Vehicle fault early warning method, system and equipment based on multi-source data fusion and medium

Abstract

The invention belongs to the technical field of fault early warning, and discloses a vehicle fault early warning method, system and device based on multi-source data fusion and a medium, and the method comprises the steps: obtaining the multi-source operation historical data of a vehicle within a preset continuous historical duration; preprocessing the multi-source operation historical data to obtain a plurality of historical data sample sequences, and constructing a training set by using the plurality of historical data sample sequences; constructing a state estimation model based on a deep learning algorithm, wherein the state estimation model is used for performing feature extraction and feature fusion on the plurality of historical data sample sequences; training a state estimation model according to the training set to obtain a trained state estimation model; when the trained state estimation model inputs multi-source real-time operation data, a real-time state estimation result is output, and the real-time state estimation result is used for carrying out fault analysis and early warning on the vehicle. According to the invention, the accuracy and timeliness of fault diagnosis can be improved, and potential abnormal risks can be found.

Description

Technical Field

[0001] The present invention belongs to the technical field of fault warning, and particularly relates to a vehicle fault warning method, system, device and medium based on multi-source data fusion. Background Technique

[0002] With the rapid development of the new energy vehicle industry, the ownership of new energy vehicles is indeed getting higher and higher, which naturally makes the safety performance of new energy vehicles the focus of public and industry attention.

[0003] The key points of the safety performance of new energy vehicles mainly include: battery safety, electrical system safety, charging safety, etc.; therefore, fault warning of new energy vehicles is an important means to ensure the safe operation of vehicles, improve user experience and extend service life. The fault warning system can monitor various performance indicators of the vehicle in real time, discover potential fault hazards in advance, and prompt the driver or maintenance personnel to take corresponding measures through warnings.

[0004] At present, the conventional warning of the vehicle operation state is to compare the relevant operation data of the vehicle with the reference standard data. If the relevant operation data exceeds the reference standard data, it means that the vehicle is in an abnormal state. However, this method has limitations. For example, the performance of various data of the vehicle varies under different working conditions, and there is a non-linear relationship between different operation data; the conventional warning method is not sufficient to judge the influence of these non-linear relationships, resulting in poor fault diagnosis effect and inability to discover potential risks. Summary of the Invention

[0005] The purpose of the present invention is to provide a vehicle fault warning method, system, device and medium based on multi-source data fusion to solve the problem that the conventional warning method is not sufficient to judge the influence of these non-linear relationships, resulting in poor fault diagnosis effect and inability to discover potential risks.

[0006] To achieve the above purpose, the present invention adopts the following technical solutions: In the first aspect, the present invention provides a vehicle fault warning method based on multi-source data fusion, and the method includes: Obtain multi-source operation historical data of the vehicle within a preset continuous historical duration; Preprocess the multi-source operation historical data to obtain multiple historical data sample sequences, and construct a training set with the multiple historical data sample sequences; Construct a state estimation model based on a deep learning algorithm, and the state estimation model is used to extract features and fuse features for the multiple historical data sample sequences; Train the state prediction model according to the training set to obtain a trained state prediction model; when the trained state prediction model inputs multi-source real-time operation data, it outputs a real-time state prediction result, and the real-time state prediction result is used for fault analysis and warning of the vehicle.

[0007] Preferably, the method further includes: Obtain the current multi-source real-time operation data of the vehicle; Input the current multi-source real-time operation data of the vehicle into the trained state prediction model, and the trained state prediction model outputs a real-time state prediction result; Conduct fault analysis on the real-time state prediction result to obtain a vehicle fault result; Generate a warning signal according to the vehicle fault result and perform visual warning with the warning signal.

[0008] Preferably, the multi-source operation historical data at least includes: vehicle state data and battery state; The battery state at least includes: battery output voltage, battery output current, battery charge, battery temperature, and single-cell battery voltage; The vehicle state data at least includes: driving speed, motor speed, and driving mileage; The multiple historical data sample sequences are a vehicle state sequence and a battery state sequence. The vehicle state sequence at least includes: a driving speed sequence, a motor speed sequence, and a driving mileage sequence; the battery state sequence at least includes: a battery output voltage sequence, a battery output current sequence, a battery charge sequence, a battery temperature sequence, and a single-cell battery voltage sequence.

[0009] Preferably, the state prediction model includes: a convolutional network unit, an attention mechanism unit, and a long short-term memory unit; The convolutional network unit takes the battery state sequence as input, and is used for non-linearly fitting the battery state sequence. The convolutional network unit outputs a voltage prediction value, and the voltage prediction value is the response output voltage of the single-cell battery; The attention mechanism unit takes the battery state sequence and the vehicle state sequence as input, and is used for mining the correlation between the battery state sequence and the vehicle state sequence. The attention mechanism unit outputs a correlation degree; The long short-term memory unit takes the voltage prediction value output by the convolutional network unit and the correlation degree output by the attention mechanism unit as input, and is used for fusing the voltage prediction value and the correlation degree and extracting the time dependence between the voltage prediction value and the correlation degree.

[0010] Preferably, the real-time state prediction result is the current voltage prediction value of the single-cell battery, and the current voltage prediction value of the single-cell battery is the current response output voltage of the single-cell battery; Perform fault analysis on the real-time state estimation result to obtain the vehicle fault result, including: Obtain the voltage measurement values of each single battery during the discharge stage; According to the current voltage prediction value and the voltage measurement value during the discharge stage, construct a discharge error sequence, where the discharge error sequence includes multiple voltage deviations, and calculate the average deviation of the discharge error sequence; Obtain the real-time current of each single battery during the discharge stage, and construct a current point set according to the real-time current; According to the current point set and the discharge error sequence, construct a current-voltage distribution map; each real-time current and the voltage deviation corresponding to the real-time current form a distribution point in the current-voltage distribution map; According to the current-voltage distribution map, determine the abnormal conditions of each single battery, and use the abnormal conditions of each single battery as the vehicle fault result.

[0011] Preferably, according to the current-voltage distribution map, determining the abnormal conditions of each single battery includes: Based on a preset algorithm, calculate the distribution distance between each distribution point in the current-voltage distribution map and the origin; Judge whether the distribution distance is greater than the threshold distance. If so, mark the distribution point as abnormal to indicate that the single battery corresponding to the voltage deviation of the distribution point is in an abnormal state; if not, mark the distribution point as normal to indicate that the single battery corresponding to the voltage deviation of the distribution point is in a normal state.

[0012] Preferably, the method further includes: determining the threshold distance, including: Screen the real-time current corresponding to the voltage deviation equal to the preset deviation, and mark it as the threshold current; Traverse all distribution points in the current-voltage distribution map, find the distribution point corresponding to the threshold current, and mark it as the threshold point; Based on a preset algorithm, calculate the distance between the threshold point and the origin to obtain the threshold distance.

[0013] In a second aspect, the present invention provides a vehicle fault warning system based on multi-source data fusion for implementing the above-mentioned vehicle fault warning method based on multi-source data fusion. The system includes: A data acquisition module for acquiring multi-source operation historical data of the vehicle within a preset continuous historical duration; A data processing module for preprocessing the multi-source operation historical data to obtain multiple historical data sample sequences, and constructing a training set with the multiple historical data sample sequences; A model construction module for constructing a state estimation model based on a deep learning algorithm, where the state estimation model is used for feature extraction and feature fusion of multiple historical data sample sequences; A model training module, configured to train a state prediction model according to a training set to obtain a trained state prediction model; when the trained state prediction model inputs multi-source real-time operation data, it outputs a real-time state prediction result, and the real-time state prediction result is used for fault analysis and early warning of the vehicle.

[0014] In a third aspect, the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the above-mentioned vehicle fault early warning method based on multi-source data fusion is implemented.

[0015] In a fourth aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, the above-mentioned vehicle fault early warning method based on multi-source data fusion is implemented.

[0016] Beneficial effects: The present invention uses multi-source operation historical data to construct a training set. At this time, the training set contains data related to different operation states of the vehicle. Using the training set to train the state prediction model can establish a non-linear mapping relationship between the multi-source operation historical data as the input and the output of the state prediction model, and establish the correlation between various different-dimensional data and each state of the vehicle; when the multi-source real-time operation data is input into the trained state prediction model, the state prediction model can accurately predict the current state of the vehicle, and then analyze the current state of the vehicle to determine whether the vehicle will have a fault. If a fault occurs, a fault early warning is given in time, which can improve the accuracy and timeliness of fault diagnosis. Moreover, due to the establishment of the correlation between various different-dimensional data and each state of the vehicle, potential abnormal risks can also be discovered. Description of the drawings

[0017] The drawings are used to provide a further understanding of the embodiments of the present invention, and constitute a part of the specification. Together with the following specific embodiments, they are used to explain the embodiments of the present invention, but do not constitute a limitation to the embodiments of the present invention. In the drawings: Figure 1 is a flowchart of a vehicle fault early warning method based on multi-source data fusion provided by an embodiment of the present invention; Figure 2 is a block diagram of a vehicle fault early warning system based on multi-source data fusion provided by an embodiment of the present invention. Detailed implementation manners

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the present invention will be briefly introduced below in conjunction with the accompanying drawings and the description of the embodiments or the prior art. Obviously, the following description of the structures of the accompanying drawings is only some embodiments of the present invention. For those of ordinary skill in the art, other accompanying drawings can be obtained based on these drawings without creative efforts. It should be noted here that the description of these embodiments is used to help understand the present invention, but does not constitute a limitation on the present invention.

[0019] Embodiment 1 Figure 1 is a flowchart of a vehicle fault warning method based on multi-source data fusion provided by an embodiment of the present invention. As Figure 1 shown, this embodiment provides a vehicle fault warning method based on multi-source data fusion, and the method includes: Step S10: Obtain multi-source operation historical data of the vehicle within a preset continuous historical duration; in this embodiment, taking a new energy vehicle as an example, the multi-source operation historical data of the training set can be data throughout the life cycle of the new energy vehicle, or relevant multi-source operation historical data extracted during the operation period of the new energy vehicle; the multi-source operation historical data in this embodiment mainly includes: vehicle state data and battery state and other types of historical data, where the vehicle state mainly includes driving speed, motor speed, and driving mileage, etc., and the battery state mainly includes battery output voltage, battery output current, battery charge, battery temperature, and single-cell battery voltage, etc.

[0020] Step S20: Preprocess the multi-source operation historical data to obtain multiple historical data sample sequences, and construct a training set with the multiple historical data sample sequences.

[0021] In this embodiment, for each type of data, it is necessary to preprocess it. The preprocessing includes: missing value processing, normalization processing, etc. After preprocessing each type of data, linearly transform the relevant data to between [0,1] to unify the dimension of the data for subsequent processing.

[0022] Therefore, the multiple historical data sample sequences are a vehicle state sequence and a battery state sequence. The vehicle state sequence at least includes: a driving speed sequence, a motor speed sequence, and a driving mileage sequence; the battery state sequence at least includes: a battery output voltage sequence, a battery output current sequence, a battery charge sequence, a battery temperature sequence, and a single-cell battery voltage sequence.

[0023] Step S30: Construct a state prediction model based on a deep learning algorithm. The state prediction model is used to extract features and fuse features from multiple historical data sample sequences.

[0024] In this embodiment, the state prediction model includes: a convolutional network unit, an attention mechanism unit, and a long short-term memory unit; wherein, the output of the convolutional network unit and the output of the attention mechanism unit are used as the input of the long short-term memory unit.

[0025] Step S40: Train the state prediction model according to the training set to obtain a trained state prediction model; when the trained state prediction model inputs multi-source real-time operation data, it outputs a real-time state prediction result, and the real-time state prediction result is used for fault analysis and warning of the vehicle.

[0026] In this embodiment, the state prediction model can be used to predict the state of the new energy vehicle battery pack, and the real-time state prediction result is the current voltage prediction value of each single battery in the battery pack. At this time, the current voltage prediction value of the single battery is the current response output voltage of the single battery.

[0027] Therefore, for the state prediction model, the convolutional network unit of the state prediction model takes the battery state sequence as the input and is used to perform non-linear fitting on the battery state sequence. The convolutional network unit outputs a voltage prediction value, and the voltage prediction value is the response output voltage of the single battery, that is, when the new energy vehicle is running, the response output voltages of the individual single batteries in the battery pack of the new energy vehicle; in the training stage, the average value of the voltage values of all the single batteries in the battery pack in the early stage (the cycle when the new energy vehicle just starts to drive, for example, one month; because the consistency between healthy batteries is better in the early stage, the voltage responses of each single battery in the same discharge segment are the same) is used as the output value of the convolutional neural network (that is, the label of the training set).

[0028] In this embodiment, the response output voltage of the single battery is related to parameters such as the current open-circuit voltage, battery loop current value, internal resistance, and capacitance. Since the internal resistance, capacitance, and other parameters of the single battery will change after multiple charge and discharge cycles (that is, as the battery usage time increases, the performance of each battery cell changes to varying degrees), and these parameters cannot directly calculate the response output voltage of the single battery; therefore, the convolutional neural network unit can extract relevant features from historical data and use the fully connected layer of the convolutional neural network unit to fit the above non-linear relationship; according to the current battery state, the response output voltage of the single battery can be indirectly predicted; then, using the attention mechanism unit and the long short-term memory unit, learn the influence of different vehicle operating conditions on the response output voltage to improve the accuracy of battery voltage estimation.

[0029] The attention mechanism unit of the state prediction model takes the battery state sequence and the vehicle state sequence as the input and is used to mine the correlation between the battery state sequence and the vehicle state sequence. The attention mechanism unit outputs a correlation degree.

[0030] The long short-term memory unit of the state prediction model takes the voltage prediction value output by the convolutional network unit and the correlation degree output by the attention mechanism unit as inputs, and is used to fuse the voltage prediction value and the correlation degree and extract the temporal dependence between the voltage prediction value and the correlation degree.

[0031] In this embodiment, the higher the importance of the output features of the convolutional network unit, the closer the output of its corresponding attention mechanism unit is to 1; conversely, the lower the importance of the output features of the convolutional network unit, the closer the output of its corresponding attention mechanism unit is to 0; the importance of the features is reflected by the magnitude of the value, thus completing the discrimination of important features.

[0032] In this embodiment, after the output of the convolutional network unit and the output of the attention mechanism unit are fused, they are used as the input of the long short-term memory unit, and the long short-term memory unit is used for sequence modeling to obtain the final prediction result. The long short-term memory unit extracts coarse-grained features from the significant fine-grained features extracted by the convolutional neural network based on the attention mechanism. While refining the features of each dimension, it prevents the problems of memory loss and gradient dispersion caused by too long a step size. This architecture can capture the temporal dependence of the effective features extracted after the convolutional operation is optimized by the attention mechanism, realize the fusion of coarse-grained and fine-grained features, and comprehensively characterize the time-series data of the battery state.

[0033] As a further optimization of this embodiment, the method further includes: Step a10: Obtain the current multi-source real-time operation data of the vehicle; among them, the current multi-source real-time operation data also includes: battery output voltage, battery output current, battery charge, battery temperature, as well as single-cell battery voltage, driving speed, motor speed, and driving mileage and other data.

[0034] Step a20: Input the current multi-source real-time operation data of the vehicle into the trained state prediction model, and the trained state prediction model outputs a real-time state prediction result; the real-time state prediction result at this time is the response output voltage of each single-cell battery.

[0035] Step a30: Conduct a fault analysis on the real-time state prediction result to obtain the vehicle fault result; specifically, it includes the following steps: Step a301: Obtain the voltage measurement values of each single-cell battery during the discharge stage.

[0036] Step a302: Construct a discharge error sequence based on the current voltage estimated value and the voltage measurement value during the discharge stage. The discharge error sequence includes multiple voltage deviations, and calculate the average deviation of the discharge error sequence; In this embodiment, since the response output voltage of each single battery has been estimated in step a20, the response output voltage of each single battery is the reference value, the voltage measurement value of each single battery during the discharge stage is the actual value, and subtracting the reference value from the actual value of each single battery can obtain the voltage deviation of this single battery.

[0037] Step a303: Obtain the real-time current of each single battery during the discharge stage, and construct a current point set based on the real-time current; In this embodiment, under different operating conditions of the vehicle (acceleration or braking deceleration), the variation range of the real-time current value data is relatively large; then the real-time current value has a significant impact on the current single battery voltage, and there is a non-linear relationship between the real-time current and voltage of the lithium-ion battery; Therefore, it is necessary to jointly evaluate the abnormal conditions of the current single battery through the real-time current and voltage deviation.

[0038] Step a304: Construct a current-voltage distribution map according to the current point set and the discharge error sequence; Each real-time current and the voltage deviation corresponding to this real-time current form a distribution point in the current-voltage distribution map. In this embodiment, taking the real-time current as the horizontal coordinate of the current-voltage distribution map and the voltage deviation as the vertical coordinate of the current-voltage distribution map, construct a two-dimensional coordinate distribution map, which is the current-voltage distribution map; The real-time current and the corresponding voltage deviation of each single battery are used as a coordinate point in the two-dimensional coordinate distribution map, that is, the distribution point.

[0039] Since the real-time current distribution is uneven, it is necessary to map the real-time current to between [0, 1] to obtain the current point set. Map the real-time current exceeding the preset current to 0, and map the real-time current lower than the preset current according to the following formula: PI = 1 - (Is / 10a); where Is is the real-time current, PI is the mapped current value, and a is the smallest number of the highest digit of the real-time current (for example, if the real-time current value is a two-digit number, then a is 10).

[0040] Step a305: Determine the abnormal conditions of each single battery according to the current-voltage distribution map, and use the abnormal conditions of each single battery as the vehicle fault result; Therefore, it is possible to judge whether each single battery fails according to the current-voltage distribution map.

[0041] Step a40: Generate a warning signal according to the vehicle fault result, and perform visual warning with the warning signal.

[0042] As a further optimization of this embodiment, in step a305, determining the abnormal conditions of each single battery according to the current-voltage distribution map includes: Step a30501: Calculate the distribution distance between each distribution point in the current-voltage distribution diagram and the origin based on a preset algorithm. The preset algorithm can be the Euclidean distance algorithm or the Mahalanobis distance algorithm. In this embodiment, the Mahalanobis distance algorithm is preferably used. Step a30502: Determine whether the distribution distance is greater than the threshold distance. If so, mark the distribution point as abnormal to indicate that the single battery corresponding to the voltage deviation of the distribution point is in an abnormal state; if not, mark the distribution point as normal to indicate that the single battery corresponding to the voltage deviation of the distribution point is in a normal state.

[0043] In this embodiment, the method further includes: determining the threshold distance, including: Step b10: Screen the real-time current corresponding to a preset deviation in voltage deviation and mark it as the threshold current. The preset deviation is 10 mV, which can be flexibly selected according to actual needs. Therefore, the real-time current corresponding to 10 mV is selected as the threshold current. Step b20: Traverse all distribution points in the current-voltage distribution diagram, find the distribution point corresponding to the threshold current, and mark it as the threshold point. Step b30: Calculate the distance between the threshold point and the origin based on a preset algorithm to obtain the threshold distance. The preset algorithm can be the Euclidean distance algorithm or the Mahalanobis distance algorithm. In this embodiment, the Mahalanobis distance algorithm is preferably used.

[0044] In this embodiment, when the distribution distance between any distribution point and the origin is equal to the threshold distance, it indicates that the battery performance has a significant deviation at this time. The more the distribution distance exceeds, the more obvious the abnormality is, and it is more likely to cause abnormalities during the charging and discharging process of the battery pack, and the battery pack has a fault of inaccurate power.

[0045] The present invention uses multi-source operation historical data to construct a training set. At this time, the training set contains data related to different operating states of the vehicle. Using the training set to train the state prediction model can establish a non-linear mapping relationship between the multi-source operation historical data as the input and the output of the state prediction model, and establish the correlation between various different-dimensional data and each state of the vehicle. When the multi-source real-time operation data is input into the trained state prediction model, the state prediction model can accurately predict the current state of the vehicle. Then, by analyzing the current state of the vehicle, it can be determined whether the vehicle will have a fault. If a fault occurs, a fault warning is given in a timely manner, which can improve the accuracy and timeliness of fault diagnosis. And because the correlation between various different-dimensional data and each state of the vehicle is established, potential abnormal risks can also be discovered.

[0046] Embodiment 2 Figure 2It is a block diagram of a vehicle fault warning system based on multi-source data fusion provided by an embodiment of the present invention. As Figure 2 shown, this embodiment provides a vehicle fault warning system based on multi-source data fusion, which is used to implement the vehicle fault warning method based on multi-source data fusion in the first embodiment. The system includes: A data acquisition module, which is used to acquire multi-source operation historical data of the vehicle within a preset continuous historical duration; A data processing module, which is used to preprocess the multi-source operation historical data to obtain multiple historical data sample sequences, and construct a training set with the multiple historical data sample sequences; A model construction module, which is used to construct a state prediction model based on a deep learning algorithm. The state prediction model is used to perform feature extraction and feature fusion on the multiple historical data sample sequences; A model training module, which is used to train the state prediction model according to the training set to obtain a trained state prediction model; when the trained state prediction model inputs multi-source real-time operation data, it outputs a real-time state prediction result, and the real-time state prediction result is used to perform fault analysis and warning on the vehicle.

[0047] This embodiment also provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the vehicle fault warning method based on multi-source data fusion in the first embodiment.

[0048] This embodiment also provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the vehicle fault warning method based on multi-source data fusion in the first embodiment.

[0049] The present invention uses multi-source operation historical data to construct a training set. At this time, the training set contains data related to different operation states of the vehicle. Using the training set to train the state prediction model can establish a non-linear mapping relationship between the multi-source operation historical data as the input and the output of the state prediction model, and establish the correlation between various different-dimensional data and each state of the vehicle; when the multi-source real-time operation data is input into the trained state prediction model, the state prediction model can accurately predict the current state of the vehicle, and then analyze the current state of the vehicle to determine whether the vehicle will have a fault. If a fault occurs, a fault warning is given in time, which can improve the accuracy and timeliness of fault diagnosis. And because the correlation between various different-dimensional data and each state of the vehicle is established, potential abnormal risks can also be discovered.

[0050] Those skilled in the art should understand that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0051] The present application is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, and the combination of flows and / or blocks in the flowchart and / or block diagram, can be realized by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a system for realizing the functions specified in one Figure 1 one flow or multiple flows and / or blocks Figure 1 a system for realizing the functions specified in one block or multiple blocks.

[0052] The above are only the embodiments of the present application and are not used to limit the present application. For those skilled in the art, various changes and modifications can be made to the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A vehicle fault warning method based on multi-source data fusion, characterized in that: The method comprises: Obtain multi-source operation history data of the vehicle within a preset continuous historical period; Preprocessing multi-source operation historical data to obtain multiple historical data sample sequences, and constructing a training set with the multiple historical data sample sequences; Building a state prediction model based on a deep learning algorithm, the state prediction model is used to extract and fuse features from multiple historical data sample sequences; The state prediction model is trained according to the training set to obtain a trained state prediction model; Obtain the vehicle's current multi-source real-time operating data; Inputting the vehicle's current multi-source real-time operating data into a trained state prediction model, wherein the trained state prediction model outputs a real-time state prediction result; Perform fault analysis on the real-time status prediction results to obtain vehicle fault results; Generate early warning signals based on vehicle failure results, and use the early warning signals for visual early warning; The real-time state estimation result is a current voltage estimation value of the single cell, and the current voltage estimation value of the single cell is a current response output voltage of the single cell; Perform fault analysis on the real-time status prediction results to obtain vehicle fault results, including: Obtaining the voltage measurement value of each single battery during the discharge stage; Constructing a discharge error sequence according to the current voltage estimation value and the voltage measurement value in the discharge stage, wherein the discharge error sequence includes a plurality of voltage deviations, and calculating a deviation average value of the discharge error sequence; Obtain the real-time current of each single battery during the discharge stage, and construct a current point set based on the real-time current; A current-voltage distribution diagram is constructed according to the current point set and the discharge error sequence; each real-time current and the voltage deviation corresponding to the real-time current constitute a distribution point in the current-voltage distribution diagram; According to the current-voltage distribution diagram, the abnormal condition of each single battery is determined, and the abnormal condition of each single battery is used as the vehicle failure result.

2. The vehicle fault warning method based on multi-source data fusion according to claim 1 is characterized in that: The multi-source operation history data includes at least: vehicle status data and battery status; The battery status includes at least: battery output voltage, battery output current, battery charge, battery temperature and single cell voltage; The vehicle status data at least includes: driving speed, motor speed and mileage; The multiple historical data sample sequences are vehicle status sequences and battery status sequences. The vehicle status sequence includes at least a driving speed sequence, a motor speed sequence, and a driving mileage sequence; the battery status sequence includes at least a battery output voltage sequence, a battery output current sequence, a battery charge sequence, a battery temperature sequence, and a single cell voltage sequence.

3. The vehicle fault early warning method based on multi-source data fusion according to claim 2 is characterized in that: The state estimation model includes: a convolutional network unit, an attention mechanism unit and a long short-term memory unit; The convolutional network unit takes the battery state sequence as input and is used to perform nonlinear fitting on the battery state sequence. The convolutional network unit outputs a voltage estimate, and the voltage estimate is a response output voltage of a single battery; The attention mechanism unit takes the battery state sequence and the vehicle state sequence as input, and is used to mine the correlation between the battery state sequence and the vehicle state sequence, and the attention mechanism unit outputs the correlation degree; The long short-term memory unit takes the voltage estimate output by the convolutional network unit and the correlation degree output by the attention mechanism unit as input, and is used to fuse the voltage estimate and the correlation degree and extract the time dependency between the voltage estimate and the correlation degree.

4. The vehicle fault early warning method based on multi-source data fusion according to claim 1 is characterized in that: According to the current-voltage distribution diagram, determine the abnormal conditions of each single battery, including: Based on a preset algorithm, the distribution distance between each distribution point and the origin in the current-voltage distribution graph is calculated; Determine whether the distribution distance is greater than the threshold distance. If so, mark the distribution point as abnormal to indicate that the single cell corresponding to the voltage deviation of the distribution point is in an abnormal state; if not, mark the distribution point as normal to indicate that the single cell corresponding to the voltage deviation of the distribution point is in a normal state.

5. The vehicle fault warning method based on multi-source data fusion according to claim 4 is characterized in that: The method further includes determining a threshold distance, comprising: Screening the real-time current corresponding to the voltage deviation equal to the preset deviation and marking it as the threshold current; Traverse all distribution points in the current-voltage distribution diagram, find the distribution point corresponding to the threshold current, and mark it as the threshold point; Based on a preset algorithm, the distance between the threshold point and the origin is calculated to obtain the threshold distance.

6. A vehicle fault warning system based on multi-source data fusion, used to implement the vehicle fault warning method based on multi-source data fusion according to any one of claims 1 to 5, characterized in that: The system comprises: A data acquisition module is used to acquire multi-source operation history data of the vehicle within a preset continuous history time; A data processing module is used to pre-process multi-source operation historical data to obtain multiple historical data sample sequences, and to construct a training set with the multiple historical data sample sequences; A model building module, used to build a state prediction model based on a deep learning algorithm, wherein the state prediction model is used to extract and fuse features of multiple historical data sample sequences; The model training module is used to train the state prediction model according to the training set to obtain a trained state prediction model; when the trained state prediction model inputs multi-source real-time operation data, it outputs a real-time state prediction result, and the real-time state prediction result is used for fault analysis and early warning of the vehicle.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the vehicle fault warning method based on multi-source data fusion described in any one of claims 1 to 5 is implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the vehicle fault warning method based on multi-source data fusion described in any one of claims 1 to 5 is implemented.

Images