A method and system for predicting the remaining driving range of an electric vehicle

Abstract

The application provides a kind of electric vehicle remaining driving range prediction method and system, for electric vehicle in low temperature environment, road working condition under the poor quality driving data, according to the different data types, by different preprocessing mode, make it become time sequence normal, not increase not lack, no abnormal point data;Data reconstruction is carried out using sliding window method, each driving event is divided into multiple short driving segments for feature extraction;From traffic conditions, driving behavior, vehicle state, external environment four aspects extract features;The extracted features are input into the time domain convolutional neural network for training, and a multi-element cross-fusion range prediction model of man-vehicle-road condition-environment is constructed, and finally the remaining driving range prediction is realized.The application can accurately predict the driving range of the vehicle, improve the use experience of electric vehicles, and provide assistance for the development of charging infrastructure, energy management optimization and the continuous development of electric vehicle technology.

Description

Technical Field

[0001] The present invention belongs to the field of remaining mileage prediction, and specifically relates to a method and system for predicting the remaining mileage of an electric vehicle, and more particularly to a method and system for predicting the remaining mileage of an electric vehicle that integrates multiple factors of people, vehicles, road conditions and the environment. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Predicting the remaining driving range of electric vehicles is crucial for improving user experience, trip planning, and driving safety. It helps users accurately understand the battery charge status, optimally plan charging times and charging station locations, and avoid mid-route stops or difficulties caused by low battery. Battery performance degrades significantly in cold environments, making these difficulties more likely and increasing the difficulty of predicting remaining driving range. Furthermore, predicting remaining driving range can optimize driving behavior by adjusting driving style and speed, extending range, improving energy efficiency, and reducing unnecessary emissions. In autonomous driving and intelligent transportation systems, remaining driving range prediction facilitates intelligent decision-making and route planning, ensuring efficient and intelligent driving of electric vehicles. Furthermore, remaining driving range prediction aids in the planning and layout of charging facilities, ensuring that electric vehicle drivers can easily find charging stations while driving.

[0004] In summary, the prediction of remaining mileage can not only improve the user experience of electric vehicles, but also promote the construction of charging infrastructure, energy management optimization and the sustainable development of electric vehicle technology.

[0005] When it comes to the problem of remaining mileage prediction, scholars at home and abroad usually conduct research based on model and data-driven methods.

[0006] The model-based method generally proceeds from the following two aspects. The state of charge (SOC) reflects the remaining power of the battery pack. Therefore, the remaining mileage of the electric vehicle can be simply predicted from the mapping relationship between the state of charge of the power battery and the vehicle mileage. Common SOC estimation methods are based on characterization parameters, ampere-hour integral, model-based estimation and data-driven methods. The state of energy (SOE) directly describes the energy supply capacity of the power battery and is more suitable for predicting the remaining mileage than SOC. However, the SOE estimation process is relatively complicated and the implementation cost is high, and there is relatively little research on SOE. Both of the above methods require the study of battery models and vehicle energy consumption to predict future vehicle operating conditions, and the cost is high in order to achieve a certain vehicle mileage prediction accuracy.

[0007] Data-driven methods use vehicle operating data to develop prediction models. In existing research, data-driven methods have superior prediction accuracy compared to other methods, but their interpretability is poor, particularly when predicting implicit vehicle changes such as vehicle speed and the battery's state of charge (SOC), state of exhaust gas (SOE), and state of hydration (SOH). However, few researchers have attempted to directly predict the remaining range of electric vehicles using data-driven methods. More studies have indirectly predicted the remaining range by predicting the mileage of an electric vehicle at two different points in time and then calculating the difference. This method requires two predictions, resulting in cumulative errors and making it difficult to guarantee prediction accuracy and stability. Summary of the Invention

[0008] In order to solve the above problems, the present invention proposes a method and system for predicting the remaining mileage of electric vehicles. The present invention can accurately predict the vehicle mileage, provide support for improving the user experience of electric vehicles, promote the construction of charging infrastructure, optimize energy management and the sustainable development of electric vehicle technology.

[0009] According to some embodiments, the present invention adopts the following technical solutions:

[0010] A method for predicting the remaining mileage of an electric vehicle comprises the following steps:

[0011] Obtaining driving data of electric vehicles under target temperature environment and road conditions;

[0012] Preprocessing the acquired driving data;

[0013] Reconstruct the preprocessed data and extract driving segments using the sliding window method;

[0014] Based on the driving segment, relevant features are extracted from four aspects: driving behavior, vehicle status, traffic conditions, and external environment;

[0015] The model constructed by training the extracted features is used to obtain a mileage prediction model that integrates multiple factors including people, vehicles, road conditions, and the environment.

[0016] The remaining mileage prediction model is used to realize the prediction of the remaining mileage by utilizing the multi-element cross-integration model of people, vehicles, road conditions and environment.

[0017] As an optional implementation manner, the driving data includes vehicle number, data collection time, vehicle status, charging status, operating mode, vehicle speed, mileage, total voltage, total current, SOC, DC-DC status, insulation resistance, accelerator pedal travel value, brake pedal travel value, gear brake status, driving status, region city, province, highest voltage battery subsystem number, highest voltage battery cell code, highest battery cell voltage value, lowest voltage battery subsystem number, lowest voltage battery cell code, lowest battery cell voltage value, highest temperature subsystem number, highest temperature probe number, highest temperature value, lowest temperature subsystem number, lowest temperature probe number, lowest temperature value, highest alarm level, general alarm flag, number of drive motors, longitude information and latitude information;

[0018] Obtain historical meteorological data based on data collection time, longitude information, latitude information, region city and province to obtain vehicle external temperature data.

[0019] As an optional implementation, the process of pre-processing the acquired driving data includes: re-ordering the time-disordered data;

[0020] For duplicate data, delete it directly according to its timestamp;

[0021] For missing data, linear interpolation algorithm is used to fill in the missing data;

[0022] For abnormal point data, a multi-stage filtering algorithm combining median filtering and wavelet soft threshold noise reduction is used for processing.

[0023] As a further step, the processing of the multi-stage filtering algorithm includes: performing maximum-minimum normalization processing on the data;

[0024] Calculate the standard deviation and local difference of the normalized data;

[0025] Calculate the volatility index of the data based on the standard deviation and local deviation;

[0026] If the volatility index of a certain type of data is less than the set value, median filtering is used, otherwise wavelet soft threshold noise reduction is used.

[0027] As a further step, the process of wavelet soft threshold denoising includes: decomposing the signal into low-frequency and high-frequency components using wavelet transform;

[0028] For each wavelet coefficient in the high-frequency part, soft threshold processing is performed;

[0029] The processed low-frequency part and high-frequency part are merged through inverse wavelet transform to reconstruct the signal.

[0030] As an optional implementation, the process of reconstructing the pre-processed data includes: removing all charging data based on the vehicle state; traversing the extracted data, treating data points where the absolute value of the SOC difference between two adjacent data points is less than or equal to a set value as the same driving event; checking the SOC at the end of the driving event; if the SOC is less than or equal to a threshold, retaining the data segment; otherwise, deleting it;

[0031] The sliding window method is used to divide each driving event into multiple short driving segments of set duration.

[0032] As an optional implementation method, the process of extracting relevant features from four aspects, namely driving behavior, vehicle status, traffic conditions, and external environment, includes:

[0033] The characteristics of driving behavior include the accelerator ratio, brake ratio, average accelerator pedal travel value, and average brake pedal travel value in a short driving segment;

[0034] The characteristics of the vehicle state include the starting SOC, ending SOC, ΔSOC, average voltage, average current, average maximum cell temperature, and average minimum cell temperature within a short driving segment;

[0035] Traffic condition characteristics include maximum speed, minimum speed, average speed, and average acceleration within a short driving segment;

[0036] Characteristics of the external environment include the air temperature outside the vehicle during short driving segments.

[0037] As an optional implementation, the constructed model is a time-domain convolutional neural network model, and the model structure includes an input layer, a first hidden layer, a second hidden layer and an output layer, wherein the first hidden layer and the second hidden layer each include multiple residual modules, each residual module includes a convolution operation layer, an activation function and a Dropout operation layer connected in sequence, and a fully connected layer is arranged before the output layer.

[0038] As an optional implementation method, in the process of training the constructed model using the extracted features, the feature data is divided into one part as training data and the other part as test data, and the mean square error is used as the loss function to measure the gap between the predicted value and the true value. By calculating the gradient of the loss function with respect to the model parameters, back propagation and optimization are performed, and the optimization algorithm is used to update the parameters, minimize the loss function, and realize the update of the network parameters of the model.

[0039] A system for predicting remaining mileage of an electric vehicle, comprising:

[0040] a data acquisition module configured to acquire driving data of the electric vehicle under target temperature environment and road conditions;

[0041] a preprocessing module, configured to preprocess the acquired driving data;

[0042] A data reconstruction module is configured to reconstruct the preprocessed data and extract driving segments using a sliding window method;

[0043] a feature extraction module configured to extract relevant features from four aspects: driving behavior, vehicle status, traffic conditions, and external environment based on the driving segment;

[0044] A model training module is configured to train the constructed model using the extracted features to obtain a mileage prediction model that integrates multiple factors including human, vehicle, road condition, and environment;

[0045] The prediction module is configured to use the multi-element cross-integration mileage prediction model of human-vehicle-road condition-environment to realize the remaining mileage prediction.

[0046] A computer-readable storage medium is used to store computer instructions, and when the computer instructions are executed by a processor, the steps in the above method are completed.

[0047] An electronic device includes a memory and a processor, and computer instructions stored in the memory and executed on the processor. When the computer instructions are executed by the processor, the steps in the above method are completed.

[0048] An electric vehicle comprises the above-mentioned electronic device or system.

[0049] Compared with the prior art, the present invention has the following beneficial effects:

[0050] The present invention aims to convert the poor-quality driving data of electric vehicles in low-temperature environments and road conditions into data with normal time sequence, no additions or omissions, and no abnormal points through different preprocessing methods according to different data types.

[0051] The present invention adopts a sliding window method to reconstruct data, dividing each driving event into a plurality of short driving segments for feature extraction.

[0052] This paper addresses the inability of raw data to reflect traffic conditions and driving behavior. It proposes specific metrics to quantify these two aspects, extracting features from four aspects: traffic conditions, driving behavior, vehicle status, and the external environment. These extracted features are then fed into a time-domain convolutional neural network for training, constructing a mileage prediction model that integrates multiple factors—driver, vehicle, road conditions, and the environment—to ultimately predict the remaining driving range.

[0053] This method is challenging to predict the remaining driving range of electric vehicles in low-temperature environments, as low temperatures significantly affect battery performance, energy management, and uncertainty in vehicle energy consumption, while also increasing data noise and model complexity. However, this technology is crucial for alleviating user range anxiety, improving safety, optimizing energy management, and route planning, while also promoting the improvement and widespread adoption of electric vehicle technology.

[0054] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] The accompanying drawings, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0056] Figure 1 is a specific flow chart of remaining mileage prediction in one embodiment;

[0057] Figure 2 This is a flow chart of a multi-stage preprocessing method for poor-quality data according to an embodiment;

[0058] Figure 3 This is a schematic diagram of a time-domain convolutional neural network structure according to an embodiment;

[0059] Figure 4 This is a structural diagram of a remaining mileage prediction model according to an embodiment. DETAILED DESCRIPTION

[0060] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0061] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present invention belongs.

[0062] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or combinations thereof.

[0063] In the absence of conflict, the embodiments and features in the embodiments of this application can be combined with each other.

[0064] Example 1

[0065] A method for predicting the remaining mileage of electric vehicles based on the multi-integration of human, vehicle, road condition and environment. Figure 1 As shown in the figure, for the poor quality driving data of electric vehicles in low temperature environment and road conditions, different preprocessing methods are used according to different data types to make it into data with normal time series, no additions or omissions, and no abnormal points; the sliding window method is used to reconstruct the data, and each driving event is divided into multiple short driving segments for feature extraction; the original data cannot reflect the traffic conditions and driving behavior, so specific indicators for quantifying the above two are proposed, and features are extracted from four aspects: traffic conditions, driving behavior, vehicle status, and external environment; the extracted features are input into the time domain convolutional neural network for training, and a multi-element cross-fusion mileage prediction model of human-vehicle-road condition-environment is constructed to finally realize the remaining mileage prediction.

[0066] The following is a detailed introduction in steps:

[0067] Step 1: Obtain real vehicle data.

[0068] In this example, driving data of a certain brand of electric vehicles is collected from January to March 2023. The dataset mainly includes the following variables: vehicle unique number on the platform, data collection time, vehicle status, charging status, operating mode, vehicle speed, mileage, total voltage, total current, SOC, DC-DC status, insulation resistance, accelerator pedal travel value, brake pedal travel value, gear brake status, driving status, region, city, province, highest voltage battery subsystem number, highest voltage battery cell code, highest battery cell voltage value, lowest voltage battery subsystem number, lowest voltage battery cell code, lowest battery cell voltage value, highest temperature subsystem number, highest temperature probe number, highest temperature value, lowest temperature subsystem number, lowest temperature probe number, lowest temperature value, highest alarm level, general alarm flag, number of drive motors, longitude information, latitude information, etc.

[0069] In this embodiment, the historical meteorological data is obtained by using the Python meteostat library according to the data collection time, longitude information, latitude information, region city, and province to obtain the vehicle external temperature data.

[0070] Step 2: Data cleaning. The data set has the following characteristics: large data size, with the number of samples reaching millions of rows; high data feature dimensions, large amount of information; many data anomalies; and missing key information. Based on this, this embodiment proposes a multi-level preprocessing method for poor quality data. The process is as follows: Figure 2 shown.

[0071] Specifically:

[0072] Reorder time-disordered data;

[0073] For duplicate data, delete it directly according to its timestamp;

[0074] For missing data, linear interpolation algorithm is used to fill in the missing data;

[0075] For abnormal point data, a multi-stage filtering algorithm combining median filtering and wavelet soft threshold noise reduction is used. The specific algorithm is as follows;

[0076] First, we analyze the data for volatility. To ensure that different features (such as vehicle speed, SOC, current, voltage, temperature, etc.) are at the same order of magnitude, we perform maximum-minimum normalization to achieve standardization and ensure that they have the same influence. The formula is as follows:

[0077]

[0078] Among them, x is a feature in the original data, x ′ is the normalized data, min(x) is the minimum value of feature x, and max(x) is the maximum value of feature x.

[0079] Calculate the standard deviation and local difference of the standardized data. The formula is as follows:

[0080]

[0081]

[0082] Δx i =|x i -x i-1 |

[0083] Where σ is the standard deviation, μ is the mean, N is the data length, and x i is the i-th value in the data, Δx i is the absolute difference between the ith data point and the previous data point.

[0084] Volatility indicators are used to measure the frequency and magnitude of changes in a feature in a time series. Volatility indicators include the relative ratio of the standard deviation (i.e., the ratio of the standard deviation to the mean) and the average of the local differences.

[0085]

[0086] It is the relative ratio of the standard deviation. A relatively large standard deviation and a relatively small mean are characterized by strong volatility. The larger the ratio, the more frequent the signal changes.

[0087]

[0088] It is the average of the local differences. This indicator indicates the degree of change of the signal at adjacent time points. Signals with large fluctuations will have larger local differences.

[0089] After calculation and analysis, it was found that the characteristics of vehicle speed, SOC, voltage, temperature, etc. have small fluctuations and are suitable for median filtering; the characteristics of current, etc. have large fluctuations and are suitable for wavelet soft threshold noise reduction.

[0090] Use wavelet transform to decompose the signal x(t) into low-frequency part (approximate coefficient) and high-frequency part (detail coefficient). Use discrete wavelet transform (DWT) to decompose the signal, the formula is as follows:

[0091]

[0092] Among them, A j is the approximate coefficient at scale j, D k is the detail coefficient at scale k, ψ j (t) and φ k (t) are the scale- and time-position-dependent wavelet basis functions, respectively.

[0093] For each wavelet coefficient D k , the formula for soft threshold processing is:

[0094]

[0095]

[0096] in, is the detail coefficient after soft threshold processing, λ is the threshold, σ is the standard deviation of Gaussian white noise, N is the signal length, sign(D k ) is D k The sign of indicates the direction of the coefficient (positive or negative).

[0097] The processed approximate coefficients and detail coefficients are combined through inverse wavelet transform (IDWT) to reconstruct the signal The formula is as follows:

[0098]

[0099] in: is the approximate coefficient, and A j same, is the detail coefficient after threshold processing.

[0100] Step 3: Data reconstruction. After pretreatment, the data still contains a large amount of redundant information, and now the effective information is extracted and the data is reconstructed. First, according to the vehicle state, all charging data is removed. After traversing the screened data, since the SOC will not change in the same driving event, for two adjacent data points, if the absolute value of the SOC difference is not greater than 1, the two data points are considered to be in the same driving event, and the SOC at the end of the driving event is checked. If it is not greater than 10, the data is retained, and all data with SOC less than 10 is discarded.

[0101] Of course, in other embodiments, the above threshold value can take other values.

[0102] In the original data, only the cumulative driving mileage is provided, and the remaining driving mileage is not provided, so calibration is required. The remaining driving mileage calibration formula is as follows:

[0103] s k (t)=l k (t end )-l k (t)

[0104] Where s k (t) represents the remaining driving mileage of the kth driving event at time t, l k (t) represents the cumulative driving mileage of the kth driving event at time t, and t end represents the time at the end of the driving event.

[0105] Then, the sliding window method is used to divide each driving event into multiple 30s short driving segments.

[0106] Similarly, in other embodiments, the length of the short driving segment can also be changed.

[0107] Step 4: Feature extraction.

[0108] In the reconstructed data, the features of each short driving segment are extracted from the following four aspects, a total of 16: traffic conditions, driving behavior, vehicle state, and external environment.

[0109] The features that can reflect the traffic conditions are: maximum speed, minimum speed, average speed, and average acceleration;

[0110] The features that can reflect the driving behavior are: accelerator pedal ratio, brake pedal ratio, average accelerator pedal travel value, and average brake pedal travel value;

[0111] The features that can reflect the vehicle state are: starting SOC, ending SOC, ΔSOC, average voltage, average current, average maximum cell temperature, and average minimum cell temperature;

[0112] Features that can reflect the characteristics of the external environment include: the temperature outside the vehicle.

[0113] The above features can be directly or calculated based on the reconstructed data.

[0114] Step 5: Construct the training set and the test set.

[0115] In this embodiment, the selected data set has a total of 91 driving events meeting the above screening conditions, which are divided into a training set and a test set according to an 8:2 ratio.

[0116] In other embodiments, the number of driving events meeting the screening conditions, the division ratio, and the like can be changed.

[0117] Step 6: Construct a time-domain convolutional neural network (TCN) remaining driving range prediction model, and the specific process is as follows:

[0118] The input feature matrix X ∈ R n×m , where n represents the number of samples, and m is the number of features of each sample.

[0119] Convolution operation: an expansion convolution is used, which can expand the receptive field of the convolution by setting the expansion coefficient d, so that each layer of convolution can cover a larger range of input data without increasing the number of convolution kernels. The structure is as shown in Figure 3 The formula of the expansion convolution is:

[0120]

[0121] where x(t-d·k) is the input data at time t and the time delay d·k of the convolution kernel; d is the expansion coefficient, which determines the interval size between each convolution operation; ω k is the weight of the convolution kernel; b is the bias term; h conv (t) is the output term.

[0122] An activation function is connected to the convolution layer to introduce nonlinearity, so that the model can better fit complex time series relationships. The formula of the ReLU activation function is:

[0123] h relu (h conv )=max(0,h conv )

[0124] After the activation function, a Dropout operation is performed, which randomly sets the output of neurons with a probability of p to zero, which is enabled during training, and the specific formula is:

[0125] h drop [j]=h relu [j]*m j

[0126] Among them, m j ∈{0,1} is a randomly generated mask, and each h relu [j] With the probability of p being 0, h drop [j] is the output of dropout.

[0127] Residual connections are used to prevent the vanishing gradient problem and speed up training. The formula for residual connections is:

[0128] h i =f(h i-1 ,W 1×1 )+h drop

[0129] Among them, h i is the current residual block output, W 1×1 is the weight matrix of the 1×1 convolution kernel, h i-1 is the output of the previous residual block, F(h i-1 ,W 1×1 ) is the 1×1 convolution operation on the output of the previous residual block.

[0130] The fully connected layer is used to map the extracted features to the final prediction output. The fully connected layer converts the output of the convolutional layer into a scalar:

[0131] y pred =W fc ·h end +b fc

[0132] Among them, y pred is the final prediction output of the model, i.e. the remaining mileage; h end is the output of the last residual block; W fc and b fc are the weights and biases of the fully connected layer respectively.

[0133] The final network structure is as follows Figure 4 shown.

[0134] During model training, the mean square error (MSE) is used as the loss function L(θ) to measure the gap between the predicted value and the true value:

[0135]

[0136] Where n is the number of samples; y i and y pred,i are the true value and predicted value of the i-th sample respectively;

[0137] By calculating the gradient of the loss function relative to the model parameters, backpropagation and optimization are performed, and the Adam optimization algorithm is used to update the parameters to minimize the loss function. The goal of optimization is to continuously update the network parameters θ so that the loss function L(θ) is minimized:

[0138]

[0139] Where η is the learning rate, is the correction of the first moment of the gradient, is the correction of the second-order moment of the gradient, and ∈ is a very small constant that is not 0.

[0140] Step 7: Obtain real-time vehicle data, preprocess and reconstruct it, or use the trained time-domain convolutional neural network (TCN) remaining mileage prediction model to make predictions based on the test set.

[0141] It will be understood by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0142] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0143] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0144] These computer program instructions can also be loaded into a computer or other programmable data processing devices, so that a series of operational steps are generated to realize the computer-implemented processes in the computer or other programmable devices, and the instructions executed in the computer or other programmable devices provide processes for implementing the functions specified in the flowchart Figure 1 one flow or a plurality of flows and / or the functions specified in the block Figure 1 one flow or a plurality of flows and / or the functions specified in the block

[0145] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. Various modifications and changes can be made to the present application by those skilled in the art without departing from the spirit and principles of the present application. Any modifications, equivalent replacements, improvements, etc. made by those skilled in the art without creative efforts should be included in the protection scope of the present application.

Claims

1. A method for predicting the remaining mileage of an electric vehicle, characterized in that: The following steps are involved: Obtaining driving data of electric vehicles under target temperature environment and road conditions; Preprocessing the acquired driving data; Reconstruct the preprocessed data and extract driving segments using the sliding window method; Based on the driving segment, relevant features are extracted from four aspects: driving behavior, vehicle status, traffic conditions, and external environment; The model constructed by training the extracted features is used to obtain a mileage prediction model that integrates multiple factors including people, vehicles, road conditions, and the environment. The remaining mileage prediction model is used to predict the remaining mileage by using the multi-element cross-integration model of people, vehicles, road conditions and environment; In the process of preprocessing the acquired driving data, a multi-stage filtering algorithm combining median filtering and wavelet soft threshold noise reduction is used to process abnormal point data. The processing process of the multi-stage filtering algorithm includes: performing maximum-minimum normalization processing on the data; Calculate the standard deviation and local difference of the normalized data; Calculate the volatility index of the data based on the standard deviation and local deviation; If the volatility index of a certain type of data is less than the set value, median filtering is used, otherwise wavelet soft threshold noise reduction is used; The data reconstruction process for the pre-processed data includes: removing all charging data based on the vehicle status; traversing the extracted data, treating the data points where the absolute value of the SOC difference between two adjacent data points is less than or equal to a set value as the same driving event; checking the SOC at the end of the driving event; if it is equal to the threshold, the data segment is retained; otherwise, it is deleted; The sliding window method is used to divide each driving event into multiple driving segments of set duration.

2. The method for predicting the remaining mileage of an electric vehicle according to claim 1, wherein: The driving data includes vehicle number, data collection time, vehicle status, charging status, operating mode, vehicle speed, mileage, total voltage, total current, SOC, DC-DC status, insulation resistance, accelerator pedal travel value, brake pedal travel value, gear brake status, driving status, region, city, province, highest voltage battery subsystem number, highest voltage battery cell code, highest battery cell voltage value, lowest voltage battery subsystem number, lowest voltage battery cell code, lowest battery cell voltage value, highest temperature subsystem number, highest temperature probe number, highest temperature value, lowest temperature subsystem number, lowest temperature probe number, lowest temperature value, highest alarm level, general alarm flag, number of drive motors, longitude information and latitude information; Based on the data collection time, longitude information, latitude information, region city and province, historical meteorological data is determined to obtain vehicle external temperature data.

3. The method for predicting the remaining mileage of an electric vehicle according to claim 1, wherein: The process of preprocessing the acquired driving data includes: reordering the temporally chaotic data; For duplicate data, delete it directly according to its timestamp; For missing data, linear interpolation algorithm is used to fill in the missing data; For abnormal point data, a multi-stage filtering algorithm combining median filtering and wavelet soft threshold noise reduction is used for processing.

4. The method for predicting the remaining mileage of an electric vehicle according to claim 1, wherein: The process of wavelet soft threshold denoising includes: using wavelet transform to decompose the signal into low-frequency and high-frequency parts; For each wavelet coefficient in the high-frequency part, soft threshold processing is performed; The processed low-frequency part and high-frequency part are merged through inverse wavelet transform to reconstruct the signal.

5. The method for predicting the remaining mileage of an electric vehicle according to claim 1, wherein: The process of extracting relevant features from four aspects: driving behavior, vehicle status, traffic conditions, and external environment includes: The characteristics of driving behavior include accelerator ratio, brake ratio, average accelerator pedal travel value, and average brake pedal travel value; The characteristics of the vehicle state include starting SOC, ending SOC, ΔSOC, average voltage, average current, average maximum cell temperature, and average minimum cell temperature; Traffic condition characteristics include maximum speed, minimum speed, average speed, and average acceleration; Characteristics of the external environment include the air temperature outside the vehicle.

6. The method for predicting the remaining mileage of an electric vehicle according to claim 1, wherein: The constructed model is a time-domain convolutional neural network model, and the model structure includes an input layer, a first hidden layer, a second hidden layer and an output layer, wherein the first hidden layer and the second hidden layer each include multiple residual modules, each residual module includes a convolution operation layer, an activation function and a Dropout operation layer connected in sequence, and a fully connected layer is arranged before the output layer.

7. The method for predicting the remaining mileage of an electric vehicle according to claim 1, wherein: In the process of training the constructed model using the extracted features, the feature data is divided into two parts, one as training data and the other as test data. The mean square error is used as the loss function to measure the gap between the predicted value and the true value. By calculating the gradient of the loss function relative to the model parameters, back propagation and optimization are performed. The optimization algorithm is used to update the parameters, minimize the loss function, and update the network parameters of the model.

8. A system for predicting remaining mileage of an electric vehicle, characterized in that: include: a data acquisition module configured to acquire driving data of the electric vehicle under target temperature environment and road conditions; a preprocessing module, configured to preprocess the acquired driving data; A data reconstruction module is configured to reconstruct the preprocessed data and extract driving segments using a sliding window method; a feature extraction module configured to extract relevant features from four aspects: driving behavior, vehicle status, traffic conditions, and external environment based on the driving segment; A model training module is configured to train the constructed model using the extracted features to obtain a mileage prediction model that integrates multiple factors including human, vehicle, road condition, and environment; A prediction module is configured to use the multi-element cross-integration mileage prediction model of the driver, vehicle, road condition and environment to predict the remaining mileage; In the process of preprocessing the acquired driving data, a multi-stage filtering algorithm combining median filtering and wavelet soft threshold noise reduction is used to process abnormal point data. The processing process of the multi-stage filtering algorithm includes: performing maximum-minimum normalization processing on the data; Calculate the standard deviation and local difference of the normalized data; Calculate the volatility index of the data based on the standard deviation and local deviation; If the volatility index of a certain type of data is less than the set value, median filtering is used, otherwise wavelet soft threshold noise reduction is used; The data reconstruction process for the pre-processed data includes: removing all charging data based on the vehicle status; traversing the extracted data, treating the data points where the absolute value of the SOC difference between two adjacent data points is less than or equal to a set value as the same driving event; checking the SOC at the end of the driving event; if it is equal to the threshold, the data segment is retained; otherwise, it is deleted; The sliding window method is used to divide each driving event into multiple driving segments of set duration.

9. A computer-readable storage medium, characterized in that: Used to store computer instructions, which, when executed by a processor, complete the steps of the method according to any one of claims 1 to 7.

10. An electronic device, characterized in that: The method comprises a memory and a processor, and computer instructions stored in the memory and executed on the processor, wherein the steps of the method according to any one of claims 1 to 7 are completed when the computer instructions are executed by the processor.

11. An electric vehicle, characterized in that: The method comprises the electronic device according to claim 10 or the system according to claim 8, or adopts the steps in the method according to any one of claims 1 to 7 to perform remaining mileage prediction.

Images