Automobile driving condition construction method based on user big data information

Abstract

The invention discloses an automobile running condition construction method based on user big data information. The method comprises the following steps: 1) acquiring a historical automobile running state based on the user running big data information; 2) preprocessing the historical vehicle running state, and constructing a vehicle running condition database; 3) processing the data in the vehicle driving condition database by using principal component analysis and K-means clustering analysis to obtain different clustering results; 4) constructing a to-be-selected operation condition library of each clustering result through random combination of short-stroke fragments; 5) carrying out weight analysis on the characteristic parameters of the automobile running state based on a CRITIC method; 6) performing an optimal decision on each operation condition in the to-be-selected operation condition library based on a TOPSIS method to obtain an optimal operation condition of each clustering result; and 7) splicing the optimal operation conditions of different clustering results to obtain the final driving condition of the vehicle. According to the method, on the basis of ensuring the reasonability of working condition construction, the driving working condition conforming to the specific region is constructed based on the user big data information and the regional characteristics.

Description

Technical Field

[0001] The present invention relates to the technical field of automobile operating conditions, and in particular to a method for constructing automobile driving operating conditions based on user big data information. Background Art

[0002] The vehicle driving condition is a continuous speed-time curve that describes the driving state characteristics of a certain type of vehicle on a certain road in a certain area. It is usually represented by a short-stroke kinematic segment. A short-stroke kinematic segment refers to a time interval that includes four driving states: idling, acceleration / deceleration, and constant speed. The best way to divide the kinematic segment is to start from the beginning of the idle state and end at the moment before the next idle state begins, that is, a short stroke, such as Figure 1 shown.

[0003] The operating states are divided into the following four categories based on the differences in acceleration and speed:

[0004] Acceleration state: vehicle acceleration a>0.15m / s 2 Continuous driving process.

[0005] Deceleration state: vehicle acceleration a<-0.15m / s 2 Continuous driving process.

[0006] Idle state: vehicle speed v<0.5km / h, and acceleration |a|≤0.15m / s 2 Continuous driving process.

[0007] Uniform speed state: the car's speed v> 0.5km / h, and the acceleration |a| ≤ 0.15m / s 2 The non-idling continuous driving process.

[0008] Existing methods for constructing automobile driving conditions include the VA matrix-based construction method and the "speed-acceleration" dual-parameter clustering Markov chain construction method. The VA matrix construction method is usually used for objects with specific driving areas, routes, and driving requirements, such as urban buses or passenger vehicles on specific routes. When user big data does not have typical driving characteristics, it is difficult to accurately obtain its typical VA distribution characteristics, which leads to insufficient accuracy of the solution results. When constructing the working condition based on the Markov chain construction method, since the operating state at the next moment is randomly generated based on the state transition probability matrix using Monte Carlo simulation, it may cause the vehicle's driving speed to fluctuate repeatedly within a short period of time, that is, the speed-time curve is obviously jagged. This is also inconsistent with the user's actual driving conditions, and to a certain extent increases the operational difficulty of the working condition. Summary of the Invention

[0009] The purpose of the present invention is to provide a method for constructing a vehicle driving condition based on user big data information, comprising the following steps:

[0010] 1) Based on user driving big data information, obtain historical vehicle operation status.

[0011] 2) Preprocess the historical vehicle operating status to generate short trip segments and then build a vehicle driving condition database.

[0012] 3) Use principal component analysis and K-means cluster analysis to process the data in the vehicle driving condition database and obtain different clustering results.

[0013] 4) For each clustering result, calculate the characteristic parameter error value between each short-trip segment and the overall clustering result, sort each short-trip segment in ascending order of error value, select several short-trip segments with the highest order, and randomly combine the selected short-trip segments to form an operating condition, thereby constructing a library of candidate operating conditions for each clustering result.

[0014] 5) Based on the CRITIC method, the characteristic parameters of the vehicle operating status are weighted and the weights of each characteristic parameter are obtained.

[0015] 6) Combining the weights of each characteristic parameter, the TOPSIS method is used to make the optimal decision for each operating condition in the candidate operating condition library, and the optimal operating condition for each clustering result is obtained.

[0016] 7) The optimal operating conditions of each clustering result are spliced ​​together to obtain the final driving condition of the vehicle.

[0017] Furthermore, the vehicle operating state includes acceleration and speed.

[0018] Furthermore, the preprocessing includes data elimination processing, data interpolation processing, and data filtering and denoising processing.

[0019] Furthermore, in step 3), the steps for obtaining different clustering results are as follows:

[0020] 3.1) Construct the characteristic parameter matrix A m×n , as shown below:

[0021]

[0022] Where m represents the total number of short-distance segments. n represents the total number of characteristic parameters. pq Represents the qth characteristic parameter of the pth short-run segment, 1≤p≤m, 1≤q≤n.

[0023] 3.2) For the characteristic parameter matrix A m×n Perform standardization to obtain the standardized matrix Zm×n , as shown below:

[0024]

[0025]

[0026] Where z pq It represents the qth characteristic parameter of the pth short-run segment after normalization. Represents the mean of the qth feature parameter. s q represents the variance of the qth feature parameter.

[0027] 3.3) According to the normalized matrix Z m×n The covariance matrix C is calculated as follows:

[0028]

[0029]

[0030] Where u and y are the number of columns, 1≤u≤n, 1≤y≤n, and Cov(u,y) represents the covariance between the u-th column and the y-th column in the standardized matrix Z. p 、y p They represent the p-th value in the u-th column and the y-th column in the normalized matrix Z respectively. They represent the means of the u-th column and the y-th column in the standardized matrix Z respectively.

[0031] 3.4) Calculate the eigenvalues ​​and unit eigenvectors of the covariance matrix C as follows:

[0032] Cv q =λ q v q (6)

[0033] Where λ q is the qth eigenvalue, λ q ≥0. v q is the qth eigenvalue λ q The corresponding unit eigenvector. 1≤q≤n.

[0034] 3.5) For the eigenvalue λ q Sort in descending order and calculate the cumulative contribution rate CR of the first r0 eigenvalues ​​in the sort, as shown below:

[0035]

[0036] Where q0 and q1 both represent the eigenvalue numbers. Indicates the eigenvalue ranked at the q0th position. Indicates the eigenvalue ranked at the q1th position. r0 is the number of eigenvalues ​​when the cumulative contribution rate CR is greater than or equal to 80%.

[0037] 3.6) According to the unit eigenvector v corresponding to the first r0 eigenvalues ​​of the sort q , construct the projection matrix As shown below:

[0038]

[0039] 3.7) Project the normalized data into the principal component space to obtain the dimensionally reduced data B, as shown below:

[0040]

[0041] Where B is the data after dimensionality reduction, and each row in B corresponds to the r0-dimensional principal component score of a short trip segment.

[0042] 3.8) The first r0 principal components are selected as the objects of cluster analysis, and the r0-dimensional principal component scores are used as the clustering basis. K-means clustering method is used for clustering to obtain different clustering results.

[0043] The sum of squared errors of the K-means clustering method is as follows:

[0044]

[0045] Where SSE is the sum of squared errors. i1 is the cluster number. k is the number of clusters. j1 is the sample number. m1 is the sample dimension. is the j1th sample of the i1th class. is the mean of class i1.

[0046] Furthermore, the duration of the operating conditions in the candidate operating condition library is as follows:

[0047]

[0048] Where i represents the clustering result index, and o represents the operating condition index. io represents the duration of the oth operating condition in the candidate operating condition library of the i-th clustering result. δ is the error limit.

[0049] Among them, the time length Tl of the i-th clustering result in the final driving condition of the vehicle is i As shown below:

[0050]

[0051] Where j represents the short-run segment index, n jRepresents the total number of short trip segments. T0 is the duration of the vehicle’s final driving condition. ij represents the duration of the jth short trip segment in the i-th clustering result. k0 represents the total number of clustering results.

[0052] Furthermore, in step 5), the steps for obtaining the weights of the characteristic parameters are as follows:

[0053] 5.1) Define w groups of operating conditions in the candidate operating condition library as decision objects, y0 characteristic parameters as decision indicators, make a single-objective decision on the decision object with the minimum error, and obtain the corresponding decision result and use the decision results Calculate the values ​​of another y0-1 decision indicators Then construct the evaluation matrix X, where i2=1,2,…,w, j2=1,2,…,y0.

[0054] 5.2) The evaluation matrix X is normalized and the normalized matrix X′ is calculated as follows:

[0055]

[0056] Where, It is the value of the i2th decision object under the j2th decision indicator after normalization.

[0057] The benefit indicator is a positive decision indicator, and the cost indicator is a negative decision indicator.

[0058] 5.3) Calculate the contrast strength between decision indicators as follows:

[0059]

[0060] Where, is the standard deviation of the j2th decision indicator. is the mean value of the j2th decision indicator after standardization.

[0061] 5.4) Calculate the correlation coefficient between the k1th decision indicator and the j2th decision indicator and conflict quantitative index values As shown below:

[0062]

[0063]

[0064] Where h is the decision object index, are the values ​​of the hth decision object under the k1th and j2th decision indicators after normalization. are the means of the k1th and j2th decision indicators after standardization, and k1≠j2.

[0065] 5.5) Calculate the information content of the decision indicator as follows:

[0066]

[0067] Where, Represents the information content of the j2th decision indicator.

[0068] 5.6) Calculate the weight of the decision indicator as follows:

[0069]

[0070] Where, represents the weight of the j2th decision indicator. is the information amount of the k1th decision indicator.

[0071] Furthermore, in step 6), the steps for obtaining the optimal operating conditions for each clustering result are as follows:

[0072] 6.1) Based on the characteristic parameter values ​​corresponding to each operating condition, a comprehensive evaluation decision matrix R1 is constructed as follows:

[0073]

[0074] Where, is the j2th characteristic parameter value corresponding to the i2th operating condition, i2=1,2,…,w, w is the total number of operating conditions, j2=1,2,…,y0, y0 is the total number of characteristic parameters.

[0075] 6.2) Based on the normalization method, the comprehensive evaluation decision matrix R1 is normalized to obtain the regularized matrix R1′, as shown below:

[0076]

[0077] Where, is the j2th characteristic parameter value corresponding to the i2th operating condition after normalization.

[0078] 6.3) Combine the weights of each feature parameter and weight the regularization matrix R1′ to obtain the decision matrix V, as shown below:

[0079]

[0080] Where, is the j2th characteristic parameter value corresponding to the i2th operating condition after weighting. Represents the weight of the j2th feature parameter.

[0081] 6.4) Construct the optimal solution set A + and the worst solution set A - , as shown below:

[0082]

[0083] Where, is the optimal value under the j2th characteristic parameter. is the worst value under the j2th characteristic parameter.

[0084] 6.5) Calculate the sequence corresponding to the i2th operating condition in the decision matrix V to the optimal solution set A + and the worst solution set A - Euclidean distance and As shown below:

[0085]

[0086] 6.6) Calculate TOPSIS relative fit As shown below:

[0087]

[0088] 6.7) Relative fit from TOPSIS The maximum value is selected, and the corresponding operating condition is the optimal operating condition for each clustering result.

[0089] Furthermore, when the characteristic parameter is a benefit indicator, the optimal value Take the maximum value and the worst value Take the minimum value.

[0090] When the characteristic parameter is a cost indicator, the optimal value Take the minimum value, the worst value Take the maximum value.

[0091] Furthermore, the characteristic parameters include average speed, average driving speed, average acceleration, average deceleration, uniform speed time ratio, idle time ratio, acceleration time ratio, deceleration time ratio, speed standard deviation, and acceleration standard deviation.

[0092] Furthermore, the final vehicle driving condition is used to test and evaluate the vehicle's energy consumption, economy, power, and reliability and durability of the vehicle system.

[0093] The technical effect of the present invention is unquestionable. The present invention proposes a short-trip random combination comprehensive decision-making method based on CRITIC-TOPSIS to construct a typical vehicle driving condition based on user driving big data information.

[0094] Based on real user big data, the present invention uses principal component analysis and clustering algorithms to statistically analyze driving data and construct a library of candidate driving conditions. Furthermore, the method uses the Criteria Importance Through Intercriteria Correlation (CRITIC) method based on the objective attributes of the data to objectively weight the defined characteristic parameters. The Technique for Order Preference by Similarity to an Ideal Solution (TOPSIS) is then used to fusion the weighted analysis results of the proposed random driving condition combinations to make the optimal decision, thereby obtaining the optimal vehicle driving condition curve that conforms to the statistical parameter characteristics of the real user big data.

[0095] The comparison results between the working conditions constructed by the present invention and the overall database show that the constructed working conditions can accurately reflect the driving characteristics of passenger cars in Lianshi, Chongqing, and can be used as a representative working condition for subsequent analysis. Comparison with working conditions in other regions and the China Light-duty Vehicle Test Cycle-passenger (CLTC-P) and New European Driving Cycle (NEDC) standard working conditions show that there are obvious regional differences in the driving characteristics of different cities in China, and the NEDC working condition is difficult to accurately reflect China's urban traffic conditions. The constructed working condition is relatively close to China's CLTC-P working condition but not completely identical, which shows that the proposed method, on the basis of ensuring the rationality of the constructed working condition, constructs a driving condition that conforms to a specific region in China based on user big data information and regional characteristics.

[0096] The operating conditions established by this invention lay the foundation for evaluating vehicle energy consumption, economy, and power, as well as testing and evaluating the reliability and durability of automotive systems. Subsequently, the constructed driving conditions will be used to conduct user road vehicle durability load spectrum testing and collection experiments, further verifying the effectiveness and matching accuracy of the constructed operating conditions from the perspective of structural damage, etc.

[0097] Compared to driving profiles constructed using existing methods and the characteristic parameters of raw big data, the vehicle driving profile constructed by this invention demonstrates superior reliability and superiority. The construction of vehicle driving profile curves for specific regions and objects has important practical applications. Establishing such driving profiles can lay the foundation for evaluating vehicle energy consumption, economy, and power, as well as testing and evaluating the reliability and durability of vehicle systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0098] Figure 1 It is a schematic diagram of a short-stroke working condition segment;

[0099] Figure 2 Flow chart of the method of the present invention;

[0100] Figure 3 Schematic diagram of the characteristic value curve of each principal component;

[0101] Figure 4 Schematic diagram of the sum of squared elbow errors (SSE);

[0102] Figure 5 Schematic diagram of clustering results;

[0103] Figure 6 Schematic diagram of comprehensive contribution coefficient of non-inferior solution; Figure 6 (a) is a schematic diagram of the comprehensive contribution coefficient of clustering condition I; Figure 6 (b) is a schematic diagram of the comprehensive contribution coefficient of clustering condition II; Figure 6 (c) is a schematic diagram of the comprehensive contribution coefficient of clustering condition III;

[0104] Figure 7 A schematic diagram of a vehicle driving condition curve constructed for the present invention;

[0105] Figure 8 Schematic diagram of vehicle operating condition curve constructed for other methods; Figure 8 (a) Schematic diagram of the results of cluster Markov chain construction method; Figure 8 (b) Schematic diagram of the results of the VA matrix construction method. DETAILED DESCRIPTION

[0106] The present invention will be further described below with reference to the following examples, but it should not be understood that the scope of the present invention is limited to the following examples. Without departing from the above technical ideas of the present invention, various substitutions and modifications can be made according to common technical knowledge and customary means in the art, and all should be included in the scope of protection of the present invention.

[0107] Example 1:

[0108] See also Figures 2 to 8 , a method for constructing a vehicle driving condition based on user big data information, comprising the following steps:

[0109] 1) Based on user driving big data information, obtain historical vehicle operation status.

[0110] 2) Preprocess the historical vehicle operating status to generate short trip segments and then build a vehicle driving condition database.

[0111] 3) Use principal component analysis and K-means cluster analysis to process the data in the vehicle driving condition database and obtain different clustering results.

[0112] 4) For each clustering result, calculate the characteristic parameter error value between each short-trip segment and the overall clustering result, sort each short-trip segment in ascending order of error value, select several short-trip segments with the highest order, and randomly combine the selected short-trip segments to form an operating condition, thereby constructing a library of candidate operating conditions for each clustering result.

[0113] 5) Based on the CRITIC method, the characteristic parameters of the vehicle operating status are weighted and the weights of each characteristic parameter are obtained.

[0114] 6) Combining the weights of each characteristic parameter, the TOPSIS method is used to make the optimal decision for each operating condition in the candidate operating condition library, and the optimal operating condition for each clustering result is obtained.

[0115] 7) The optimal operating conditions of each clustering result are spliced ​​together to obtain the final driving condition of the vehicle.

[0116] Example 2:

[0117] A method for constructing automobile driving conditions based on user big data information, the main technical content of which is shown in Example 1. Furthermore, the automobile operating status includes acceleration and speed.

[0118] Example 3:

[0119] The method for constructing automobile driving conditions based on user big data information, the main technical content of which can be found in any one of Examples 1 to 2. Furthermore, the preprocessing includes data elimination processing, data interpolation processing, and data filtering and denoising processing.

[0120] Example 4:

[0121] The method for constructing automobile driving conditions based on user big data information has the following main technical contents as shown in any one of Examples 1 to 3. Furthermore, in step 3), the steps for obtaining different clustering results are as follows:

[0122] 3.1) Construct the characteristic parameter matrix A m×n , as shown below:

[0123]

[0124] Where m represents the total number of short-distance segments. n represents the total number of characteristic parameters. pq Represents the qth characteristic parameter of the pth short-run segment, 1≤p≤m, 1≤q≤n.

[0125] 3.2) For the characteristic parameter matrix A m×n Perform standardization to obtain the standardized matrix Z m×n , as shown below:

[0126]

[0127]

[0128] Where z pq It represents the qth characteristic parameter of the pth short-run segment after normalization. Represents the mean of the qth feature parameter. s q represents the variance of the qth feature parameter.

[0129] 3.3) According to the normalized matrix Z m×n The covariance matrix C is calculated as follows:

[0130]

[0131]

[0132] Where u and y are the number of columns, 1≤u≤n, 1≤y≤n, and Cov(u,y) represents the covariance between the u-th column and the y-th column in the standardized matrix Z. p 、y D They represent the p-th value in the u-th column and the y-th column in the normalized matrix Z respectively. They represent the means of the u-th column and the y-th column in the standardized matrix Z respectively.

[0133] 3.4) Calculate the eigenvalues ​​and unit eigenvectors of the covariance matrix C as follows:

[0134] Cv q =λ q v q (6)

[0135] Where λ q is the qth eigenvalue, λ q ≥0. v q is the qth eigenvalue λ q The corresponding unit eigenvector. 1≤q≤n.

[0136] 3.5) For the eigenvalue λ q Sort in descending order and calculate the cumulative contribution rate CR of the first r0 eigenvalues ​​in the sort, as shown below:

[0137]

[0138] Where q0 and q1 both represent the eigenvalue numbers. Indicates the eigenvalue ranked at the q0th position. Indicates the eigenvalue ranked at the q1th position. r0 is the number of eigenvalues ​​when the cumulative contribution rate CR is greater than or equal to 80%.

[0139] 3.6) According to the unit eigenvector v corresponding to the first r0 eigenvalues ​​of the sort q , construct the projection matrix As shown below:

[0140]

[0141] 3.7) Project the normalized data into the principal component space to obtain the dimensionally reduced data B, as shown below:

[0142]

[0143] Where B is the data after dimensionality reduction, and each row in B corresponds to the r0-dimensional principal component score of a short trip segment.

[0144] 3.8) The first r0 principal components are selected as the objects of cluster analysis, and the r0-dimensional principal component scores are used as the clustering basis. K-means clustering method is used for clustering to obtain different clustering results.

[0145] The sum of squared errors of the K-means clustering method is as follows:

[0146]

[0147] Where SSE is the sum of squared errors. i1 is the cluster number. k is the number of clusters. j1 is the sample number. m1 is the sample dimension. is the j1th sample of the i1th class. is the mean of class i1.

[0148] Example 5:

[0149] The method for constructing automobile driving conditions based on user big data information has the following main technical contents as shown in any one of Examples 1 to 4. Furthermore, the duration of the operating conditions in the candidate operating condition library is as follows:

[0150]

[0151] Where i represents the clustering result index, and o represents the operating condition index. io represents the duration of the oth operating condition in the candidate operating condition library of the i-th clustering result. δ is the error limit.

[0152] Among them, the time length Tl of the i-th clustering result in the final driving condition of the vehicle is i As shown below:

[0153]

[0154] Where j represents the short-run segment index, n jRepresents the total number of short trip segments. T0 is the duration of the vehicle’s final driving condition. ij represents the duration of the jth short trip segment in the i-th clustering result. k0 represents the total number of clustering results.

[0155] Example 6:

[0156] The method for constructing a vehicle driving condition based on user big data information has the following main technical contents as shown in any one of Examples 1 to 5. Furthermore, in step 5), the steps for obtaining the weights of the characteristic parameters are as follows:

[0157] 5.1) Define w groups of operating conditions in the candidate operating condition library as decision objects, y0 characteristic parameters as decision indicators, make a single-objective decision on the decision object with the minimum error, and obtain the corresponding decision result and use the decision results Calculate the values ​​of another y0-1 decision indicators Then construct the evaluation matrix X, where i2=1,2,…,w, j2=1,2,…,y0.

[0158] 5.2) The evaluation matrix X is normalized and the normalized matrix X′ is calculated as follows:

[0159]

[0160] Where, It is the value of the i2th decision object under the j2th decision indicator after normalization.

[0161] The benefit indicator is a positive decision indicator, and the cost indicator is a negative decision indicator.

[0162] 5.3) Calculate the contrast strength between decision indicators as follows:

[0163]

[0164] Where, is the standard deviation of the j2th decision indicator. is the mean value of the j2th decision indicator after standardization.

[0165] 5.4) Calculate the correlation coefficient between the k1th decision indicator and the j2th decision indicator and conflict quantitative index values As shown below:

[0166]

[0167] Where h is the decision object index, are the values ​​of the hth decision object under the k1th and j2th decision indicators after normalization. are the means of the k1th and j2th decision indicators after standardization, and k1≠j2.

[0168] 5.5) Calculate the information content of the decision indicator as follows:

[0169]

[0170] Where, Represents the information content of the j2th decision indicator.

[0171] 5.6) Calculate the weight of the decision indicator as follows:

[0172]

[0173] Where, represents the weight of the j2th decision indicator. is the information amount of the k1th decision indicator.

[0174] Example 7:

[0175] The method for constructing automobile driving conditions based on user big data information has the following main technical contents as shown in any one of Examples 1 to 6. Furthermore, in step 6), the steps for obtaining the optimal operating condition for each clustering result are as follows:

[0176] 6.1) Based on the characteristic parameter values ​​corresponding to each operating condition, a comprehensive evaluation decision matrix R1 is constructed as follows:

[0177]

[0178] Where, is the j2th characteristic parameter value corresponding to the i2th operating condition, i2=1,2,…,w, w is the total number of operating conditions, j2=1,2,…,y0, y0 is the total number of characteristic parameters.

[0179] 6.2) Based on the normalization method, the comprehensive evaluation decision matrix R1 is normalized to obtain the regularized matrix R1′, as shown below:

[0180]

[0181] Where, is the j2th characteristic parameter value corresponding to the i2th operating condition after normalization.

[0182] 6.3) Combine the weights of each feature parameter and weight the regularization matrix R1′ to obtain the decision matrix V, as shown below:

[0183]

[0184] Where, is the j2th characteristic parameter value corresponding to the i2th operating condition after weighting. Represents the weight of the j2th feature parameter.

[0185] 6.4) Construct the optimal solution set A + and the worst solution set A - , as shown below:

[0186]

[0187] Where, is the optimal value under the j2th characteristic parameter. is the worst value under the j2th characteristic parameter.

[0188] 6.5) Calculate the sequence corresponding to the i2th operating condition in the decision matrix V to the optimal solution set A + and the worst solution set A - Euclidean distance and As shown below:

[0189]

[0190] 6.6) Calculate TOPSIS relative fit As shown below:

[0191]

[0192] 6.7) Relative fit from TOPSIS The maximum value is selected, and the corresponding operating condition is the optimal operating condition for each clustering result.

[0193] Example 8:

[0194] The method for constructing automobile driving conditions based on user big data information, the main technical content of which is shown in any one of embodiments 1 to 7, further, when the characteristic parameter is a benefit indicator, the optimal value Take the maximum value and the worst value Take the minimum value.

[0195] When the characteristic parameter is a cost indicator, the optimal value Take the minimum value, the worst value Take the maximum value.

[0196] Example 9:

[0197] A method for constructing automobile driving conditions based on user big data information, the main technical content of which is shown in any one of Examples 1 to 8. Furthermore, the characteristic parameters include average speed, average driving speed, average acceleration, average deceleration, uniform speed time ratio, idle time ratio, acceleration time ratio, deceleration time ratio, speed standard deviation, and acceleration standard deviation.

[0198] Example 10:

[0199] A method for constructing automobile driving conditions based on user big data information, the main technical content of which can be found in any one of Examples 1 to 9. Furthermore, the final driving conditions of the vehicle are used to test and evaluate the vehicle's energy consumption, economy, power, and reliability and durability of the vehicle system.

[0200] Example 11:

[0201] See also Figures 2 to 8 , a method for constructing automobile driving conditions based on user big data information, the main technical contents include:

[0202] 1. User driving big data processing

[0203] 1.1 Data collection and preprocessing

[0204] For user-operated vehicles, GPS devices were installed to collect speed and time signals from normal operating vehicles. The sampling frequency was set to 0.5 Hz. The test collected continuous status information from 18 vehicles for approximately one month. After collation and statistical processing, approximately 1.57 million data sets were generated. The large data database formed by the raw vehicle operation data collected by the on-board data collector cannot be directly used to establish driving conditions. During the data collection process, signal discontinuities due to GPS signal loss, long periods of idling, data anomalies caused by drastic acceleration / deceleration, or special traffic interventions are common. Therefore, further preprocessing of the raw data is required, including the removal of abnormal or invalid data, interpolation of discontinuous data segments, and data filtering and denoising. This generates a valid speed-time series database sufficient for constructing vehicle driving conditions. Furthermore, the sampling frequency of the data acquisition equipment was 0.5 Hz, which does not meet the basic requirements for establishing driving conditions. GB / T 38146.1-2019 stipulates that the sampling frequency for vehicle driving conditions in China is 1 Hz. This means that data at a rate of at least 1 Hz should be used to depict the transient characteristics of vehicle operation to ensure accurate representation of driving conditions. Therefore, the database is resampled using linear interpolation to make the data more representative and enhance the credibility of subsequent driving condition reconstruction results.

[0205] 1.2 Definition of short-stroke kinematic segments and their characteristic parameters

[0206] A kinematic segment is a data interval that includes four driving states: constant speed, acceleration, deceleration, and idling. The best way to divide a kinematic segment is to start from the beginning of the idling state and end at the moment before the next idling state begins, that is, a short trip, such as Figure 1 As shown in Figure 2, a short-stroke kinematic segment usually contains multiple driving states due to the limitations of road traffic conditions and the interference of other vehicles, and exhibits different operating state characteristics at different times.

[0207] Based on the pre-processed operating big data information, the driving conditions are segmented and extracted. The operating conditions are divided into the following four categories based on the differences in acceleration and speed:

[0208] (1) Acceleration state: vehicle acceleration a>0.15m / s 2 Continuous driving process.

[0209] (2) Deceleration state: vehicle acceleration a<-0.15m / s 2 Continuous driving process.

[0210] (3) Idle state: vehicle speed v < 0.5 km / h, and acceleration |a| ≤ 0.15 m / s 2 Continuous driving process.

[0211] (4) Uniform speed: The vehicle's speed v> 0.5 km / h, and the acceleration |a| ≤ 0.15 m / s 2 The non-idling continuous driving process.

[0212] Based on Matlab software programming, the defined short-stroke kinematic segments were extracted. According to the above principles, a total of approximately 18,127 short-stroke kinematic segments were extracted. For the vehicle's operating status, time and speed parameters alone cannot fully describe its complex driving conditions. It is usually necessary to add characteristic parameters of the short-stroke kinematic segments to further describe its operating characteristics. Common characteristic parameters of short-stroke kinematic segments include average speed, average driving speed, average acceleration / deceleration, average deceleration, idle / acceleration / deceleration time ratio, and speed / acceleration standard deviation. Furthermore, to accurately describe the status and characteristics of each kinematic segment, this paper additionally considers the key characteristic parameter of speed range proportion, dividing the speed range into five main speed ranges with 20 km / h as the boundary, to better reflect the driving information of each short-stroke kinematic segment. A total of 14 characteristic parameters are defined, as shown in Table 1, which constitute the indicators of the vehicle driving condition characteristic evaluation system.

[0213] Table 1 Definition of characteristic parameters of short-stroke kinematic segments

[0214]

[0215] 2. Automobile operating condition construction method based on CRITIC-TOPSIS comprehensive decision-making

[0216] 2.1 Construction of operating condition library based on short-stroke kinematic segments

[0217] 2.1.1 Principal Component Analysis

[0218] The dimension of the feature parameter matrix is ​​high, and directly analyzing it has high computational complexity and computational cost requirements. In addition, there is a certain correlation between some feature parameters, so feature dimensionality reduction can be performed. Principal Component Analysis (PCA) is a commonly used data dimensionality reduction analysis method, which is used in image processing, mathematical analysis, mathematical modeling and other fields. The PCA method can transform multiple original feature parameters with certain correlations into new independent feature parameter comprehensive indicators (i.e., principal components). The PCA method retains the most important features of high-dimensional data to streamline the data in the database, thereby reducing the time for clustering analysis operations on the feature parameter matrix. The specific steps of the PCA method are as follows:

[0219] (1) Center standardization

[0220] For m short-stroke operating conditions, the characteristic parameter matrix A is constructed corresponding to n characteristic parameters. m×n :

[0221]

[0222] For matrix A m×n The elements in are normalized:

[0223]

[0224] This gives the normalized matrix Z m×n , expressed as:

[0225]

[0226] (2) Calculate the covariance matrix

[0227] The covariance matrix C is calculated from the standardized matrix:

[0228]

[0229] C is a symmetric square matrix, where:

[0230]

[0231] Where Cov(x,y) is the covariance between the xth row and yth column of the covariance matrix C.

[0232] (3) Calculate the eigenvalues ​​and unit eigenvectors of the covariance matrix C as follows:

[0233] Cv q =λ q v q (6)

[0234] Where λ q ≥0 is the qth eigenvalue; v q is the corresponding unit eigenvector; 1≤q≤n;

[0235] (4) Sort the eigenvalues ​​in descending order and calculate the cumulative contribution rate CR of the first r0 principal components, as shown below:

[0236]

[0237] Where q0 represents the eigenvalue number; represents the eigenvalue ranked at the q0th position; λ q represents the qth eigenvalue; r0 is the number of eigenvalues ​​when the cumulative contribution rate CR is greater than or equal to 80%;

[0238] (5) The eigenvectors (principal component directions) corresponding to the first r0 eigenvalues ​​constitute the projection matrix (m×r0 dimensions), as shown below:

[0239]

[0240] (6) Project the data into the principal component space to obtain the dimension-reduced data B, as shown below:

[0241]

[0242] Where, each row of B corresponds to the r0-dimensional principal component score of a sample.

[0243] 2.1.2 Cluster analysis

[0244] The first r principal components selected are used as the basic data for subsequent cluster analysis, and their score matrix is ​​calculated as the clustering object. Among many algorithms, K-means is the most commonly used unsupervised clustering algorithm. K-means clustering selects the number of clusters k and the initial cluster center, calculates the minimum distance between each sample data and the cluster center, and assigns the sample data to the nearest cluster center according to the distance. It continuously iterates and selects new cluster centers and adjusts each data category. In K-means clustering, as the number of clusters k increases, the sample division will be more refined, the degree of aggregation of each cluster will gradually increase, and the sum of squared errors (SSE) will gradually decrease. The first r0 principal components are the objects of cluster analysis, and the r0-dimensional principal component scores are used as the clustering basis. K-means clustering is used for clustering to obtain different clustering results;

[0245] The sum of squared errors of the K-means clustering method is as follows:

[0246]

[0247] Where SSE is the sum of squared errors; i1 is the cluster number; k is the number of clusters; j1 is the sample number; m1 is the sample dimension; is the j1th sample of the i1th class; is the mean of cluster i1. When k is less than the true number of clusters, the SSE decreases significantly because increasing k increases the degree of aggregation within each cluster. When k approaches the true number of clusters, the decrease in SSE decreases sharply with further increases in k, and then levels off as k continues to increase. The elbow method is often used to select the k value.

[0248] 2.1.3 Driving Condition Library Construction Method

[0249] According to the definition of the K-means clustering algorithm, the smaller the distance between a sample in each category and its cluster center, the more the sample can reflect the characteristics of this category. Based on the application requirements, a target driving condition of an electric vehicle with a length of approximately 1200 seconds is constructed. Based on the proportion of the total driving time of category i (i = 1, 2…k) in the total driving time, the time required for the simulated condition of category i can be calculated, as shown in formula (8):

[0250]

[0251] Where: Tl i is the length of time that the i-th operating condition is in the final synthesized vehicle driving condition, t ij is the duration of the jth short trip segment in the i-th operating condition.

[0252] The clustered short-stroke segments are sorted by clustering conditions by calculating the distance between the short-stroke segments and the cluster center. For each clustering condition, the top 100 candidate short-stroke kinematic segments closest to the cluster center are selected. These 100 candidate segments are randomly combined and spliced ​​into operating conditions. The operating conditions are then combined with the length screening criterion of formula (9) to retain all operating conditions that meet the time length requirements to form a library of candidate operating conditions.

[0253]

[0254] Where: L io is the duration of the oth operating condition in the i-th candidate operating condition library. In engineering applications, δ is generally set to 5%.

[0255] 2.2 Characteristic parameter weight analysis based on CRITIC method

[0256] In the traditional direct construction method, the operating condition with the smallest total error of characteristic parameters in the candidate operating condition library is directly taken as the final result. Although this method is simple, it ignores the relationship between different characteristic parameters and the importance of the error of each characteristic parameter. Therefore, in order to objectively quantify the importance of each characteristic parameter, the CRITIC method is applied to perform weight analysis of characteristic parameters based on the candidate operating condition library. As an objective weight analysis method without subjective intervention, the CRITIC method objectively calculates the weight coefficient by measuring the contrast intensity and conflict index of each target. Its contrast intensity draws on the idea of ​​the standard deviation method. If the difference between all evaluation indexes of the same target is greater (that is, the larger the standard deviation), the greater the amount of information it contains and the more important the target is. The conflict between targets is measured by the correlation coefficient. If the positive correlation between the two targets is strong, the conflict is low. Based on the characteristic parameter relative error matrix (that is, decision object) of the w groups of candidate operating conditions in the candidate operating condition library corresponding to the user big data, according to the defined y characteristic parameters (that is, decision indicators), single-target decision-making (minimum error) is carried out in turn, and the corresponding decision result w is used. i Calculate the values ​​of the other y-1 characteristic parameters and record them as x ij (i=1, ..., wj=1, ..., y-1) All recorded values ​​constitute the evaluation matrix X used for the CRITIC weight analysis method.

[0257] (1) Data standardization. The larger the value of the positive decision-making indicator (benefit indicator), the better; the smaller the value of the negative decision-making indicator (cost indicator), the better. That is:

[0258]

[0259] (2) Calculate the contrast strength between indicators:

[0260]

[0261] Where, σ j is the standard deviation of the jth term.

[0262] (3) Calculate the correlation coefficient r ij and conflict quantitative index value l j :

[0263]

[0264]

[0265] Where, l j It is the conflict measure of other indicators of the jth indicator.

[0266] (4) Calculate the information content of the indicators. The objective weight of each indicator is comprehensively measured by the comparison intensity and conflict. The information content C contained in the jth evaluation indicator j Calculated as:

[0267] C j =σ j ×l j ,i≠jj=1,2,…,y(17)

[0268] (5) Calculate the weight of the indicator ω j :

[0269]

[0270] Where C j The larger the value, the greater the amount of information contained in the jth evaluation index, and the greater the relative importance of the index, that is, the weight ω j The bigger.

[0271] 2.3 Determination of optimal operating conditions based on TOPSIS comprehensive decision-making

[0272] The TOPSIS method (ranking by closest ideal solutions) is a multi-objective evaluation method proposed by Hwang et al. in 1981. Its core idea is to define a set of positive and negative ideal solutions and use Euclidean distance to measure the degree of fit or distance between an evaluation object and the positive and negative ideal solutions, thereby ranking them. The calculation steps are as follows:

[0273] (1) The comprehensive evaluation decision matrix R is constructed based on the 14 characteristic parameters corresponding to each candidate working condition:

[0274]

[0275] Where x wy It represents the value of the wth candidate working condition under the yth evaluation index.

[0276] (2) The comprehensive evaluation decision matrix R is normalized based on the normalization method to obtain the regularized matrix R′:

[0277]

[0278] (3) Combine the weight coefficient ω of each characteristic parameter j , weight the matrix R′ to obtain the new decision matrix V:

[0279] V=(v ij ) w×y =(ω j ×m ij ) w×y (twenty one)

[0280] (4) Optimal solution set A + and the worst solution set A - are the minimum and maximum deviations corresponding to each test criterion, respectively.

[0281]

[0282] (5) Calculate the Euclidean distance from the sequence corresponding to the i-th variable in the decision matrix to the positive and negative ideal solution sets and :

[0283]

[0284] Where, The smaller the value, the closer the sequence corresponding to the i-th variable is to the positive ideal solution set; The smaller the value, the closer the sequence corresponding to the i-th variable is to the negative ideal solution set. Otherwise, it is farther away.

[0285] (6) Define the TOPSIS relative fit T between the sequence corresponding to the i-th variable and the ideal solution set i The larger the value is, the greater the comprehensive contribution of the evaluation object to each evaluation index is.

[0286]

[0287] Based on the above theory, a comprehensive decision-making method for short-trip random combination working conditions based on CRITIC-TOPSIS is proposed. First, for the candidate operating condition library constructed under each clustering result, the defined characteristic parameters are objectively weighted analyzed by the CRITIC method; secondly, the TOPSIS method is used to make the optimal decision on the weight analysis results of each working condition in the candidate operating condition library, and the compromise optimal operating condition of each clustering result is obtained. Finally, the optimal operating conditions in each clustering result are linearly combined to obtain the final complete driving condition curve, so that the constructed vehicle driving condition is consistent with the statistical parameter characteristics and driving status of the actual user big data to the greatest extent. The method flow is as follows Figure 2 shown.

[0288] Example 12:

[0289] The method for constructing a vehicle driving condition based on user big data information, the main technical content of which is shown in Example 11. Further, an example of constructing a vehicle driving condition is as follows:

[0290] 3.1 Kinematic short-stroke segment analysis

[0291] 3.1.1 Feature Parameter Matrix

[0292] Based on the 14 defined characteristic parameters, the characteristic parameter values ​​of each short-stroke kinematic segment are calculated and the characteristic parameter matrix is ​​constructed, as shown in Table 2.

[0293] 3.1.2 PCA Dimensionality Reduction

[0294] There are 14 characteristic parameters defined, but there is a linear correlation between the 5 characteristic parameters corresponding to the speed interval ratio, so only the first 4 of them need to be selected, so the number of principal components is 13. The principal component analysis of the short-stroke kinematic segment is performed to obtain the characteristic value, contribution rate and cumulative contribution rate of each principal component, as shown in Table 3 and Figure 3 shown.

[0295] Table 2 Characteristic parameter matrix of short-stroke kinematic segments

[0296]

[0297] Table 3 Eigenvalues, contribution rates and cumulative contribution rates of each principal component

[0298]

[0299] According to the above principle of screening the number of principal components, when the cumulative contribution rate of the nth principal component exceeds 80%, it indicates that the first n principal components can represent the main information of the original data. Figure 3As can be seen, by the fourth principal component, the cumulative contribution rate is close to 85%. In addition, the eigenvalues ​​of the first four principal components are all greater than 1, but from the fifth principal component onwards, the eigenvalues ​​drop significantly and fall far below 1. Therefore, the first four principal components are selected as the elements for subsequent cluster analysis.

[0300] Table 4 Principal component score coefficient matrix

[0301]

[0302] Table 5 Principal component score matrix

[0303]

[0304] After PCA dimensionality reduction, the principal component score coefficient matrix represents the correlation between the original data feature parameters and each principal component. A higher score indicates a greater correlation, meaning that the principal component better represents the information about that feature parameter. The principal component score coefficient matrix is ​​shown in Table 4. It can be seen that the first principal component primarily reflects average speed, average driving speed, idle time ratio, speed standard deviation, and the proportion of the 0-20 km / h speed range; the second principal component primarily reflects the acceleration time ratio, deceleration time ratio, the proportion of the 20-40 km / h speed range, and the proportion of the 60-80 km / h speed range; the third principal component primarily reflects average acceleration and average deceleration; and the fourth principal component primarily reflects information such as the acceleration standard deviation.

[0305] The principal component score matrix is ​​calculated from the principal component score coefficient matrix and the standardized characteristic parameter matrix, as shown in Table 5. The principal component score matrix will be used as the basis for cluster analysis in the future.

[0306] 3.1.3 Clustering of Short-Range Kinematic Segments

[0307] Various existing standard working conditions show that the types of cluster working conditions are usually divided into 3-5 types. That is, the number of clusters k can be set in the range of 3-5, which is more appropriate. Combined with the elbow method, the SSE value is calculated as the number of clusters k changes ( Figure 4 ), so the number of clusters k is set to 3 to balance the number of clustering conditions and computational efficiency.

[0308] Table 6 Comparative analysis of characteristic values ​​of various clusters

[0309]

[0310] Based on the K-means clustering method, cluster analysis is performed based on the principal component score matrix. In order to intuitively show the clustering effect, the sample scatter distribution of the first three principal components is plotted, as shown in the figure below. Figure 5As shown in the figure, the boundaries between adjacent clusters are clear. The sample points of the first two clusters are relatively concentrated, while the sample points of the third cluster have a larger span and are more dispersed. In the coordinate directions of the first and third principal components, the center points of each class are relatively large, while the spacing in the coordinate direction of the second principal component is slightly smaller. Overall, the scattered point patterns of each class can be well distinguished, and the clustering results are relatively ideal. The three clusters obtained are labeled as Cluster Condition I, Condition II, and Condition III. Table 6 shows the characteristic parameter values ​​of the three cluster conditions obtained by clustering. In Cluster Condition I, the average speed is low, the difference between the average speed and the average driving speed is large, and the idling ratio is the highest, reflecting the low-speed driving state of the vehicle in the city center with heavy traffic flow. In Condition II, the average speed is moderate, and the acceleration, deceleration, constant speed, and idling time are relatively uniform, reflecting the medium-speed driving state of the vehicle, such as in national / provincial highways and other operating areas with good traffic flow. In working condition III, the average driving speed is high, the idling time is short, and the average acceleration difference is small, which can reflect that the vehicle is driving at a relatively high speed, that is, driving on a suburban expressway or highway.

[0311] 3.1.4 Build an operating condition library

[0312] Combining the methods described in 2.1.3, a library of candidate operating conditions was constructed for each clustering result. Finally, the operating condition libraries constructed for clustering conditions I, II, and III are shown in Table 7. The data volumes for the candidate operating conditions are 4,335, 268,044, and 668, respectively.

[0313] Table 7: Candidate operating condition library

[0314]

[0315] 3.2 Determination of optimal vehicle driving conditions

[0316] 3.2.1 Characteristic parameter weight analysis based on CRITIC method

[0317] Taking the candidate operating condition library of each clustering condition as the decision object, single parameter target decision is carried out in sequence based on the defined characteristic parameters. The corresponding values ​​of the remaining characteristic parameters (i.e., decision criteria) are calculated and recorded using the corresponding decision results (i.e., decision objects), and the evaluation matrix X for the CRITIC weight analysis method is constructed.

[0318] Table 8 Weight coefficient ω j

[0319]

[0320] Furthermore, the weight analysis of characteristic parameters is performed by combining the aforementioned CRITIC method with the evaluation matrix to calculate the weight coefficient ω of each characteristic parameter. j, as shown in Table 8. It can be seen that in the low-speed driving condition corresponding to Cluster Condition I, the idle / deceleration time ratio and acceleration standard deviation have higher weights, indicating a greater impact. This is consistent with the frequent start-stop traffic conditions in real urban congestion. In the medium-speed driving condition corresponding to Cluster Condition II, the vehicle speed of 40-60 km / h accounts for a larger proportion of weights, which is also consistent with the common traffic conditions on real urban main roads. In the high-speed driving condition corresponding to Cluster Condition III, the weights are relatively balanced, which is also consistent with the relatively stable and seldom changing characteristics of real highway driving.

[0321] 3.2.2 Determination of the optimal operating condition combination based on the TOPSIS method

[0322] According to the characteristic parameter values ​​corresponding to the candidate operating condition library of each cluster condition, the decision matrix R is constructed. And according to the calculated weight coefficient ω of each characteristic parameter j , the contribution ranking is performed using the TOPSIS method, and the comprehensive contribution T of the relative characteristic parameters of each candidate operating condition under each clustering condition is calculated. i , the results are as follows Figure 6 shown.

[0323] Depend on Figure 6 It can be seen that Cluster Conditions I, II, and III have the highest comprehensive contribution coefficients corresponding to the 7th, 1st, and 1st candidate conditions, respectively, indicating that they have a higher degree of comprehensive adaptability to the 14 characteristic parameters. Therefore, the 7th, 1st, and 1st candidate conditions were ultimately selected as the optimal decision results for Cluster Conditions I, II, and III, respectively. The specific values ​​of their corresponding characteristic parameters are shown in Table 9.

[0324] Table 9 Characteristic parameter values ​​corresponding to the comprehensive decision results of clustering conditions I-III

[0325]

[0326] 3.2.3 Working condition construction results

[0327] Through calculation, the optimal decision-making conditions for cluster conditions I, II, and III correspond to time lengths of 260s, 431s, and 534s, respectively. The complete operating condition curve constructed by linear combination is as follows: Figure 7 As shown, the total time length is 1225s.

[0328] 4.1 Results of other working condition construction methods

[0329] 4.1.1 Cluster Markov Chain Construction Method

[0330] Clustering Markov chain construction method: The basic principle of working condition construction based on Markov chain is to regard the vehicle speed-time series as a random process without aftereffect, that is, the vehicle speed at the next moment is only related to the vehicle speed at the current moment and its transition probability, and has nothing to do with other moments.

[0331] On the basis of cluster analysis, the “speed-acceleration” dual-parameter Markov chain method is used to establish the curves of three clustering conditions, and then the complete vehicle driving condition curve is obtained by splicing. Figure 8 As shown in (a).

[0332] 4.1.2 V-A Matrix Construction Method

[0333] The VA matrix method expresses the probability of occurrence of the overall data and the speed and acceleration of the short-stroke skewness in a joint probability distribution to obtain the VA joint probability distribution matrix.

[0334] Based on cluster analysis, the speed data is bounded at 10 km / h and the acceleration data is bounded at 0.5 m / s. 2 The intervals are divided into two groups, and the joint probability distribution of VA of big data for each clustering condition and the joint probability distribution of VA of each short trip are solved. The similarity coefficients of the VA matrix of all short trips corresponding to various clustering conditions and the VA matrix of the overall data are calculated. Short trip segments with similarity coefficients greater than 0.8 are selected and randomly combined under the time length constraint to select the class condition with the smallest total error. The three selected class conditions are spliced ​​together to obtain the complete vehicle operation condition curve, as shown in Figure 8 (b) shown.

[0335] 4.2 Comparative Analysis

[0336] 4.2.1 Construction working condition verification and error analysis

[0337] By comparing with actual operation big data, the effectiveness and relative error of the working condition construction are confirmed.

[0338] Table 10 Error analysis of construction operating conditions

[0339]

[0340] As shown in Table 10, the relative errors between the constructed driving condition characteristic parameters and the original big data characteristic parameters are all within 10%, and the average relative error of each characteristic parameter value is only 3.23%, indicating that the constructed driving condition can better represent the user's actual driving status.

[0341] Furthermore, in order to explore the impact of regional differences on operating conditions, this embodiment compares and analyzes the operating conditions constructed in Chongqing with the operating conditions constructed in other regions and standard operating conditions. In order to reflect regional differences, the operating conditions of Shanghai, Beijing, and Chengdu in eastern, northern, and western China, as well as China's passenger car CLTC-P conditions and European NEDC conditions were selected. The results are shown in Table 11. It can be found that due to China's vast territory, different regions have obvious differences in economic development, road infrastructure construction levels, etc., resulting in large differences in the average speed and average driving speed of the constructed operating conditions. In addition, there are also different degrees of differences between different regions and CLTC-P and NEDC conditions.

[0342] Compared to the NEDC cycle, the constructed Chongqing cycle has lower average speeds and average travel speeds, indicating that traffic flow in the Chongqing region is slightly more complex than in Europe. However, it is closer to the CLTC-P cycle commonly used in China, indicating that the constructed cycle better reflects actual traffic flow conditions in China.

[0343] Table 11 Comparison of the constructed driving condition with other conditions

[0344]

[0345] 4.2.2 Analysis of rationality and applicability of construction conditions

[0346] Further comparison Figure 7 and Figure 8 It can be seen that both the method proposed in this embodiment and the existing method construct relatively standard vehicle driving condition curves. However, a specific analysis of the distribution characteristics and characteristic parameter errors of each driving condition curve reveals that:

[0347] (1) The vehicle operating condition curve constructed by the method proposed in this embodiment has smaller overall and average errors in characteristic parameters than existing methods. The smaller characteristic parameter errors indicate a higher degree of fit between the constructed driving condition and actual user operation big data. The constructed condition more accurately reflects the overall characteristics and can be used as a representative condition for subsequent energy consumption, emissions, and durability testing analyses.

[0348] (2) The proposed method has better user big data correlation effect and universality than the VA matrix construction method. The VA matrix construction method is usually used to construct driving conditions for objects with specific driving areas, routes, and driving requirements, such as urban buses or passenger vehicles on specific routes. When user big data does not have typical driving characteristics, it is difficult to accurately obtain its typical VA distribution characteristics, which leads to insufficient accuracy of the solution results.

[0349] (3) The vehicle operating condition curve constructed by the proposed method has stronger practical maneuverability. When constructing the operating condition based on the Markov chain construction method, since the operating state at the next moment is randomly generated based on the state transition probability matrix using Monte Carlo simulation, it may cause the vehicle speed to fluctuate repeatedly within a short period of time, such as Figure 8 As shown in (b), the speed-time curve exhibits a distinct jagged shape. This is inconsistent with actual user driving conditions and, to a certain extent, increases the operational difficulty of the operating condition. Compared to the driving condition curves constructed using the clustered Markov chain method, the short-stroke kinematic segments used in the proposed method are all derived from the original user database and are continuously changing time segments, which truly reflect the driving maneuvers and other conditions in the user's operating conditions. Therefore, the driving condition curves constructed based on these short-stroke kinematic segments greatly ensure the driver's operability during subsequent testing, that is, they are more reproducible.

Claims

1. A method for constructing automobile driving conditions based on user big data information, characterized in that: The following steps are involved: 1) Based on user driving big data information, obtain historical vehicle operation status. 2) Preprocess historical vehicle operating status to generate short trip segments and then build a vehicle driving condition database; 3) Use principal component analysis and K-means cluster analysis to process the data in the vehicle driving condition database and obtain different clustering results; 4) For each clustering result, calculate the characteristic parameter error between each short-trip segment and the overall clustering result, sort the short-trip segments in ascending order of error value, select several short-trip segments with the highest ranking, and randomly combine the selected short-trip segments to form an operating condition, thereby constructing a library of candidate operating conditions for each clustering result; 5) Based on the CRITIC method, the characteristic parameters of the vehicle operating state are weighted and the weight of each characteristic parameter is obtained; 6) Combining the weights of each characteristic parameter, the TOPSIS method is used to make the optimal decision for each operating condition in the candidate operating condition library, and the optimal operating condition for each clustering result is obtained; 7) The optimal operating conditions of each clustering result are spliced ​​together to obtain the final driving condition of the vehicle.

2. The method for constructing automobile driving conditions based on user big data information according to claim 1, characterized in that: The vehicle running state includes acceleration and speed.

3. The method for constructing automobile driving conditions based on user big data information according to claim 1, characterized in that: The preprocessing includes data elimination processing, data interpolation processing, and data filtering and denoising processing.

4. The method for constructing automobile driving conditions based on user big data information according to claim 1, characterized in that: In step 3), the steps to obtain different clustering results are as follows: 3.1) Construct the characteristic parameter matrix A m×n , as shown below: Where m represents the total number of short-trip segments; n represents the total number of characteristic parameters; a pq Represents the qth characteristic parameter of the pth short-distance segment, 1≤p≤m, 1≤q≤n; 3.2) For the characteristic parameter matrix A m×n Perform standardization to obtain the standardized matrix Z m×n , as shown below: Where z pq represents the qth characteristic parameter of the pth short-run segment after normalization; represents the mean of the qth characteristic parameter; s q represents the variance of the qth feature parameter; 3.3) According to the normalized matrix Z m×n The covariance matrix C is calculated as follows: Where u and y are the number of columns, 1≤u≤n, 1≤y≤n, Cov(u,y) represents the covariance between the u-th column and the y-th column in the standardized matrix Z; u p 、y p Respectively represent the p-th value of the u-th column and the y-th column in the normalized matrix Z; Respectively represent the means of the u-th column and the y-th column in the standardized matrix Z; 3.4) Calculate the eigenvalues ​​and unit eigenvectors of the covariance matrix C as follows: Cv q =λ q v q (6) Where λ q is the qth eigenvalue, λ q ≥0;v q is the qth eigenvalue λ q The corresponding unit eigenvector; 1≤q≤n; 3.5) For the eigenvalue λ q Sort in descending order and calculate the cumulative contribution rate CR of the first r0 eigenvalues ​​in the sort, as shown below: In the formula, q0 and q1 both represent the eigenvalue numbers; Indicates the eigenvalue ranked at q0; represents the eigenvalue ranked in the q1th position; r0 is the number of eigenvalues ​​when the cumulative contribution rate CR is greater than or equal to 80%; 3.6) According to the unit eigenvector v corresponding to the first r0 eigenvalues ​​of the sort q , construct the projection matrix As shown below: 3.7) Project the normalized data into the principal component space to obtain the dimensionally reduced data B, as shown below: Where B is the data after dimensionality reduction, and each row in B corresponds to the r0-dimensional principal component score of a short trip segment; 3.8) The first r0 principal components in the ranking are used as the objects of cluster analysis, and the r0-dimensional principal component scores are used as the clustering basis. K-means clustering method is used to cluster and obtain different clustering results; The sum of squared errors of the K-means clustering method is as follows: Where SSE is the sum of squared errors; i1 is the cluster number; k is the number of clusters; j1 is the sample number; m1 is the sample dimension; is the j1th sample of the i1th class; is the mean of class i1.

5. The method for constructing automobile driving conditions based on user big data information according to claim 1, characterized in that: The duration of the operating conditions in the candidate operating condition library is as follows: Where i represents the clustering result index, o represents the operating condition index; L io represents the duration of the oth operating condition in the candidate operating condition library of the i-th clustering result; δ is the error limit; Among them, the time length Tl of the i-th clustering result in the final driving condition of the vehicle is i As shown below: Where j represents the short-run segment index, n j represents the total number of short trip segments; T0 is the length of the vehicle's final driving condition; t ij represents the duration of the jth short trip segment in the i-th clustering result; k0 represents the total number of clustering results.

6. The method for constructing automobile driving conditions based on user big data information according to claim 1, characterized in that: In step 5), the steps for obtaining the weight of each characteristic parameter are as follows: 5.1) Define w groups of operating conditions in the candidate operating condition library as decision objects, y0 characteristic parameters as decision indicators, make a single-objective decision on the decision object with the minimum error, and obtain the corresponding decision result and use the decision results Calculate the values ​​of another y0-1 decision indicators Then construct the evaluation matrix X, where i2 = 1, 2, ..., w, j2 = 1, 2, ..., y0; 5.2) The evaluation matrix X is normalized and the normalized matrix X′ is calculated as follows: Where, is the value of the i2th decision object under the j2th decision indicator after normalization; The benefit indicator is a positive decision indicator, and the cost indicator is a negative decision indicator; 5.3) Calculate the contrast strength between decision indicators as follows: Where, is the standard deviation of the j2th decision indicator; is the mean value of the j2th decision indicator after standardization; 5.4) Calculate the correlation coefficient between the k1th decision indicator and the j2th decision indicator and conflict quantitative index values As shown below: Where h is the decision object index, are the values ​​of the hth decision object under the k1th and j2th decision indicators after normalization; are the means of the k1th and j2th decision indicators after standardization, and k1≠j2; 5.5) Calculate the information content of the decision indicator as follows: Where, represents the information content of the j2th decision indicator; 5.6) Calculate the weight of the decision indicator as follows: Where, represents the weight of the j2th decision indicator; is the information amount of the k1th decision indicator.

7. The method for constructing automobile driving conditions based on user big data information according to claim 1, characterized in that: In step 6), the steps for obtaining the optimal operating conditions for each clustering result are as follows: 6.1) Based on the characteristic parameter values ​​corresponding to each operating condition, a comprehensive evaluation decision matrix R1 is constructed as follows: Where, is the j2th characteristic parameter value corresponding to the i2th operating condition, i2=1,2,…,w, w is the total number of operating conditions, j2=1,2,…,y0, y0 is the total number of characteristic parameters; 6.2) Based on the normalization method, the comprehensive evaluation decision matrix R1 is normalized to obtain the regularized matrix R1′, as shown below: Where, is the j2th characteristic parameter value corresponding to the i2th operating condition after normalization; 6.3) Combine the weights of each feature parameter and weight the regularization matrix R1′ to obtain the decision matrix V, as shown below: Where, is the j2th characteristic parameter value corresponding to the i2th operating condition after weighting; represents the weight of the j2th feature parameter; 6.4) Construct the optimal solution set A + and the worst solution set A - , as shown below: Where, is the optimal value under the j2th characteristic parameter; is the worst value under the j2th characteristic parameter; 6.5) Calculate the sequence corresponding to the i2th operating condition in the decision matrix V to the optimal solution set A + and the worst solution set A - Euclidean distance and As shown below: 6.6) Calculate TOPSIS relative fit As shown below: 6.7) Relative fit from TOPSIS The maximum value is selected, and the corresponding operating condition is the optimal operating condition for each clustering result.

8. The method for constructing automobile driving conditions based on user big data information according to claim 7, characterized in that: When the characteristic parameter is a benefit indicator, the optimal value Take the maximum value and the worst value Take the minimum value; When the characteristic parameter is a cost indicator, the optimal value Take the minimum value, the worst value Take the maximum value.

9. The method for constructing automobile driving conditions based on user big data information according to claim 1, characterized in that: The characteristic parameters include average speed, average driving speed, average acceleration, average deceleration, uniform speed time ratio, idling time ratio, acceleration time ratio, deceleration time ratio, speed standard deviation, and acceleration standard deviation.

10. The method for constructing automobile driving conditions based on user big data information according to claim 1, characterized in that: The final vehicle driving condition is used to test and evaluate the vehicle's energy consumption, economy, power, and reliability and durability of the vehicle system.

Images