A vehicle operating mode transfer planning method for fuel consumption optimization
By dividing and clustering historical data into regions, the vehicle operating condition transfer path is planned as a directed graph shortest path search, which solves the problem of fuel consumption optimization under transient vehicle operating conditions and achieves fuel consumption reduction.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- TIANJIN UNIV
- Filing Date
- 2023-06-20
- Publication Date
- 2026-05-29
AI Technical Summary
Existing technologies struggle to effectively optimize fuel consumption under transient vehicle conditions, especially when acceleration and deceleration conditions are frequent, resulting in insufficient fuel consumption optimization methods.
By processing historical data offline, performing regional division and clustering, the vehicle operating condition transfer path is planned as a directed graph shortest path search. The directed graph shortest path algorithm is used to optimize the operating condition transfer path and select the path with the lowest fuel consumption.
It simplifies vehicle operating condition transfer planning and optimizes fuel consumption, reducing fuel consumption during vehicle transfers between different operating conditions.
Smart Images
Figure CN116777094B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of vehicle control, and more particularly to a vehicle operating condition transition planning method for fuel consumption optimization. Background Technology
[0002] With the continuous introduction of vehicle-related regulations and the increasing awareness of the public, fuel-efficient driving is becoming an inevitable trend. Studies have shown that the main operating conditions of automobiles in my country include idling, acceleration, deceleration, and constant speed. Constant speed driving accounts for less than 30% of the time, while acceleration and deceleration account for as much as 54.78%. Even when a vehicle is driving at a constant speed, the engine's operating conditions are not entirely constant. Therefore, research on transient operating condition control of vehicles is crucial.
[0003] To optimize fuel consumption under transient operating conditions, vehicle operating condition transition paths can be planned. Given the initial and final operating conditions, different operations may lead to different operating condition transition paths, and fuel consumption changes with the path. Selecting an appropriate operating condition transition path can reduce vehicle fuel consumption. Summary of the Invention
[0004] To address the aforementioned problems, this invention proposes a vehicle operating condition transfer planning method for fuel consumption optimization. This method divides historical data into regions based on different operating conditions through offline processing. Each region is treated as a vertex in a directed graph, and the transfer cost between regions is used as the distance between vertices. This transforms vehicle operating condition transfer path planning into a directed graph shortest path search, simplifying the vehicle operating condition transfer planning process and optimizing fuel consumption.
[0005] A vehicle operating condition transition planning method for fuel consumption optimization includes the following steps:
[0006] S1. Collect driving data based on relevant vehicle test conditions, including vehicle driving information and engine information, for offline analysis and processing;
[0007] S2. Select reasonable operating condition characteristic variables, and perform clustering and partitioning on the corresponding driving history data after normalization;
[0008] S3. Based on driving history data, extract information from each region of the cluster to obtain the relationship between each region and the transfer cost;
[0009] S4. Construct a directed graph of work condition transitions based on the relationships between each interval, and calculate all minimum cost transition paths between different start and end regions accordingly; wherein: the process of constructing the directed graph of work condition transitions is as follows:
[0010] Convert each region into vertices in a directed graph;
[0011] And the distance between vertices in the directed graph will be calculated using the region transfer cost;
[0012] The optimal fuel consumption path is searched based on the directed graph of the working condition transition: all nodes of the directed graph are divided into two parts, one part stores all starting points and the other part stores all ending points; the shortest path algorithm is used to search for the shortest path between all starting and ending points, which is the path with the lowest fuel consumption between the starting and ending working conditions, and the corresponding directed graph shortest path search table is constructed.
[0013] S5. Based on the current actual working conditions and target working conditions, search for all possible transfer paths between the two working conditions in historical data, and determine the areas traversed under different transfer paths according to the partitioning method in step S2, thereby forming different transfer paths with fixed start and end points in the directed graph; determine the areas to which the current working condition and target working condition belong, and use these as the start and end nodes in the directed graph; find the minimum cost transition path according to step S4, compare the directed graph paths formed by the statistical areas under different transfer paths with the minimum cost transition path, select the one that is the same as the optimal transfer path between the actual working condition and the target working condition, and issue the corresponding signal to execute.
[0014] Furthermore, the clustering and partitioning method in step S3 can obtain the region corresponding to each working condition, and construct the relationship between each region based on the continuously changing working conditions in historical driving data:
[0015] If there are a large number of continuous operating conditions transitioning between the two regions, it indicates that the two regions are strongly connected and there is a transfer relationship, so the transfer cost of the regions needs to be calculated.
[0016] If a transfer relationship exists between the two intervals, calculate the density distribution of all operating points in each interval and determine the core operating point of that interval; further divide the endpoint region into multiple smaller intervals and determine their probability densities; calculate the transfer fuel consumption from all operating conditions in the starting region to all operating conditions in the smaller interval; multiply the probability density of each smaller interval by the transfer fuel consumption and sum them as the fuel consumption transfer cost between the two intervals; that is:
[0017]
[0018] Where: fc represents the fuel consumption during the two-range transition, fc i Let pi be the fuel consumption for the transfer from the initial region to the i-th cell, and let pi be the probability density between the i-th cells.
[0019] Furthermore, in step S2, reasonable operating condition characteristic variables are selected, and a clustering and partitioning process is performed after normalizing the driving history data:
[0020] Selection of operating condition characteristic variables: Select characteristic variables related to vehicle fuel consumption, and conduct correlation analysis on different characteristics to set a reasonable combination scheme of characteristic variables;
[0021] Historical driving data normalization: The selected features X = (X1, X2, X3, X4, ...) in the historical driving data are processed to ensure they fall within a specific interval.
[0022]
[0023] Where: x is the normalized variable data, μ is the mean of the relevant variables, and σ is the standard deviation of the relevant variables;
[0024] Clustering partitioning:
[0025] The number of cluster partitions is determined using the following formula:
[0026]
[0027] Wherein: the sum of squared errors (SSE) is used as the dividing criterion; the SSE corresponding to the number of different classes (k) is calculated; when the rate of decrease in SSE suddenly slows down, it is considered to be the optimal value of k; x represents the normalized variable data, m i C is the average value of class i. i Let k be the set of class i, and k be the number of classes;
[0028] The cluster partitions are divided into categories using the following formula:
[0029]
[0030] Wherein: using Euclidean distance as the dividing criterion, the distance d from each working condition to the k central working conditions is calculated; if the distance is the smallest, the working condition is assigned to the category to which the central working condition belongs; x i x j Here are the variable data for different operating conditions, where o represents the number of operating condition features, and x represents the variable data for different operating conditions. ik x jk These are the different characteristics of the working conditions.
[0031] This invention can also be implemented through the following technical solutions:
[0032] During actual vehicle operation, information on the current operating condition and required operating condition of the vehicle is acquired in real time and used as the initial operating condition and target operating condition, respectively. Based on the historical driving data collected in step S1, all transfer paths with the starting point operating condition as the initial operating condition and the ending point operating condition as the target operating condition are searched. Based on the clustering and partitioning method in step S2, the region to which each operating condition point belongs under the corresponding transfer path is calculated, thereby obtaining different transfer paths with fixed starting and ending points in the directed graph.
[0033] The current working condition and the region to which the required working condition belong are calculated according to the clustering and partitioning method in step S2 as the starting node and the final node in the directed graph, and the shortest path is obtained by searching the shortest path table of the directed graph.
[0034] The obtained shortest path is compared with the different transition paths in the directed graph of the previous sub-step. The same path is selected as the optimal transition path, and the corresponding working condition timing data is constructed and sent to the vehicle electronic control unit to realize the path.
[0035] Beneficial effects
[0036] The beneficial effects of this invention are: It proposes a vehicle operating condition path transfer planning method based on historical data to address the issue of varying fuel consumption under different operating conditions. This method divides the historical data into regions representing different operating conditions, treating each region as a vertex in a directed graph, and using the transfer cost between regions as the distance between vertices, thus transforming vehicle operating condition transfer path planning into a directed graph shortest path search. This not only simplifies the vehicle operating condition transfer planning process but also optimizes fuel consumption while better reflecting actual vehicle conditions. Attached Figure Description
[0037] Figure 1 This is a flowchart of the method of the present invention.
[0038] Figure 2 This invention relates to a schematic diagram of class classification changes. Detailed Implementation
[0039] The following is in conjunction with the appendix Figure 1 The technical solution of the present invention will be described in further detail below, but the scope of protection of the present invention is not limited to the following description.
[0040] like Figure 1 As shown, a vehicle operating condition transition planning method for fuel consumption optimization includes the following steps:
[0041] Step 101: Design vehicle test conditions and collect information during the driving process for offline analysis, including but not limited to:
[0042] Driving information: speed, acceleration, accelerator pedal opening, brake pedal, gear position, and steering wheel angle, etc.
[0043] Engine information: Speed, torque, intake-related information, valve opening and closing information, and fuel consumption.
[0044] Step 102: Perform working condition clustering and partitioning based on historical data, including feature variable selection sub-steps, data preprocessing sub-steps, and clustering and partitioning sub-steps.
[0045] The data collected in step 101 involves various states of vehicle and engine operation, but not all variables are significantly related to vehicle fuel consumption. Therefore, it is necessary to select appropriate feature variables for combination to best reflect changes in fuel consumption, i.e., to maximize their correlation with fuel consumption. In this case, the feature combination of engine speed, torque, and intake air volume is selected for illustration, i.e., the three features in X = (X1, X2, X3).
[0046] Because the collected data vary in size, normalization is necessary before use to eliminate the adverse effects of outlier samples and ensure that data features have a uniform measurement scale. In this case, the normalization method used is as follows: the selected feature X = (X1, X2, X3) from the driving history data is processed to fall within a specific interval:
[0047]
[0048] x represents the normalized variable data, μ represents the mean of the related variables, and σ represents the standard deviation of the related variables.
[0049] After data normalization, unsupervised clustering methods are used to divide the data into clusters. In this case, an improved K-means algorithm is used for clustering.
[0050] The first step is to determine the number of classes, using the sum of squared errors (SSE) from the "elbow rule":
[0051]
[0052] x represents the normalized variable data, m i C is the average value of class i. i Let k be the set of class i, and k be the number of classes.
[0053] As the value of k increases, the SSE gradually decreases. The optimal k value is considered to be when the rate of SSE decrease suddenly slows down. At this point, the system is divided into k categories with k central operating conditions. The SSE variation process in this case is as follows: Figure 2 As shown, the value of k can be 6.
[0054] The second step is to classify the categories. Using Euclidean distance as the basis for classification, k central working conditions are first randomly determined. Then, the position of the working condition center is continuously updated, and the distance d from each working condition to the k central working conditions is calculated. If the distance is the smallest, the working condition is classified into the category to which the central working condition belongs.
[0055]
[0056] x i x j Here are the variable data for different operating conditions, where o represents the number of operating condition features, and x represents the variable data for different operating conditions.ik x jk These are the different characteristics of the working conditions.
[0057] Step 103: Construct interval relationships and costs based on historical data. Using the clustering and partitioning method in Step 102, the regions corresponding to each operating condition can be obtained. If a large number of consecutive operating conditions transition between two regions, it indicates a strong connection between the two intervals, considered as a transfer relationship. If only a very small number of consecutive operating conditions transition between the two intervals, it indicates a weak connection, which can be ignored. If a transfer relationship exists between the two intervals, calculate the density distribution of all operating condition points in each region and determine the core operating condition point of that interval. Divide the endpoint region into n smaller intervals and determine their probability densities. Calculate the transfer fuel consumption from all operating conditions in the starting region to all operating conditions in each smaller interval. Multiply the probability density of each smaller interval by the transfer fuel consumption and sum them as the fuel consumption transfer cost between the two intervals.
[0058]
[0059] fc represents the fuel consumption during the two-range transition. i Let pi be the fuel consumption for the transfer from the initial region to the i-th cell, and let pi be the probability density between the i-th cells.
[0060] Step 104: Perform shortest path search based on the directed graph of work condition transitions, including the following sub-steps:
[0061] The first step is to construct a directed graph of work condition transition based on the relationships between the regions and the transition costs in step 3: After dividing the selected work condition into regions, each region is regarded as a vertex in the directed graph. The weight (i.e., distance) between each vertex is defined by the region transition cost, and a weighted directed graph can be obtained. At this time, the actual work condition path planning problem is transformed into a directed graph shortest path search problem.
[0062] The second step is to search for the optimal fuel consumption path based on the directed graph of the working condition transition: divide all nodes of the directed graph into two parts, one part stores all starting points and the other part stores all ending points. In this case, the Floyd-Warshall algorithm is used to search for the shortest path between any starting point and ending point: starting from the starting point, each vertex is used as the medium k in turn. Then, for each pair of starting and ending points, check whether there is a path that passes through k and is shorter than the known path. If it exists, update it; otherwise, keep the original path.
[0063] The shortest path algorithm is used to search for the shortest possible path between all start and end points. This shortest path is the path with the lowest fuel consumption during the operation mode transfer. All shortest paths are stored in a table for online querying.
[0064] Step 105: Select the optimal working condition transition path based on historical data, including the following sub-steps:
[0065] The first step is to obtain all paths between the actual working condition and the target working condition. Based on the Euclidean distance formula in step 2, the regions to which the two working conditions belong are determined. Then, all transfer paths are searched in the continuously changing working conditions in the historical driving data. Finally, the regions traversed by the transfer paths are counted to form different transfer paths with fixed start and end points in the directed graph.
[0066] The second step is to select the optimal path. The regions where the actual working conditions and the target working conditions are located are respectively used as the starting node and the final node of the directed graph shortest path search method in step 4. The shortest path found in the shortest path storage table is compared with the transfer regions corresponding to all transfer paths in the first step. The same path is selected as the optimal transfer path. If multiple paths match, the one with the shorter transfer time is selected.
[0067] The third step is to implement the optimal transfer path. After determining the optimal transfer path, the timing changes of engine speed, torque, and other quantities under this path are also determined. Based on the engine's own control, relevant commands such as the target speed and torque are sent to the vehicle's electronic control unit to change the engine's operating state, thus completing the implementation of each operating condition in the optimal transfer path.
[0068] This invention proposes a working condition transition planning method based on clustering and shortest path, which transforms the actual working condition path planning problem into a directed graph shortest path search problem. This not only simplifies the vehicle working condition transition planning process, but also optimizes fuel consumption while better reflecting the actual conditions of the vehicle.
[0069] The technical means disclosed in this invention are not limited to the algorithms disclosed in the above embodiments, but also include technical solutions composed of any combination of similar methods of the above algorithms. It should be noted that all other embodiments obtained by making improvements without departing from the principles of this invention, without creative effort, should fall within the scope of protection of this invention.
Claims
1. A vehicle operating condition transition planning method for fuel consumption optimization, characterized in that, Includes the following steps: S1. Collect driving data based on relevant vehicle test conditions, including vehicle driving information and engine information, for offline analysis and processing; S2. Filter the obtained driving history data and select the corresponding operating condition characteristic variables for clustering and partitioning; S3. Based on driving history data, extract information from each region of the cluster to obtain the relationship between each region and the transfer cost; S4. Construct a directed graph of work condition transitions based on the relationships between each interval, and calculate all minimum cost transition paths between different start and end regions accordingly; wherein: the process of constructing the directed graph of work condition transitions is as follows: Convert each region into vertices in a directed graph; And the distance between vertices in the directed graph will be calculated using the region transfer cost; The optimal fuel consumption path is searched based on the directed graph of the working condition transition: all nodes of the directed graph are divided into two parts, one part stores all starting points and the other part stores all ending points; the shortest path algorithm is used to search for the shortest path between all starting and ending points, which is the path with the lowest fuel consumption between the starting and ending working conditions, and the corresponding directed graph shortest path search table is constructed. S5. Based on the current actual working conditions and target working conditions, search for all possible transfer paths between the two working conditions in the historical data, and determine the areas traversed under different transfer paths according to the partitioning method in step S2, thereby forming different transfer paths with fixed start and end points in the directed graph; determine the areas to which the current working condition and target working condition belong, and use these as the start and end nodes in the directed graph; find the minimum cost transition path according to step S4; compare the directed graph paths formed by the statistical areas under different transfer paths with the minimum cost transition path, select the one that is the same as the optimal transfer path between the actual working condition and the target working condition, and issue the corresponding signal to execute.
2. The vehicle operating condition transition planning method for fuel consumption optimization according to claim 1, characterized in that, In step S3, the region corresponding to each working condition can be obtained based on the clustering partitioning method, and the relationship between each region is constructed based on the continuously changing working conditions in the historical driving data: If there are a large number of continuous operating conditions transitioning between the two regions, it indicates that the two regions are strongly connected and there is a transfer relationship, so the transfer cost of the regions needs to be calculated. If a transfer relationship exists between the two intervals, calculate the density distribution of all operating points in each interval and determine the core operating point of that interval; further divide the endpoint region into multiple smaller intervals and determine their probability densities; calculate the transfer fuel consumption from all operating conditions in the starting region to all operating conditions in the smaller interval; multiply the probability density of each smaller interval by the transfer fuel consumption and sum them as the fuel consumption transfer cost between the two intervals; that is: in: fc Fuel consumption for the two-section transfer, fc i For the initial region to the th i Fuel consumption during inter-community transfers pi For the first i Probability density between neighborhoods.
3. The vehicle operating condition transition planning method for fuel consumption optimization according to claim 1, characterized in that, Step S2 involves selecting appropriate operating condition characteristic variables and performing clustering and partitioning processes after normalizing the historical driving data. Selection of operating condition characteristic variables: Select characteristic variables related to vehicle fuel consumption, and conduct correlation analysis on different characteristics to set a reasonable combination scheme of characteristic variables; Normalization processing of driving history data: This involves normalizing the selected historical driving data... X =( X1 , X2 , X3 , X4 The features (, ...) are processed to fall into a specific interval: in ：x For normalized variable data, μ The average value of the relevant variables. σ The standard deviation of the relevant variables; Clustering partitioning: The number of cluster partitions is determined using the following formula: ； Specifically, the sum of squared errors was used as the criterion for dividing the classes, and the number of distinct classes was calculated. k The corresponding sum of squared errors SSE ,when SSE A sudden slowdown in the rate of decline is considered optimal. k value; x For the normalized variable data, m i For class i The average value, C i For category i The set, k The number of classes; The cluster partitions are divided into categories using the following formula: ； Wherein: Euclidean distance is used as the dividing line, and the distance to each working condition is calculated. k Distance between the central working conditions d If the distance is the smallest, then the working condition is classified into the category to which the central working condition belongs; x i , x j For variable data under different working conditions, o The number of operating condition characteristics, x ik , x jk These are the different characteristics of the working conditions.
4. A vehicle operating condition transition planning method for fuel consumption optimization according to any one of claims 1-3, characterized in that, During actual vehicle operation, information on the current operating condition and required operating condition of the vehicle is acquired in real time and used as the initial operating condition and target operating condition, respectively. Based on the historical driving data collected in step S1, all transfer paths with the starting point operating condition as the initial operating condition and the ending point operating condition as the target operating condition are searched. Based on the clustering and partitioning method in step S2, the region to which each operating condition point belongs under the corresponding transfer path is calculated, thereby obtaining different transfer paths with fixed starting and ending points in the directed graph. The current working condition and the region to which the required working condition belong are calculated according to the clustering and partitioning method in step S2 as the starting node and the final node in the directed graph, and the shortest path is obtained by searching the shortest path table of the directed graph. The obtained shortest path is compared with the different transition paths in the directed graph of the previous sub-step. The same path is selected as the optimal transition path, and the corresponding working condition timing data is constructed and sent to the vehicle electronic control unit to realize the path.