 1

[0095]

[0096] Note: The collection of data in the table has been authorized by relevant users and complies with relevant laws and regulations.

[0097] It should be noted that the average accelerator pedal travel, standard deviation of accelerator pedal travel, average accelerator pedal change rate, and maximum accelerator pedal change rate in Table 1 above represent pedal information. Furthermore, the driving modes corresponding to each historical time, such as the first mode, the second mode, the third mode, and the fourth mode, may be the same or different.

[0098] Exemplarily, the first mode in the above driving modes is the soothing mode, the second mode in the driving mode is the comfort mode (also called the "standard mode"), the third mode in the driving mode is the aggressive mode (also called the "sports mode"), and the fourth mode in the driving mode is the dedicated high-speed mode.

[0099] The following is an example to illustrate the generation of the aforementioned weight matrix: First, the driving behavior data is converted into an adjacency matrix and a node feature matrix respectively. The adjacency matrix is ​​used to define the graph structure corresponding to the driving behavior data. The node feature matrix is ​​used to describe the attribute information of the nodes in the graph structure. In the aforementioned graph structure, the node is the driving mode in the driving behavior data, and the attribute information of the node is the pedal information corresponding to the historical time in the corresponding driving mode. After determining the adjacency matrix and the node feature matrix, the adjacency matrix and the node feature matrix can be processed by the target convolutional network to generate a weight matrix.

[0100] To further extract effective information from the driving behavior data that reflects or influences changes in the battery charge in the vehicle battery system, and to avoid introducing excessive redundant information into the third feature matrix obtained by the aforementioned fusion due to the low correlation between the driving behavior data and the endurance of the vehicle battery system, in one embodiment, after generating a weight matrix using a graph convolutional network, the second remaining battery charge is determined using the following method:

[0101] First, the first characteristic matrix and the second characteristic matrix are concatenated to achieve fusion of the first characteristic matrix and the second characteristic matrix to obtain a third characteristic matrix.

[0102] Then, a dot product is performed on the third feature matrix and the weight matrix to calibrate the feature information in the third feature matrix through the weight matrix to obtain a seventh feature matrix, wherein the third feature matrix and the weight matrix have the same dimension.

[0103] Finally, the second remaining power can be obtained based on the seventh feature matrix. Specifically, the seventh feature matrix is ​​subjected to feature mapping to obtain the eighth feature matrix. The eighth feature matrix is ​​then input into the fully connected layer to obtain the second remaining power at the target time.

[0104] An embodiment is provided below to further illustrate the aforementioned method for predicting the power level of the vehicle-mounted power battery system.

[0105] Please refer to Figure 4 , the historical data consists of the power data and driving behavior data corresponding to the historical time. The first target model is composed of a gated recurrent network and a first convolutional network. The gated recurrent network is a Bi-LSTM, and the first convolutional network includes a Ghost module, a first convolutional layer, and a global average pooling layer in sequence. Among them, the first convolutional layer is a 1-dimensional convolutional layer. Therefore, in this embodiment, the aforementioned first target model, Bi-LSTM, GCN (Graph Convolutional Network), feature mapping block, and fully connected layer can constitute the second target model.

[0106] When historical data is input into the second target model, since the historical data includes power data with a first identifier and driving behavior data with a second identifier, the power data can be input into the first target model based on the first identifier. In the first target model, the first convolutional network first processes the power data using the Ghost module and inputs the output fifth feature matrix into the first convolutional layer. The first convolutional layer outputs the sixth feature matrix, which is input into the global average pooling layer to obtain the first feature matrix. The dimension of this first feature matrix is ​​m1.

[0107] At the same time, in the first target model, the first remaining power corresponding to the historical time in the power data is processed by Bi-LSTM to obtain the second feature matrix. The Bi-LSTM consists of two forward LSTMs and backward LSTMs with opposite time directions. The structure of the Bi-LSTM can be found in Figure 5 The forward LSTM processes the input information in chronological order (t=1→N), and the backward LSTM processes the input information in reverse chronological order (t=N→1). Figure 5 As shown, the input vector sequence X is X=[x 1 , x 2 ,…,x N ], the input vector sequence X is passed through the forward LSTM to generate the forward hidden state sequence h → ;h → =[(h → ) 1 ,(h → ) 2 ,…,(h → ) N The input vector sequence x is passed through the backward LSTM to generate the backward hidden state sequence h ← ;h ← =(h ← ) 1 ,(h ← ) 2 ,…,(h ← ) N ]. In the output layer, the forward hidden state vector (h → ) n , and the backward hidden state in the backward hidden state sequence (h ← ) n Corresponding fusion, finally forming the output vector sequence The output vector sequence Corresponding to the aforementioned second characteristic matrix, in this embodiment, the dimension of the second characteristic matrix is ​​m2.

[0108] The first characteristic matrix and the second characteristic matrix are concatenated to obtain a third characteristic matrix of dimension m1+m2.

[0109] Next, corresponding to the power data with the first identifier, driving behavior data can be extracted from the historical data based on the second identifier. Then, according to the preset rules, the driving behavior data is converted into an adjacency matrix and a node feature matrix. In the graph structure corresponding to the adjacency matrix and the node feature matrix, the nodes correspond to a soothing driving mode, an intense driving mode, a moderate driving mode, and a dedicated high-speed driving mode, respectively. Moreover, the attribute information in each node is the pedal information of multiple historical times. The pedal information includes at least the accelerator pedal change rate, the average accelerator pedal change rate, the maximum accelerator pedal change rate, the average accelerator pedal stroke, the standard deviation of the accelerator pedal stroke, and the energy consumption per 100 kilometers.

[0110] Thus, the adjacency matrix and node feature matrix are input into the target GCN to obtain a weight matrix. The dimension of this weight matrix is ​​m3, where m3 = m1 + m2.

[0111] Finally, the third feature matrix and the weight matrix are multiplied by points, and the seventh feature matrix obtained by the point multiplication is input into the feature mapping block and the fully connected layer, finally obtaining the second remaining power at the target time. It can be understood that the target time is after the historical time and the interval between it and the latest historical time in the historical time is less than or equal to the preset time interval.

[0112] Furthermore, the second target model of the second remaining power obtained by processing historical data can be obtained by pre-training. During the training stage, in the to-be-trained model corresponding to the second target model, except for the post-positioned loss function module, the structure of the to-be-trained model and the connection relationship between the substructures therein are consistent with the second target model. The loss function module is used to evaluate the degree of deviation between the output of the to-be-trained model and the true label through the loss function therein, so as to calculate the gradient information of each parameter using the backpropagation algorithm. These gradient information can be in the form of gradient signals in the to-be-trained model, and can be passed back layer by layer through the chain rule to drive the optimizer (such as SGD, Adam, etc.) to dynamically adjust the network weights in the to-be-trained model along the negative gradient direction, so that as the iteration steps of the to-be-trained model accumulate, the deviation between the output of the to-be-trained model and the true label gradually decreases, thereby achieving improved accuracy, until the to-be-trained model meets the preset conditions and obtains the second target model. The preset condition can be that the iteration step meets the first threshold, or the loss function is less than the respective second threshold.

[0113] Furthermore, when training the target model corresponding to the second target model, the hyperparameters are set as follows: the number of training samples per batch is 28,000, the total number of training epochs is 100, and the Adam optimizer is used. The initial learning rate is set to 0.1. The learning rate is updated by multiplying it by 0.8 every 10 training epochs. In addition, the number of layers of the first convolutional layer in the first convolutional network in the second target model is set to 2, the output dimensions are 16 and 32 respectively, the convolution kernel size is 3, and the stride is set to SAME; the output dimension of the Bi-LSTM network is set to 32. The number of convolutional layers in the GCN is set to 2, and the GCN output feature dimension is set to 64.

[0114] Then, the actual driving data of any four E3 small pure electric vehicles during their journey from March 15, 2022 to December 1, 2022 are collected as a validation dataset to verify the performance of the above second objective model.

[0115] The second target model was set up using TensorFlow (version 2.5.0). This model was run and trained using CUDA (Compute Unified Device Architecture) version 11.3, accelerated on an NVIDIA Graphics Processing Unit (NVIDIA) GPU. Performance test data for the second target model based on these settings is shown in Table 2.

[0116] Table 2

[0117] MAE RMSE R-squared Runtimes(s) 0.0052 0.0064 98.95％ 229.25

[0118] In Table 2, MAE (Mean Absolute Error) is used to measure the error between the predicted value and the true value.

[0119] RMSE (Root Mean Square Error) is obtained by taking the square root of the average of the squares of the errors between the predicted value and the true value.

[0120] R-squared (Coefficient of Determination) indicates how well the independent variable explains the dependent variable. The R-squared for this second target model is 98.95%, close to 1, indicating a good fit to the historical data.

[0121] Runtimes: The runtime of the second target model is 229.25 seconds, which has the advantage of high efficiency.

[0122] The performance test data for the second objective model in Table 2 demonstrates that, when predicting the battery capacity of an onboard battery system based on historical user data, the second objective model provided by the present embodiment can effectively improve the accuracy of power prediction. Furthermore, the second objective model boasts high processing efficiency, enabling real-time feedback and updating of power prediction data.

[0123] It should be noted that in the embodiments of the present application, the collection of personal data such as user data or driving data of the vehicle used by the user (such as historical data containing battery data, pedal information, etc.) is authorized by the user and complies with relevant laws and regulations.

[0124] It should be understood that although Figure 2 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. In addition, Figure 2 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0125] Based on the same inventive concept, Figure 6 As shown, the embodiment of the present application provides a device for determining the power level of an on-board battery system, including: a history module 601, a feature module 602, and a power level module 603, wherein:

[0126] The history module 601 is used to obtain historical data.

[0127] Wherein, the historical data includes power data, and the power data includes a first remaining power corresponding to a historical time;

[0128] A feature module 602 is configured to process the power data using a first convolutional network in a first target model to obtain a first feature matrix, and to process the power data using a gated recurrent network in the first target model to obtain a second feature matrix;

[0129] The power module 603 is configured to fuse the first characteristic matrix and the second characteristic matrix, and determine a second remaining power at a target time based on a third characteristic matrix obtained by the fusion, wherein the target time is after the historical time.

[0130] In one embodiment, the first convolutional network includes a Ghost module, a first convolutional layer, and a global pooling layer; the feature module 602 is specifically configured to:

[0131] The power data is processed by the Ghost module to obtain a fifth characteristic matrix; the fifth characteristic matrix is ​​processed by the first convolutional layer to obtain a sixth characteristic matrix; and the sixth characteristic matrix is ​​processed by the global pooling layer to obtain the first characteristic matrix.

[0132] In one embodiment, the feature module 602 is specifically configured to:

[0133] The power data is processed and compressed by the second convolutional layer in the Ghost module to obtain an intrinsic feature matrix; the intrinsic feature matrix is ​​linearly transformed to obtain a Ghost feature matrix; and the intrinsic feature matrix and the Ghost feature matrix are concatenated to obtain the fifth feature matrix.

[0134] In one embodiment, the power module 603 is specifically used to:

[0135] Perform feature mapping processing on the third feature matrix to obtain a fourth feature matrix; input the fourth feature matrix into a fully connected layer to obtain the second remaining power.

[0136] In one embodiment, the power module 603 is further configured to:

[0137] The first characteristic matrix and the second characteristic matrix are concatenated to obtain the third characteristic matrix; a dot product is performed on the third characteristic matrix and the weight matrix to obtain a seventh characteristic matrix; and the second remaining power is obtained based on the seventh characteristic matrix.

[0138] In one embodiment, the historical data also includes driving behavior data of the vehicle, and the driving behavior data corresponds to the historical time; the device also includes a weight module for generating a weight matrix corresponding to the driving behavior data based on a target graph convolutional network; wherein the weight matrix is ​​used to perform feature calibration processing on the third feature matrix to obtain the second remaining power.

[0139] In one embodiment, the weight module is specifically used to convert the driving behavior data into an adjacency matrix and a node feature matrix respectively; and process the adjacency matrix and the node feature matrix through the target graph convolutional network to generate the weight matrix.

[0140] For the specific definition of the power determination device of the vehicle battery system, please refer to the definition of the power determination method of the vehicle battery system above, which will not be repeated here. The various modules in the above-mentioned power determination device of the vehicle battery system can be implemented in whole or in part by software, hardware, and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the electronic device in the form of hardware, or can be stored in the memory of the electronic device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0141] Based on the same inventive concept, see Figure 7 The present application also provides an electronic device. In one embodiment, the electronic device may include a memory 701, a communication module 703, and one or more processors 702 as shown in the figure.

[0142] The memory 701 is used to store computer programs executed by the processor 702. The memory 701 may mainly include a program storage area and a data storage area. The program storage area may store an operating system; the data storage area may store various operating instruction sets.

[0143] Memory 701 may be a volatile memory, such as random-access memory (RAM); a non-volatile memory, such as read-only memory, flash memory, a hard disk drive (HDD), or a solid-state drive (SSD); or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but is not limited thereto. Memory 701 may be a combination of the above memories.

[0144] The processor 702 may include one or more central processing units (CPUs) or digital processing units, etc. The processor 702 is configured to implement the above-mentioned method for determining the power level of the vehicle-mounted battery system when calling the computer program stored in the memory 701 .

[0145] The communication module 703 is used to communicate with terminal devices, site devices or other network devices.

[0146] The specific connection medium between the memory 701, the communication module 703 and the processor 702 is not limited in the embodiment of the present application. Figure 7 In the embodiment, the memory 701 and the processor 702 are connected via a bus 704. Figure 7 The connections between the other components are shown in bold lines, which are only for illustration and are not intended to be limiting. The bus 704 can be divided into an address bus, a data bus, a control bus, etc. For ease of description, Figure 7 The diagram shows a single thick line, but this does not indicate that there is only one bus or one type of bus.

[0147] The memory 701 stores a computer storage medium, which stores computer-executable instructions for implementing the method for determining the power level of an onboard battery system according to an embodiment of the present application. The processor 702 is configured to execute the method for determining the power level of an onboard battery system according to each embodiment of the computer-executable instructions.

[0148] In one embodiment, when the computer executable instructions are executed by the processor 702, the following steps may be implemented:

[0149] Acquire historical data; wherein the historical data includes power data, and the power data includes a first remaining power corresponding to a historical time;

[0150] Processing the power data through a first convolutional network in a first target model to obtain a first feature matrix, and processing the power data through a gated recurrent network in the first target model to obtain a second feature matrix;

[0151] The first characteristic matrix and the second characteristic matrix are fused, and based on a third characteristic matrix obtained by the fusion, a second remaining power at a target time is determined; wherein the target time is after the historical time.

[0152] Those skilled in the art will understand that Figure 7 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the electronic device to which the solution of the present application is applied. The specific electronic device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0153] Based on the same inventive concept, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the following steps are implemented:

[0154] Acquire historical data; wherein the historical data includes power data, and the power data includes a first remaining power corresponding to a historical time;

[0155] Processing the power data through a first convolutional network in a first target model to obtain a first feature matrix, and processing the power data through a gated recurrent network in the first target model to obtain a second feature matrix;

[0156] The first characteristic matrix and the second characteristic matrix are fused, and based on a third characteristic matrix obtained by the fusion, a second remaining power at a target time is determined; wherein the target time is after the historical time.

[0157] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0158] Based on the same inventive concept, an embodiment of the present application further provides a computer program product, including a computer program, which, when executed by a processor, implements any of the above-mentioned methods for determining the power level of a vehicle-mounted battery system.

[0159] The program code for executing the computer program product of the present application may be written in any combination of one or more programming languages, and the program code may be executed entirely on the user device, partially on the user device, as an independent software package, partially on the user device and partially on a remote device, or entirely on the remote device.

[0160] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0161] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0162] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0163] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of user-operated steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable device to implement the process. Figure 1 a process or multiple processes and / or boxes Figure 1 The steps for the function specified in one or more boxes.

[0164] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art could make various modifications and improvements without departing from the spirit of the present application, all of which fall within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be determined by the appended claims.

Claims

1. A method for determining the power level of a vehicle-mounted battery system, characterized in that: include: Acquire historical data; wherein the historical data includes power data, and the power data includes a first remaining power corresponding to a historical time; Processing the power data through a first convolutional network in a first target model to obtain a first feature matrix, and processing the power data through a gated recurrent network in the first target model to obtain a second feature matrix; The first characteristic matrix and the second characteristic matrix are fused, and based on a third characteristic matrix obtained by the fusion, a second remaining power at a target time is determined; wherein the target time is after the historical time.

2. The method according to claim 1, wherein The first convolutional network includes a Ghost module, a first convolutional layer and a global pooling layer; The first convolutional network in the first target model processes the power data to obtain a first feature matrix, including: Processing the power data through the Ghost module to obtain a fifth characteristic matrix; Processing the fifth characteristic matrix through the first convolutional layer to obtain a sixth characteristic matrix; The sixth feature matrix is ​​processed by the global pooling layer to obtain the first feature matrix.

3. The method according to claim 2, wherein The fifth characteristic matrix is ​​obtained by processing the power data through the Ghost module, including: Processing the power data through the second convolutional layer in the Ghost module and compressing it to obtain an intrinsic feature matrix; Performing linear transformation on the inherent characteristic matrix to obtain a Ghost characteristic matrix; The intrinsic characteristic matrix and the Ghost characteristic matrix are concatenated to obtain the fifth characteristic matrix.

4. The method according to claims 1 to 3, characterized in that The historical data also includes driving behavior data of the vehicle, and the driving behavior data corresponds to the historical time; After obtaining the historical data, the method further includes: Based on a target graph convolutional network, a weight matrix corresponding to the driving behavior data is generated; wherein the weight matrix is ​​used to perform feature calibration processing on the third feature matrix to obtain the second remaining power.

5. The method according to claim 4, wherein The generating of a weight matrix corresponding to the driving behavior data based on the target graph convolutional network includes: Converting the driving behavior data into an adjacency matrix and a node feature matrix respectively; The adjacency matrix and the node feature matrix are processed by the target graph convolutional network to generate the weight matrix.

6. The method according to claim 4, wherein The fusing the first characteristic matrix and the second characteristic matrix and determining a second remaining power at a target time based on a third characteristic matrix obtained by the fusion includes: concatenating the first characteristic matrix and the second characteristic matrix to obtain the third characteristic matrix; Performing a dot product process on the third characteristic matrix and the weight matrix to obtain a seventh characteristic matrix; The second remaining power is obtained based on the seventh characteristic matrix.

7. A device for determining the power level of a vehicle-mounted battery system, characterized in that: include: A history module, configured to obtain historical data; wherein the historical data includes power data, and the power data includes a first remaining power corresponding to a historical time; a feature module, configured to process the power data through a first convolutional network in a first target model to obtain a first feature matrix, and to process the power data through a gated recurrent network in the first target model to obtain a second feature matrix; The power module is configured to fuse the first characteristic matrix and the second characteristic matrix, and determine a second remaining power at a target time based on a third characteristic matrix obtained by the fusion; wherein the target time is after the historical time.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and configured to run on the processor, wherein: When the processor executes the computer program, the method according to any one of claims 1 to 6 is implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images