Vehicle charge state prediction method and prediction device

Abstract

The invention relates to the technical field of artificial intelligence, and provides a vehicle charge state prediction method and prediction device. The method comprises the following steps: respectively carrying out graph construction on acquired voltage data and vehicle speed data of current driving of a vehicle to obtain a first voltage graph and a first vehicle speed graph, and updating a voltage characteristic matrix in the first voltage graph, updating a vehicle speed characteristic matrix in the first vehicle speed map to enable the updated voltage characteristic matrix to include voltage data characteristics and interaction characteristics of the voltage data and the vehicle speed data, and updating the vehicle speed characteristic matrix in the first vehicle speed map to enable the updated vehicle speed characteristic matrix to include vehicle speed data characteristics and interaction characteristics of the voltage data and the vehicle speed data, and then inputting the updated voltage characteristic matrix and the vehicle speed characteristic matrix into the trained prediction network to obtain two prediction results, and finally determining one of the two prediction results as a vehicle SOC result based on the current working condition information of the vehicle. The method can improve the accuracy of vehicle SOC prediction.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a vehicle state of charge prediction method and prediction device. Background Art

[0002] Electric vehicles rely on battery packs and energy management systems to drive the vehicle to operate efficiently, so that the vehicle can truly achieve zero pollutant emissions. It is one of the key development areas in today's automotive industry. The state of charge (SOC) parameter is a key parameter that affects the endurance and overall performance of electric vehicles. Research and analysis of SOC parameters has become a hot issue in improving the performance of electric vehicles. To this end, a large number of researchers have tested the factors that affect the changes in SOC parameters through operating condition experiments, and established state parameter models for real vehicle testing to find ways to improve SOC performance. However, there are many factors affecting the SOC parameters and the parameters change in a complex manner, making it difficult to accurately construct an accurate mathematical calculation model, which limits the research accuracy. Summary of the invention

[0003] In view of this, an embodiment of the present application provides a vehicle state of charge prediction method and prediction device to solve the problem of inaccurate vehicle state of charge prediction in the prior art.

[0004] According to a first aspect of an embodiment of the present application, a method for predicting a vehicle state of charge is provided, comprising:

[0005] Obtain the voltage data and speed data of the vehicle during this trip, as well as the vehicle's current operating condition information;

[0006] Performing graph construction on the voltage data and the vehicle speed data respectively to obtain a first voltage graph and a first vehicle speed graph, wherein the first voltage graph at least includes a first voltage characteristic matrix, and the first vehicle speed graph at least includes a first vehicle speed characteristic matrix;

[0007] The first voltage characteristic matrix in the first voltage map is updated to obtain a second voltage characteristic matrix, and the first vehicle speed characteristic matrix in the first vehicle speed map is updated to obtain a second vehicle speed characteristic matrix, wherein the second voltage characteristic matrix includes at least voltage data characteristics and interaction characteristics between voltage data and vehicle speed data, and the second vehicle speed characteristic matrix includes at least vehicle speed data characteristics and interaction characteristics between voltage data and vehicle speed data;

[0008] Inputting the second voltage characteristic matrix and the second vehicle speed characteristic matrix into the trained prediction network respectively to obtain a first state of charge prediction result and a second state of charge prediction result;

[0009] The first state of charge prediction result or the second state of charge prediction result is determined as a vehicle state of charge prediction result based on the operating condition information.

[0010] According to a second aspect of an embodiment of the present application, a vehicle charge state prediction device is provided, comprising:

[0011] An acquisition module is configured to acquire voltage data and vehicle speed data of the vehicle during this driving, as well as information on the current working condition of the vehicle;

[0012] A map construction module is configured to construct maps of the voltage data and the vehicle speed data respectively to obtain a first voltage map and a first vehicle speed map, wherein the first voltage map at least includes a first voltage characteristic matrix, and the first vehicle speed map at least includes a first vehicle speed characteristic matrix;

[0013] a feature updating module configured to update a first voltage feature matrix in the first voltage map to obtain a second voltage feature matrix, and to update a first vehicle speed feature matrix in the first vehicle speed map to obtain a second vehicle speed feature matrix, wherein the second voltage feature matrix includes at least voltage data features and interaction features between voltage data and vehicle speed data, and the second vehicle speed feature matrix includes at least vehicle speed data features and interaction features between voltage data and vehicle speed data;

[0014] A prediction module is configured to input the second voltage characteristic matrix and the second vehicle speed characteristic matrix into the trained prediction network respectively to obtain a first state of charge prediction result and a second state of charge prediction result;

[0015] A determination module is configured to determine, based on the operating condition information, that the first state of charge prediction result or the second state of charge prediction result is a vehicle state of charge prediction result.

[0016] Compared with the prior art, the embodiment of the present application has the following beneficial effects: the embodiment of the present application constructs the graphs of the voltage data and the speed data of the vehicle during the current driving respectively, obtains the first voltage graph and the first speed graph, updates the voltage feature matrix in the first voltage graph, so that the updated voltage feature matrix includes the voltage data feature and the interactive feature of the voltage data and the speed data, updates the speed feature matrix in the first speed graph, so that the updated speed feature matrix includes the speed data feature and the interactive feature of the voltage data and the speed data, and then inputs the updated voltage feature matrix and the speed feature matrix into the trained prediction network to obtain two prediction results, and finally determines one of the two prediction results as the vehicle SOC result based on the current working condition information of the vehicle. The method introduces the speed data to predict the state of charge of the vehicle, fully considers the different effects of the vehicle on energy consumption at different speeds, and at the same time, by calculating the interactive features of the voltage data and the speed data, the prediction model can be compensated with reference to the speed parameters when using the battery's own data for prediction, thereby improving the accuracy of the vehicle SOC prediction, and thus providing more accurate reference information for the performance optimization, mileage improvement and vehicle energy management of the vehicle battery management system.

[0017] Furthermore, when predicting the state of charge of a vehicle, the embodiment of the present application can obtain the interactive features of voltage data and speed data at a better ratio under different working conditions by pre-training and saving the voltage enhancement coefficient and vehicle speed enhancement coefficient under different working conditions. That is, the technical solution of the embodiment of the present application improves the accuracy of vehicle SOC prediction by constructing a difference enhancement mechanism, using the prediction results to compare the confidence of the respective graph nodes of vehicle speed and voltage, and reversely enhancing the relevant features according to the confidence difference, thereby laying a foundation for the subsequent vehicle energy management allocation and further improving the prediction accuracy.

[0018] Furthermore, when predicting the vehicle state of charge, the embodiment of the present application pre-trains and saves a mapping relationship table between operating conditions and voltage data and vehicle speed data, and thus can determine whether to use the voltage prediction result or the vehicle speed prediction result as the vehicle SOC prediction result under different operating conditions, thereby further improving the vehicle SOC prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0020] Figure 1 It is a flow chart of a vehicle charge state prediction method provided in an embodiment of the present application.

[0021] Figure 2 It is a flow chart of a method for respectively constructing a graph of voltage data and a vehicle speed data to obtain a first voltage graph and a first vehicle speed graph, provided in an embodiment of the present application.

[0022] Figure 3 It is a flowchart of a method for screening the connection relationship of each voltage data feature to obtain the updated connection relationship between N voltage data features provided in an embodiment of the present application.

[0023] Figure 4 It is a flowchart of a method for screening the connection relationship of each vehicle speed data feature to obtain the updated connection relationship between N vehicle speed data features provided in an embodiment of the present application.

[0024] Figure 5 It is a structural diagram of the implementation diagram provided in the embodiment of the present application to construct a network model.

[0025] Figure 6 It is a flowchart of a method for updating a first voltage characteristic matrix in a first voltage map to obtain a second voltage characteristic matrix provided in an embodiment of the present application.

[0026] Figure 7 It is a flow chart of a method for updating a first vehicle speed characteristic matrix in a first vehicle speed map to obtain a second vehicle speed characteristic matrix provided in an embodiment of the present application.

[0027] Figure 8 It is a structural diagram of an interactive graph convolutional network for updating a feature matrix provided in an embodiment of the present application.

[0028] Fig. 9 It is a flowchart diagram of another method provided by an embodiment of the present application for updating a first voltage characteristic matrix in a first voltage map to obtain a second voltage characteristic matrix.

[0029] Fig.10 It is a flowchart diagram of another method provided by an embodiment of the present application for updating a first vehicle speed characteristic matrix in a first vehicle speed map to obtain a second vehicle speed characteristic matrix.

[0030] Fig.11 It is a flow chart of a method for determining a voltage enhancement factor and a vehicle speed enhancement factor provided in an embodiment of the present application.

[0031] Fig.12 It is a result schematic diagram of a difference enhancement network for enhancing a feature matrix provided in an embodiment of the present application.

[0032] Fig.13It is a flowchart of a method for determining a first state of charge prediction result or a second state of charge prediction result as a vehicle state of charge prediction result based on operating condition information provided in an embodiment of the present application.

[0033] Fig.14 It is a flow chart of a method for determining a mapping relationship table between operating conditions, voltage data, and vehicle speed data provided in an embodiment of the present application.

[0034] Fig.15 It is a schematic diagram of a vehicle charge state prediction device provided in an embodiment of the present application.

[0035] Fig.16 It is a schematic diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0036] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application may also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present application.

[0037] A method and device for predicting a vehicle state of charge according to an embodiment of the present application will be described in detail below with reference to the accompanying drawings.

[0038] As mentioned above, there are many factors affecting SOC parameters and the parameters vary in a complex manner, making it difficult to accurately construct an accurate mathematical calculation model, which limits the accuracy of the research. With the development of computer technology, deep learning-based methods have made great progress in SOC prediction. Deep learning-based methods usually require less prior knowledge and the internal characteristics of the battery, and mainly rely on the relationship between input and output without in-depth understanding of the internal mechanisms and characteristics of the system. Deep learning methods collect and analyze large amounts of input and output data to discover patterns and laws, thereby making predictions.

[0039] However, the current deep learning-based methods only consider the battery's own parameters, such as current, voltage, and temperature, when predicting SOC. Other SCO changes caused by non-battery parameters are rarely considered. In addition, SOC prediction is often affected by factors such as model uncertainty, measurement error, and initial error. Effective auxiliary information is continuously introduced to compensate for these effects.

[0040] In view of this, the embodiment of the present application introduces vehicle speed data as supplementary input data for SOC prediction. By analyzing the vehicle speed data, the energy consumption patterns of the vehicle at different speeds can be obtained, and these patterns can be combined with the charging and discharging behavior of the battery to improve the SOC prediction algorithm. Furthermore, vehicle speed prediction can also be used as auxiliary information to compensate for SOC. For example, when the vehicle speed is high, the battery may be in a high-power discharge state, resulting in a faster rate of decrease in SOC. By combining vehicle speed and battery parameters, errors in SOC prediction can be more accurately compensated.

[0041] Further, the embodiment of the present application provides a method for predicting the state of charge of a vehicle, by respectively constructing the graphs of the voltage data and the speed data of the vehicle during the current driving, obtaining a first voltage graph and a first speed graph, and updating the voltage feature matrix in the first voltage graph so that the updated voltage feature matrix includes the voltage data feature and the interactive feature of the voltage data and the speed data, and updating the speed feature matrix in the first speed graph so that the updated speed feature matrix includes the speed data feature and the interactive feature of the voltage data and the speed data, and then inputting the updated voltage feature matrix and the speed feature matrix into the trained prediction network to obtain two prediction results, and finally determining one of the two prediction results as the vehicle SOC result based on the current working condition information of the vehicle. The method introduces the speed data to predict the state of charge of the vehicle, fully considering the different effects of the vehicle on energy consumption at different speeds, and at the same time, by calculating the interactive features of the voltage data and the speed data, the prediction model can be compensated by referring to the speed parameters when using the battery's own data for prediction, thereby improving the accuracy of the prediction of the state of charge of the vehicle, and thus providing more accurate reference information for the performance optimization, mileage improvement and vehicle energy management of the vehicle battery management system.

[0042] Figure 1 FIG. 1 is a flow chart of a method for predicting a vehicle state of charge provided by an embodiment of the present application. Figure 1 As shown, the method comprises the following steps:

[0043] In step S101, the voltage data and speed data of the vehicle during this driving, as well as the current operating condition information of the vehicle are obtained.

[0044] In step S102, graphs are constructed for the voltage data and the vehicle speed data to obtain a first voltage graph and a first vehicle speed graph.

[0045] The first voltage map at least includes a first voltage characteristic matrix, and the first vehicle speed map at least includes a first vehicle speed characteristic matrix.

[0046] In step S103, the first voltage characteristic matrix in the first voltage map is updated to obtain a second voltage characteristic matrix, and the first vehicle speed characteristic matrix in the first vehicle speed map is updated to obtain a second vehicle speed characteristic matrix.

[0047] The second voltage characteristic matrix includes at least voltage data characteristics and interaction characteristics between voltage data and vehicle speed data, and the second vehicle speed characteristic matrix includes at least vehicle speed data characteristics and interaction characteristics between voltage data and vehicle speed data.

[0048] In step S104, the second voltage characteristic matrix and the second vehicle speed characteristic matrix are respectively input into the trained prediction network to obtain a first state of charge prediction result and a second state of charge prediction result.

[0049] In step S105 , the first state of charge prediction result or the second state of charge prediction result is determined as the vehicle state of charge prediction result based on the operating condition information.

[0050] In an embodiment of the present application, the vehicle charge state prediction method can be executed by a vehicle-mounted terminal, or by a server connected to the vehicle or other terminal with certain computing capabilities.

[0051] In the embodiment of the present application, the voltage data and speed data of the vehicle during this driving, as well as the current working condition of the vehicle can be obtained first. In one example, when the vehicle receives a predicted SOC instruction while driving, the voltage data and speed data of the vehicle during this driving can be obtained from the vehicle battery management system, the vehicle computer and other on-board units, and further, the current working condition of the vehicle can be obtained.

[0052] In the embodiment of the present application, the acquired voltage data and vehicle speed data can be respectively constructed into graphs to obtain a first voltage graph and a first vehicle speed graph. The first voltage graph includes at least a first voltage characteristic matrix, and the first vehicle speed graph includes at least a first vehicle speed characteristic matrix. The specific implementation method of the graph construction is described in detail below and will not be repeated here.

[0053] In the embodiment of the present application, the first voltage characteristic matrix in the first voltage map can be updated to obtain the second voltage characteristic matrix, and the first vehicle speed characteristic matrix in the first vehicle speed map can be updated to obtain the second vehicle speed characteristic matrix. The second voltage characteristic matrix includes at least voltage data characteristics and interaction characteristics between voltage data and vehicle speed data, and the second vehicle speed characteristic matrix includes at least vehicle speed data characteristics and interaction characteristics between voltage data and vehicle speed data. In this way, it is possible to obtain prediction network input data including voltage data characteristics and interaction characteristics between voltage data and vehicle speed data, and prediction network input data including vehicle speed data characteristics and interaction characteristics between voltage data and vehicle speed data.

[0054] In the embodiment of the present application, the second voltage feature matrix and the second vehicle speed feature matrix are respectively input into the trained prediction network, that is, the prediction network input data of the voltage data feature and the interactive feature of the voltage data and the vehicle speed data, and the prediction network input data including the vehicle speed data feature and the interactive feature of the voltage data and the vehicle speed data are respectively input into the trained prediction network, and the first SOC prediction result and the second SOC prediction result can be obtained. Finally, based on the working condition information, the first SOC prediction result or the second SOC prediction result can be determined as the vehicle SOC prediction result.

[0055] According to the technical solution provided by the embodiment of the present application, the voltage data and speed data of the vehicle that have been obtained during this driving are respectively constructed to obtain a first voltage map and a first speed map, and the voltage feature matrix in the first voltage map is updated so that the updated voltage feature matrix contains the voltage data feature and the interactive feature between the voltage data and the speed data, and the speed feature matrix in the first speed map is updated so that the updated speed feature matrix contains the speed data feature and the interactive feature between the voltage data and the speed data, and then the updated voltage feature matrix and the speed feature matrix are input into the trained prediction network to obtain two prediction results, and finally one of the two prediction results is determined to be the vehicle SOC result based on the current working condition information of the vehicle. The method introduces speed data to predict the state of charge of the vehicle, fully considering the different effects of the vehicle on energy consumption at different speeds, and by calculating the interactive features of voltage data and speed data, the prediction model can be compensated by referring to the speed parameters when using the battery's own data for prediction, thereby improving the accuracy of the prediction of the state of charge of the vehicle, and thus providing more accurate reference information for the performance optimization, mileage improvement and vehicle energy management of the vehicle battery management system.

[0056] Figure 2 1 is a flow chart of a method for constructing a first voltage map and a first vehicle speed map by respectively constructing a map of voltage data and a vehicle speed data provided by an embodiment of the present application. Figure 2 As shown, the method comprises the following steps:

[0057] In step S201, feature extraction networks are used to extract features from voltage data and vehicle speed data respectively, to obtain N voltage data features and the connection relationship between the N voltage data features, and the connection relationship between the N vehicle speed data features and the N vehicle speed data features.

[0058] Wherein, N is a positive integer greater than 1.

[0059] In step S202, the connection relationship of each voltage data feature is screened to obtain the updated connection relationship between the N voltage data features, and the connection relationship of each vehicle speed data feature is screened to obtain the updated connection relationship between the N vehicle speed data features.

[0060] Among them, each updated voltage data feature is connected with k other voltage data features, and each updated vehicle speed data feature is connected with k other vehicle speed data features, k is a positive integer less than or equal to K, and K is a positive integer greater than 1 and less than N.

[0061] In step S203, a graph is used to construct a network, and a second voltage graph is constructed using the connection relationship between the N voltage data features and the updated N voltage data features, and a second vehicle speed graph is constructed using the connection relationship between the N vehicle speed data features and the updated N vehicle speed data features.

[0062] The second voltage map at least includes a third voltage characteristic matrix, and the second vehicle speed map at least includes a third vehicle speed characteristic matrix.

[0063] In step S204, a graph convolutional network is used to update the third voltage feature matrix in the second voltage map to the first voltage feature matrix to obtain a first voltage map, and the third vehicle speed feature matrix in the second vehicle speed map is updated to the first vehicle speed feature matrix to obtain a first vehicle speed map.

[0064] In an embodiment of the present application, when constructing the first voltage map, the voltage data can be firstly extracted using a feature extraction network to obtain N voltage data features and the connection relationship between the N voltage data features. Then, the connection relationship of each voltage data feature is screened to obtain the connection relationship between the updated N voltage data features. Next, a graph is constructed using a network to construct a second voltage map using the connection relationship between the N voltage data features and the updated N voltage data features. Finally, a graph convolutional network is used to update the third voltage feature matrix in the second voltage map to the first voltage feature matrix to obtain the first voltage map.

[0065] In an embodiment of the present application, when constructing the first vehicle speed map, the feature extraction network can be first used to extract features from the vehicle speed data to obtain N vehicle speed data features and the connection relationship between the N vehicle speed data features. Then, the connection relationship of each vehicle speed data feature is screened to obtain the connection relationship between the updated N vehicle speed data features. Next, a graph construction network is used to construct a second vehicle speed map using the connection relationship between the N vehicle speed data features and the updated N vehicle speed data features. Finally, a graph convolutional network is used to update the third vehicle speed feature matrix in the second vehicle speed map to the first vehicle speed feature matrix to obtain the first vehicle speed map.

[0066] Figure 3 1 is a flow chart of a method for screening the connection relationship of each voltage data feature to obtain the updated connection relationship between N voltage data features provided by an embodiment of the present application. Figure 3 As shown, the method comprises the following steps:

[0067] In step S301, a first similarity between a target voltage data feature and each voltage data feature connected thereto is obtained.

[0068] The target voltage data feature is one of the N voltage data features.

[0069] In step S302, in response to the number of voltage data features connected to the target voltage data features being greater than K, only the connection relationships between the K target voltage data features with the largest first similarities and other voltage data features are retained to obtain updated connection relationships between the target voltage data features.

[0070] In step S303, N voltage characteristic data are traversed to obtain the updated connection relationship between the N voltage data characteristics.

[0071] In an embodiment of the present application, when screening the connection relationship of each voltage data feature, the first similarity between the target voltage data feature and each voltage data feature connected thereto can be obtained first, and the target voltage data feature is one of the N voltage data features. Then, when the number of voltage data features connected to the target voltage data feature is greater than K, only the connection relationship between the K target voltage data features with the largest first similarity and other voltage data features is retained to obtain the updated connection relationship between the target voltage data features. It can be understood that if the number of voltage data features connected to the target voltage data feature is less than or equal to K, no processing can be performed on the connection relationship of the target voltage data feature, that is, all the connection relationships of the target voltage data feature are retained. Finally, the N voltage feature data are traversed to obtain the connection relationship between the N updated voltage data features.

[0072] Figure 4 1 is a flow chart of a method for screening the connection relationship of each vehicle speed data feature to obtain the updated connection relationship between N vehicle speed data features provided by an embodiment of the present application. Figure 4 As shown, the method comprises the following steps:

[0073] In step S401, a second similarity between the target vehicle speed data feature and each vehicle speed data feature connected thereto is obtained.

[0074] The target vehicle speed data feature is one vehicle speed data feature among the N vehicle speed data features.

[0075] In step S402, in response to the number of vehicle speed data features connected to the target vehicle speed data features being greater than K, only the connection relationships between the K target vehicle speed data features with the second largest similarities and other vehicle speed data features are retained to obtain updated connection relationships between the target vehicle speed data features.

[0076] In step S403, N vehicle speed feature data are traversed to obtain the connection relationship between the N updated vehicle speed data features.

[0077] In the embodiment of the present application, when screening the connection relationship of each vehicle speed data feature, the first similarity between the target vehicle speed data feature and each vehicle speed data feature connected thereto can be obtained first, and the target vehicle speed data feature is one of the N vehicle speed data features. Then, when the number of speed data features connected to the target vehicle speed data feature is greater than K, only the connection relationship between the K target vehicle speed data features with the largest first similarity and other vehicle speed data features is retained to obtain the updated connection relationship of the target vehicle speed data feature. It can be understood that if the number of speed data features connected to the target vehicle speed data feature is less than or equal to K, then no processing may be performed on the connection relationship of the target vehicle speed data feature, that is, all the connection relationships of the target vehicle speed data feature are retained. Finally, the N vehicle speed feature data are traversed to obtain the connection relationship between the N updated vehicle speed data features.

[0078] In the embodiment of the present application, when updating the third voltage characteristic matrix, the product of the third voltage characteristic matrix and the transposed matrix of the third voltage characteristic matrix can be first determined as the adjacency matrix of the voltage data. Then, the adjacency matrix of the voltage data is symmetrically normalized. Finally, a preset activation function is used to activate the product of the normalized adjacency matrix of the voltage data, the third voltage characteristic matrix, and the preset voltage weight matrix to obtain the first voltage characteristic matrix.

[0079] Similarly, when updating the third vehicle speed characteristic matrix, the product of the third vehicle speed characteristic matrix and the transposed matrix of the third vehicle speed characteristic matrix can be first determined as the adjacency matrix of the vehicle speed data. Then, the adjacency matrix of the vehicle speed data is symmetrically normalized. Finally, a preset activation function is used to activate the product of the normalized adjacency matrix of the vehicle speed data, the third vehicle speed characteristic matrix, and the preset vehicle speed weight matrix to obtain the first vehicle speed characteristic matrix. Among them, the voltage weight matrix and the vehicle speed weight matrix can be set according to actual needs.

[0080] Figure 5 Schematic diagram of the structure of the network model provided by the embodiment of the present application. Figure 5As shown, the model may include two input units for inputting battery voltage data and vehicle speed data respectively. The battery voltage data is first fed into the graph construction network, and the feature extraction network in the graph construction network, such as Convolutional Neural Networks (CNN), is used to extract features of the battery voltage data to obtain N voltage data features and the connection relationship between the N voltage data features. In one example, the CNN may be a one-dimensional CNN, and the extracted voltage feature matrix may be represented by H1=Convld(X1), where X1 is the voltage data, and H1 may include N voltage data features, each voltage data feature is a node in the voltage graph, and the node may have a connection relationship with other nodes in the voltage graph, and the number of connection relationships is determined by the feature extraction network when performing feature extraction.

[0081] Since the introduction of more nodes may generate redundant information and noise, a Top-K sorting mechanism is also set in the graph construction network to screen the connection relationship of each node. In one example, the similarity between the nodes in the voltage graph can be obtained, and then the K value is set, for example, K=5, and the top 5 connection relationships with the greatest similarity between the target node and other nodes are retained, for example, set to 1, and the remaining connection relationships are set to 0, and the connection relationship screening can be completed, thereby greatly reducing the amount of calculation and avoiding the introduction of redundant information that leads to the first model accuracy. Finally, the graph construction network constructs the voltage graph G1=(V1, A1, H1), where V1 is the set of nodes in the graph, A1 is the adjacency matrix,

[0082] In the same way, the second speed graph G2=(V2, A2, H2) can be constructed, where V2 is the set of nodes in the graph, H2=Convld(X2), X2 is the speed data, A2 is the adjacency matrix,

[0083]

[0084] In the embodiment of the present application, the characteristic information of the voltage graph structure G1=(V1, A1, H1) and the vehicle speed graph G2=(V2, A2, H2) can also be deeply mined through the graph convolution network. The mining process can be: Among them, H i 1 is the voltage feature matrix after aggregation by graph convolutional network, To perform symmetric normalization on the adjacency matrix, W i is the weight matrix, ReLU is the activation function, i is equal to 1 or 2. When i is equal to 1, each parameter is the corresponding parameter of the voltage data. When i is equal to 2, each parameter is the corresponding parameter of the vehicle speed data.

[0085] The voltage graph after aggregation using graph convolutional network is G1=(V1, A1, H1 1 ), the speed map is G2=(V2,A2,H2 1 ).

[0086] Figure 6 1 is a flow chart of a method for updating a first voltage characteristic matrix in a first voltage map to obtain a second voltage characteristic matrix provided by an embodiment of the present application. Figure 6 As shown, the method comprises the following steps:

[0087] In step S601, the first voltage characteristic matrix is ​​multiplied by the first vehicle speed characteristic matrix to obtain an interaction characteristic matrix of the first voltage characteristic matrix and the first vehicle speed characteristic matrix.

[0088] In step S602, the first voltage characteristic matrix is ​​added to the interaction characteristic matrix to obtain a second voltage characteristic matrix.

[0089] In the embodiment of the present application, the first voltage characteristic matrix and the first vehicle speed characteristic matrix may be first multiplied to obtain an interaction characteristic matrix of the first voltage characteristic matrix and the first vehicle speed characteristic matrix, and then the first voltage characteristic matrix and the interaction characteristic matrix may be added to obtain a second voltage characteristic matrix.

[0090] Figure 7 1 is a flow chart of a method for updating a first vehicle speed characteristic matrix in a first vehicle speed map to obtain a second vehicle speed characteristic matrix provided by an embodiment of the present application. Figure 7 As shown, the method comprises the following steps:

[0091] In step S701, the first voltage characteristic matrix is ​​multiplied by the first vehicle speed characteristic matrix to obtain an interaction characteristic matrix of the first voltage characteristic matrix and the first vehicle speed characteristic matrix.

[0092] In step S702, the first vehicle speed feature matrix is ​​added to the interaction feature matrix to obtain a second vehicle speed feature matrix.

[0093] In the embodiment of the present application, the first voltage characteristic matrix and the first vehicle speed characteristic matrix may be first multiplied to obtain an interaction characteristic matrix of the first voltage characteristic matrix and the first vehicle speed characteristic matrix, and then the first vehicle speed characteristic matrix and the interaction characteristic matrix may be added to obtain a second vehicle speed characteristic matrix.

[0094] Figure 8 is a schematic diagram of the structure of an interactive graph convolutional network for updating a feature matrix provided in an embodiment of the present application. Figure 8 As shown in Figure 1, the correlation between nodes can be captured by product interaction, that is, the interaction between nodes is realized by product operation. Figure 5The number of nodes in the voltage graph and the speed graph obtained by constructing the network model shown in the figure is the same. The nodes at the corresponding positions of the speed and voltage can be combined in pairs to obtain the characteristic vector (V 1j , V 2j ), where j = 1, 2, ..., N, where N is the number of nodes in the graph. Furthermore, the number of elements in each pair of nodes is the same, so for the feature vectors of each pair of nodes, their corresponding elements are multiplied to obtain a new feature vector, which can represent the interaction between the voltage data feature and the vehicle speed data feature. In this way, the correlation between vehicle speed and voltage can be captured and integrated into the feature representation.

[0095] The interaction feature is denoted as H C , then the interaction feature H C Respectively with the voltage characteristics H1 before interaction 1 and vehicle speed characteristic H2 1 Add them together to get the voltage feature H1 that integrates the interactive features. 2 And the vehicle speed feature H2 fused with interactive features 2 That is, H1 2 =H1 1 +H C , H2 2 =H2 1 +H C .

[0096] Since the original features of voltage and speed contain the information of the voltage data and speed data respectively, and the interactive features capture the correlation between them, adding them together can make the model use this information more comprehensively and better consider the relationship between speed and voltage, thereby improving the accuracy of SOC energy consumption prediction.

[0097] In the embodiment of the present application, since the effects of vehicle speed and voltage on SOC are different under different working conditions, in order to further capture the differential impacts, a differential enhancement mechanism can be used to compare the confidence levels of the respective graph nodes of vehicle speed and voltage through prediction results, and reversely enhance the relevant features based on the confidence difference, thereby improving the accuracy of SOC prediction and laying the foundation for subsequent vehicle energy management allocation.

[0098] Fig. 9 FIG. 1 is a flow chart of another method for updating the first voltage characteristic matrix in the first voltage map to obtain the second voltage characteristic matrix provided by an embodiment of the present application. Fig. 9 As shown, the method comprises the following steps:

[0099] In step S901, the first voltage characteristic matrix is ​​multiplied by the first vehicle speed characteristic matrix to obtain an interaction characteristic matrix of the first voltage characteristic matrix and the first vehicle speed characteristic matrix.

[0100] In step S902, the first voltage characteristic matrix and the interaction characteristic matrix are added to obtain a fourth voltage characteristic matrix.

[0101] In step S903, a voltage enhancement coefficient is obtained based on the operating condition information.

[0102] In step S904, the fourth voltage characteristic matrix is ​​multiplied by the voltage enhancement coefficient to obtain a second voltage characteristic matrix.

[0103] In the embodiment of the present application, when updating the first voltage characteristic matrix, the first voltage characteristic matrix can be first multiplied by the first vehicle speed characteristic matrix to obtain an interactive characteristic matrix of the first voltage characteristic matrix and the first vehicle speed characteristic matrix. Then, the first voltage characteristic matrix and the interactive characteristic matrix are added to obtain a fourth voltage characteristic matrix. Next, the voltage enhancement coefficient is obtained based on the operating condition information, and the fourth voltage characteristic matrix is ​​multiplied by the voltage enhancement coefficient to obtain the second voltage characteristic matrix.

[0104] Fig.10 FIG. 1 is a flow chart of another method for updating the first vehicle speed characteristic matrix in the first vehicle speed map to obtain the second vehicle speed characteristic matrix provided by an embodiment of the present application. Fig.10 As shown, the method comprises the following steps:

[0105] In step S1001, the first voltage characteristic matrix is ​​multiplied by the first vehicle speed characteristic matrix to obtain an interaction characteristic matrix of the first voltage characteristic matrix and the first vehicle speed characteristic matrix.

[0106] In step S1002, the first vehicle speed feature matrix is ​​added to the interaction feature matrix to obtain a fourth vehicle speed feature matrix.

[0107] In step S1003, a vehicle speed enhancement coefficient is obtained based on the operating condition information.

[0108] In step S1004, the fourth vehicle speed characteristic matrix is ​​multiplied by the vehicle speed enhancement coefficient to obtain a second vehicle speed characteristic matrix.

[0109] In the embodiment of the present application, when updating the first vehicle speed characteristic matrix, the first voltage characteristic matrix and the first vehicle speed characteristic matrix may be first multiplied to obtain an interactive characteristic matrix of the first voltage characteristic matrix and the first vehicle speed characteristic matrix. Then, the first vehicle speed characteristic matrix and the interactive characteristic matrix are added to obtain a fourth vehicle speed characteristic matrix. Next, the vehicle speed enhancement coefficient is obtained based on the working condition information, and the fourth vehicle speed characteristic matrix is ​​multiplied by the vehicle speed enhancement coefficient to obtain a second vehicle speed characteristic matrix.

[0110] In the embodiment of the present application, the voltage enhancement coefficient and the vehicle speed enhancement coefficient can be determined when training the prediction network.

[0111] Fig.11 is a flow chart of a method for determining a voltage enhancement factor and a vehicle speed enhancement factor provided in an embodiment of the present application. Fig.11 As shown, the method comprises the following steps:

[0112] In step S1101, a training data set is obtained.

[0113] The training data set includes voltage training data, vehicle speed training data and state of charge labels, and the state of charge labels include actual state of charge values ​​and operating condition labels.

[0114] In step S1102, the initialization voltage enhancement coefficient and the vehicle speed enhancement coefficient are set.

[0115] In step S1103, graphs are constructed for the voltage training data and the vehicle speed training data respectively to obtain a training voltage graph and a training vehicle speed graph.

[0116] The training voltage map at least includes a first training voltage characteristic matrix, and the training vehicle speed map at least includes a first training vehicle speed characteristic matrix.

[0117] In step S1104, the first training voltage feature matrix in the training voltage map is updated to obtain a second training voltage feature matrix, and the first training vehicle speed feature matrix in the training vehicle speed map is updated to obtain a second training vehicle speed feature matrix.

[0118] In step S1105, a second training voltage feature matrix and a second training vehicle speed feature matrix are respectively input into the prediction network to obtain a first training state of charge prediction result and a second training state of charge prediction result.

[0119] In step S1106 , a first loss of the first training SOC prediction result relative to the actual SOC value in the SOC label and a second loss of the second training SOC prediction result relative to the actual SOC value in the SOC label are calculated respectively.

[0120] In step S1107, in response to determining that both the first loss and the second loss are smaller than a preset threshold, and the first loss is smaller than the second loss, the initialization voltage enhancement coefficient is increased by a preset value to obtain an updated voltage enhancement coefficient.

[0121] In step S1108, the updated voltage enhancement coefficient is multiplied by the first training voltage feature matrix to obtain an updated first training voltage feature matrix, the updated first training voltage feature matrix is ​​input into the prediction network to obtain an updated first training state of charge prediction result, and the updated first loss is calculated.

[0122] In step S1109, in response to determining that both the first loss and the second loss are less than the preset threshold, and the second loss is less than the first loss, the initialized vehicle speed enhancement coefficient is increased by a preset value to obtain an updated vehicle speed enhancement coefficient.

[0123] In step S1110, the updated vehicle speed enhancement coefficient is multiplied by the first training vehicle speed feature matrix to obtain an updated first training vehicle speed feature matrix, the updated first training vehicle speed feature matrix is ​​input into the prediction network to obtain an updated second training state of charge prediction result, and the updated second loss is calculated.

[0124] In step S1111, the steps of updating the voltage enhancement coefficient or the vehicle speed enhancement coefficient, using the updated voltage enhancement coefficient to obtain an updated first voltage characteristic matrix or using the updated vehicle speed enhancement coefficient to obtain an updated first vehicle speed characteristic matrix, inputting the updated first voltage characteristic matrix into the prediction network to obtain a prediction result and calculating an updated first loss or inputting the updated first vehicle speed characteristic matrix into the prediction network to obtain a prediction result and calculating an updated second loss are iteratively performed until at least one of the updated first loss and the updated second loss is less than or equal to a preset threshold, and it is determined that the prediction model converges under the target operating condition.

[0125] The target operating condition is the same operating condition as the operating condition label in the state of charge label.

[0126] In step S1112, the updated voltage enhancement coefficient and the updated vehicle speed enhancement coefficient when the model converges are determined to be the voltage enhancement coefficient and the vehicle speed enhancement coefficient under the target operating condition.

[0127] In an embodiment of the present application, a training data set may be first acquired, the training data set including voltage training data, vehicle speed training data and a state of charge label, the state of charge label including an actual state of charge value and an operating condition label.

[0128] Then, the initialization voltage enhancement coefficient and the vehicle speed enhancement coefficient are set, and the voltage training data and the vehicle speed training data are respectively constructed to obtain a training voltage graph and a training vehicle speed graph. The training voltage graph includes at least a first training voltage feature matrix, and the training vehicle speed graph includes at least a first training vehicle speed feature matrix. The first training voltage feature matrix in the training voltage graph is updated to obtain a second training voltage feature matrix, and the first training vehicle speed feature matrix in the training vehicle speed graph is updated to obtain a second training vehicle speed feature matrix.

[0129] Next, the second training voltage feature matrix and the second training vehicle speed feature matrix are respectively input into the prediction network to obtain the first training state of charge prediction result and the second training state of charge prediction result. The first loss of the first training state of charge prediction result relative to the actual state of charge value in the state of charge label and the second loss of the second training state of charge prediction result relative to the actual state of charge value in the state of charge label are respectively calculated.

[0130] If the first loss and the second loss are both less than the preset threshold value, and the first loss is less than the second loss, the initialization voltage enhancement coefficient is increased by a preset value to obtain an updated voltage enhancement coefficient. The updated voltage enhancement coefficient is multiplied by the first training voltage feature matrix to obtain an updated first training voltage feature matrix, and the updated first training voltage feature matrix is ​​input into the prediction network to obtain an updated first training state of charge prediction result, and the updated first loss is calculated.

[0131] If both the first loss and the second loss are less than the preset threshold, and the second loss is less than the first loss, the initialized vehicle speed enhancement coefficient is increased by a preset value to obtain an updated vehicle speed enhancement coefficient. The updated vehicle speed enhancement coefficient is multiplied by the first training vehicle speed feature matrix to obtain an updated first training vehicle speed feature matrix, and the updated first training vehicle speed feature matrix is ​​input into the prediction network to obtain an updated second training state of charge prediction result, and the updated second loss is calculated.

[0132] Iteratively execute the steps of updating the voltage enhancement coefficient or the vehicle speed enhancement coefficient, obtaining an updated first voltage characteristic matrix using the updated voltage enhancement coefficient or obtaining an updated first vehicle speed characteristic matrix using the updated vehicle speed enhancement coefficient, inputting the updated first voltage characteristic matrix into the prediction network to obtain a prediction result and calculating an updated first loss or inputting the updated first vehicle speed characteristic matrix into the prediction network to obtain a prediction result and calculating an updated second loss, until at least one of the updated first loss and the updated second loss is less than or equal to a preset threshold, and determining that the prediction model converges under the target operating condition. The target operating condition is the same operating condition as the operating condition label in the state of charge label.

[0133] Finally, the updated voltage enhancement coefficient and the updated vehicle speed enhancement coefficient when the model converges are determined as the voltage enhancement coefficient and the vehicle speed enhancement coefficient under the target operating condition.

[0134] Fig.12 Schematic diagram of the result of the difference enhancement network for enhancing the feature matrix provided in the embodiment of the present application. Fig.12 As shown, we can first integrate the interactive features H C Voltage characteristic H1 2 and fusion interaction feature H C Vehicle speed characteristic H2 2 All nodes in the prediction network, such as a two-layer multilayer perceptron (MLP), each output a predicted SOC, H1 2 The prediction results and H2 2 The prediction results are recorded as y s1 (j) and s2 (j) , the true label corresponding to the data is y real (j) Then choose the absolute error L as the loss function, through L(y si ,y real )The calculated loss is: in, and Respectively represent H1 2 and H2 2 The error value between the predicted SCO and the true label, j represents the jth node.

[0135] Then and The corresponding combination is performed, and the error values ​​predicted by the voltage data features and the vehicle speed data features at the jth node are combined to obtain a set of values ​​containing two predicted error values. The set of error values ​​is normalized to obtain two numbers between [0, 1]. The closer to 0 means the smaller the error. If the error predicted by the voltage at the jth node is closer to 0 than the error predicted by the vehicle speed, it means that the voltage data is more relevant to the task in the prediction of the jth node and has a higher confidence. Then the error is input into the vehicle speed and voltage interaction graph convolution module to perform the voltage feature H1 before the interaction. 1 The corresponding j-th node is enhanced. The enhancement operation can be performed by generating an enhancement coefficient Implementation, normalize() represents the normalization operation performed above, multiplying the enhancement coefficient by the voltage feature before the interaction, thereby increasing the proportion of voltage in the feature aggregation and making its predicted value at the jth node more accurate.

[0136] The enhancement coefficient of each node can be adaptively obtained, and the voltage or speed characteristics corresponding to each node can be enhanced to improve the overall prediction accuracy. In other words, the formula can be used and The enhanced voltage characteristic matrix is ​​calculated and vehicle speed feature matrix Among them, p = 1, 2, ..., N, q = 1, 2, ..., N, and p is not equal to q. Then the enhanced feature matrix is ​​input into the MLP again, and each of them outputs a predicted SOC, calculates the mean absolute error, and determines the predicted SOC with the smallest mean absolute error as the final output.

[0137] Fig.13 1 is a flow chart of a method for determining a first state of charge prediction result or a second state of charge prediction result as a vehicle state of charge prediction result based on operating condition information provided in an embodiment of the present application. Fig.13 As shown, the method comprises the following steps:

[0138] In step S1301, a mapping relationship table between operating conditions, voltage data and vehicle speed data is obtained, where the mapping relationship table is used to indicate input data for predicting the state of charge under different operating conditions, where the input data is voltage data or vehicle speed data.

[0139] In step S1302, in response to determining in the mapping relationship table that the input data is voltage data based on the operating condition information, the first state of charge prediction result is determined to be a vehicle state of charge prediction result.

[0140] In step S1303 , in response to determining in the mapping relationship table that the input data is vehicle speed data based on the operating condition information, the second state of charge prediction result is determined to be a vehicle state of charge prediction result.

[0141] In the embodiment of the present application, a pre-stored mapping table of working conditions, voltage data and vehicle speed data can be obtained. The mapping table is determined when the prediction network is trained and can be saved in the trained prediction network or saved in the cloud to obtain other storage space. Further, the mapping table is used to indicate the input data for predicting the state of charge under different working conditions, and the input data is voltage data or vehicle speed data.

[0142] In the embodiment of the present application, when the input data is determined to be voltage data in the mapping relationship table based on the operating condition information, the first state of charge prediction result can be determined to be the vehicle state of charge prediction result. On the other hand, when the input data is determined to be vehicle speed data in the mapping relationship table based on the operating condition information, the second state of charge prediction result can be determined to be the vehicle state of charge prediction result.

[0143] Fig.141 is a flow chart of a method for determining a mapping relationship table between an operating condition and voltage data and a vehicle speed data provided in an embodiment of the present application. Fig.14 As shown, the method comprises the following steps:

[0144] In step S1401, a training data set is obtained.

[0145] The training data set includes voltage training data, vehicle speed training data and state of charge labels, and the state of charge labels include actual state of charge values ​​and operating condition labels.

[0146] In step S1402, graphs are constructed for the voltage training data and the vehicle speed training data to obtain a training voltage graph and a training vehicle speed graph.

[0147] The training voltage map at least includes a first training voltage characteristic matrix, and the training vehicle speed map at least includes a first training vehicle speed characteristic matrix.

[0148] In step S1403, the first training voltage feature matrix in the training voltage map is updated to obtain a second training voltage feature matrix, and the first training vehicle speed feature matrix in the training vehicle speed map is updated to obtain a second training vehicle speed feature matrix.

[0149] In step S1404, a second training voltage feature matrix and a second training vehicle speed feature matrix are respectively input into the prediction network to obtain a first training state of charge prediction result and a second training state of charge prediction result.

[0150] In step S1405 , a first loss of the first training SOC prediction result relative to the actual SOC value in the SOC label and a second loss of the second training SOC prediction result relative to the actual SOC value in the SOC label are calculated respectively.

[0151] In step S1406, in response to determining that both the first loss and the second loss are less than a preset threshold, the first training voltage feature matrix or the first vehicle speed feature training matrix is ​​updated, and the updated first training voltage feature matrix is ​​input into the prediction network to obtain an updated prediction result and calculate the updated first loss, or the updated first vehicle speed feature training matrix is ​​input into the prediction network to obtain an updated prediction result and calculate the updated second loss, until the updated first loss is less than or equal to the preset threshold, or the updated second loss is less than or equal to the preset threshold.

[0152] In step S1407 , in response to determining that the updated first loss is less than or equal to a preset threshold, it is determined that the input data for predicting the state of charge under the vehicle operating condition corresponding to the operating condition label is voltage data.

[0153] In step S1408 , in response to determining that the updated second loss is less than or equal to a preset threshold, it is determined that the input data for predicting the state of charge under the vehicle operating condition corresponding to the operating condition label is the vehicle speed data.

[0154] In step S1409, all vehicle operating conditions are traversed to obtain a mapping relationship table between the operating conditions and the voltage data and the vehicle speed data.

[0155] In an embodiment of the present application, a training data set may be first acquired, the training data set including voltage training data, vehicle speed training data and a state of charge label, the state of charge label including an actual state of charge value and an operating condition label.

[0156] Then, the voltage training data and the vehicle speed training data are respectively constructed to obtain a training voltage graph and a training vehicle speed graph. The training voltage graph includes at least a first training voltage feature matrix, and the training vehicle speed graph includes at least a first training vehicle speed feature matrix. The first training voltage feature matrix in the training voltage graph is updated to obtain a second training voltage feature matrix, and the first training vehicle speed feature matrix in the training vehicle speed graph is updated to obtain a second training vehicle speed feature matrix.

[0157] Next, the second training voltage feature matrix and the second training vehicle speed feature matrix are respectively input into the prediction network to obtain the first training state of charge prediction result and the second training state of charge prediction result. The first loss of the first training state of charge prediction result relative to the actual state of charge value in the state of charge label and the second loss of the second training state of charge prediction result relative to the actual state of charge value in the state of charge label are respectively calculated.

[0158] Iteratively execute the steps of updating the first training voltage feature matrix or the first vehicle speed feature training matrix in response to determining that the first loss and the second loss are both less than a preset threshold, and inputting the updated first training voltage feature matrix into the prediction network to obtain an updated prediction result and calculate the updated first loss, or inputting the updated first vehicle speed feature training matrix into the prediction network to obtain an updated prediction result and calculate the updated second loss, until the updated first loss is less than or equal to the preset threshold, or the updated second loss is less than or equal to the preset threshold.

[0159] When it is determined that the updated first loss is less than or equal to the preset threshold, it is determined that the input data for predicting the state of charge under the vehicle operating condition corresponding to the operating condition label is voltage data. Alternatively, when it is determined that the updated second loss is less than or equal to the preset threshold, it is determined that the input data for predicting the state of charge under the vehicle operating condition corresponding to the operating condition label is vehicle speed data. All vehicle operating conditions are traversed to obtain a mapping relationship table between the operating conditions and the voltage data and the vehicle speed data.

[0160] The technical solution of the embodiment of the present application provides an energy consumption prediction method based on enhanced vehicle speed prediction differences, while considering the impact of battery voltage parameters and vehicle speed data under different driving conditions on energy consumption prediction, designing a vehicle speed and battery voltage parameter interactive graph convolution module, and then constructing a difference enhancement mechanism, comparing the confidence of each graph node of vehicle speed and voltage through the prediction results, and reversely enhancing related features according to the confidence difference, thereby improving the accuracy of SOC prediction and laying the foundation for subsequent vehicle energy management allocation.

[0161] Furthermore, the technical solution of the embodiment of the present application first selects the battery voltage parameters of the electric vehicle and the speed data of the vehicle under different driving conditions, deeply mines the respective feature information in the form of graph convolution, and constructs graphs for the voltage and speed respectively, while considering the speed and battery data information. At the same time, a speed and voltage interactive graph convolution module is designed. The correlation between nodes can be captured through product interaction. The obtained interactive features of speed and voltage are added to the features before interaction respectively, retaining the information of the original features, and predicting the SOC respectively, so as to better distinguish the most critical node information affecting the SOC prediction. A difference enhancement mechanism is also constructed. The confidence of each graph node of speed and voltage is compared through the prediction results, and the relevant features are reversely enhanced according to the confidence difference, thereby improving the accuracy of SOC prediction and laying the foundation for subsequent vehicle energy management allocation.

[0162] All the above optional technical solutions can be arbitrarily combined to form optional embodiments of the present application, which will not be described one by one here.

[0163] The following is an embodiment of the device of the present application, which can be used to execute the embodiment of the method of the present application. For details not disclosed in the embodiment of the device of the present application, please refer to the embodiment of the method of the present application.

[0164] Fig.15 Schematic diagram of a vehicle charge state prediction device provided in an embodiment of the present application. Fig.15 As shown, the device comprises:

[0165] The acquisition module 1501 is configured to acquire the voltage data and speed data of the vehicle during this driving, as well as the current operating condition information of the vehicle.

[0166] The graph construction module 1502 is configured to construct graphs for the voltage data and the vehicle speed data respectively to obtain a first voltage graph and a first vehicle speed graph, wherein the first voltage graph at least includes a first voltage characteristic matrix, and the first vehicle speed graph at least includes a first vehicle speed characteristic matrix.

[0167] The feature update module 1503 is configured to update the first voltage feature matrix in the first voltage map to obtain a second voltage feature matrix, and to update the first vehicle speed feature matrix in the first vehicle speed map to obtain a second vehicle speed feature matrix, wherein the second voltage feature matrix at least includes voltage data features and interaction features between voltage data and vehicle speed data, and the second vehicle speed feature matrix at least includes vehicle speed data features and interaction features between voltage data and vehicle speed data.

[0168] The prediction module 1504 is configured to input the second voltage characteristic matrix and the second vehicle speed characteristic matrix into the trained prediction network respectively to obtain a first state of charge prediction result and a second state of charge prediction result.

[0169] The determination module 1505 is configured to determine, based on the operating condition information, that the first state of charge prediction result or the second state of charge prediction result is a vehicle state of charge prediction result.

[0170] According to the technical solution provided by the embodiment of the present application, the voltage data and speed data of the vehicle that have been obtained during this driving are respectively constructed to obtain a first voltage map and a first speed map, and the voltage feature matrix in the first voltage map is updated so that the updated voltage feature matrix contains the voltage data feature and the interactive feature between the voltage data and the speed data, and the speed feature matrix in the first speed map is updated so that the updated speed feature matrix contains the speed data feature and the interactive feature between the voltage data and the speed data, and then the updated voltage feature matrix and the speed feature matrix are input into the trained prediction network to obtain two prediction results, and finally one of the two prediction results is determined to be the vehicle SOC result based on the current working condition information of the vehicle. The method introduces speed data to predict the state of charge of the vehicle, fully considering the different effects of the vehicle on energy consumption at different speeds, and by calculating the interactive features of voltage data and speed data, the prediction model can be compensated by referring to the speed parameters when using the battery's own data for prediction, thereby improving the accuracy of the prediction of the state of charge of the vehicle, and thus providing more accurate reference information for the performance optimization, mileage improvement and vehicle energy management of the vehicle battery management system.

[0171] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0172] Fig.16 Schematic diagram of an electronic device provided in an embodiment of the present application. Fig.16As shown, the electronic device 16 of this embodiment includes: a processor 1601, a memory 1602, and a computer program 1603 stored in the memory 1602 and executable on the processor 1601. When the processor 1601 executes the computer program 1603, the steps in the above-mentioned method embodiments are implemented. Alternatively, when the processor 1601 executes the computer program 1603, the functions of the modules / units in the above-mentioned device embodiments are implemented.

[0173] The electronic device 16 may be a desktop computer, a notebook, a PDA, a cloud server, or other electronic device. The electronic device 16 may include, but is not limited to, a processor 1601 and a memory 1602. Those skilled in the art will appreciate that Fig.16 The electronic device 16 is merely an example and does not limit the electronic device 16 , and may include more or less components than those shown in the figure, or different components.

[0174] Processor 1601 can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0175] The memory 1602 may be an internal storage unit of the electronic device 16, for example, a hard disk or memory of the electronic device 16. The memory 1602 may also be an external storage device of the electronic device 16, for example, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card (FlashCard), etc. equipped on the electronic device 16. The memory 1602 may also include both an internal storage unit of the electronic device 16 and an external storage device. The memory 1602 is used to store computer programs and other programs and data required by the electronic device.

[0176] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units.

[0177] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A method for predicting a vehicle state of charge, characterized in that: include: Obtain the voltage data and speed data of the vehicle during this trip, as well as the vehicle's current operating condition information; Performing graph construction on the voltage data and the vehicle speed data respectively to obtain a first voltage graph and a first vehicle speed graph, wherein the first voltage graph at least includes a first voltage characteristic matrix, and the first vehicle speed graph at least includes a first vehicle speed characteristic matrix; The first voltage characteristic matrix in the first voltage map is updated to obtain a second voltage characteristic matrix, and the first vehicle speed characteristic matrix in the first vehicle speed map is updated to obtain a second vehicle speed characteristic matrix, wherein the second voltage characteristic matrix includes at least voltage data characteristics and interaction characteristics between voltage data and vehicle speed data, and the second vehicle speed characteristic matrix includes at least vehicle speed data characteristics and interaction characteristics between voltage data and vehicle speed data; Inputting the second voltage characteristic matrix and the second vehicle speed characteristic matrix into the trained prediction network respectively to obtain a first state of charge prediction result and a second state of charge prediction result; The first state of charge prediction result or the second state of charge prediction result is determined as a vehicle state of charge prediction result based on the operating condition information.

2. The method according to claim 1, characterized in that The step of constructing graphs of the voltage data and the vehicle speed data to obtain a first voltage graph and a first vehicle speed graph comprises: Using a feature extraction network to perform feature extraction on the voltage data and the vehicle speed data respectively, to obtain a connection relationship between N voltage data features and the N voltage data features, and a connection relationship between N vehicle speed data features and the N vehicle speed data features, where N is a positive integer greater than 1; Screening the connection relationship of each voltage data feature to obtain the connection relationship between the N updated voltage data features, screening the connection relationship of each vehicle speed data feature to obtain the connection relationship between the N updated vehicle speed data features, wherein each updated voltage data feature is connected to k other voltage data features, and each updated vehicle speed data feature is connected to k other vehicle speed data features, k is a positive integer less than or equal to K, and K is a positive integer greater than 1 and less than N; Using a graph to construct a network, using the connection relationship between the N voltage data features and the updated N voltage data features to construct a second voltage graph, and using the connection relationship between the N vehicle speed data features and the updated N vehicle speed data features to construct a second vehicle speed graph, the second voltage graph at least includes a third voltage feature matrix, and the second vehicle speed graph at least includes a third vehicle speed feature matrix; A graph convolutional network is used to update the third voltage feature matrix in the second voltage map to the first voltage feature matrix to obtain a first voltage map, and the third vehicle speed feature matrix in the second vehicle speed map is updated to the first vehicle speed feature matrix to obtain a first vehicle speed map.

3. The method according to claim 2, characterized in that The screening of the connection relationship of each voltage data feature to obtain the updated connection relationship between the N voltage data features includes: Acquire a first similarity between a target voltage data feature and each voltage data feature connected thereto, wherein the target voltage data feature is one of the N voltage data features; In response to the number of voltage data features connected to the target voltage data features being greater than K, only the connection relationships between the K target voltage data features with the largest first similarities and other voltage data features are retained to obtain updated connection relationships between the target voltage data features; Traversing the N voltage characteristic data to obtain a connection relationship between the updated N voltage data characteristics; The connection relationship of each vehicle speed data feature is screened to obtain the updated connection relationship between the N vehicle speed data features, including: Acquire a second similarity between a target vehicle speed data feature and each vehicle speed data feature connected thereto, wherein the target vehicle speed data feature is one of the N vehicle speed data features; In response to the number of vehicle speed data features connected to the target vehicle speed data features being greater than K, only the connection relationships between the K target vehicle speed data features with the second largest similarity and other vehicle speed data features are retained to obtain an updated connection relationship between the target vehicle speed data features; The N vehicle speed feature data are traversed to obtain the connection relationship between the updated N vehicle speed data features.

4. The method according to claim 2, characterized in that: The updating the third voltage feature matrix in the second voltage map to the first voltage feature matrix using a graph convolutional network includes: Determine that the product of the third voltage characteristic matrix and the transposed matrix of the third voltage characteristic matrix is ​​an adjacency matrix of voltage data; Performing a symmetric normalization operation on the adjacency matrix of the voltage data; Using a preset activation function, an activation calculation is performed on the product of the adjacency matrix of the normalized voltage data, the third voltage feature matrix, and a preset voltage weight matrix to obtain a first voltage feature matrix; The method of using a graph convolutional network to update the third vehicle speed feature matrix in the second vehicle speed map to the first vehicle speed feature matrix includes: Determine the product of the third vehicle speed characteristic matrix and the transposed matrix of the third vehicle speed characteristic matrix as the adjacency matrix of the vehicle speed data; Performing a symmetric normalization operation on the adjacency matrix of the vehicle speed data; Using a preset activation function, an activation calculation is performed on the product of the adjacency matrix of the normalized vehicle speed data, the third vehicle speed feature matrix, and a preset vehicle speed weight matrix to obtain a first vehicle speed feature matrix.

5. The method according to claim 1, characterized in that The updating of the first voltage characteristic matrix in the first voltage map to obtain the second voltage characteristic matrix includes: Multiplying the first voltage characteristic matrix by the first vehicle speed characteristic matrix to obtain an interaction characteristic matrix of the first voltage characteristic matrix and the first vehicle speed characteristic matrix; Adding the first voltage characteristic matrix to the interaction characteristic matrix to obtain the second voltage characteristic matrix; The updating of the first vehicle speed characteristic matrix in the first vehicle speed map to obtain the second vehicle speed characteristic matrix includes: Multiplying the first voltage characteristic matrix by the first vehicle speed characteristic matrix to obtain an interaction characteristic matrix of the first voltage characteristic matrix and the first vehicle speed characteristic matrix; The first vehicle speed feature matrix is ​​added to the interaction feature matrix to obtain the second vehicle speed feature matrix.

6. The method according to claim 5, characterized in that The updating of the first voltage characteristic matrix in the first voltage map to obtain the second voltage characteristic matrix includes: Multiplying the first voltage characteristic matrix by the first vehicle speed characteristic matrix to obtain an interaction characteristic matrix of the first voltage characteristic matrix and the first vehicle speed characteristic matrix; Adding the first voltage characteristic matrix to the interaction characteristic matrix to obtain a fourth voltage characteristic matrix; Acquire a voltage enhancement factor based on the operating condition information; Multiplying the fourth voltage characteristic matrix by the voltage enhancement coefficient to obtain the second voltage characteristic matrix; The updating of the first vehicle speed characteristic matrix in the first vehicle speed map to obtain the second vehicle speed characteristic matrix includes: Multiplying the first voltage characteristic matrix by the first vehicle speed characteristic matrix to obtain an interaction characteristic matrix of the first voltage characteristic matrix and the first vehicle speed characteristic matrix; Adding the first vehicle speed characteristic matrix to the interaction characteristic matrix to obtain a fourth vehicle speed characteristic matrix; acquiring a vehicle speed enhancement coefficient based on the operating condition information; The fourth vehicle speed characteristic matrix is ​​multiplied by the vehicle speed enhancement coefficient to obtain the second vehicle speed characteristic matrix.

7. The method according to claim 6, characterized in that The voltage enhancement coefficient and the vehicle speed enhancement coefficient are determined when training the prediction network; Determine the voltage enhancement factor and vehicle speed enhancement factor, including: Acquire a training data set, wherein the training data set includes voltage training data, vehicle speed training data, and a state of charge label, wherein the state of charge label includes an actual state of charge value and an operating condition label; Set the initialization voltage enhancement factor and vehicle speed enhancement factor; Performing graph construction on the voltage training data and the vehicle speed training data respectively to obtain a training voltage graph and a training vehicle speed graph, wherein the training voltage graph includes at least a first training voltage feature matrix, and the training vehicle speed graph includes at least a first training vehicle speed feature matrix; The first training voltage characteristic matrix in the training voltage map is updated to obtain a second training voltage characteristic matrix, and the first training vehicle speed characteristic matrix in the training vehicle speed map is updated to obtain a second training vehicle speed characteristic matrix; Inputting the second training voltage feature matrix and the second training vehicle speed feature matrix into the prediction network respectively to obtain a first training state of charge prediction result and a second training state of charge prediction result; respectively calculating a first loss of the first training state of charge prediction result relative to the actual state of charge value in the state of charge label, and a second loss of the second training state of charge prediction result relative to the actual state of charge value in the state of charge label; In response to determining that both the first loss and the second loss are less than a preset threshold, and the first loss is less than the second loss, increasing the initialization voltage enhancement coefficient by a preset value to obtain an updated voltage enhancement coefficient; Multiplying the updated voltage enhancement coefficient by the first training voltage feature matrix to obtain an updated first training voltage feature matrix, inputting the updated first training voltage feature matrix into the prediction network to obtain an updated first training state of charge prediction result, and calculating an updated first loss; In response to determining that both the first loss and the second loss are less than a preset threshold, and the second loss is less than the first loss, increasing the initialized vehicle speed enhancement coefficient by a preset value to obtain an updated vehicle speed enhancement coefficient; Multiplying the updated vehicle speed enhancement coefficient by the first training vehicle speed feature matrix to obtain an updated first training vehicle speed feature matrix, inputting the updated first training vehicle speed feature matrix into the prediction network to obtain an updated second training state of charge prediction result, and calculating an updated second loss; Iteratively executing the steps of updating the voltage enhancement coefficient or the vehicle speed enhancement coefficient, obtaining an updated first voltage characteristic matrix using the updated voltage enhancement coefficient or obtaining an updated first vehicle speed characteristic matrix using the updated vehicle speed enhancement coefficient, inputting the updated first voltage characteristic matrix into the prediction network to obtain a prediction result and calculating an updated first loss or inputting the updated first vehicle speed characteristic matrix into the prediction network to obtain a prediction result and calculating an updated second loss, until at least one of the updated first loss and the updated second loss is less than or equal to the preset threshold, and determining that the prediction model converges under the target operating condition, wherein the target operating condition is the same operating condition as the operating condition label in the state of charge label; The updated voltage enhancement coefficient and the updated vehicle speed enhancement coefficient when the model converges are determined to be the voltage enhancement coefficient and the vehicle speed enhancement coefficient under the target operating condition.

8. The method according to claim 1, characterized in that The determining, based on the operating condition information, that the first state of charge prediction result or the second state of charge prediction result is a vehicle state of charge prediction result includes: Acquire a mapping relationship table between operating conditions and voltage data and vehicle speed data, wherein the mapping relationship table is used to indicate input data for predicting the state of charge under different operating conditions, wherein the input data is voltage data or vehicle speed data; In response to determining that the input data is voltage data in the mapping relationship table based on the operating condition information, determining that the first state of charge prediction result is the vehicle state of charge prediction result; In response to determining in the mapping relationship table that the input data is vehicle speed data based on the operating condition information, the second state of charge prediction result is determined to be the vehicle state of charge prediction result.

9. The method according to claim 8, characterized in that The mapping relationship table between the working condition and the voltage data and the vehicle speed data is determined when training the prediction network; Determine the mapping relationship table between the working condition and the voltage data and the vehicle speed data, including: Acquire a training data set, wherein the training data set includes voltage training data, vehicle speed training data, and a state of charge label, wherein the state of charge label includes an actual state of charge value and an operating condition label; Performing graph construction on the voltage training data and the vehicle speed training data respectively to obtain a training voltage graph and a training vehicle speed graph, wherein the training voltage graph includes at least a first training voltage feature matrix, and the training vehicle speed graph includes at least a first training vehicle speed feature matrix; The first training voltage characteristic matrix in the training voltage map is updated to obtain a second training voltage characteristic matrix, and the first training vehicle speed characteristic matrix in the training vehicle speed map is updated to obtain a second training vehicle speed characteristic matrix; Inputting the second training voltage feature matrix and the second training vehicle speed feature matrix into the prediction network respectively to obtain a first training state of charge prediction result and a second training state of charge prediction result; respectively calculating a first loss of the first training state of charge prediction result relative to the actual state of charge value in the state of charge label, and a second loss of the second training state of charge prediction result relative to the actual state of charge value in the state of charge label; Iteratively executing the steps of updating the first training voltage feature matrix or the first vehicle speed feature training matrix in response to determining that both the first loss and the second loss are less than a preset threshold, inputting the updated first training voltage feature matrix into the prediction network, obtaining an updated prediction result and calculating an updated first loss, or inputting the updated first vehicle speed feature training matrix into the prediction network, obtaining an updated prediction result and calculating an updated second loss, until the updated first loss is less than or equal to the preset threshold, or the updated second loss is less than or equal to the preset threshold; In response to determining that the updated first loss is less than or equal to the preset threshold, determining that the input data for predicting the state of charge under the vehicle operating condition corresponding to the operating condition label is voltage data; In response to determining that the updated second loss is less than or equal to the preset threshold, determining that the input data for predicting the state of charge under the vehicle operating condition corresponding to the operating condition label is vehicle speed data; All vehicle operating conditions are traversed to obtain a mapping relationship table between the operating conditions and voltage data and vehicle speed data.

10. A vehicle charge state prediction device, characterized in that: include: An acquisition module is configured to acquire voltage data and vehicle speed data of the vehicle during this driving, as well as information on the current working condition of the vehicle; A map construction module is configured to construct maps of the voltage data and the vehicle speed data respectively to obtain a first voltage map and a first vehicle speed map, wherein the first voltage map at least includes a first voltage characteristic matrix, and the first vehicle speed map at least includes a first vehicle speed characteristic matrix; a feature updating module configured to update a first voltage feature matrix in the first voltage map to obtain a second voltage feature matrix, and to update a first vehicle speed feature matrix in the first vehicle speed map to obtain a second vehicle speed feature matrix, wherein the second voltage feature matrix includes at least voltage data features and interaction features between voltage data and vehicle speed data, and the second vehicle speed feature matrix includes at least vehicle speed data features and interaction features between voltage data and vehicle speed data; A prediction module is configured to input the second voltage characteristic matrix and the second vehicle speed characteristic matrix into the trained prediction network respectively to obtain a first state of charge prediction result and a second state of charge prediction result; A determination module is configured to determine, based on the operating condition information, that the first state of charge prediction result or the second state of charge prediction result is a vehicle state of charge prediction result.

Images