Vehicle driving duration prediction method and device, computer equipment and storage medium

Abstract

The invention relates to the field of new energy vehicles, and discloses a vehicle driving duration prediction method and device, computer equipment and a storage medium, and the method comprises the steps: obtaining the starting characteristics of a vehicle in a current driving cycle; screening out a target driving cycle cluster associated with the starting feature from driving cycle clusters of a historical driving cycle set; and predicting the driving duration of the current driving cycle based on the starting characteristics and the target driving cycle cluster to obtain predicted driving duration. According to the invention, the problem that the existing new energy vehicle cannot carry out personalized driving fragment time consumption prediction when the user does not use navigation is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of new energy vehicles, in particular to a vehicle driving duration prediction method and device, a computer device and a storage medium. BACKGROUND

[0002] With the global energy structure adjustment, electric vehicles gradually become the development trend in the future. However, electric vehicles face problems such as high energy consumption and reduced range in actual use, which seriously affect user experience. User driving driving segment analysis results show that short driving driving segments account for a very high proportion, including daily commuting, surrounding shopping and other scenarios. In such short driving process, the battery thermal management system often over-responds, resulting in a significant increase in energy consumption loss. In order to optimize the energy consumption performance of electric vehicles, especially for short driving scenarios, a method is needed to predict the driving duration to identify short driving driving segments and take appropriate battery thermal management optimization strategies to reduce energy consumption loss and improve the range of electric vehicles.

[0003] Currently, the vehicle driving segment time is mainly estimated by the navigation system. This method requires the user to actively open the mobile network and use the navigation function, and relies on real-time traffic information. This solution is only suitable for scenarios where the user actively uses navigation and inputs complete driving segment information, and cannot cover the case where the user does not use navigation, resulting in limited coverage of driving segment prediction and lack of personalized prediction ability for user driving habits. SUMMARY

[0004] Therefore, the embodiments of the present application provide a vehicle driving duration prediction method and device, a computer device and a storage medium to solve the problem that the existing new energy vehicles cannot perform personalized driving segment time prediction when the user does not use navigation.

[0005] In a first aspect, the embodiments of the present application provide a vehicle driving duration prediction method, which comprises: obtaining a start-up feature of a vehicle in a current driving cycle; selecting a target driving cycle cluster associated with the start-up feature from driving cycle clusters of a historical driving cycle set; predicting the driving duration of the current driving cycle based on the start-up feature and the target driving cycle cluster, to obtain a predicted driving duration.

[0006] Further, the selecting a target driving cycle cluster associated with the start-up feature from driving cycle clusters of a historical driving cycle set comprises: obtaining historical start-up features of each driving cycle cluster in the historical driving cycle set; matching the start-up feature with each historical start-up feature respectively to obtain a matching result; determining a target driving cycle cluster from the set of historical driving cycles based on the matching result.

[0007] Further, before obtaining the historical start-up features of each driving cycle cluster in the set of historical driving cycles, the method further comprises: obtaining time stamps of state switching signals of the vehicle in a historical time period; dividing the historical time period into a plurality of driving segments based on the time stamps of the state switching signals, and obtaining start-up data and driving data generated by the vehicle in each driving segment; performing clustering processing on the driving segments based on the start-up data and the driving data to obtain at least one driving cycle cluster, and generating a historical start-up feature corresponding to each driving cycle cluster according to the start-up data.

[0008] Further, the dividing of the historical time period into a plurality of driving segments based on the time stamps of the state switching signals comprises: dividing the historical time period into a plurality of initial time periods according to the time stamps of the state switching signals; identifying an interval duration between adjacent initial time periods; when the interval duration is less than a duration threshold, merging the adjacent initial time periods into one driving segment; or when the interval duration is greater than or equal to the duration threshold, taking the adjacent initial time periods as one driving segment respectively.

[0009] Further, the performing of clustering processing on the driving segments based on the start-up data and the driving data to obtain at least one driving cycle cluster comprises: calculating segment distances between each driving segment based on the start-up data and the driving data; determining clustering parameters according to the segment distances between the driving segments, and dividing the driving segments into a plurality of cluster samples according to the clustering parameters; constructing at least one driving cycle cluster using the cluster samples.

[0010] Further, the matching of the start-up feature with each historical start-up feature to obtain a matching result comprises: calculating similarities between the start-up feature and the historical start-up features of each driving cycle cluster; sorting the driving cycle clusters based on the similarities to obtain a cycle cluster sequence; selecting a preset number of candidate driving cycle clusters from the cycle cluster sequence as the matching result.

[0011] Further, the method comprises the following steps of: obtaining driving data of each driving segment in the target driving cycle cluster; generating historical driving features of the target driving cycle cluster by using the driving data; calculating the predicted driving duration of the current driving cycle based on the starting feature and the historical driving features.

[0012] In a second aspect, an embodiment of the present application provides a vehicle driving duration prediction device, and the device comprises: an obtaining module, configured to obtain a starting feature of a vehicle in a current driving cycle; a screening module, configured to screen a target driving cycle cluster associated with the starting feature from driving cycle clusters in a historical driving cycle set; a prediction module, configured to predict a driving duration of the current driving cycle based on the starting feature and the target driving cycle cluster, to obtain a predicted driving duration.

[0013] In a third aspect, an embodiment of the present application provides a computer device, which comprises a memory and a processor, the memory and the processor are communicatively connected, the memory stores computer instructions, and the processor executes the computer instructions to perform the method of the first aspect or any of the corresponding embodiments.

[0014] In a fourth aspect, an embodiment of the present application provides a computer readable storage medium, which stores computer instructions, and the computer instructions are used to make a computer execute the method of the first aspect or any of the corresponding embodiments.

[0015] The method provided by the embodiment of the present application has the following beneficial effects: The method provided by the embodiment of the present application can accurately capture the key state information at the time of starting the vehicle by obtaining the starting feature of the vehicle in the current driving cycle in real time, and provides real-time input data basis for subsequent prediction; the target driving cycle cluster associated with the current starting feature is intelligently screened from the driving cycle clusters in the historical driving cycle set, the historical driving mode most similar to the current driving segment is effectively identified, and the precise matching of the user driving habit is realized; the comprehensive prediction based on the starting feature and the target driving cycle cluster not only fully considers the real-time state of the current driving segment, but also integrates the historical driving rules of the user, so that the predicted driving duration with high personalization and continuously optimized accuracy is generated without relying on external navigation, and the precision of the short driving segment identification and the response efficiency of the battery thermal management strategy are significantly improved. BRIEF DESCRIPTION OF DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the description of the embodiments or the prior art. Obviously, the drawings described below are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor based on these drawings.

[0017] Figure 1 is a flowchart of a vehicle driving duration prediction method according to an embodiment of the present application; Figure 2 is a flowchart of another vehicle driving duration prediction method according to an embodiment of the present application; Figure 3 is a model construction and prediction flowchart according to an embodiment of the present application; Figure 4 is a structural block diagram of a vehicle driving duration prediction device according to an embodiment of the present application; Figure 5 is a hardware structure diagram of a computer device according to an embodiment of the present application. DETAILED DESCRIPTION

[0018] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the following will combine the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are some of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0019] According to the embodiments of the present application, vehicle driving duration prediction methods, devices, computer devices and storage media are provided. It should be noted that the steps shown in the flowchart of the drawings can be executed in a computer system such as a group of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0020] In the present embodiment, a vehicle driving duration prediction method is provided, Figure 1 is a flowchart of a vehicle driving duration prediction method according to an embodiment of the present application, as Figure 1 shown, the flow includes the following steps: Step S101, obtaining the starting feature of the vehicle in the current driving cycle.

[0021] In the embodiments of the present application, in order to realize personalized driving segment time consumption prediction, it is necessary to capture and quantify the start-up feature at the beginning of the current driving segment in real time, wherein the start-up feature is used to represent the starting time mode of the current driving segment, for example, the user's travel time habit is reflected by the departure time. The way to obtain the start-up feature specifically includes: Firstly, the vehicle-mounted sensor network (such as CAN bus) continuously listens to and collects switching signals related to the running state of the vehicle. The state switching signal can be understood as a vehicle usage mode signal, which is a symbol describing the current running state of the vehicle, such as driving, starting (or ignition), and engine off, which represents the key state switching of the vehicle power system and power management.

[0022] Then, the original signals that may exist redundantly or out of order are preprocessed to form regular time series data. The core of preprocessing is to remove redundancy and sort: arrange all collected signals in ascending order according to their timestamps to generate a continuous event stream; for multiple state switching signals appearing at the same timestamp, remove redundancy based on the preset priority rule and retain the signal with the highest priority. The priority rule is: driving > starting > engine off. For example, if two state switching signals of starting and driving are received at the same millisecond, only the driving signal is retained because driving is a higher-order running state.

[0023] Finally, after forming the regular time series, the departure time of the start-up feature is extracted from it. The departure time refers to the time point when the vehicle enters the driving state from the non-driving state (such as engine off or starting) in the current driving cycle. In order to eliminate the influence of date changes and focus on the daily travel rules, this time point is converted to the number of minutes based on 0 o'clock of the current day. For example, if the vehicle first enters the driving state at 8:30 am, the feature value of the current start-up data is 8 * 60 + 30 = 510 minutes. Finally, the calculated departure time constitutes the start-up feature of the current driving cycle. This feature serves as the input of the current driving segment identification and prediction model, providing a data basis for subsequent query of similar historical driving segments and prediction. The whole process is completed locally on the vehicle side without user intervention and external network and navigation information.

[0024] Step S102, screening a target driving cycle cluster associated with the start-up feature from the driving cycle clusters of the historical driving cycle set.

[0025] In the embodiments of the present application, the historical starting features of each driving cycle cluster in the historical driving cycle set are first obtained, and then the starting feature of the current driving cycle is matched with each historical starting feature. The similarity (such as the similarity score based on the Euclidean distance) between the two is calculated, and all driving cycle clusters are sorted in descending order of similarity to form a cycle cluster sequence. Finally, a preset number of candidate driving cycle clusters are selected from the sequence as the matching result, and they are determined as the target driving cycle cluster. A similarity threshold can also be set, and only when the highest similarity exceeds the threshold, the matching is considered successful, otherwise a backup prediction strategy can be triggered, such as using the average time consumption of all driving segments in the recent period as the prediction value.

[0026] In step S103, the driving time of the current driving cycle is predicted based on the starting feature and the target driving cycle cluster, and a predicted driving time is obtained.

[0027] In the embodiments of the present application, the driving data of each driving segment in the target driving cycle cluster is used to generate the historical driving feature (such as the average driving time) of the cluster. Then, the starting feature (such as the departure time) of the current driving cycle is combined with the historical driving feature, and the predicted driving time is obtained through a preset calculation model. In addition, a weighted average algorithm can also be used to calculate the driving time of each historical driving segment by weighting, and the historical driving segment closest to the current time is given a higher weight, so as to generate a more accurate predicted driving time.

[0028] In the embodiments of the present application, the target driving cycle cluster associated with the starting feature is selected from the driving cycle clusters of the historical driving cycle set, including the following steps A1-A3: Step A1, obtaining the historical starting feature of each driving cycle cluster in the historical driving cycle set.

[0029] Specifically, the representative features of each typical driving mode are extracted from the historical driving mode library constructed, so as to compare with the current driving segment: First, the locally stored historical driving cycle set is accessed, which is constructed by clustering the historical driving segments and contains at least one driving cycle cluster. Each driving cycle cluster is essentially a set of similar driving segments (i.e. single historical driving cycle), which are classified into the same category because of the proximity in the feature space constituted by the starting features (such as historical starting data), representing the repeated regular driving modes, such as "commuting driving segment around 8 am on weekdays" or "driving segment to the mall in the afternoon on weekends".

[0030] It should be noted that the historical start-up feature is not a feature of a specific driving segment in the cluster, but a statistical representation of the overall trend of the cluster, aiming to capture the commonality of the driving mode, for example, the average (or median) of the start time of all driving segments in the cluster as the representative start time of the cluster, and the statistical value of the start time constitutes the historical start-up feature of the driving cycle cluster, which is a feature vector summarizing the typical start time of the entire cluster. The historical start-up features of all driving cycle clusters are cached in the memory as a query table for fast matching. These features serve as a benchmark for measuring the similarity of the current driving segment to the historical mode, providing direct and efficient data support for the next real-time matching operation.

[0031] Step A2, match the start-up feature with each historical start-up feature to obtain a matching result.

[0032] In the embodiments of the present application, the start-up feature is matched with each historical start-up feature to obtain a matching result, including the following steps A21-A23: Step A21, calculate the similarity between the start-up feature and the historical start-up feature of each driving cycle cluster.

[0033] Specifically, the start-up feature (such as the start time) of the current driving cycle is compared with the historical start-up feature (i.e. the representative start time of each cluster) of each driving cycle cluster in the historical driving cycle set one by one. This comparison can be done by calculating the distance between the current start-up feature vector and the historical start-up feature vector of each driving cycle cluster, such as the Euclidean distance. The closer the two features, the smaller the distance value, indicating a higher similarity. In actual operation, the calculated distance value can be directly used (in which case a minimization judgment needs to be made), or the distance value can be converted into a standardized similarity score between 0 and 1, such as by taking the inverse of the distance or using a Gaussian kernel function to convert, so that a higher score represents a higher similarity.

[0034] Step A22, sort the driving cycle clusters based on the similarity to obtain a cycle cluster sequence.

[0035] Specifically, the similarity value of each driving cycle cluster in the historical driving cycle set to the current start-up feature is obtained. Subsequently, all driving cycle clusters are arranged in descending order using the similarity value as the sorting key, indicating that the driving cycle cluster with the highest similarity score that is most similar to the current driving segment will be ranked first, while the least similar will be ranked last. Through the sorting operation, all driving cycle clusters are converted into an ordered list, i.e. a cycle cluster sequence. The cycle cluster sequence directly reflects the association priority of each historical driving mode to the current driving segment, forming a clear echelon from the most matching to the least matching, which is used to select a preset number of candidate driving cycle clusters, ensuring that the target driving cycle cluster for prediction is the most relevant historical reference under the current circumstances.

[0036] Step A23, select a preset number of candidate driving cycle clusters from the cycle cluster sequence as the matching result.

[0037] Specifically, based on a predefined preset number (a fixed value or a dynamically calculated proportion based on the total length of the cycle cluster sequence), a specified number of driving cycle clusters are sequentially selected from the top of the cycle cluster sequence. The driving cycle cluster with the highest similarity is selected as the candidate driving cycle cluster, which collectively constitutes the matching result. The matching mechanism of the start-up feature and the historical start-up feature ensures that the filtered historical patterns have the strongest relevance to the current driving segment. At the same time, by introducing multiple candidate clusters (rather than a single best match), a more robust reference data foundation is provided for the subsequent prediction step, for example, a weighted average can be performed based on the prediction results of multiple candidate clusters, thereby improving the accuracy and stability of the final predicted driving duration. Finally, the preset number of candidate driving cycle clusters will be used as the matching result to determine the target driving cycle cluster.

[0038] By calculating the similarity between the start-up feature and each historical start-up feature, a quantitative comparison between the current driving segment and the historical patterns is achieved; based on the similarity, the driving cycle clusters are sorted to form a cycle cluster sequence, establishing a clear priority relationship; by selecting a preset number of candidate driving cycle clusters from the sequence as the matching result, both the accuracy of the matching and the risk of overfitting are avoided through quantity control, making the finally determined target driving cycle cluster both representative and stable.

[0039] Step A3, determine the target driving cycle cluster from the historical driving cycle set based on the matching result.

[0040] Specifically, based on the matching result containing a preset number of candidate driving cycle clusters, the target driving cycle cluster is determined from the entire historical driving cycle set. In one implementation, the candidate driving cycle cluster with the highest similarity ranking (i.e., located at the top of the cycle cluster sequence) in the matching result can be directly determined as the target driving cycle cluster, as it represents the most similar historical pattern to the current driving segment. In another implementation, the target driving cycle cluster can be constructed as a composite model composed of the top K (K is the preset number) candidate driving cycle clusters, which will jointly calculate the historical driving characteristics of these clusters in subsequent prediction, for example, performing a weighted average of their prediction results, and the weights are determined according to the respective calculated similarity scores.

[0041] By obtaining the historical starting features of each driving cycle cluster in the historical driving cycle set, a complete reference benchmark is provided for subsequent matching. The current starting feature is matched with each historical starting feature, effectively identifying the most relevant historical mode for the current driving segment. Based on the matching result, the target driving cycle cluster is determined to ensure that the historical data relied on for prediction has high relevance, improving the accuracy and personalization of prediction.

[0042] In the embodiments of the present application, as shown in Figure 2 Before obtaining the historical starting features of each driving cycle cluster in the historical driving cycle set, the method further includes: Step S201, obtaining the time stamp of the state switching signal generated by the vehicle in the historical time period.

[0043] In the embodiments of the present application, when running locally at the vehicle end, all state switching signals generated by the vehicle in a preset historical time period (e.g. the past 30 days or the past 100 driving segments) are continuously monitored and recorded. The state switching signal refers to an event signal collected by the vehicle sensor network, which represents a fundamental change in the vehicle operating mode. Typical signal types include but are not limited to: switching from the off state to the start state (i.e. the vehicle is started), switching from the start state to the driving state (i.e. the vehicle starts moving), and switching from the driving state to the start or off state (i.e. the vehicle stops). Whenever such a signal is generated, the event is captured and a time stamp is recorded. The time stamp can be used as a time marker to record the minute of the state switching event. All state switching signals captured in the historical time period and their corresponding time stamps are stored as a raw event sequence arranged in chronological order.

[0044] Step S202, dividing the historical time period into multiple driving segments based on the time stamp of the state switching signal, and obtaining the starting data and driving data generated by the vehicle in each driving segment.

[0045] In the embodiments of the present application, based on the time stamp sequence of the state switching signal, each complete driving cycle is first identified. A typical driving cycle starts from the vehicle switching from the off state to the start state, then may enter the driving state, and finally ends from the driving or start state switching back to the off state. According to the key state switching point, the historical time period is preliminarily divided into multiple initial time periods. In the division process, the relationship between adjacent initial time periods is judged: when the interval length (such as the difference between the end time stamp of the previous period and the start time stamp of the next period) between adjacent initial time periods is less than a preset length threshold (such as 3 minutes), it is determined that this is a short interruption (such as waiting for a red light) in the driving segment, and thus the two initial time periods are combined into one driving segment; otherwise, when the interval length is greater than or equal to the length threshold, it is determined as two independent driving segments, and the initial time periods are respectively regarded as independent driving segments. For each finally determined driving segment, the start-up data thereof is extracted, including at least the departure time (converted to minutes from 0 o'clock on the same day) of the segment. In addition, the week information and season identifier of the driving segment can also be recorded as auxiliary features, which can be used for model training or data analysis. In step S203, the driving segments are clustered based on the start-up data and the driving data to obtain at least one driving cycle cluster, and the historical start-up features corresponding to each driving cycle cluster are generated according to the start-up data.

[0046] In the embodiments of the present application, based on the start-up data and the driving data (departure time and driving time) of all driving segments, the segment distance between each driving segment is first calculated to construct a distance matrix; a density-based clustering algorithm (such as DBSCAN) is used to determine the neighborhood radius eps parameter (such as taking the 90% quantile) by analyzing the distance distribution of all driving segments, and the minimum sample number min_samples is set to 1 / 10 (rounded down) of the sample number of the full data set, to determine the clustering parameters; the driving segments are divided into cluster samples and noise samples according to the clustering parameters; a corresponding driving cycle cluster is constructed for each cluster sample set, and the historical start-up features of the driving cycle cluster are generated by calculating the average value of the departure times of all driving segments in the cluster, which are used as the statistical representation of the driving mode. In addition, the number of samples in each driving cycle cluster, the feature variance, and other statistical information can also be recorded to evaluate the representativeness of the driving mode.

[0047] By obtaining the state switching signal time stamps in the historical time period, the original time sequence information of the vehicle use is recorded completely; the driving segments are divided based on the time stamps to obtain the start-up data and the driving data, and the continuous running data is converted into structured driving segment units; the driving cycle clusters are generated by clustering based on the start-up data and the driving data, and the historical start-up features are calculated, which realizes the automatic induction and feature extraction of the user driving mode and lays a solid data foundation for personalized prediction.

[0048] In the embodiments of the present application, the historical time period is divided into multiple driving segments based on the time stamp of the state switching signal, including: Case 1: The historical time period is divided into multiple initial time periods according to the time stamp of the state switching signal; the interval duration between adjacent initial time periods is identified; when the interval duration is less than the duration threshold, the adjacent initial time periods are merged into one driving segment.

[0049] Specifically, the present case aims to merge the driving segments that should belong to the same trip due to a short interruption, so as to restore the complete driving cycle. First, according to the time stamp of the state switching signal, the historical time period is roughly cut into multiple initial time periods, each of which usually starts when the vehicle enters the driving state and ends when the vehicle enters the engine-off state. Then, the interval duration between adjacent initial time periods is calculated, i.e. the time difference between the end time stamp of the previous time period and the start time stamp of the next time period. When it is identified that the interval duration is less than a preset duration threshold (for example, 3 minutes), it is determined that the interval is only a short parking (such as waiting for a red light or temporarily picking up and dropping off passengers), and is not the real end of the driving segment. Therefore, the two (or more, if the condition is continuously met) adjacent initial time periods are logically merged to form a longer driving segment. The departure time of the merged driving segment is taken from the start time of the first merged segment, and the driving duration is the sum of the driving durations of all the merged segments and the intermediate interval duration, so as to reflect the complete time consumption of the user's single trip.

[0050] Case 2: The historical time period is divided into multiple initial time periods according to the time stamp of the state switching signal; the interval duration between adjacent initial time periods is identified; when the interval duration is greater than or equal to the duration threshold, the adjacent initial time periods are respectively taken as one driving segment.

[0051] Specifically, the present case aims to identify and retain independent driving segments, and ensure that each analysis is based on a complete and independent trip. As in Case 1, multiple initial time periods are first generated according to the time stamp of the state switching signal. After calculating and identifying the interval duration between adjacent initial time periods, when the interval duration is greater than or equal to a preset duration threshold (for example, 3 minutes), it is determined that the interval represents a significant pause between two independent trips (for example, the vehicle has been turned off and the user has left the vehicle to handle business). This indicates that the previous initial time period represents a complete driving segment that has ended, and the next initial time period is the beginning of a new driving segment. Therefore, the two time periods will not be merged, but will be retained as two independent driving segments. Each independent driving segment will have its own departure time and driving duration, and will enter the subsequent clustering analysis as a separate sample, ensuring that each record in the historical driving database corresponds to an independent trip event.

[0052] The initial time period is divided according to the time stamp of the state switching signal, and the preliminary driving segment boundary is established; the time characteristics of the driving segment interruption are quantified by identifying the interval length between adjacent initial time periods; and the driving segments are intelligently merged or separated according to the comparison result of the interval length and the threshold value, which not only effectively filters short-time interference such as temporary parking, but also retains the integrity of the independent driving segments, ensuring the accuracy and representativeness of the data used for analysis.

[0053] In the embodiments of the present application, the driving segments are clustered based on the start-up data and the driving data to obtain at least one driving cycle cluster, including the following steps B1-B3: Step B1, calculating the segment distance between each driving segment based on the start-up data and the driving data.

[0054] Specifically, the start-up data and the driving data of the driving segment to be processed are obtained, that is, two core features of each segment: the departure time (tstart) ) and the driving duration (tduration ). The two feature values of each driving segment are regarded as coordinate points in a two-dimensional space. Then, the segment distance between each two different driving segments in the two-dimensional feature space is calculated. The segment distance is usually calculated by using the Euclidean distance, that is: For any two driving segments i and j, the distance is:

[0055] wherein, is the segment distance between the driving segment i and the driving segment j, is the departure time of the driving segment i, is the departure time of the driving segment j, is the driving duration of the driving segment i, is the driving duration of the driving segment j.

[0056] The smaller the segment distance is, the more similar the two driving segments are in the departure time and the total time consumption, and the more likely they belong to the same commuting mode (for example, both are driving segments that depart at 8 o'clock in the morning on weekdays and take about 30 minutes); on the contrary, the larger the segment distance is, the greater the mode difference between the two driving segments. By calculating the segment distance between all driving segments, a distance matrix is constructed, which describes the mutual relationship between all driving segment samples in the data set and is the core input driving the clustering algorithm.

[0057] Step B2, determining the clustering parameters according to the segment distance between the driving segments, and dividing the driving segments into multiple cluster samples according to the clustering parameters.

[0058] Specifically, based on the segment distance matrix between all driving segments, first determine the clustering parameters required by the clustering algorithm (such as DBSCAN), including at least the neighborhood radius eps and the minimum sample number min_samples. Among them, the eps parameter is determined by analyzing the distribution of the "third nearest distance" of all driving segments, specifically calculating the 90th percentile of the sorted segment distance, this method can adapt to the data distribution density; and the min_samples parameter is set to 1 / 10 (rounded down) of the total data set sample number, so that the minimum sample number required to form an effective cluster can be adjusted adaptively with the size of the data set, avoiding the small data set from being unable to form an effective cluster due to the threshold being too high. After determining the clustering parameters, run the clustering algorithm, judge whether each driving segment is a core point, a boundary point or a noise point based on eps and min_samples, so as to divide all driving segments into different clusters or mark them as noise. Finally, all driving segments divided into clusters are marked as cluster samples, and those not belonging to any cluster are regarded as noise samples and excluded.

[0059] As an example, assuming there are n=100 driving segments, and the driving segment is a two-dimensional feature vector (including departure time and driving duration). First, calculate the segment distance of each driving segment to all other driving segments (a total of 99 distance values). For each driving segment i, sort the segment distances of the driving segment i and other driving segments from small to large, and take the 3rd value (i.e. the third nearest distance), for example, the segment distance list of driving segment i is sorted as [0.2, 0.5, 0.8, 1.3,...], and its third nearest distance is 0.8. Here K=4 is a preset empirical parameter, and selecting the third nearest distance can exclude the noise interference of the nearest neighbor and better reflect the data distribution characteristics. Ensure n≥4 to meet the calculation condition. Next, aggregate the third nearest distances of all 100 driving segments to form a list containing 100 values (for example, [0.6, 0.8, 0.5, 1.2,...]). Reorder the list from small to large to get an ordered list (for example, [0.5, 0.6, 0.8, 1.2,...]). Then, calculate the 90th percentile (i.e. the 90th percentile value) of the sorted list, assuming the value is 1.0, then set the eps parameter to 1.0. At the same time, the min_samples parameter is calculated as floor(100 / 10)=10. Finally, use the DBSCAN clustering algorithm with eps=1.0 and min_samples=10 as parameters to divide the driving segments into cluster samples and noise samples.

[0060] Step B3, using the cluster samples to construct at least one driving cycle cluster.

[0061] Specifically, a corresponding driving cycle cluster is created for each independent cluster sample. This driving cycle cluster is not only a list of driving segments, but also a statistically meaningful model entity. In the construction process, the statistical central values of the start data (such as the departure time) and the driving data (such as the driving duration) of all driving segments in the cluster are calculated, for example, the average value of the departure time and the average value of the driving duration are calculated respectively. These two statistical values together constitute the historical start feature of the driving cycle cluster, which is the pattern representation of the entire cluster. Finally, the multiple driving cycle clusters constructed (each cluster represents a repeated driving pattern, such as early morning rush hour commuting or weekend shopping) are included in the historical driving cycle set. This set is the personalized model library for driving segment time consumption prediction, which enables driving duration prediction without navigation based on the user's own historical behavior data.

[0062] The segment distance between driving segments is calculated based on start data and driving data, which quantifies the similarity difference between different driving segments; the clustering parameters are determined according to the segment distance distribution, so that the algorithm can adapt to different sizes and qualities of data sets; the driving cycle cluster is constructed using cluster samples, which classifies driving segments with similar features into a unified driving mode, forming a driving cycle cluster set with clear statistical characteristics, providing a reliable pattern library for personalized prediction.

[0063] In the embodiments of the present application, the driving duration of the current driving cycle is predicted based on the start feature and the target driving cycle cluster, to obtain the predicted driving duration, including the following steps C1-C3: Step C1, obtaining the driving data of each driving segment in the target driving cycle cluster.

[0064] Specifically, after determining the target driving cycle cluster most relevant to the current driving segment, the cluster sample corresponding to the target driving cycle cluster is accessed. The driving data corresponding to each driving segment in the cluster sample is obtained one by one. The driving data specifically refers to the core feature calculated and stored in the historical record of each driving segment: driving duration ( ), that is, the total number of minutes actually consumed from the start to the end of the historical driving segment. Through this operation, the mode concept of the target driving cycle cluster is restored to the collection of all specific and real historical driving segment time consumption data, providing the original data basis for statistical feature extraction.

[0065] Step C2, generating the historical driving feature of the target driving cycle cluster using the driving data.

[0066] Specifically, statistical analysis is performed on the driving data of all driving segments belonging to the target driving cycle cluster to generate the historical driving feature of the cluster. The historical driving feature is a scalar value representing the typical time consumption of the target driving cycle cluster, which is usually obtained by calculating the arithmetic mean of the driving durations of all driving segments in the cluster. For example, if the target driving cycle cluster contains 5 historical driving segments with driving durations of 28 minutes, 30 minutes, 32 minutes, 29 minutes, and 31 minutes, the calculated historical driving feature (average driving duration) is 30 minutes. The historical driving feature represents the expected or typical driving segment time consumption under a specific driving mode (e.g., "commuting on weekdays at 8 am"), which will be used as a reference value to combine with the start-up feature of the current driving segment to calculate the predicted driving duration of the current driving segment.

[0067] Step C3, based on the start-up feature and the historical driving feature, calculate the predicted driving duration of the current driving cycle.

[0068] Specifically, the start-up feature of the current driving cycle (e.g., the departure time Tstart) and the historical driving feature of the target driving cycle cluster (e.g., the driving duration of the typical feature of the cluster) are used as inputs to calculate the predicted driving duration of the current driving cycle through a predefined or trained calculation model.

[0069] In one embodiment, the historical driving feature (average driving duration) of the target driving cycle cluster is directly used as the output of the predicted driving duration. It is determined that the current driving segment is highly similar to the target historical mode, and therefore its time consumption is expected to be consistent with the average time consumption of the mode.

[0070] In another embodiment, a linear regression model is constructed, which takes the departure time in the start-up feature as input and uses the historical driving feature as a reference or bias term. The model is quickly fitted on the internal data of the target driving cycle cluster, and the specific departure time is used to fine-tune the reference predicted value. The output is the calculated predicted driving duration, which can be directly used to trigger the corresponding battery thermal management optimization strategy to achieve precise control of energy consumption.

[0071] In another embodiment, a pre-trained prediction model is constructed, which is trained on the data of historical driving cycle clusters using machine learning algorithms such as decision tree, random forest, LightGBM, etc. When predicting, the departure time of the current driving cycle is input into the pre-trained model, and the model outputs the predicted driving duration. The specific training process of the prediction model includes: obtaining a commuting data set Dcluster from the historical driving cycle clusters, each sample containing a departure time Tstart and a driving duration Tdrive; stratified sampling the driving cycle clusters obtained after clustering according to the week, dividing them into a training set Dtrain and a validation set Dvalidation, with a sample size ratio of 8:2, to ensure the consistency of data distribution in different time periods; training a regression model M based on supervised learning (such as decision tree, RandomForest, LightGBM, etc.) using the training set Dtrain, with the model input being the departure time Tstart and the output being the driving duration Tdrive; during the training process, the model continuously adjusts the parameters to minimize the difference between the predicted value and the true value through optimization algorithms such as gradient descent; the validation set Dvalidation is used to verify the trained model M, and evaluation indicators such as root mean square error (RMSE) and coefficient of determination (R²) are used to evaluate the performance of the model; the GridSearch method is used to optimize the model hyperparameters, and the hyperparameters to be optimized are arranged and combined within the value range, and each combination is trained and validated to finally select the hyperparameter combination with the best validation performance (i.e., the optimal evaluation indicator); the optimal hyperparameter combination is used to retrain the model on the entire commuting data set to obtain the final prediction model for real-time prediction tasks.

[0072] As an example, as shown in Figure 3 The model construction and prediction process includes: first, obtaining vehicle historical driving data through the data acquisition and storage link to form a full driving cycle data set (commuting data set); then using the DBSCAN clustering algorithm to discover patterns in the full data set, generating a commuting data set representing different driving patterns; then training the model based on the commuting data set, dividing the data set into a training set and a validation set, combining the GridSearch hyperparameter optimization method, iteratively optimizing the model parameters, and evaluating the model performance through the validation set; finally, applying the trained optimal model to real-time prediction to output the prediction results.

[0073] By acquiring driving data from each driving segment within the target driving cycle cluster, complete historical time consumption information for this driving mode was obtained. Historical driving features were generated using the driving data, and typical time consumption patterns of this mode were extracted. Based on the current start features and historical driving features, the predicted driving time was calculated. By organically combining real-time driving segment features with historical mode features, the individual characteristics of the current driving segment were considered, while the statistical advantages of historical patterns were incorporated, thereby generating highly accurate and personalized prediction results.

[0074] This embodiment also provides a vehicle travel time prediction device, which is used to implement the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0075] This embodiment provides a vehicle travel time prediction device, such as... Figure 4 As shown, it includes: Module 41 is used to acquire the vehicle's start-up characteristics in the current driving cycle; Filtering module 42 is used to filter out target driving cycle clusters associated with startup features from driving cycle clusters in the historical driving cycle set; The prediction module 43 is used to predict the driving time of the current driving cycle based on the starting features and the target driving cycle cluster, so as to obtain the predicted driving time.

[0076] In this embodiment of the application, the screening module 42 includes: The acquisition submodule is used to acquire the historical startup characteristics of each driving cycle cluster in the historical driving cycle set; The matching submodule is used to match the startup features with each historical startup feature to obtain the matching results; The determination submodule is used to determine the target driving cycle cluster from the historical driving cycle set based on the matching results.

[0077] In this embodiment, the matching submodule is specifically used to calculate the similarity between the startup features and the historical startup features of each driving cycle cluster; sort the driving cycle clusters based on the similarity to obtain a cycle cluster sequence; and select a preset number of candidate driving cycle clusters from the cycle cluster sequence as the matching result.

[0078] In this embodiment of the application, the apparatus further includes: The first processing module is used to obtain the timestamp of the state switching signal of the vehicle within the historical time period; The second processing module is used to divide the historical time period into multiple driving segments based on the timestamp of the state switching signal, and to acquire the start-up data and driving data generated by the vehicle in each driving segment. The third processing module is used to cluster driving segments based on startup data and driving data to obtain at least one driving cycle cluster, and generate historical startup features corresponding to each driving cycle cluster based on startup data.

[0079] In this embodiment of the application, the second processing module is specifically used to divide the historical time period into multiple initial time periods according to the timestamp of the state switching signal; identify the interval length between adjacent initial time periods; when the interval length is less than the duration threshold, merge the adjacent initial time periods into one driving segment; or, when the interval length is greater than or equal to the duration threshold, treat the adjacent initial time periods as one driving segment respectively.

[0080] In this embodiment of the application, the third processing module is specifically used to calculate the segment distance between each driving segment based on the startup data and driving data; determine the clustering parameters according to the segment distance between the driving segments, and divide the driving segments into multiple cluster samples according to the clustering parameters; and construct at least one driving cycle cluster using the cluster samples.

[0081] In this embodiment of the application, the prediction module 43 is specifically used to acquire driving data of each driving segment in the target driving cycle cluster; generate historical driving features of the target driving cycle cluster using the driving data; and calculate the predicted driving time of the current driving cycle based on the start features and historical driving features.

[0082] Please see Figure 5 , Figure 5 This is a schematic diagram of the structure of a computer device provided in an optional embodiment of the present invention, such as... Figure 5 As shown, the computer device includes one or more processors 10, memory 20, and interfaces for connecting the components, including high-speed interfaces and low-speed interfaces. The components communicate with each other via different buses and can be mounted on a common motherboard or otherwise installed as needed. The processors can process instructions executed within the computer device, including instructions stored in or on memory to display graphical information of a GUI on external input / output devices (such as display devices coupled to the interfaces). In some alternative implementations, multiple processors and / or multiple buses can be used with multiple memories and multiple memory modules, if desired. Similarly, multiple computer devices can be connected, each providing some of the necessary operations (e.g., as a server array, a group of blade servers, or a multiprocessor system).

[0083] Processor 10 may be a central processing unit, a network processor, or a combination thereof. Processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The programmable logic device may be a complex programmable logic device (CAMP), a field-programmable gate array (FPGA), a general-purpose array logic (GPA), or any combination thereof.

[0084] The memory 20 stores instructions executable by at least one processor 10 to cause at least one processor 10 to perform the method shown in the above embodiments.

[0085] The memory 20 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the computer device as shown by a landing page for an app. Furthermore, the memory 20 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some alternative embodiments, the memory 20 may optionally include memory remotely located relative to the processor 10, which can be connected to the computer device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0086] The memory 20 may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as flash memory, hard disk or solid-state drive; the memory 20 may also include a combination of the above types of memory.

[0087] The computer device also includes a communication interface 30 for communicating with other devices or communication networks.

[0088] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code, which, when accessed and executed by the computer, processor, or hardware, implements the methods shown in the above embodiments.

[0089] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method for predicting vehicle travel time, characterized in that, The method includes: Obtain the vehicle's startup characteristics in the current driving cycle; Filter out the target driving cycle clusters associated with the activation features from the driving cycle clusters of the historical driving cycle set; Based on the starting characteristics and the target driving cycle cluster, the driving time of the current driving cycle is predicted to obtain the predicted driving time.

2. The method according to claim 1, characterized in that, The step of selecting the target driving cycle cluster associated with the activation feature from the driving cycle clusters of the historical driving cycle set includes: Obtain the historical startup characteristics of each driving cycle cluster in the historical driving cycle set; The startup feature is matched with each of the historical startup features to obtain the matching results; The target driving cycle cluster is determined from the historical driving cycle set based on the matching results.

3. The method according to claim 2, characterized in that, The step of matching the startup feature with each of the historical startup features to obtain the matching result includes: Calculate the similarity between the startup features and the historical startup features of each of the driving cycle clusters; The driving cycle clusters are sorted based on the similarity to obtain a cycle cluster sequence; A preset number of candidate driving cycle clusters are selected from the cycle cluster sequence as the matching results.

4. The method according to claim 2, characterized in that, Before obtaining the historical startup characteristics of each driving cycle cluster in the historical driving cycle set, the method further includes: Obtain the timestamp of the state transition signal of the vehicle within the historical time period; Based on the timestamp of the state switching signal, the historical time period is divided into multiple driving segments, and the start-up data and driving data generated by the vehicle in each driving segment are obtained. Based on the startup data and the driving data, the driving segments are clustered to obtain at least one driving cycle cluster, and historical startup features corresponding to each driving cycle cluster are generated according to the startup data.

5. The method according to claim 4, characterized in that, The historical time period is divided into multiple driving segments based on the timestamp of the state switching signal, including: The historical time period is divided into multiple initial time periods based on the timestamp of the state switching signal; Identify the duration of the interval between adjacent initial time periods; When the interval duration is less than the duration threshold, the adjacent initial time periods are merged into one driving segment; or, when the interval duration is greater than or equal to the duration threshold, the adjacent initial time periods are each treated as a driving segment.

6. The method according to claim 4, characterized in that, The process of clustering the driving segments based on the startup data and the driving data to obtain at least one driving cycle cluster includes: Calculate the segment distance between each of the driving segments based on the startup data and the driving data; Clustering parameters are determined based on the segment distances between the driving segments, and the driving segments are divided into multiple cluster samples based on the clustering parameters; At least one driving cycle cluster is constructed using the cluster sample.

7. The method according to claim 1, characterized in that, The step of predicting the driving time of the current driving cycle based on the starting characteristics and the target driving cycle cluster to obtain the predicted driving time includes: Obtain driving data for each driving segment in the target driving cycle cluster; The driving data is used to generate historical driving characteristics of the target driving cycle cluster; Based on the startup characteristics and the historical driving characteristics, the predicted driving time of the current driving cycle is calculated.

8. A vehicle travel time prediction device, characterized in that, The device includes: The acquisition module is used to acquire the vehicle's startup characteristics in the current driving cycle; A filtering module is used to filter out target driving cycle clusters associated with the activation feature from driving cycle clusters in the historical driving cycle set; The prediction module is used to predict the driving time of the current driving cycle based on the starting characteristics and the target driving cycle cluster, so as to obtain the predicted driving time.

9. A computer device, characterized in that, include: A memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, the processor executing the computer instructions to perform the method of any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to perform the method of any one of claims 1 to 7.

Images