Intelligent data processing method and system for intelligent connected vehicles

Abstract

The present invention relates to an intelligent data processing method and system for intelligent networked vehicles, belonging to the technical field of intelligent networked vehicles. The intelligent data processing method and system for intelligent networked vehicles include: collecting historical data during the driving process of new energy vehicles, obtaining power change, mileage, battery temperature and battery health data, performing fixed window segmentation processing on all historical data, and obtaining a data window containing a unit mileage energy consumption rate sequence; the system can subdivide it into two modes of "normal temperature high-speed cruising" and "low temperature high-speed cruising" according to different battery temperatures; even if the fluctuation characteristics of the energy consumption sequence are similar, due to the difference in efficiency factors, they actually represent different energy consumption states. Under complex driving conditions, the clustering results can also accurately reflect the actual energy consumption mode of the vehicle, and improve the accuracy of the prediction of the cruising range of intelligent networked vehicles.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent networked vehicles, and in particular to an intelligent data processing method and system for intelligent networked vehicles. Background Art

[0002] With the rapid development of new energy vehicles, accurately predicting the vehicle's range has become a key issue. The current mainstream range prediction method is mainly based on analysis of historical driving data. A common method is to use a clustering algorithm to analyze the vehicle's power-mileage historical relationship curve, divide these curves into different energy consumption mode categories, and then predict the range based on the clustering results; specifically, the existing technology usually adopts the following steps: first, obtain the vehicle's historical power data and corresponding mileage data, and establish a power-mileage relationship curve; then use an algorithm to cluster these curves into several modes such as high power consumption, medium power consumption, and low power consumption; finally, match the corresponding energy consumption mode according to the current driving data to predict the remaining range. This method takes into account the energy consumption differences under different driving conditions and improves the accuracy of range prediction to a certain extent.

[0003] In the prior art, when calculating the distance between the power-mileage curves, the clustering algorithm treats the entire curve as an indivisible clustering object. This method ignores the fact that the same curve may exhibit different energy consumption characteristics at different stages, such as different driving states such as urban congestion and high-speed cruising, and only focuses on the simple energy consumption changes between adjacent sampling points and fails to effectively identify the energy consumption pattern conversion characteristics on a larger time scale. As a result, under complex driving conditions, the clustering results cannot accurately reflect the vehicle's actual energy consumption pattern, thereby affecting the accuracy of the range prediction. Summary of the invention

[0004] Based on this, it is necessary to provide an intelligent data processing method and system for intelligent connected vehicles to address the problem that under complex driving conditions, the clustering results cannot accurately reflect the actual energy consumption pattern of the vehicle, thereby affecting the accuracy of the cruising range prediction.

[0005] An intelligent data processing method and system for an intelligent connected vehicle, comprising:

[0006] S1: Collect historical data of new energy vehicles during driving, obtain data on power change, mileage, battery temperature and battery health, perform fixed window segmentation processing on all historical data, and obtain a data window containing a unit mileage energy consumption rate sequence;

[0007] S2: For the energy consumption rate sequence, preliminary optimization is performed based on DTW distance calculation and energy consumption fluctuation characteristic analysis. The optimization factor a is obtained by calculating the fluctuation degree and stability, and the distance metric D1 after preliminary optimization is obtained;

[0008] S3: Construct the efficiency feature vector F according to the battery temperature and health, design an adaptive adjustment mechanism based on the arctan function, and calculate the final optimization distance D based on the preliminary optimization distance D1;

[0009] S4: Use the k-means clustering algorithm to perform cluster analysis on the optimized distance metric, determine the optimal number of clusters k through the elbow rule, and obtain clustering results that can reflect the characteristics of different driving conditions;

[0010] S5: Match the current driving window to the closest cluster category based on the optimized distance metric, select the most similar historical samples within the category for weighted average calculation, and obtain an accurate remaining range prediction result;

[0011] The sampling frequency in S1 is set to once per second, the recorded power change is ΔE, the mileage is Δs, the battery temperature and the battery health, and the unit mileage energy consumption rate of each sampling point is e i =ΔE i / Δs i , the length of the window is T, T is 10 minutes, and the number of windows is 600, and the data points used for clustering are obtained:

[0012] P = {e_i}i∈[1,600]

[0013] The e_i is the energy consumption rate per unit mileage of each sampling point in the window, and each of the windows contains an energy consumption rate sequence for subsequent distance metric optimization calculation.

[0014] In one embodiment, the DTW distance between the two window energy consumption rate sequences is d, and the optimized distance calculation formula is:

[0015] D1 = d(1 + α).

[0016] In one embodiment, the calculation formula of the adjustment factor α is:

[0017] α=log(1+Σ|e i+1 –e i | / e avg )exp(-n stable / n)

[0018]

[0019] Among them, the e i is the energy consumption rate of the i-th sampling point in the window, and the e i +1 is the energy consumption rate of the i+1th sampling point in the window, and the e avgis the average energy consumption rate in the window, n_{stable} is the number of stable sampling points in the window whose energy consumption rate change is less than the threshold δ, n is the total number of sampling points in the window, I() is an indicative function, which takes 1 when the condition is met, otherwise takes 0, and δ is the stability threshold.

[0020] In one embodiment, the fluctuation quantization term log(1+Σ|e i+1 –e i | / e avg ) reflects the fluctuation degree of energy consumption by calculating the cumulative amount of energy consumption rate changes between adjacent sampling points in the window, and dividing by e avg The normalization process is performed to eliminate the influence of different energy consumption levels and to take the logarithmic compression effect of large fluctuations. The stability weight term exp(-n stable / n) stable / n represents the proportion of sampling points with stable energy consumption in the window. This ratio is mapped to the interval (0,1] through an exponential function. When the proportion of stable points is high, the exp term approaches 0, thereby weakening the influence of the fluctuation term. When the proportion of stable points is low, the exp term approaches 1.

[0021] In one embodiment, when the window energy consumption rate changes, the Σ|e i+1 –e i | value is large, and due to unstable energy consumption, n stable / n is small, the exp term approaches 1, and finally a larger α value is obtained. When the window energy consumption rate is relatively stable, the Σ|e i+1 –e i | is small, and due to the stable energy consumption, n stable As the / n value is larger, the exp term approaches 0, and eventually a smaller α value is obtained.

[0022] In one of the embodiments, the d is a basic metric of DTW, the (1+α) term can dynamically adjust the original distance, and the α value is always non-negative.

[0023] In one embodiment, the complete distance calculation formula is:

[0024]

[0025] Wherein, D is the final optimized distance value, D1 is the preliminary optimized distance in step a, F1 and F2 are the efficiency feature vectors of the two windows to be compared, λ is the adaptive adjustment coefficient, and the efficiency feature vector F=[T / T max ,H / 100] contains two components: T is the battery temperature, through the T maxNormalized, H is the battery health, the adaptive adjustment coefficient Wherein, σ(e) is the standard deviation of the energy consumption rate in the window, k is the proportionality coefficient and its initial value is set to 1.5, and ε is a small positive number used to avoid the denominator being zero.

[0026] In one embodiment, the arctan function maps any real number to the interval (-π / 2, π / 2), and the arctan function is divided by π and then added with 0.5 to ensure that the final adjustment factor range is within the interval (0, 1).

[0027] In one of the embodiments, the proportionality coefficient k determines the degree of influence of the efficiency characteristic difference on the final distance, and its value needs to be adjusted according to the actual application scenario.

[0028] In one embodiment, the current travel window W current Assign it to the closest cluster category C based on the optimized distance metric i And it selects p = 10 historical windows closest to the current window as the reference sample set S = {W1, W2, ..., W p}, the final remaining mileage prediction is calculated using the following formula:

[0029]

[0030] Where: R is the predicted remaining mileage, SOC is the current battery remaining power percentage, d i The reference sample W i With the current window W current The inverse of the optimization distance, s i The reference sample W i The corresponding actual mileage.

[0031] 1. During the clustering process, the power-mileage curve will show different energy consumption characteristics at different driving stages. Although the DTW distance can handle the nonlinear alignment of time series and effectively identify similar energy consumption change patterns, it mainly focuses on the overall shape matching of the sequence, which leads to the neglect of energy consumption fluctuation characteristics. For example, when the energy consumption rate sequences of two windows both show an upward trend, even if one sequence is a smooth increase and the other sequence is a fluctuating increase, the distance will judge them as relatively similar energy consumption patterns because of the similar overall trends. In this case, it is difficult to accurately reflect the differences in volatility and stability of energy consumption sequences by relying solely on the distance.

[0032] 2. By introducing key parameters such as battery temperature and battery health to further adjust the distance calculation, not only the similarity of energy consumption sequences is considered, but also the factors affecting energy efficiency are integrated so that the clustering results can more accurately reflect the energy consumption characteristics under different conditions. For example, when the energy consumption sequences of two windows are similar but the battery temperatures are very different, the optimized distance will increase accordingly, thereby dividing them into different categories. This classification method is more conducive to the system's understanding and prediction of the vehicle's energy consumption performance under different conditions. For the same high-speed cruising condition, the system can subdivide it into two modes, "normal temperature high-speed cruising" and "low temperature high-speed cruising", according to the different battery temperatures. Even if the fluctuation characteristics of the energy consumption sequences are similar, due to the differences in efficiency factors, they actually represent different energy consumption states. This subdivision is crucial to improving the accuracy of the range prediction. Under complex driving conditions, the clustering results can also accurately reflect the vehicle's true energy consumption mode and improve the accuracy of the range prediction of intelligent connected vehicles. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0034] Figure 1 It is a schematic diagram of the system flow of the present invention. DETAILED DESCRIPTION

[0035] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0036] It should be noted that when a component is referred to as being "fixed to" or "disposed on" another component, it may be directly on the other component or there may be a central component. When a component is considered to be "connected to" another component, it may be directly connected to the other component or there may be a central component at the same time. The terms "vertical", "horizontal", "upper", "lower", "left", "right" and similar expressions used in the specification of the present invention are for illustrative purposes only and do not represent the only implementation method.

[0037] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0038] In the present invention, unless otherwise clearly specified and limited, a first feature being “above” or “below” a second feature may mean that the first feature is directly in contact with the second feature, or the first feature and the second feature are in contact indirectly through an intermediate medium. Moreover, a first feature being “above”, “above”, and “above” a second feature may mean that the first feature is directly above or obliquely above the second feature, or simply means that the first feature is higher in level than the second feature. A first feature being “below”, “below”, and “below” a second feature may mean that the first feature is directly below or obliquely below the second feature, or simply means that the first feature is lower in level than the second feature.

[0039] Unless otherwise defined, all technical and scientific terms used in the specification of the present invention have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used in the specification of the present invention includes any and all combinations of one or more related listed items.

[0040] Combine the following Figure 1 The present invention describes an intelligent data processing method and system for an intelligent connected vehicle.

[0041] In one embodiment, data collection and preprocessing;

[0042] The present invention first obtains historical data of different new energy vehicles of the same model during driving. The sampling frequency is set to once per second, and the power change ΔE, mileage Δs, battery temperature and battery health are recorded. The collected raw data is preprocessed to calculate the energy consumption rate per unit mileage e at each sampling point. i =ΔE i / Δs i Then, a fixed window method with a window length T of 10 minutes is used to segment the data and remove data segments less than 10 minutes. For each window containing 600 sampling points, the data points used for clustering are obtained:

[0043] P = {e_i}i∈[1,600]

[0044] Among them, e_i is the energy consumption rate per unit mileage of each sampling point in the window. Each window contains an energy consumption rate sequence, which is used for subsequent distance metric optimization calculation. The data used for clustering here only contains the energy consumption rate per unit mileage calculated from the battery change and the mileage. The battery temperature and battery health data are used for subsequent feature calculations.

[0045] In one embodiment, distance metric optimization and cluster analysis based on energy consumption characteristics;

[0046] Big step logic:

[0047] In the prior art, the problem of the k-means clustering algorithm when calculating the distance between the power-mileage curves is that the entire curve is treated as an indivisible clustering object, ignoring that the same curve may show different energy consumption characteristics at different stages; it focuses on the simple energy consumption changes between adjacent sampling points and fails to effectively identify the energy consumption mode conversion characteristics on a larger time scale. In response to the above problem, in step a, the problem of energy consumption fluctuation feature identification is solved by obtaining the optimization factor α. The degree of fluctuation and stability of the energy consumption rate sequence in the analysis window are analyzed to achieve dynamic adjustment of the DTW distance. The distance calculation can effectively distinguish between the energy consumption modes of steady changes and fluctuating changes, and more accurately reflect the characteristics of different driving conditions. In step b, a deep optimization based on efficiency characteristics is obtained for the problem of the impact of battery status on energy efficiency. By combining the adaptive adjustment mechanism of the two key parameters of battery temperature and health, the distance calculation can reflect the differences in energy consumption characteristics under different battery states. This design not only takes into account the similarity of energy consumption sequences, but also integrates the key factors affecting energy efficiency, so that the clustering results can more accurately reflect the energy consumption differences in actual driving scenarios.

[0048] Finally, in step c, the k-means clustering algorithm is used to perform cluster analysis on the optimized distance metric, and the optimal number of clusters is determined by the elbow rule.

[0049] a. Obtain the preliminary optimization factor a of the distance metric based on the energy consumption fluctuation characteristics.

[0050] Detailed logic:

[0051] During the k-means clustering process, the battery-mileage curve will show different energy consumption characteristics in different driving stages. Although the DTW distance can handle the nonlinear alignment of time series and effectively identify similar energy consumption change patterns, it mainly focuses on the overall shape matching of the sequence, which causes it to ignore the energy consumption fluctuation characteristics. For example, when the energy consumption rate sequences of the two windows both show an upward trend, even if one of the sequences is a smooth increase and the other sequence is a fluctuating increase, the DTW distance will judge them as relatively similar energy consumption patterns because of the similar overall trends. In this case, relying solely on the DTW distance is difficult to accurately reflect the differences in volatility and stability of the energy consumption series.

[0052] To address this problem, step a designs an optimization scheme based on energy consumption fluctuation characteristics: the distance calculation is adjusted by analyzing the degree of fluctuation and stability of the energy consumption rate sequence within the window. Specifically, even if two sequences show similar change trends after DTW alignment, if one of the sequences fluctuates relatively violently, the energy consumption rate change between adjacent sampling points is large, and the stability is poor, then the distance value between it and the other window should be increased; while for energy consumption sequences with gentle fluctuations, the energy consumption rate change between adjacent sampling points is small and the stability is good, then the distance value should be kept relatively small. This distance adjustment mechanism based on energy consumption fluctuation characteristics can make up for the shortcomings of DTW in characterizing energy consumption change characteristics and improve the accuracy of clustering.

[0053] Formula logic:

[0054] In view of the above-described problems, a distance optimization formula based on energy consumption fluctuation characteristics is proposed. The optimized distance calculation formula is:

[0055] D1=d(1+α)

[0056] Where D1 is the optimized distance value, d is the DTW distance between the two window energy consumption rate sequences, and α is the adjustment factor based on the energy consumption fluctuation characteristics. The calculation formula of the adjustment factor α is:

[0057] α=log(1+Σ|e i+1 –e i | / e avg )exp(-n stable / n)

[0058]

[0059] where e i is the energy consumption rate of the i-th sampling point in the window, e i +1 is the energy consumption rate of the i+1th sampling point in the window, e avgis the average energy consumption rate in the window, n_{stable} is the number of stable sampling points in the window where the energy consumption rate change is less than the threshold δ, and n is the total number of sampling points in the window. I() is an indicative function, which takes 1 when the condition is met and 0 otherwise; δ is the stability threshold, which is set to a small proportion of the average energy consumption rate, here set to 3%.

[0060] The fluctuation quantification term log(1+Σ|e i+1 –e i | / e avg ) reflects the fluctuation of energy consumption by calculating the cumulative amount of energy consumption rate changes between adjacent sampling points in the window. avg The purpose of normalization is to eliminate the influence of different energy consumption levels, to compress the influence of large fluctuations by logarithmic compression, and to avoid distortion of distance calculation caused by extreme values. stable / n) stable / n represents the proportion of sampling points with stable energy consumption in the window. This ratio is mapped to the interval (0,1] through an exponential function. When the proportion of stable points is high, the exp term approaches 0, thereby reducing the influence of the fluctuation term; when the proportion of stable points is low, the exp term approaches 1, retaining the influence of the fluctuation term.

[0061] Through the above design, when the energy consumption in the window fluctuates, such as in frequent acceleration and deceleration scenarios, Σ|e i+1 –e i | is large, and due to unstable energy consumption, n stable / n is small, making the exp term approach 1, and finally getting a larger α value, which significantly increases the distance between this window and other windows. On the contrary, when the energy consumption in the window is relatively stable, such as in the constant speed cruising scenario, Σ|e i+1 –e i | is small, and due to the stable energy consumption, n stable A larger / n value makes the exp term approach 0, and eventually a smaller α value is obtained, and the distance change is not obvious.

[0062] For the overall optimization, the DTW distance d is kept as the basic metric, and the original distance is dynamically adjusted through the (1+α) term, where the α value is always non-negative, ensuring that the optimized distance will not be less than the original distance; finally, the value of α is subject to the dual constraints of energy consumption fluctuation and stability, avoiding the deviation that may be caused by a single feature. In other words, even if the energy consumption rate sequences of the two windows show similar overall trends after DTW alignment, if their fluctuation characteristics are significantly different, the optimized distance will increase accordingly, which will help to divide them into different clusters and improve the recognition ability of different driving conditions.

[0063] Let's take a typical scenario as an example. Assume there are three energy consumption rate sequences, all of which show a trend from low energy consumption to high energy consumption. Sequence A shows a steady rise, the energy consumption rate changes slowly and continuously, the change between adjacent sampling points is small, and most sampling points are stable, so Σ|e i+1 –e i | / e avg The value is small and n stable / n is close to 1, resulting in a very small α value. Although the overall trend of sequence B is similar to that of A and the distance d after DTW alignment is smaller, the rising process is accompanied by small fluctuations, making Σ|e i+1 –e i | / e avg As the value increases, n stable / n decreases, resulting in a significant increase in the α value. Sequence C also shows an upward trend, but it is manifested in an obvious fluctuating upward pattern. The energy consumption rate change between adjacent sampling points is larger, and the proportion of stable points is lower, so the largest α value is obtained.

[0064] In this case, although the three sequences may show similar overall trends and similar original DTW distances d after DTW alignment, the optimized distance D1 will clearly distinguish them due to the difference in α values. This distance adjustment mechanism based on fluctuation characteristics enables the clustering algorithm to identify subtle differences in the energy consumption sequence during the change process, thereby more accurately reflecting the characteristics of different driving conditions. This distance optimization scheme based on energy consumption fluctuation characteristics provides a more reasonable distance metric for subsequent clustering analysis. And when the energy consumption characteristics of the two windows are exactly the same, α = 0, and the optimized distance is equal to the original DTW distance; when the difference in the energy consumption characteristics of the two windows is greater, the α value is larger, and the optimized distance increases accordingly. This monotonicity ensures that the optimized distance still meets the basic requirements of the clustering algorithm for distance measurement. At the same time, by considering the energy consumption fluctuation characteristics, a distance measurement method that is more in line with the characteristics of actual driving conditions is provided.

[0065] b. Deep optimization of distance metric based on the energy efficiency characteristics of the battery working status.

[0066] Detailed logic:

[0067] After completing the preliminary optimization based on the energy consumption fluctuation characteristics in step a, it was found that even for vehicles of the same model in similar driving modes, the actual energy consumption performance would be significantly different due to different battery states and environmental conditions, which involved the combined effects of different service years of each vehicle, different battery aging, and different regions. Although step a has been able to identify the similarity of energy consumption sequences through DTW distance and fluctuation feature optimization, it does not consider issues that affect the energy conversion efficiency of electric vehicles. For example, when driving in cold weather, due to the low battery temperature, even if the driving mode is exactly the same, the vehicle's energy consumption level will be significantly higher than that at normal temperature. This is not only reflected in the improvement of the overall energy consumption rate, but also in the reduction of energy recovery efficiency. Similarly, when the battery health (SOH) gradually decreases with the use time, the energy consumption performance under the same working conditions will also change. These factors directly affect the energy utilization efficiency of electric vehicles and are important parameters that cannot be ignored for accurately predicting the cruising range.

[0068] To address this problem, step b designs an optimization scheme based on efficiency influencing factors: by introducing key parameters such as battery temperature and battery health to further adjust the distance calculation. Not only the similarity of the energy consumption sequence is considered, but also the factors affecting the energy efficiency are integrated so that the clustering results can more accurately reflect the energy consumption characteristics under different conditions. For example, when the energy consumption sequences of two windows are similar but the battery temperatures are very different, the optimized distance will increase accordingly, thereby dividing them into different categories. This classification method is more conducive to the system's understanding and prediction of the vehicle's energy consumption performance under different conditions.

[0069] The above optimization design can establish a more systematic and detailed energy consumption mode classification. For the same high-speed cruising condition, the system can subdivide it into two modes, "normal temperature high-speed cruising" and "low temperature high-speed cruising" according to the different battery temperatures; even if the fluctuation characteristics of the energy consumption sequence are similar, they actually represent different energy consumption states due to differences in efficiency factors. This subdivision is crucial to improving the accuracy of mileage prediction.

[0070] The above optimization design also takes into account the dynamics of the energy consumption characteristics of electric vehicles. Battery temperature is not a static parameter, it will change as the vehicle continues to drive. For example, when driving in cold weather, the energy efficiency is relatively poor due to the low battery temperature; but as the driving time increases, the battery temperature gradually rises, and the energy efficiency will improve accordingly. By considering these dynamically changing efficiency factors in cluster analysis, the system can better understand and predict the energy consumption performance of the vehicle at different stages. In addition, the optimization scheme is also highly adaptable. The energy consumption characteristics of electric vehicles will vary significantly in different seasons and different climatic conditions. By incorporating efficiency influencing factors into distance calculations, the system can automatically adapt to these changes and provide a more reliable basis for the prediction of the range under different conditions. In addition, the optimized distance metric not only focuses on "how the vehicle drives" and is reflected through the energy consumption sequence characteristics, but also considers "under what conditions the vehicle drives" and is reflected through efficiency influencing factors. This enables the clustering results to more accurately reflect the actual driving scenario and provide more reliable data support for the prediction of the range.

[0071] Formula logic:

[0072] In view of the above problems, the present invention designs a deep optimization formula based on efficiency influencing factors. The complete distance calculation formula is:

[0073]

[0074] Where D is the final optimized distance value, D1 is the initial optimized distance in step a, F1 and F2 are the efficiency feature vectors of the two windows to be compared, and λ is the adaptive adjustment coefficient. Efficiency feature vector F = [T / T max ,H / 100] contains two components: T is the battery temperature (through T max Normalized), H is the battery health (SOH) value. Adaptive adjustment coefficient Where σ(e) is the standard deviation of the energy consumption rate within the window, k is the proportionality coefficient, the initial value is set to 1.5, and ε is a small positive number to avoid the denominator being zero.

[0075] The following is a detailed explanation of the formula: In this optimization formula, the arctan function maps any real number to the interval (-π / 2,π / 2), divides it by π, and then adds 0.5 to ensure that the final adjustment factor range is in the interval (0,1). This ensures that the optimized distance will not be less than half of the original distance, and also avoids the distance value from being over-magnified. Secondly, the nonlinear characteristics of the arctan function make the adjustment effect relatively mild when the efficiency characteristics are small; when the difference is significant, the adjustment effect increases rapidly. This characteristic is consistent with the characteristics of the actual energy consumption efficiency changes of electric vehicles.

[0076] The battery temperature T reflects the current working state of the battery and directly affects the energy conversion efficiency. max Normalization is performed to make it comparable with the health level in terms of numerical range; the battery health level H (SOH) reflects the aging degree of the battery, affecting its charging and discharging efficiency and capacity performance, and the value itself is in the range of 0-100 and does not require additional normalization. The combination of these two characteristic dimensions constitutes a basic description of the energy efficiency of electric vehicles.

[0077] The adaptive adjustment coefficient λ uses the inverse of the standard deviation σ(e) of the energy consumption rate in the window and adjusts it through the proportional coefficient k as the adjustment coefficient, thus realizing adaptive adjustment for different degrees of energy consumption fluctuation. When the energy consumption rate in the window fluctuates greatly, the standard deviation increases and the λ value decreases, thereby weakening the impact of the efficiency characteristic difference on the distance; when the energy consumption rate is relatively stable, the standard deviation is small and the λ value increases, at which time the difference in efficiency characteristics will have a more significant adjustment effect. It ensures that the efficiency difference will not be excessively amplified in the case of drastic fluctuations in energy consumption, and can more sensitively capture changes in efficiency characteristics when energy consumption is relatively stable.

[0078] The above design shows good adaptability in different scenarios. When comparing two working conditions with similar battery temperatures and similar health, the value of ||F1-F2|| will be very small. After being mapped by the arctan function, the final adjustment factor is close to 0.5, which basically maintains the distance value calculated in step a. When comparing a low temperature working condition and a normal temperature working condition, due to their significant differences in the efficiency feature dimension, the value of ||F1-F2|| is large, and the final adjustment factor will be significantly greater than 0.5, further increasing the distance between them and effectively enhancing the distinction between different efficiency states. It should be noted that the above optimization factors maintain good numerical stability. First, by introducing a small positive number ε in the calculation of λ, the problem of zero division when the standard deviation is 0 is avoided; secondly, the boundedness of the arctan function ensures that the adjustment factor will not have extreme values; finally, the components in the eigenvector F have the same numerical range after normalization, and through the adaptive adjustment of λ, a reasonable balance of feature differences is achieved. These characteristics make the optimization formula computationally stable in practical applications.

[0079] The proportionality coefficient k determines the degree of influence of the difference in efficiency characteristics on the final distance, and its value needs to be adjusted according to the actual application scenario. A larger k value will enhance the distinguishing effect of the efficiency characteristics and is suitable for scenarios that need to be particularly sensitive to efficiency differences; a smaller k value will weaken the influence of the efficiency characteristics and is more suitable for scenarios that focus on the driving mode itself. In practical applications, the optimal k value is determined, including but not limited to, through methods such as cross-validation. Note: The k here is different from the clustering parameter k below.

[0080] In addition, the design of the optimization formula also takes into account the scalability in practical applications. Although the feature vector F currently mainly contains two dimensions: battery temperature and health, the framework can be easily expanded to include other factors that affect energy efficiency. For example, if the influence of ambient temperature needs to be considered, the feature vector can be expanded to F = [T / T max ,H / 100,Te / Te max ], where Te is the ambient temperature, Te max is the corresponding normalized parameter. This scalable design ensures that the optimization scheme can be continuously improved with different needs and in-depth understanding of the energy consumption characteristics of electric vehicles.

[0081] c. k-means clustering analysis based on optimized distance.

[0082] Detailed logic:

[0083] After completing the above distance optimization, the present invention uses the k-means clustering algorithm to perform cluster analysis on the optimized distance metric. The elbow rule is used to select the k value. Specifically, we calculate the total sum of square errors of the clustering results under different k values. As the k value increases, the SSE will decrease accordingly, but at a certain k value, the downward trend of the SSE will slow down significantly, forming an "elbow". The k value corresponding to this "elbow" is the optimal number of clusters. This selection method can not only ensure the effectiveness of clustering, but also avoid over-segmentation.

[0084] The clustering process uses the standard k-means algorithm process:

[0085] Randomly select k windows as initial cluster centers.

[0086] Each window is assigned to the nearest cluster center according to the optimized distance metric.

[0087] Update the center point of each class.

[0088] Repeat steps 2 and 3 until convergence or the maximum number of iterations is reached.

[0089] Since all of the above technologies are existing methods or algorithms, no specific details are described here.

[0090] Summary of major steps:

[0091] At this point, the optimization design of the distance metric is completed.

[0092] In one embodiment, a dynamic prediction of the cruising range is performed based on the clustering results.

[0093] After completing k-means clustering, the present invention dynamically predicts the remaining mileage of the electric vehicle based on the clustering results. The operating condition characteristics of the current driving window are most likely to continue into the recent driving process, so the future energy consumption is predicted based on the clustering category to which the current window belongs. This prediction method based on operating condition similarity can effectively capture the impact of driving behavior and road conditions on energy consumption, thereby providing a more accurate range estimation.

[0094] Specifically, for the current driving window W current , assign it to the closest cluster category C based on the optimized distance metric i In this category, we select p = 10 historical windows closest to the current window as the reference sample set S = {W1, W2, ..., W p The final remaining mileage prediction is calculated using the following formula:

[0095]

[0096] Where: R is the predicted remaining mileage; SOC is the current battery remaining power percentage; d i The reference sample W i With the current window W current The inverse of the optimization distance; s i The reference sample W i The above formula selects the p most similar historical operating conditions in the same cluster for weighted averaging. The weight is designed based on the inverse of the optimization distance, so that the more similar the historical samples are to the current operating conditions, the greater the influence.

[0097] So far, the present invention is completed.

[0098] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0099] The above-described embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the present invention. It should be pointed out that, for a person of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the attached claims.

Claims

1. An intelligent data processing method and system for intelligent connected vehicles, characterized in that: include: S1: Collect historical data of new energy vehicles during driving, obtain data on power change, mileage, battery temperature and battery health, perform fixed window segmentation processing on all historical data, and obtain a data window containing a unit mileage energy consumption rate sequence; S2: For the energy consumption rate sequence, preliminary optimization is performed based on DTW distance calculation and energy consumption fluctuation characteristic analysis. The optimization factor a is obtained by calculating the fluctuation degree and stability, and the distance metric D1 after preliminary optimization is obtained; S3: Construct the efficiency feature vector F according to the battery temperature and health, design an adaptive adjustment mechanism based on the arctan function, and calculate the final optimization distance D based on the preliminary optimization distance D1; S4: Use the k-means clustering algorithm to perform cluster analysis on the optimized distance metric, determine the optimal number of clusters k through the elbow rule, and obtain clustering results that can reflect the characteristics of different driving conditions; S5: Match the current driving window to the closest cluster category based on the optimized distance metric, select the most similar historical samples within the category for weighted average calculation, and obtain an accurate remaining range prediction result; The sampling frequency in S1 is set to once per second, the recorded power change is ΔE, the mileage is Δs, the battery temperature and the battery health, and the unit mileage energy consumption rate of each sampling point is e i =ΔE i / Δs i , the length of the window is T, T is 10 minutes, and the number of windows is 600, and the data points used for clustering are obtained: P = {e_i}i∈[1,600] The e_i is the energy consumption rate per unit mileage of each sampling point in the window, and each of the windows contains an energy consumption rate sequence for subsequent distance metric optimization calculation.

2. The intelligent data processing method and system for intelligent connected vehicles according to claim 1, characterized in that: The DTW distance between the two window energy consumption rate sequences is d, and the optimized distance calculation formula is: D1 = d(1 + α).

3. The intelligent data processing method and system for intelligent connected vehicles according to claim 1, characterized in that: The calculation formula of the adjustment factor α is: α=log(1+Σ|e i+1 -that i | / e avg )exp(-n stable / n Among them, the e i is the energy consumption rate of the i-th sampling point in the window, and the e i +1 is the energy consumption rate of the i+1th sampling point in the window, and the e avg is the average energy consumption rate in the window, n_{stable} is the number of stable sampling points in the window whose energy consumption rate change is less than the threshold δ, n is the total number of sampling points in the window, I() is an indicative function, which takes 1 when the condition is met, otherwise takes 0, and δ is the stability threshold.

4. The intelligent data processing method and system for intelligent connected vehicles according to claim 3, characterized in that: The fluctuation quantification term log(1+Σ|e i+1 –e i | / e avg ) reflects the fluctuation degree of energy consumption by calculating the cumulative amount of energy consumption rate changes between adjacent sampling points in the window, and the energy consumption fluctuation degree is reflected by dividing by e avg The normalization process is performed to eliminate the influence of different energy consumption levels and to take the logarithmic compression effect of large fluctuations. The stability weight term exp(-n stable / n) stable / n represents the proportion of sampling points with stable energy consumption in the window. This ratio is mapped to the interval (0,1] through an exponential function. When the proportion of stable points is high, the exp term approaches 0, thereby weakening the influence of the fluctuation term. When the proportion of stable points is low, the exp term approaches 1.

5. The intelligent data processing method and system for intelligent connected vehicles according to claim 3, characterized in that: When the window energy consumption rate changes, the Σ|e i+1 –e i | value is large, and due to unstable energy consumption, n stable / n is small, the exp term approaches 1, and finally a larger α value is obtained. When the window energy consumption rate is relatively stable, the Σ|e i+1 –e i | is small, and due to the stable energy consumption, n stable As the / n value is larger, the exp term approaches 0, and eventually a smaller α value is obtained.

6. The intelligent data processing method and system for intelligent connected vehicles according to claim 1, characterized in that: The d is the basic metric of DTW, the (1+α) term can dynamically adjust the original distance, and the α value is always non-negative.

7. The intelligent data processing method and system for intelligent connected vehicles according to claim 1, characterized in that: The complete distance calculation formula is: Wherein, D is the final optimized distance value, D1 is the preliminary optimized distance in step a, F1 and F2 are the efficiency feature vectors of the two windows to be compared, λ is the adaptive adjustment coefficient, and the efficiency feature vector F=[T / T max ,H / 100] contains two components: T is the battery temperature, through the T max Normalized, H is the battery health, the adaptive adjustment coefficient Wherein, σ(e) is the standard deviation of the energy consumption rate in the window, k is the proportionality coefficient and its initial value is set to 1.5, and ε is a small positive number used to avoid the denominator being zero.

8. The intelligent data processing method and system for intelligent connected vehicles according to claim 7, characterized in that: The arctan function maps any real number to the interval (-π / 2, π / 2), and the arctan function is divided by π and then added with 0.5 to ensure that the final adjustment factor range is within the interval (0, 1).

9. The intelligent data processing method and system for intelligent connected vehicles according to claim 7, characterized in that: The proportionality coefficient k determines the degree of influence of the efficiency characteristic difference on the final distance, and its value needs to be adjusted according to the actual application scenario.

10. The intelligent data processing method and system for intelligent connected vehicles according to claim 1, characterized in that: The current driving window W current Assign it to the closest cluster category C based on the optimized distance metric i And it selects p = 10 historical windows closest to the current window as the reference sample set S = {W1, W2, ..., W p }, the final remaining mileage prediction is calculated using the following formula: Where: R is the predicted remaining mileage, SOC is the current battery remaining power percentage, d i The reference sample W i With the current window W current The inverse of the optimization distance, s i The reference sample W i The corresponding actual mileage.

Images