Method and apparatus for determining charging strategy based on vehicle behavior segment

Abstract

The application discloses a kind of based on the determination method and device of charging strategy of vehicle behavior fragment, it is related to new energy automobile technical field, wherein, the method comprises: based on target fragment dataset determines the destination of vehicle, charging place list, commuting path list, charging event list, charging control parameter, user driving behavior characteristics, risk event list, wherein, destination includes home place and work place;X triplets are generated based on target fragment dataset and charging event list;based on destination, charging place list, commuting path list, charging event list, charging control parameter, user driving behavior characteristics, risk event list, the mode label corresponding to X triplets is determined with preset electricity price data Target charging strategy of vehicle.The application solves the technical problems that the analysis effect of the operating data of new energy vehicles in the prior art is poor and the recommended charging strategy is inaccurate.

Description

Technical Field

[0001] This application relates to the field of new energy vehicle technology, and more specifically, to a method and apparatus for determining a charging strategy based on vehicle behavior segments. Background Technology

[0002] With the popularization of electric vehicles and the development of vehicle networking technology, the collection and analysis of vehicle operation data has become the foundation for applications such as user profiling, charging planning, and battery health management. Existing vehicle behavior analysis technologies include: analysis technology based on complete trajectory data, charging behavior analysis technology based on charging pile data, user profile generation technology based on statistics, and behavior prediction based on machine learning.

[0003] However, existing vehicle behavior analysis technologies ignore the data quality defects in the original segments uploaded by the actual vehicle. Furthermore, existing technologies often treat dwell time identification, charging identification, and commuting identification as parallel modules, lacking a constraint mechanism for subsequent steps based on previous results. They also lack modeling of user behavior, battery risk, charging strategies, and the relationships between the three. In addition, existing charging strategy recommendations are mostly based on grid load or electricity price, without considering the regularity of user behavior, resulting in low user acceptance and low actual adoption rate of the recommended strategies.

[0004] There is currently no effective solution to the technical problems of ineffective analysis of operational data of new energy vehicles and inaccurate recommended charging strategies. Summary of the Invention

[0005] This application provides a method and apparatus for determining a charging strategy based on vehicle behavior segments, in order to at least solve the technical problems of poor analysis of operating data of new energy vehicles and inaccurate recommended charging strategies in the prior art.

[0006] According to one aspect of this application, a method for determining a charging strategy based on vehicle behavior segments is provided, comprising: weighted summing of L segment scores for each standard segment in a standard segment dataset to obtain a confidence score for each standard segment, wherein L is a positive integer, and the L segment scores include a time continuity score, a physical consistency score, a field integrity score, and a geographical stability score, and each standard segment is used to record vehicle behavior data; removing standard segments in the standard segment dataset whose confidence scores are less than a preset confidence threshold to obtain a target segment dataset, wherein each target segment in the target segment dataset is one of the following: a charging segment, a driving segment, or a parking segment; and determining the vehicle based on the target segment dataset. The dataset includes the destination, charging location list, commuting route list, charging event list, charging control parameters, user driving behavior characteristics, and risk event list. The destination includes both home and workplace locations. Based on the target segment dataset and the charging event list, X triples are generated, where X is a positive integer. Each triple records the correspondence between a valid charging segment, a preceding parking segment, and a subsequent driving segment in the charging event list, and each triple corresponds to a pattern label. The target charging strategy for the vehicle is determined based on the destination, charging location list, commuting route list, charging event list, charging control parameters, user driving behavior characteristics, risk event list, the pattern labels corresponding to the X triples, and preset electricity price data.

[0007] Optionally, determining the vehicle's destination, charging location list, commuting route list, charging event list, charging control parameters, user driving behavior characteristics, and risk event list based on the target segment dataset includes: after determining the dwell location list based on parking segments in the target segment dataset, determining the destination point based on the time label table corresponding to the target segment dataset and the dwell location list, wherein the time label table includes derived time labels for each target segment; after clustering the charging locations of all charging segments in the target segment dataset, generating a charging location list based on the cluster center coordinates of each cluster of charging locations obtained from the clustering; determining the commuting route list based on driving segments and destination points in the target segment dataset; determining the charging event list and charging control parameters based on the charging segments in the target segment dataset; determining user driving behavior characteristics based on driving segments in the target segment dataset; and determining the risk event list based on the target segment dataset, destination points, charging location list, and preset risk fields.

[0008] Optionally, after determining the list of rest stops based on the parking segments in the target segment dataset, the destination point is determined based on the time label table and the list of rest stops corresponding to the target segment dataset. This includes: using the endpoint of each parking segment in the target segment dataset as a candidate rest stop, and calculating the kernel density weight of each parking segment centered on each candidate rest stop; performing spatial clustering of the candidate rest stops for all parking segments based on the kernel density weight of each parking segment to obtain A rest stop clusters, where A is a positive integer, and each rest stop cluster includes at least two candidate rest stops; calculating the location cluster index for each of the A rest stop clusters, where the location cluster index includes the cluster center coordinates, access frequency, average dwell time, and dwell time standard deviation; performing cluster filtering on the A rest stop clusters based on the location cluster index of each rest stop cluster, where cluster filtering is used to remove rest stop clusters whose location cluster index is less than a preset index threshold; and selecting the rest stop clusters that pass the cluster filtering. The cluster center coordinates are used as the dwelling locations. All dwelling locations are aggregated to obtain a dwelling location list. Based on the time label table corresponding to the target segment dataset and the location cluster index of the dwelling location cluster to which each dwelling location belongs in the dwelling location list, P location features are determined for each dwelling location, where P is a positive integer. The P location features include the proportion of nighttime stays, the proportion of weekday daytime stays, the normalized value of the average stay duration, the ratio of consecutive stay days to total stay days, and the proportion of times it is used as a departure point. Based on the P location features of each dwelling location, a first location score and a second location score are determined for each dwelling location. The first location score is used to represent the probability that each dwelling location is a home location, and the second location score is used to represent the probability that each dwelling location is a work location. The dwelling location with the largest first location score in the dwelling location list is designated as the home location, and the dwelling location with the largest second location score in the dwelling location list is designated as the work location. The home location and the work location are designated as the destination points.

[0009] Optionally, a commuter route list is determined based on the driving segments and destinations in the target segment dataset, including: mapping the starting point and ending point of each driving segment in the target segment dataset to a spatial grid of a preset scale to obtain X spatial paths corresponding to X driving segments, where X is a positive integer; dividing the X spatial paths into Y path sets, where Y is a positive integer less than or equal to X, and the starting points of any two spatial paths in each path set are located in the same spatial grid; calculating the path index for each path set in the Y path sets, where the path index includes frequency of occurrence, average distance, average duration, and average speed; determining the commuter score for each path set based on the path index for each path set; taking any spatial path in each path set with a commuter score greater than or equal to a preset commuter threshold as a commuter path, and summarizing the Y commuter paths corresponding to the Y path sets to obtain a commuter route list.

[0010] Optionally, the charging event list and charging control parameters are determined based on the charging segments in the target segment dataset, including: statistically analyzing the charging indicators for each charging segment in the target segment dataset, wherein the charging indicators include the vehicle's state of charge, power information, and charging period; determining the charging confidence level for each charging segment based on the charging indicators for each charging segment; identifying charging segments with a charging confidence level greater than or equal to a preset confidence threshold as valid charging segments, and summarizing all valid charging segments to obtain a charging event list; clustering the charging start times corresponding to all valid charging segments in the charging event list to obtain typical charging periods for the vehicle, wherein each typical charging period is the cluster center period of the clustered effective charging period clusters; determining the typical starting charging charge value and typical ending charging charge value for the vehicle based on the median of the state of charge corresponding to all valid charging segments in the charging event list; determining the ratio of fast to slow charging for the vehicle based on the power information corresponding to all valid charging segments in the charging event list; and using the typical charging period, typical starting charging charge value, typical ending charging charge value, and fast to slow charging ratio for the vehicle as charging control parameters.

[0011] Optionally, user driving behavior characteristics are determined based on driving segments in the target segment dataset, and a risk event list is determined based on the target segment dataset, destinations, charging location lists, and preset risk fields. This includes: dividing all driving segments in the target segment dataset into a first segment set and a second segment set based on a preset lower charge limit, wherein the vehicle's charge value in each driving segment in the first segment set is less than the preset lower charge limit at the start and / or end of driving, and the vehicle's charge value in each driving segment in the second segment set is greater than or equal to the preset lower charge limit at both the start and end of driving; determining the vehicle's driving behavior deviation index based on the driving indexes of the first and second segment sets, and based on the driving behavior... The system determines user driving behavior characteristics based on the offset index, where driving indicators include average speed, average distance, the percentage of trips ending at home / charging location, and the percentage of long-distance trips. Target segments in the target segment dataset that hit preset risk fields are designated as risk segments. For each risk segment, the associated scenarios are determined based on the list of destinations and charging locations, where the associated scenarios are one of the following: home, workplace, or charging location. A risk score is determined for each risk segment based on its associated scenarios, including pressure difference risk score, temperature difference risk score, power risk score, and recurrence risk score. Risk events are generated based on risk segments with risk scores greater than or equal to preset risk thresholds and their associated scenarios. All risk events are then aggregated to obtain a risk event list.

[0012] Optionally, X triples are generated based on the target segment dataset and the charging event list, including: determining the preceding parking segment and the following driving segment corresponding to each valid charging segment in the charging event list based on the target segment dataset, wherein the duration between the charging start time of each valid charging segment and the parking end start time of the corresponding preceding parking segment is less than a preset duration, and the duration between the charging end time of each valid charging segment and the driving start time of the corresponding following driving segment is less than a preset duration; associating each valid charging segment with the corresponding preceding parking segment and the following driving segment to obtain X triples, and determining the pattern label corresponding to each triple based on the association strength and temporal characteristics of each triple, wherein the pattern label is one of the following: charging while parking, unplugging and leaving, charging overnight at home, and replenishing power at work.

[0013] Optionally, the target charging strategy for the vehicle is determined based on the destination, charging location list, commuting route list, charging event list, charging control parameters, user driving behavior characteristics, risk event list, pattern labels corresponding to X triples, and preset electricity price data. This includes: determining the vehicle's reservation charging suitability based on the destination, charging location list, charging event list, user driving behavior characteristics, and pattern labels corresponding to X triples; if the vehicle's reservation charging suitability is greater than or equal to a preset suitability threshold, obtaining B candidate charging strategies, where B is a positive integer; determining the strategy benefit of each candidate charging strategy based on the commuting route list, charging control parameters, user driving behavior characteristics, risk event list, pattern labels corresponding to X triples, and preset electricity price data, where the strategy benefit is used to characterize off-peak electricity revenue, vehicle battery charge health, departure guarantee, and user disturbance cost; and selecting the candidate charging strategy with the highest strategy benefit as the target charging strategy.

[0014] Optionally, after determining the target charging strategy for the vehicle based on the destination, charging location list, commuting route list, charging event list, charging control parameters, user driving behavior characteristics, risk event list, pattern labels corresponding to X triples, and preset electricity price data, the method for determining the charging strategy based on vehicle behavior segments further includes: determining the evidence chain confidence of the target charging strategy based on the evidence chain used in the determination process, and determining the strategy output level of the target charging strategy based on the evidence chain confidence, wherein the evidence chain includes confidence score, location score, commuting score, risk score, and charging confidence; and / or, performing conflict detection on the determination process of the target charging strategy, and if there are conflicting operations in the determination process, performing a rollback operation on the steps with conflicting operations, and downgrading the strategy output level of the target charging strategy.

[0015] According to another aspect of this application, a device for determining a charging strategy based on vehicle behavior segments is also provided, comprising: a credibility detection unit, configured to perform a weighted summation of L segment scores for each standard segment in a standard segment dataset to obtain a credibility score for each standard segment, wherein L is a positive integer, and the L segment scores include a time continuity score, a physical consistency score, a field integrity score, and a geographical stability score, and each standard segment is used to record vehicle behavior data; a segment elimination unit, configured to eliminate standard segments in the standard segment dataset whose credibility scores are less than a preset credibility threshold to obtain a target segment dataset, wherein each target segment in the target segment dataset is one of the following: a charging segment, a driving segment, or a parking segment; and a first determination unit, configured to determine the target segment data based on the target segment data. The system consists of three parts: a first part, a second part, and a third part. The first part defines the vehicle's destination, a list of charging locations, a list of commuting routes, a list of charging events, charging control parameters, user driving behavior characteristics, and a list of risk events. The destination includes both home and workplace locations. The second part generates X triples based on the target segment dataset and the charging event list, where X is a positive integer. Each triple records the correspondence between a valid charging segment, a preceding parking segment, and a subsequent driving segment in the charging event list, and each triple corresponds to a pattern label. The third part determines the vehicle's target charging strategy based on the destination, charging location list, commuting route list, charging event list, charging control parameters, user driving behavior characteristics, risk event list, the pattern labels corresponding to the X triples, and preset electricity price data.

[0016] According to another aspect of this application, a computer program product is also provided, which stores a computer program, wherein a method for determining a charging strategy based on vehicle behavior segments, which controls the computer program product to execute any of the above-mentioned methods, is provided when the computer program is running.

[0017] According to another aspect of this application, an electronic device is also provided, wherein the electronic device includes one or more processors and a memory for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method for determining a charging strategy based on vehicle behavior segments as described above.

[0018] In this application, firstly, the L segment scores of each standard segment in the standard segment dataset are weighted and summed to obtain the credibility score of each standard segment, where L is a positive integer. The L segment scores include time continuity score, physical consistency score, field integrity score, and geographical stability score. Each standard segment is used to record vehicle behavior data. Then, standard segments with credibility scores lower than a preset credibility threshold in the standard segment dataset are removed to obtain the target segment dataset. Each target segment in the target segment dataset is one of the following: charging segment, driving segment, or parking segment. Subsequently, the vehicle's destination, charging location list, and commuting information are determined based on the target segment dataset. The dataset includes a route list, a charging event list, charging control parameters, user driving behavior characteristics, and a risk event list. Destination points include home and work locations. Then, based on the target segment dataset and the charging event list, X triples are generated, where X is a positive integer. Each triple records the correspondence between a valid charging segment, a preceding parking segment, and a subsequent driving segment in the charging event list. Each triple corresponds to a pattern label. Subsequently, based on the destination points, charging location list, commuting route list, charging event list, charging control parameters, user driving behavior characteristics, risk event list, the pattern labels corresponding to the X triples, and preset electricity price data, the vehicle's target charging strategy is determined.

[0019] As described above, this application employs a multi-stage behavioral reasoning and credibility correction approach. First, it calculates credibility by weighted summation of scores for the temporal continuity, physical consistency, field completeness, and geographical stability of standard segments, and then removes low-credibility segments to obtain the target segment dataset. Based on this dataset, it generates X triples containing the correspondence between charging, driving, and parking segments. It then integrates multi-dimensional information including destination points, charging location lists, commuting route lists, charging event lists, charging control parameters, user driving behavior characteristics, risk event lists, and pattern labels, combined with preset electricity price data, to determine the vehicle's target charging strategy. This achieves the goal of full-process correlation modeling from data cleaning and behavioral pattern recognition to strategy generation, even in the context of quality defects in vehicle operation data and error propagation from independent processing of operation data by different reasoning modules. This improves the accuracy of operation data analysis, outputs interpretable vehicle charging strategies, and increases user adoption rates of recommended charging strategies, thus solving the technical problems of poor analysis of new energy vehicle operation data and inaccurate recommended charging strategies in existing technologies. Attached Figure Description

[0020] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0021] Figure 1 This is a flowchart of an optional method for determining a charging strategy based on vehicle behavior segments according to an embodiment of this application;

[0022] Figure 2 This is a flowchart of an optional method for determining the credibility score and level of a segment according to an embodiment of this application;

[0023] Figure 3 This is a flowchart of an optional method for determining a location type according to an embodiment of this application;

[0024] Figure 4 This is a flowchart of an optional vehicle behavior analysis method based on fragment-level runtime data according to an embodiment of this application;

[0025] Figure 5 This is an architecture diagram of an optional vehicle behavior analysis system based on fragment-level operational data according to an embodiment of this application;

[0026] Figure 6 This is a schematic diagram of an optional charging strategy determination device based on vehicle behavior segments according to an embodiment of this application;

[0027] Figure 7 This is a structural block diagram of an electronic device according to an embodiment of this application. Detailed Implementation

[0028] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0029] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0030] It should also be noted that all information and data (including but not limited to information used for display and analysis) involved in this application are authorized by the user or fully authorized by all parties. For example, if there is an interface between this system and the relevant user or organization, before obtaining the relevant information, it is necessary to send a request to the aforementioned user or organization through the interface, and obtain the relevant information only after receiving consent from the aforementioned user or organization.

[0031] Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of relevant information and data involved in this application all comply with the relevant laws, regulations, and standards of the relevant regions, and necessary security measures have been taken. They do not violate public order and good morals. In addition, this application provides corresponding operation entry points for users to choose to agree to authorization or refuse authorization. If the user chooses to refuse authorization, the corresponding expert decision-making process will be initiated.

[0032] In one optional related technical embodiment, a method for analyzing vehicle operation data based on complete trajectory data is provided. This method uses GPS (Global Positioning System) frame-by-frame data to reconstruct the user's vehicle travel trajectory, identify frequently visited locations and travel patterns. This type of method relies on high-frequency data acquisition (usually above 1Hz), which results in large data volume, high transmission costs, and high requirements for vehicle terminals and communication networks.

[0033] In one optional related technical embodiment, a charging behavior analysis method based on charging pile data is provided. This method analyzes user charging habits through charging pile records. However, this type of method can only obtain data in the charging process and cannot fully reflect user vehicle usage behavior. Furthermore, it relies on data access from charging infrastructure.

[0034] In one optional related technical embodiment, a statistical user profile generation method is provided. This method generates user tags by performing statistical analysis on fragment-level data uploaded by vehicles. Such methods are mostly parallel analysis architectures, with each analysis module processing independently. They lack the constraint of previous results on subsequent steps, and errors are easily propagated.

[0035] In one alternative related technical embodiment, a machine learning-based behavior prediction method is provided. This method uses a deep learning model to predict user travel and charging behavior. However, this type of method relies on a large amount of labeled data, has high model training costs, poor interpretability, and is difficult to promote and apply in engineering practice.

[0036] In summary, the above embodiments have the following technical defects:

[0037] (1) The assumptions about data credibility are too idealistic: The above schemes mostly assume that the input data is complete and accurate, but the actual fragment-level data uploaded by the vehicle end has the following quality defects:

[0038] Missed segment: Some driving or charging segments were not correctly identified; Incorrect segment: The segment category was incorrectly determined (e.g., parking was mistakenly identified as driving); GPS drift: There is a large error in the location information; SOC (State of Charge) jump: The battery state of charge data jumps abnormally; Field missing: Some key fields are empty or abnormal.

[0039] In traditional techniques, directly analyzing data with the aforementioned quality defects would lead to unreliable results.

[0040] (2) The analysis method is a parallel architecture: The above schemes mostly treat dwell identification, charging identification, commuting identification, etc. as parallel modules and process them independently, lacking a constraint mechanism for subsequent steps based on previous results, for example:

[0041] The results of the dwell point identification are not used to constrain the charging point identification; the commuting route identification does not consider the location inference results; the errors of each module accumulate independently and cannot be mutually corrected.

[0042] The parallel architecture described above makes it easy for errors to propagate, resulting in poor overall inference stability.

[0043] (3) Lack of interpretability of output: The above scheme directly outputs user labels or conclusions, lacks confidence assessment and evidence chain display. Under fragment-level data conditions, users and engineers find it difficult to judge the reliability of the conclusions, which is not conducive to manual review and subsequent optimization.

[0044] (4) Disconnect between behavior and risk / strategy: Existing solutions analyze user behavior, battery risk, and charging strategy in isolation, lacking modeling of the relationship between the three. For example, charging habit analysis does not consider battery health risks, risk assessment is not related to specific usage scenarios, and charging strategy recommendations do not consider the regularity of user behavior. As a result, the engineering application value is limited and it is difficult to use it directly for battery design optimization and charging function development.

[0045] (5) Lack of judgment on the suitability of scheduled charging: Traditional charging recommendation schemes are mostly based on grid load or electricity price, without considering the regularity of user behavior (such as the stability of arrival time, departure time, and home charging availability), resulting in low user acceptance of the recommendation strategy and low actual adoption rate.

[0046] To address the aforementioned issues, this application provides a vehicle behavior analysis method and system based on fragment-level operational data for credibility correction and multi-stage behavioral reasoning, aiming to achieve the following technical objectives:

[0047] (1) Under the premise that the data is incomplete, discontinuous and not completely reliable, accurately restore the user's main vehicle use behavior pattern, identify and downgrade low-quality segments through data credibility layering, and reduce misjudgment.

[0048] (2) By using a multi-stage reasoning chain with credibility constraints, error propagation is reduced and reasoning stability is improved. The credibility score of the preceding step is used as the weight factor of the subsequent step, and low credibility results are automatically downgraded or skipped.

[0049] (3) Output layered labels with confidence and evidence chain to reduce the risk of misjudgment. Calculate the total confidence of the evidence chain and output four types of labels according to the threshold: fact statistics, strong inference, weak inference, and risk warning.

[0050] (4) Link behavioral analysis with risk scenarios and scheduled charging strategies to enhance the value of engineering applications. The behavioral analysis results are spatiotemporally coupled with battery risk scenarios and scheduled charging strategies, and directly used for battery design optimization and charging function development.

[0051] The present invention will now be described in detail with reference to various embodiments.

[0052] Example 1

[0053] According to an embodiment of this application, an embodiment of a method for determining a charging strategy based on vehicle behavior segments is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0054] This application provides a charging strategy generation system based on vehicle behavior segments (hereinafter referred to as the generation system) for executing the method for determining the charging strategy based on vehicle behavior segments in this application.

[0055] Figure 1 This is a flowchart of an optional method for determining a charging strategy based on vehicle behavior segments according to an embodiment of this application, such as... Figure 1 As shown, the method includes the following steps:

[0056] Step S101: The L segment scores of each standard segment in the standard segment dataset are weighted and summed to obtain the credibility score of each standard segment, where L is a positive integer, and the L segment scores include time continuity score, physical consistency score, field integrity score and geographical stability score. Each standard segment is used to record vehicle behavior data.

[0057] Figure 2This is a flowchart of an optional method for determining the credibility score and level of a segment according to an embodiment of this application, such as... Figure 2 As shown, after obtaining the standard fragments, the calculation steps for the temporal continuity score, physical consistency score, field integrity score, and geographic stability score are as follows:

[0058] (1) Calculate the time continuity score:

[0059] For each segment i, calculate the temporal continuity score. :

[0060] ;

[0061] In the above formula, : The time interval (in hours) between segment i and the previous segment; Historical average time interval (sliding window statistics, taking the most recent 30 segments); : Standard deviation of historical time intervals; Smoothing factor, set to 0.01.

[0062] (2) Calculate the physical consistency score:

[0063] For each segment i, calculate the physical consistency score. :

[0064] ;

[0065] In the above formula, This represents the rationality of SOC changes, when A score of 1.0 represents two scenarios: one where the vehicle is in "charging" (category=30) and the final SOC is greater than the initial SOC; the other where the vehicle is in "driving" (category=10) and the final SOC is less than the initial SOC. A score of 0.5 indicates that the variation in SOC is within ±2%; when A score of 0.0 represents any situation other than those mentioned above.

[0066] The power and current are consistent, and the calculation formula is as follows:

[0067] ;

[0068] in, Total energy of the segment (unit: kW, i.e., power); The average current of the segment; The average voltage of the segment. This refers to the power deviation value corresponding to the segment.

[0069] Represents consistency between speed and category, when A score of 1.0 represents two scenarios: one is that the vehicle is in "driving" status (i.e., category=10) and the average speed is... Greater than Another situation is when the vehicle status is " (i.e., category = 20 / 60) and average speed Approximately ;when A score of 0.0 indicates situations other than those mentioned above, such as the average vehicle speed in the driving segment being too slow (i.e., The vehicles in the parking segment were traveling too fast.

[0070] This represents the consistency between distance and speed, when A score of 1.0 indicates a distance deviation rate of less than 20%. This deviation rate is calculated as the absolute value of the difference between the actual distance and the theoretical distance (average speed × time length / 3600), divided by the actual distance. A score of 0.5 indicates a distance deviation rate between 20% and 50%; when A score of 0.0 indicates a distance deviation rate greater than 50%, suggesting an anomaly in the data.

[0071] Recommended weight values ​​above: .

[0072] (3) Calculate the field integrity score:

[0073] For each segment i, calculate the missing rate of key fields. :

[0074] ;

[0075] In the above formula, : Number of missing fields (values ​​are empty or abnormal); Total number of core fields (e.g., 15).

[0076] (4) Calculate the geographical stability score:

[0077] For parking segments, calculate the degree of GPS drift. :

[0078] ;

[0079] In the above formula, Standard deviation of locations within a segment (if applicable) or distance of position jump between adjacent segments; Reference standard deviation, taken as 50 meters; Smoothing factor, set to 0.01.

[0080] (5) Calculate the overall credibility score :

[0081] ;

[0082] Recommended weight values ​​above: .

[0083] (6) Credibility rating:

[0084] Each segment is classified into credibility levels based on its credibility score. :

[0085] ;

[0086] Recommended values ​​for the above-mentioned grading thresholds: .

[0087] The output data obtained through the above processing steps includes: SegmentConfidence score table, SegmentLevels confidence level results, and Segments_low_conf list of low-confidence segments.

[0088] Optionally, traditional techniques typically assume that the uploaded data is complete and accurate, and directly use it for subsequent analysis. This leads to the introduction of a large amount of noise due to data omissions, miscuts, GPS drift, or missing fields. This step constructs a multi-dimensional credibility evaluation system to examine the original data from four dimensions: time, physical, completeness, and geography, obtaining multi-dimensional segment scores. Then, the multi-dimensional segment scores are weighted and summed to ensure that the resulting credibility score can comprehensively reflect the data quality, thereby quantifying the credibility of each segment and providing an objective basis for subsequent removal of low-quality data.

[0089] Step S102: Remove standard segments from the standard segment dataset whose confidence scores are less than a preset confidence threshold to obtain the target segment dataset. Each target segment in the target segment dataset is one of the following: charging segment, driving segment, or parking segment.

[0090] Optionally, the generation system achieves data-level noise reduction by removing low-confidence segments. Low-confidence segments often contain GPS drift points, misclassified data, or abnormally abrupt SOC jumps. If they are directly entered into subsequent analysis, it will lead to incorrect clustering of dwell points, failure to identify charging events, or deviations in path identification. This step uses threshold filtering to ensure that the data entering the subsequent complex inference model is all high-quality data, reducing the risk of error propagation in subsequent multi-stage inference and improving the robustness of the overall analysis process.

[0091] Step S103: Based on the target segment dataset, determine the vehicle's destination location, charging location list, commuting route list, charging event list, charging control parameters, user driving behavior characteristics, and risk event list. The destination location includes home and work locations.

[0092] Optionally, the generation system transforms discrete vehicle-side data fragments into structured multidimensional data. By inferring destination and charging location lists, it can identify the user's travel area radius and charging network dependence, providing geographical constraints for generating personalized charging strategies. By determining the commuting route list, the system identifies the user's high-frequency travel patterns, used to predict future travel demand and time windows. Through the charging event list and charging control parameters, the system clarifies the user's charging habits and vehicle hardware boundaries, ensuring the feasibility of the strategy. Through user driving behavior characteristics, the system understands the user's driving style, used to adjust the aggressiveness or conservatism of the strategy. Through the risk event list, the system correlates battery health with user behavior, ensuring that the subsequent charging strategy recommendation process considers not only economy but also battery safety, thus providing comprehensive and multidimensional input for the final strategy generation.

[0093] Step S104: Generate X triples based on the target segment dataset and the charging event list, where X is a positive integer. Each triple is used to record the correspondence between the valid charging segment, the preceding parking segment, and the following driving segment in the charging event list. Each triple corresponds to a pattern label.

[0094] Optionally, after obtaining the target fragment dataset and the list of charging events, the generation system determines X triples and the pattern label corresponding to each triple through the following steps:

[0095] (1) Identify associated event triples:

[0096] Find the preceding parking segment (arrival_event) and the subsequent driving segment (departure_event) corresponding to each charging event (charge_event) (i.e., a valid charging segment) in the charging event list. The preceding parking segment ends within 2 hours before the charging starts in the charging event, and the subsequent driving segment begins within 2 hours after the charging ends, forming a triplet (arrival_event, charge_event, departure_event).

[0097] (2) Calculate the association strength of each triple:

[0098] For each triple n, calculate the association strength. The calculation formula is as follows:

[0099] ;

[0100] In the above formula, Evidence of arrival (if the distance between the parking and charging locations is less than 100 meters, it is counted as 1); Evidence of charging (charging confidence level) (Normalization) : Departure evidence (1 if departing within 1 hour after charging is completed).

[0101] Recommended weight values ​​above: .

[0102] (3) Classify the patterns for each triple:

[0103] Based on the association strength and fragment temporal characteristics, the pattern label for each triple is determined to be one of the following types:

[0104] Park and start charging: Charging will begin within 30 minutes of arrival;

[0105] Pull out and go: Depart within 30 minutes after charging is complete;

[0106] Overnight home charging: Charging overnight and remaining at home for ≥6 hours;

[0107] Recharging at work: Charge during weekdays and stay for ≥2 hours.

[0108] The output data obtained through the above processing steps includes: a list of association triplets, association scores, and behavior patterns.

[0109] Optionally, traditional technologies often analyze parking, charging, or driving behaviors in isolation, lacking modeling of the causal relationships between these behaviors. This step establishes a strong correlation between parking, charging, and departure by constructing triples. Furthermore, the introduction of pattern labels enables the system to identify users' charging patterns. For example, users in the "pull and go" mode are sensitive to charging wait times and are not suitable for long-term off-peak electricity reservations; users in the "overnight home charging" mode have ample time and are suitable for deep off-peak charging. This correlation modeling provides fine-grained behavioral constraints for subsequent strategy generation, making strategy recommendations more in line with users' actual lifestyles and improving user experience.

[0110] Step S105: Determine the target charging strategy for the vehicle based on the destination, charging location list, commuting route list, charging event list, charging control parameters, user driving behavior characteristics, risk event list, pattern labels corresponding to X triples, and preset electricity price data.

[0111] Optionally, by integrating the aforementioned multi-dimensional information, the generation system generates charging strategies that not only consider economy and battery health but also fully respect users' behavioral habits and safety needs. This improves the adoption rate of the target charging strategy recommended to users and increases user satisfaction. For example, in determining the target charging strategy, the generation system introduces pattern labels corresponding to triples, making the strategy highly personalized. For users with regular behavior, the system is willing to recommend aggressive off-peak charging strategies; for users with random behavior, the system recommends conservative strategies. At the same time, by combining a risk event list and user driving behavior characteristics, the system can identify high-risk users (such as those who still travel long distances when the battery is low) and mitigate risks in strategy generation by increasing the target SOC or reducing the weight of disturbance costs.

[0112] As described above, this application employs a multi-stage behavioral reasoning and credibility correction approach. First, it calculates credibility by weighted summation of scores for the temporal continuity, physical consistency, field completeness, and geographical stability of standard segments, and then removes low-credibility segments to obtain the target segment dataset. Based on this dataset, it generates X triples containing the correspondence between charging, driving, and parking segments. It then integrates multi-dimensional information including destination points, charging location lists, commuting route lists, charging event lists, charging control parameters, user driving behavior characteristics, risk event lists, and pattern labels, combined with preset electricity price data, to determine the vehicle's target charging strategy. This achieves the goal of full-process correlation modeling from data cleaning and behavioral pattern recognition to strategy generation, even in the context of quality defects in vehicle operation data and error propagation from independent processing of operation data by different reasoning modules. This improves the accuracy of operation data analysis, outputs interpretable vehicle charging strategies, and increases user adoption rates of recommended charging strategies, thus solving the technical problems of poor analysis of new energy vehicle operation data and inaccurate recommended charging strategies in existing technologies.

[0113] In one optional embodiment, before performing a weighted summation of the L segment scores for each standard segment in the standard segment dataset, the generation system parses the raw data file recording the original behavior data of the vehicle to obtain K raw segments, wherein each raw segment includes at least one of the following: an identifier field, a time field, a spatial field, a mileage / speed field, an energy field, and a battery field. Then, the K raw segments are standardized and outlier filtered to obtain standard segments in the standard segment dataset. The standardization process is used to convert the time field to a preset timestamp format and the spatial field to a preset coordinate system format, and the outlier filtering is used to remove raw segments containing outliers.

[0114] Optionally, the identifier fields are shown in List 1 below:

[0115] Table 1:

[0116]

[0117] Optionally, time-related fields are shown in List 2 below:

[0118] Table 2:

[0119]

[0120] Optionally, the spatial class fields are shown in List 3 below:

[0121] Table 3:

[0122]

[0123] Optionally, the mileage / speed class fields are shown in List 4 below:

[0124] Table 4:

[0125]

[0126] Optionally, the energy-related fields are shown in List 5 below:

[0127] Table 5:

[0128]

[0129] Optionally, the battery category fields are shown in List 6 below:

[0130] Table 6:

[0131]

[0132] Optionally, after obtaining the raw data file uploaded by the vehicle, the processing procedure for the fragment-level vehicle operation data (i.e., raw behavior data) is as follows:

[0133] (1) Perform data parsing operations: Read the original data file and extract the core fields;

[0134] (2) Perform time standardization: Convert the timestamps of heterogeneous data in the original fragments to UTC (Coordinated Universal Time) timestamps (milliseconds), and match derived time tags for each fragment. The derived time tags include:

[0135] Date: date=format(st_time_e,'yyyy-MM-dd');

[0136] Hour: hour=format(st_time_e,'HH');

[0137] Week: day_of_week=dayOfWeek(st_time_e);

[0138] Weekday / Weekend: is_workday=(day_of_weekin[1,2,3,4,5]);

[0139] Day / Night: is_night=(hour>=22orhour<6);

[0140] Morning / Evening Rush Hour: is_peak=((hour==7 or hour==8) or (hour>=17 and hour<=19));

[0141] (3) Perform spatial standardization: Convert the heterogeneous data in the original fragments to the WGS84 coordinate system and calculate the center point of each fragment coordinate system. The calculation formula is as follows:

[0142] ;

[0143] In the above formula, Longitude value of the center point of each segment's coordinates; : The longitude value of the starting point of each segment's coordinates; : The longitude value of the endpoint of each segment's coordinates; : The latitude value of the center point of each segment's coordinates; : The latitude value of the starting point of each segment's coordinates; : The latitude value of the endpoint of each segment coordinate.

[0144] (4) Perform preliminary outlier filtering:

[0145] Remove segments with obviously abnormal timestamps (end time earlier than start time), segments with latitude and longitude outside a reasonable range (latitude and longitude of 0 or outside [-180, 180]), and segments with SOC outside the range [0, 100].

[0146] The output data obtained through the above processing steps includes: the standardized fragment dataset Segments_std, the time label table TimeLabels, and the list of invalid fragments Segments_invalid.

[0147] Optionally, the generation system processes the original data files of the vehicle's original behavior data through the above steps, and the technical effects that can be achieved include: (1) It realizes the unification and standardization of heterogeneous data and solves the technical problems caused by inconsistent data formats: By performing standardization processing, the time-type fields are unified into the preset timestamp format and the spatial-type fields are unified into the preset coordinate system format, eliminating data heterogeneity. This means that when calculating the time continuity score based on the time interval and the geographical stability score based on the spatial distance in the subsequent steps, there is no need to perform complex format conversion or coordinate correction, ensuring the accurate execution of time series analysis and spatial clustering algorithms, and providing a unified data benchmark for subsequent calculation of credibility scores; (2) It removes low-quality invalid data and improves the reliability of subsequent analysis results: By performing outlier filtering, the original fragments containing obvious logical errors are directly removed, avoiding dirty data from entering the subsequent complex credibility score calculation process. Through pre-filtering, it ensures that the data entering the standard fragment dataset has basic logical validity, and improves the robustness of processing massive vehicle-side operation data.

[0148] In one optional embodiment, after determining a list of stopping locations based on parking segments in the target segment dataset, the generation system determines a destination based on a time-stamp table corresponding to the target segment dataset and the list of stopping locations. The time-stamp table includes derived time stamps for each target segment. After clustering the charging locations of all charging segments in the target segment dataset, a list of charging locations is generated based on the cluster center coordinates of each cluster of charging locations. Then, a list of commuting routes is determined based on driving segments and destinations in the target segment dataset. Next, a list of charging events and charging control parameters are determined based on the charging segments in the target segment dataset. Subsequently, user driving behavior characteristics are determined based on driving segments in the target segment dataset, and a list of risk events is determined based on the target segment dataset, destinations, the list of charging locations, and preset risk fields.

[0149] Optionally, the technical effects that the generation system can achieve through the steps in the above embodiments include: (1) improving the accuracy and interpretability of destination identification: by introducing a time tag table and combining it with a list of places of residence for spatiotemporal joint analysis, for example, using the feature of long and repetitive nighttime stays to infer home locations, and using the feature of daytime stays on weekdays to infer work locations, this reasoning based on time semantics improves the accuracy of destination identification and provides a foundation for subsequent location-based personalized strategies (such as home charging strategies); (2) realizing the standardization and deduplication of charging locations: the charging data uploaded by the vehicle end often shows differences due to GPS drift or users charging at different charging piles. To achieve dispersion, the system clusters the charging locations of all charging segments and generates a list of charging locations based on the coordinates of the cluster centers. This process not only removes duplicate charging records but also accurately represents the locations of commonly used charging stations by users through the cluster centers. Thus, when generating charging strategies, it can determine whether users have the conditions for home charging or commonly use public charging stations, thereby improving the targeting of strategy recommendations; (3) It constructs a fine-grained charging behavior profile and battery status monitoring, solving the problem of charging strategies ignoring battery health and user habits: The generation system not only generates a list of charging events but also further extracts charging control parameters (such as typical start-up SOC / stop-up SOC), which makes the system more efficient and targeted. The system can extract micro-level battery usage habits from macro-level charging behavior. At the same time, by pre-setting risk fields to identify a list of risk events, it maps the battery's health status (such as abnormal pressure difference and temperature difference) to specific spatiotemporal scenarios (such as a fast charging process). This approach of closely combining user behavior with the battery's physical state allows the subsequent generated charging strategy to not only consider economy and convenience, but also the health and safety of the battery, avoiding the recommendation of aggressive charging strategies in high-risk scenarios; (4) It realizes the construction of multi-dimensional user profiles, providing comprehensive data support for personalized charging strategies: The scheme executes the determination of commuting route list, the extraction of user driving behavior characteristics, and the determination of commuting route list in parallel. The system generates a risk event list, which includes a commuting route list revealing users' high-frequency travel patterns and distance constraints, user driving behavior characteristics revealing their driving style, and a risk event list revealing potential battery health hazards. These multi-dimensional pieces of information together form a complete user and environment profile. When determining the target charging strategy, the system can comprehensively consider factors such as whether the user frequently commutes long distances (affecting departure reliability), the user's driving style (affecting energy consumption prediction), and whether there are risks to battery operation (affecting charging power limits). This comprehensive data processing approach avoids the problems of decreased user experience or increased battery wear caused by a "one-size-fits-all" charging strategy.

[0150] In one optional embodiment, the generation system first uses the endpoint location of each parking segment in the target segment dataset as a candidate dwell location, and calculates the kernel density weight of each parking segment centered on each candidate dwell location. Then, based on the kernel density weight of each parking segment, the candidate dwell locations of all parking segments are spatially clustered to obtain A dwell location clusters, where A is a positive integer, and each dwell location cluster includes at least two candidate dwell locations. Subsequently, the cluster index of each dwell location cluster in the A dwell location clusters is calculated, where the cluster index includes the cluster center coordinates, access frequency, average dwell time, and dwell time standard deviation. Then, cluster filtering is performed on the A dwell location clusters based on the cluster index of each dwell location cluster, where cluster filtering is used to remove dwell location clusters whose cluster index is less than a preset index threshold. Finally, the cluster center coordinates of the dwell location clusters that have passed the cluster filtering are used as dwell locations, and all dwell locations are summarized to obtain a dwell location list.

[0151] Optionally, Figure 3 This is a flowchart of an optional location type determination method according to an embodiment of this application, wherein the location type is one of home, workplace, or charging location, such as... Figure 3 As shown, before determining the specific location type, the generation system needs to determine a list of locations to stay in. The steps for determining the list of locations to stay in are as follows:

[0152] (1) Extract parking segments from the target segment dataset that meet the following conditions:

[0153] The category ∈ {20, 60} (parked or turned off), time_length ≥ 300 seconds (more than 5 minutes), and the credibility level ∈ {high credibility, medium credibility}.

[0154] (2) Calculate the kernel density weights of the dwell points for each parking segment:

[0155] Candidate dwell points for each parking segment (Centered on the end of the parking segment) Calculate the kernel density weights The calculation formula is as follows:

[0156] ;

[0157] In the above formula, Locations of other parking segments; Bandwidth parameter, 100 meters recommended; Euclidean distance (meters) between two points.

[0158] (3) Perform spatial clustering operations:

[0159] Based on the kernel density weights of the dwell points of each parking segment, the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) algorithm was used to spatially cluster all the selected parking segments, with eps (neighborhood radius) of 150 meters and min_samples (minimum number of samples) of 3. Then, for each location cluster obtained by clustering, the following indicators were calculated:

[0160] Cluster center coordinates (cluster_lng, cluster_lat), visit frequency (visit_count), average stay duration (avg_duration), and standard deviation of stay duration (std_duration).

[0161] (4) Perform a filtering operation based on the statistically obtained location cluster indicators to remove location clusters that meet the following conditions:

[0162] Access frequency < 2 times; average stay time < 10 minutes; number of points within the cluster < 3.

[0163] The output data obtained through the above processing steps includes: a list of stay clusters (StayClusters) and statistical indicators for each cluster (i.e., center coordinates, access frequency, and stay duration).

[0164] Optionally, the technical effects that the generation system can achieve through the steps in the above embodiments include: (1) effectively suppressing GPS drift noise and improving the accuracy of station location positioning: by introducing kernel density weights and combining them with spatial clustering algorithms, spatially adjacent and high-density candidate station locations are merged into station location clusters. This density-based clustering method can ignore isolated drift points, thereby identifying the true station center of the vehicle and reducing the location identification deviation caused by positioning errors; (2) realizing automatic noise reduction and filtering of station locations and improving the purity of data: by statistically analyzing the location cluster indicators and performing cluster filtering, invalid location clusters with low access frequency and short stay time are automatically eliminated, improving the data quality of subsequent behavior analysis. (3) It provides multi-dimensional stay characteristics, which enhances the interpretability and accuracy of subsequent destination inference: This step not only outputs the list of places of stay, but also retains the statistical indicators of the location clusters such as visit frequency, average stay duration and standard deviation of stay duration. Among them, high visit frequency and long stay duration usually point to the residence or office of the vehicle driver, while the regularity of stay duration (low standard deviation) helps to identify regular travel patterns. When determining the destination in the future, the system can use these statistical indicators in combination with the time label table to make a weighted score. This comprehensive evaluation based on multi-dimensional location cluster statistical indicators improves the accuracy of the destination obtained by subsequent reasoning.

[0165] In one optional embodiment, the generation system first determines P location features for each location based on the time label table corresponding to the target fragment dataset and the location cluster index of the location cluster to which each location belongs in the location list. Here, P is a positive integer, and the P location features include the proportion of nighttime stays, the proportion of weekday daytime stays, the normalized value of the average stay duration, the ratio of consecutive stay days to total stay days, and the frequency of use as a departure point. Then, based on the P location features of each location, a first location score and a second location score are determined for each location. The first location score is used to characterize the probability that each location is a home location, and the second location score is used to characterize the probability that each location is a work location. Then, the location with the largest first location score in the location list is designated as a home location, the location with the largest second location score in the location list is designated as a work location, and both the home location and the work location are designated as destination points.

[0166] Optionally, after obtaining the list of lodging locations and the TimeLabels table corresponding to the target fragment dataset, such as... Figure 3 As shown, the generation system performs the following steps to determine the destination:

[0167] (1) For each cluster of dwelling points j, extract the following location features:

[0168] : Percentage of nighttime stays (number of stays between 22:00 and 6:00 / total number of stays); Percentage of stay during weekdays (number of stays between 9:00 and 18:00 on weekdays / total number of stays); : Normalized average stay duration (min(duration / 12h, 1.0)); : Number of consecutive days / Total number of days; : The percentage of times it is used as a starting point.

[0169] (2) Calculate the comprehensive location score based on the extracted location features. The calculation formula is as follows:

[0170] ;

[0171] In the above formula, the recommended weighting values ​​used when calculating the home location score (i.e., the first location score) are as follows: Recommended weighting values ​​for calculating workplace ratings (i.e., second-place ratings): .

[0172] (3) Determination of home / work location:

[0173] Home Location: Select The highest cluster (calculated using the recommendation weights of family location ratings) is required to... .

[0174] Work location: Select The highest cluster (calculated using the recommendation weights of workplace ratings) is required to... .

[0175] like Below If so, the location will be marked as "unidentified".

[0176] The output data obtained through the above processing steps includes: HomeCandidate and WorkCandidate.

[0177] Optionally, after obtaining the charging segments in the target segment dataset, the TimeLabels table corresponding to the target segment dataset, and the destination points, the generation system performs the following steps to determine the list of charging locations:

[0178] (1) Filter the charging segments in the target segment dataset (category=30);

[0179] (2) Cluster the charging locations in all charging locations and count the following indicators of the clustered charging location clusters: charging frequency, average charging time, fast / slow charging ratio, and distance from the destination.

[0180] (3) Determine the charging location: Those charging ≥ 3 times and whose distance from the destination is < 500 meters are marked as "home charging / work location charging" type;

[0181] (4) Summarize the charging locations indicated by the cluster center coordinates of the clusters of charging locations obtained by clustering, as well as the types of each charging location, to obtain a list of commonly used charging points, ChargePoints.

[0182] Optionally, the technical effects that the generation system can achieve through the steps in the above embodiments include: (1) achieving accurate location classification based on behavioral semantics: by extracting P location features such as the proportion of nighttime stays and the proportion of weekday daytime stays, the nature of the location is inferred by utilizing the temporal regularity of user behavior. For example, a high proportion of nighttime stays usually corresponds to a home, while a high proportion of weekday daytime stays usually corresponds to a workplace. This joint analysis based on "time + space" transforms abstract geographical coordinates into home and work locations with clear business semantics, significantly improving the accuracy and interpretability of location identification; (2) The robustness and objectivity of identification are improved by quantifying the inference results through a probability scoring mechanism: by introducing the first location score and the second location score, qualitative judgment is transformed into quantitative calculation. By comprehensively considering multiple dimensions of indicators such as nighttime stay, weekday stay, stay duration, continuity and departure frequency, a comprehensive scoring model is constructed, and the stay location with the highest score is finally selected as the destination, avoiding misjudgment that may be caused by a single indicator; (3) It provides a high-precision geographical constraint basis for subsequent personalized charging strategies: accurate identification of home locations enables the system to determine whether users have the conditions for home charging, thereby recommending a reservation charging strategy based on home grid load and personal habits; accurate identification of work locations helps to analyze users' commuting routes and workplace charging behavior. Through refined feature engineering and scoring mechanisms, the high precision of destination inference is ensured, directly improving the applicability and effectiveness of subsequent target charging strategy generation, and realizing closed-loop optimization from behavior analysis to strategy decision-making.

[0183] In one optional embodiment, the generation system first maps the starting point and ending point of each driving segment in the target segment dataset to a spatial grid of a preset scale, obtaining X spatial paths corresponding to X driving segments, where X is a positive integer. Then, the X spatial paths are divided into Y path sets, where Y is a positive integer less than or equal to X. The starting points of any two spatial paths in each path set are located in the same spatial grid, and the starting points of any two spatial paths are located in the same spatial grid. Then, the path index of each path set in the Y path sets is calculated, where the path index includes frequency of occurrence, average distance, average duration, and average speed. Subsequently, the commuting score of each path set is determined based on the path index of each path set. Then, any spatial path in each path set with a commuting score greater than or equal to a preset commuting threshold is taken as a commuting path. The Y commuting paths corresponding to the Y path sets are summarized to obtain a commuting path list.

[0184] Optionally, after obtaining the driving segments from the target segment dataset, the generation system performs the following steps to determine the list of commuting routes:

[0185] (1) Perform grid normalization processing on the start and end points of the driving segment:

[0186] The coordinates of the start and end points of the driving segment (i.e., the starting point and the end point of the driving segment) are normalized and mapped onto the grid. The recommended grid size is 200 meters × 200 meters. The formula used for normalization is shown below:

[0187] ;

[0188] In the above formula, : Gridded longitude values, obtained by converting the original longitude of the driving segments Divide by 0.002, round down, and then multiply by 0.002 to get the result; : Gridded latitude values, by converting the original latitude Divide by 0.002, round down, and then multiply by 0.002 to get the result.

[0189] (2) Merge the multiple paths obtained by normalization:

[0190] Travel segments with the same origin and destination grid are grouped into the same candidate path, and the frequency of occurrence, average distance, average duration, and average speed of all spatial paths included in each candidate path are statistically analyzed.

[0191] (3) Calculate the commuting probability score:

[0192] For each candidate path r, calculate the commuting probability. The calculation formula is as follows:

[0193] ;

[0194] In the above formula, Consistency of time window (frequency of morning and evening rush hours / total frequency); : Fixed origin and destination (frequency of origin / destination being home / workplace / total frequency); Distance stability (1 - standard deviation of distance / average distance); : Repetition (number of days it occurs / total number of days).

[0195] Recommended weight values ​​above: .

[0196] (4) Commuting determination:

[0197] If the commuting probability of a candidate path If so, then mark it as a "commuter route" type.

[0198] The output data obtained through the above processing steps includes: a list of fixed routes (FixedRoutes) and a list of commuting routes (CommuteRoutes) (including commuting probability scores).

[0199] Optionally, the technical effects that the generation system can achieve through the steps in the above embodiments include: (1) By spatial gridding and path clustering, the structure and simplification of driving paths are realized, solving the problems of high noise in trajectory data and difficulty in identifying patterns due to complex paths: The starting point and the ending point are mapped to a spatial grid of a preset scale, and paths with the same starting point and ending point grid are grouped into the same path set, effectively filtering out position noise at the micro level, and abstracting complex continuous trajectories into discrete and deterministic spatial paths, so that the system can clearly identify the fixed routes that users frequently travel, laying a data foundation for subsequent identification of commuting behavior; (2) By comprehensively evaluating multi-dimensional path indicators, the accuracy and robustness of commuting path identification are improved: The generation system not only counts the frequency of occurrence, but also introduces path indicators such as average distance, average duration and average speed. Combined with time label tables (such as morning and evening peak hours), the commuting attributes of the route can be further verified. Based on these indicators, The system calculates commuting scores, and only routes with scores reaching the preset commuting threshold are recognized as commuting routes. This multi-indicator comprehensive evaluation mechanism avoids the misjudgment that may be caused by judging based solely on frequency and improves the purity of the commuting route list. (3) It provides key travel constraints for personalized charging strategies and improves the practicality of the strategies: commuting distance and commuting time directly determine the user's daily basic power consumption. The regularity of commuting routes (such as fixed departure time) helps to predict the user's departure time. Based on the commuting route list, the system can more accurately estimate the user's power demand for the next day's trip, thereby ensuring the departure guarantee when generating the target charging strategy. For example, if a high-frequency, long-distance commuting route is identified, the system may recommend a higher target SOC or an earlier charging time to avoid the user's power being insufficient during the commute. Therefore, this solution enhances the matching degree between the charging strategy and the user's actual travel needs by accurately identifying commuting routes.

[0200] In one optional embodiment, the generation system first statistically analyzes the charging indicators for each charging segment in the target segment dataset. These indicators include the vehicle's state of charge (SOC), power information, and charging time period. Then, based on the SOC, the system determines the charging confidence level for each charging segment. Segments with a charging confidence level greater than or equal to a preset confidence threshold are considered valid charging segments. All valid charging segments are aggregated to obtain a charging event list. Subsequently, the charging start times corresponding to all valid charging segments in the charging event list are clustered to obtain typical charging time periods for the vehicle. Each typical charging time period is the cluster center time period of the clustered valid charging time period clusters. Then, based on the median SOC corresponding to all valid charging segments in the charging event list, the system determines the vehicle's typical starting charging load value and typical ending charging load value. Next, based on the power information corresponding to all valid charging segments in the charging event list, the system determines the vehicle's fast / slow charging ratio. Finally, the vehicle's typical charging time period, typical starting charging load value, typical ending charging load value, and fast / slow charging ratio are used as charging control parameters.

[0201] Optionally, after obtaining the charging segments in the target segment dataset, the generation system performs the following steps to determine the list of charging events and charging control parameters:

[0202] (1) Calculate the confidence level of the charging event:

[0203] For each charging segment e, calculate its charging confidence. The calculation formula is as follows:

[0204] ;

[0205] In the above formula, : SOC rising indicator (1 if end_soc > start_soc, 0 otherwise); : Negative power indicator (1 if average power < 0); : Stop indication (segment type is charging or parking); : Duration indicator (1 if duration ≥ 30 minutes).

[0206] Recommended weight values ​​above: .

[0207] (2) Filter out charging segments with low charging confidence, i.e., remove them. The charging segment is used to obtain a list of charging events.

[0208] (3) Clustering of charging segments:

[0209] Based on the charging start time, the effective charging segments in the filtered charging event list are clustered (by hour), and typical charging periods (such as 20:00-22:00) are identified based on the cluster center time of the effective charging period clusters.

[0210] (4) Extract the representative value of the initial charge SOC. The SOC representative value of the stop-charge The calculation formula is as follows:

[0211] ;

[0212] In the above formula, The initial SOC value of the charging segment; : The SOC value of the charging segment.

[0213] (5) Differentiate between fast and slow charging:

[0214] The fast charging segment and the slow charging segment are distinguished based on the average power or charging rate, and the ratio of the fast charging segment to the slow charging segment is calculated. Among them, the average charging power of the fast charging segment is ≥40kW or the charging rate is ≥1C; the average charging power of the slow charging segment is <10kW or the charging rate is <0.5C.

[0215] The output data obtained through the above processing steps includes: a list of charging events (ChargeEvents), typical starting charge (SOCTypicalStartSOC), typical ending charge (SOCTypicalEndSOC), typical charging time slots (ChargeTimeSlots), and the ratio of fast to slow charging (FastSlowRatio).

[0216] Optionally, the technical effects that the generation system can achieve through the steps in the above embodiments include: (1) Filtering noise through confidence assessment to ensure the data accuracy of the charging behavior profile: By calculating the charging confidence based on the state of charge, power information and charging period, and setting a preset confidence threshold, low-quality charging segments are effectively eliminated, and only valid charging segments are retained to form a charging event list, thus avoiding the interference of noise data on the accuracy of charging control parameters; (2) Quantifying user charging habits through statistical features to provide precise behavioral constraints for personalized strategies: Four dimensions of charging control parameters are extracted through statistical methods. These parameters together constitute the fingerprint of user charging behavior, enabling the subsequently generated target charging strategy to be more appropriate. (3) Improved the scientific nature and user acceptance of charging strategy generation: Based on the above charging control parameters, when determining the target charging strategy, the system can recommend scheduled charging near the typical charging time period of the user's habit, which is in line with the user's work and rest and reduces the user's disturbance cost; By referring to the typical start / stop charge value to set a reasonable target SOC, it avoids insufficient power (ensuring start guarantee) and avoids overcharging to damage the battery (improving health friendliness); Based on the fast and slow charging ratio and typical charging time period, the calculation of off-peak electricity benefits is optimized. This data-driven personalized parameter extraction makes the charging strategy more intelligent and humanized, and improves the accuracy of strategy recommendation and user adoption rate.

[0217] In one optional embodiment, the generation system first divides all driving segments in the target segment dataset into a first segment set and a second segment set based on a preset lower charge limit. In the first segment set, the vehicle's charge value at the start and / or end of driving is less than the preset lower charge limit, while in the second segment set, the vehicle's charge value at the start and end of driving is greater than or equal to the preset lower charge limit. Then, based on the driving indicators of the first and second segment sets, the system determines the vehicle's driving behavior deviation index, and based on the driving behavior deviation index, determines the user's driving behavior characteristics. The driving indicators include average speed, average distance, the percentage of driving segments ending at home / charging location, and the percentage of long-distance travel segments.

[0218] Optionally, after obtaining the driving segments from the target segment dataset, the generation system performs the following steps to determine the user's driving behavior characteristics:

[0219] (1) Divide the low SOC segment into the normal segment. The low SOC segment has a charge value start_soc < 30% at the start of driving or a charge value end_soc < 30% at the end of driving. The normal segment has a charge value start_soc ≥ 30% at the start of driving and a charge value end_soc ≥ 30% at the end of driving.

[0220] (2) Calculate the behavioral offset index The calculation formula is as follows:

[0221] ;

[0222] In the above formula, : Normalized value of the average velocity difference between low SOC and normal SOC; : Normalized value of the average distance difference between low SOC and normal SOC; The difference between the proportion of low SOC and the usual destination of going home / charging; The difference between the proportion of low SOC and regular long-distance travel (>50km).

[0223] Recommended weight values ​​above: .

[0224] (3) Conservative driving judgment, if The vehicle's driver will then be categorized as "low battery conservative driving".

[0225] The output data obtained through the above processing steps includes: LowSOC Behavior Index (LowSOCBehaviorIndex), Conservative Driving Flag (ConservativeDrivingFlag), and LowSOC Stats (LowSOCStats).

[0226] Optionally, the technical effects that the generation system can achieve through the steps in the above embodiments include: (1) It realizes comparative analysis of driving behavior based on battery status, revealing the psychological and behavioral patterns of users when the battery is low: By setting a lower charge limit, the driving segment is divided into a first segment set (low battery) and a second segment set (normal battery), and their driving indicators are statistically analyzed respectively. By comparing the differences between the two sets in average speed, average distance, destination ratio and long-distance travel ratio, the system can quantify the changes in user behavior when facing low battery, thereby deeply exploring the potential behavioral patterns of users; (2) It assesses the user's driving style and risk preference: By introducing a driving behavior deviation index, qualitative behavioral differences are transformed into quantitative numerical indicators. This index comprehensively considers multiple dimensions such as speed, distance, destination preference and long-distance travel, and can comprehensively reflect the degree of behavioral deviation of users when the battery is low. If the deviation index is large, it indicates that the user is driving when the battery is low. Significant changes (such as sharp deceleration, avoiding long distances) may indicate a conservative user who is more sensitive to battery anxiety. If the deviation index is small, it indicates that the user's driving style is stable and less affected by battery level. This quantitative user driving behavior characteristic is more diagnostically valuable than a simple statistical average, providing a more refined user profile for subsequent personalized strategies. (3) It provides behavioral constraints for the generation of differentiated charging strategies and improves the targeting of the strategies. For conservative users with a high driving behavior deviation index, the system can predict that they are more inclined to recharge or shorten their trip when the battery is low. Therefore, when generating the target charging strategy, the system can appropriately increase the target SOC or recommend charging in advance to ensure that the user has sufficient battery before low battery anxiety and improve the departure guarantee. For users with a low deviation index, the system can more flexibly optimize off-peak electricity benefits based on their stable driving habits, without being overly conservative, thereby achieving a more humanized and accurate charging recommendation.

[0227] In one optional embodiment, the generation system first identifies target segments in the target segment dataset that hit a preset risk field as risk segments. Then, it determines the associated scenarios for each risk segment based on a list of destinations and charging locations, where the associated scenarios are one of the following: home location, workplace, or charging location. Next, it determines a risk score for each risk segment based on its associated scenarios, where the risk scores include pressure difference risk score, temperature difference risk score, power risk score, and recurrence risk score. Subsequently, it generates risk events based on risk segments with risk scores greater than or equal to a preset risk threshold and their associated scenarios. Finally, it summarizes all risk events to obtain a risk event list.

[0228] Optionally, the system generates a list of risk events by performing the following steps:

[0229] (1) Identify risk segments:

[0230] Segments that meet any of the following conditions are selected to obtain risk segments:

[0231] Abnormal voltage difference: max_cell_volt - min_cell_volt > 50mV;

[0232] Temperature difference anomaly: max_diff_temp > 8°C;

[0233] High temperature: max_acquisition_point_temp > 45°C;

[0234] High power: abs(max_power) > 100kW;

[0235] High-rate charging: charge_c > 1.5C;

[0236] Low SOC, high load: start_soc < 20% and abs(max_power) > 50kW.

[0237] (2) Calculate the risk scenario coupling score:

[0238] For each risk segment z, calculate the risk score. The calculation formula is as follows:

[0239] ;

[0240] In the above formula, : Normalized value of differential pressure risk (differential pressure / 100mV); Temperature difference / high temperature risk normalized value (temperature difference / 15°C); : Normalized value of high power risk (power / 150kW); : Recurrence (number of historical risks in this location / scene / 5); Scenario risks (fast charging scenario 0.8, slow charging scenario 0.4, driving scenario 0.6).

[0241] Recommended weight values ​​above: .

[0242] (3) Spatiotemporal scene mapping:

[0243] Risk segments are mapped to corresponding locations, time periods, and scenario categories. Location refers to one of home, workplace, or charging station; time period refers to one of nighttime, daytime, or peak hours; and scenario category refers to one of fast charging, slow charging, driving, or parking.

[0244] (4) Risk recurrence assessment:

[0245] For risk segments in the same location / scene, if the number of risk segments is ≥3, the risk type of the risk segment mapped to the corresponding location / scene will be marked as "recurring risk".

[0246] The output data obtained through the above processing steps includes a list of risk events (RiskEvents), risk scores (RiskScores), a spatiotemporal scene mapping (RiskSceneMapping), and a list of recurrent risks (RecurrentRisks).

[0247] Optionally, the technical effects that the generation system can achieve through the steps in the above embodiments include: (1) realizing the spatiotemporal coupling of battery risks and specific user scenarios, solving the problems of isolated risk identification and lack of scenario awareness: by mapping risk fragments to the list of destinations and charging locations, the associated scenarios of each risk are determined, thereby providing a data basis for providing targeted suggestions in the future; (2) improving the accuracy of risk identification and sensitivity to repeated risks through a multi-dimensional risk scoring mechanism: this solution not only considers the magnitude of a single risk (pressure difference / temperature difference / power risk score), but also introduces a recurrence risk score, in which the recurrence risk score considers the frequency of risk occurrence under the same associated scenario. If a charging location frequently experiences pressure difference anomalies, even if the single risk score is not high, its comprehensive risk score will increase due to the high recurrence rate. This comprehensive assessment mechanism can effectively identify "systemic risk points" (such as old charging stations or specific battery batch problems), avoid over-warning of occasional and minor anomalies, and prevent underreporting of high-frequency systemic risks; (3) it provides key constraints for generating safe charging control strategies, enhances the initiative of battery health management, and the generated risk event list is directly used as input for generating charging control parameters or strategies. If a high recurrence risk score is identified in a certain related scenario (such as a specific charging station), the system can actively avoid fast charging strategies in that scenario or reduce the maximum charging power in that scenario when formulating the target charging strategy, so as to protect the battery; if a high temperature risk is identified in the user's home location, the system can optimize the nighttime charging strategy, such as adjusting the charging power or time to facilitate heat dissipation. This solution realizes the transformation from "passive alarm" to "active defense" by associating risk identification with the scenario, so that the charging strategy not only considers economy and convenience, but also reduces battery safety hazards.

[0248] In one optional embodiment, the generation system first determines the preceding parking segment and the following driving segment corresponding to each valid charging segment in the charging event list based on the target segment dataset. The duration between the charging start time of each valid charging segment and the parking end time of the corresponding preceding parking segment is less than a preset duration, and the duration between the charging end time of each valid charging segment and the driving start time of the corresponding following driving segment is less than a preset duration. Then, each valid charging segment is associated with the corresponding preceding parking segment and the following driving segment to obtain X triples. Based on the association strength and temporal characteristics of each triple, the system determines the pattern label corresponding to each triple. The pattern label is one of the following: charging while parking, unplugging and leaving, charging overnight at home, and charging at work.

[0249] Optionally, the technical effects that the generation system can achieve through the steps in the above embodiments include: (1) solving the problem of missing context caused by isolated analysis of charging events: by introducing a preset duration constraint, the effective charging segment is associated with the preceding parking segment and the following driving segment to form X triples. This association method restores the user's complete vehicle usage chain, enabling the system to understand the specific context of the charging behavior, which provides a complete data foundation for a more accurate understanding of user habits in the future; (2) realizing a fine-grained profile of user charging behavior and improving the accuracy of behavior recognition: this solution not only generates triples, but also further determines the pattern label based on the association strength and temporal characteristics. This fine-grained classification (parking and charging, unplugging and leaving, charging at home overnight, and charging at work) can be more accurate. The system accurately depicts users' charging habits. For example, the "home charging overnight" mode implies that users have the conditions for home charging and are used to charging at night, while the "workplace charging" mode implies that users rely on charging facilities at their workplace. (3) It improves the adaptability of the strategy: If the system identifies the home charging overnight mode, it can prioritize recommending the reservation charging strategy that utilizes off-peak electricity at night to maximize off-peak electricity benefits. If the system identifies the workplace charging mode, it can consider the charging availability at the workplace and suggest that users use their working hours to charge. If the system identifies the "pick up and go" mode, it indicates that users are sensitive to charging time. The system needs to make a more refined trade-off between departure guarantee and charging time. Through this strategy generation based on behavior patterns, the system can provide charging suggestions that truly conform to users' actual life patterns and car usage habits, thereby improving user acceptance and the effectiveness of strategy execution.

[0250] In one optional embodiment, the generation system first determines the vehicle's reservation charging suitability based on the destination, a list of charging locations, a list of charging events, user driving behavior characteristics, and pattern labels corresponding to X triples. Then, if the vehicle's reservation charging suitability is greater than or equal to a preset suitability threshold, B candidate charging strategies are obtained, where B is a positive integer. Next, the strategy benefit of each candidate charging strategy is determined based on the commuting route list, charging control parameters, user driving behavior characteristics, a risk event list, pattern labels corresponding to X triples, and preset electricity price data. The strategy benefit is used to characterize off-peak electricity revenue, vehicle battery charge health, departure guarantee, and user disturbance cost. Finally, the candidate charging strategy with the highest strategy benefit is selected as the target charging strategy.

[0251] Optionally, the generation system performs the following steps to determine the target charging strategy:

[0252] (1) Calculate the vehicle's scheduled charging compatibility. The calculation formula is as follows:

[0253] ;

[0254] In the above formula, : Arrival time stability (1 - standard deviation of arrival time / 2h, normalized to [0,1]); : Departure time stability (1 - standard deviation of departure time / 2h, normalized to [0,1]); Home charging availability (home charging frequency / total charging frequency); Nighttime window (percentage of days with a nighttime stay of ≥6 hours); Charging habits (typically, if the initial charging SOC is <50% and the final charging SOC is >80%, then it is 1).

[0255] Recommended weight values ​​above: .

[0256] (2) Adaptability level determination, the determination criteria are as follows:

[0257] High compatibility: Chinese adaptation: Low compatibility: .

[0258] (3) Generate the strategy payoff function. For each candidate strategy a (obtained by the combination of start time and target SOC), calculate the strategy payoff. The calculation formula is as follows:

[0259] ;

[0260] In the above formula, Off-peak electricity revenue (the percentage of electricity generated during off-peak hours according to the strategy). Health-Friendliness (Avoiding full charge earns 1.0, 80-90% earns 0.9, 90-95% earns 0.75, >95% earns 0.5); Departure assurance (the probability that the SOC will meet demand at the time of departure, estimated based on historical travel battery power distribution); User disruption cost (waiting cost caused by delayed charging, cost of 0.1 for a 1-hour delay).

[0261] Recommended weight values ​​above:

[0262] (4) Select the optimal policy from the candidate policies, i.e., select The most important strategy is to recommend the target charging strategy to the user.

[0263] The output data obtained through the above processing steps includes: ChargingAdaptabilityScore, AdaptabilityLevel, and RecommendedStrategy.

[0264] Optionally, the technical effects that the generation system can achieve through the steps in the above embodiments include:

[0265] (1) By conducting a pre-assessment of suitability, the relevance and effectiveness of the recommendation strategy are improved, and invalid recommendations are avoided: The suitability of scheduled charging is calculated based on the destination, the list of charging locations, the list of charging events, the user's driving behavior characteristics, and the pattern labels corresponding to X triples. Only when the suitability of scheduled charging is greater than or equal to the preset suitability threshold will the strategy generation stage be entered, ensuring that the system only recommends scheduled charging to users who have regular charging conditions and habits, thereby improving the adoption rate of the recommendation strategy and user satisfaction; (2) A multi-dimensional strategy benefit evaluation model is constructed: This scheme does not use "lowest electricity cost" or "highest power consumption" as the sole evaluation criteria, but defines a model that includes four The strategic benefits across dimensions include: Off-peak electricity revenue: maximizing economic benefits by combining preset electricity price data; Vehicle battery charge health: prioritizing charging intervals that are friendly to battery life (e.g., avoiding high temperatures and prolonged full charging) by combining charging control parameters and a list of risk events; Departure assurance: ensuring users have no anxiety about their journey the next day by combining commuting route lists (predicting power consumption) and user driving behavior characteristics; User disruption cost: quantifying users' tolerance for charging delays or restrictions to avoid excessive interference in users' lives. By comprehensively considering these four mutually constraining goals (e.g., low electricity costs and fast charging may conflict, high assurance and high health may conflict), the system can find a globally optimal solution, rather than a locally optimal solution.

[0266] In one optional embodiment, after determining the vehicle's target charging strategy based on the destination, charging location list, commuting route list, charging event list, charging control parameters, user driving behavior characteristics, risk event list, pattern labels corresponding to X triples, and preset electricity price data, the generation system first determines the evidence chain confidence of the target charging strategy based on the evidence chain used in the determination process, and determines the strategy output level of the target charging strategy based on the evidence chain confidence. The evidence chain includes a confidence score, location score, commuting score, risk score, and charging confidence. And / or, conflict detection is performed on the target charging strategy determination process. If conflicting operations exist in the determination process, a rollback operation is performed on the steps with conflicting operations, and the strategy output level of the target charging strategy is downgraded.

[0267] Optionally, after completing the reasoning process for the target charging strategy, the generation system can also perform the following steps:

[0268] (1) Calculate the total confidence level of the chain of evidence:

[0269] For each conclusion y (such as "home location" or "commuter user") obtained in the reasoning process of the target charging strategy, summarize the confidence of the evidence chain supporting that conclusion. :

[0270] ;

[0271] In the above formula, : The confidence level of the m-th piece of evidence (e.g., credibility score, charging confidence level, etc.); Quantity of evidence.

[0272] (2) Strategy output level Layering determination:

[0273] ;

[0274] Recommended values ​​for the above stratification thresholds: .

[0275] (3) Constructing a chain of evidence:

[0276] For each conclusion, record the evidence supporting that conclusion, and record the evidence type (e.g., "frequency of nighttime stays", "confidence level of charging"), evidence value, and confidence level of each piece of evidence.

[0277] The output data obtained through the above processing steps includes: LayeredLabels, EvidenceChains, and ConfidenceScores.

[0278] Optionally, after completing the reasoning process for the target charging strategy, the generation system can also perform the following steps:

[0279] (1) Define the types of evidence conflicts in the reasoning process, for example:

[0280] Location conflict: Home / workplace is determined to be the same location;

[0281] Behavioral conflict: High probability of commuting but no fixed route;

[0282] Charging conflict: Home chargers account for a high percentage of users, but there are no home charging points available.

[0283] (2) Determine the backoff loss for each conflict The calculation formula is as follows:

[0284]

[0285] In the above formula, Number of conflicts; Number of missing evidence; Number of logical transitions; The number of pieces of evidence that have been verified.

[0286] Recommended weight values ​​above:

[0287] (3) Triggering rollback operation:

[0288] like This triggers a rollback operation, which includes at least one of the following operations:

[0289] Downgrade the output (e.g., downgrade from "stronger inference" to "weaker inference"), mark conflict points, and suggest manual review.

[0290] The output data obtained through the above processing steps includes: CorrectedResults, Conflicts, and ManualReviewSuggestions.

[0291] Optionally, the technical effects that the generation system can achieve through the steps in the above embodiments include: (1) increasing users' trust in the recommendation results: by constructing an evidence chain containing credibility score, location score, commuting score, risk score, and charging confidence, and calculating the confidence of the evidence chain, the generation of each target charging strategy is verifiable. This evidence chain-based output enhances the transparency of the generated target charging strategy, which helps users understand and accept the recommendation; (2) effectively curbing the propagation of erroneous reasoning and improving the robustness of the system through conflict detection and rollback mechanisms; by introducing a conflict detection mechanism to identify conflicting operations in the determination process. And execute rollback operation to correct or abandon conflicting intermediate conclusions, avoiding the direct result of incorrect conclusions leading to incorrect target charging strategies. This self-correction mechanism improves the stability of the system under data noise or abnormal scenarios; (3) By dynamically adjusting the strategy output level, risk-controllable hierarchical recommendation is achieved, and user experience is optimized: when there is a conflicting operation or the confidence of the evidence chain is low, a downgrade operation is executed to reduce the strategy level. This hierarchical mechanism not only ensures efficient decision-making when the data is reliable, but also reminds users to make manual confirmation by reducing the output level when the data is questionable, thus avoiding the decline in user experience caused by blindly executing unreliable strategies.

[0292] Figure 4 This is a flowchart of an optional vehicle behavior analysis method based on fragment-level operational data according to an embodiment of this application, such as... Figure 4 As shown, the method includes the following steps:

[0293] Step S1, Segment-level vehicle operation data access and standardization: Access the original segment data uploaded by the vehicle, parse the core fields (segment identifier, time, space, mileage speed, energy, battery, etc.), unify the time format to UTC timestamp, unify the spatial coordinates to WGS84 coordinate system, derive time labels (date, hour, weekday, weekday / weekend, daytime / nighttime, morning peak / evening peak), and remove obviously abnormal segments.

[0294] Step S2, Data Quality Review and Credibility Stratification: For each segment, a credibility score is calculated based on four dimensions: temporal continuity, physical consistency, field completeness, and geographical stability. The scores are weighted and summed to obtain a comprehensive credibility score. Based on a threshold, the segments are classified into three levels: high credibility, medium credibility, and low credibility. Low-credibility segments are automatically downgraded or skipped in subsequent steps.

[0295] Step S3, Spatiotemporal Dwell Point Generation and Clustering: Based on high-confidence and medium-confidence parking segments, spatial clustering is performed using kernel density weights, GPS drift points are merged, dwell point clusters are generated, and statistical information such as the center coordinates, access frequency, and average dwell time of each cluster are calculated.

[0296] Step S4, Home / Workplace / Frequently Used Charging Point Candidate Inference: Based on the nighttime stay characteristics, weekday stay characteristics, stay duration characteristics, repetition characteristics, and departure characteristics of the lodging point cluster, calculate the comprehensive location score and infer the candidate home and workplace locations. Cluster the charging locations to identify frequently used charging points.

[0297] Step S5, Fixed Path Extraction and Commuting Probability Scoring: The origin and destination of the driving segment are normalized by grid, similar paths are merged, and commuting probability scores are calculated based on time window consistency, origin and destination fixity, distance stability, and repeatability to identify fixed paths and commuting paths.

[0298] Step S6, Typical charging event identification and threshold extraction: Calculate the confidence of charging events (four factors: SOC rise, negative power, dwell time, and duration) for charging segments, filter low-confidence events, use the median to estimate typical starting SOC and typical ending SOC, and identify typical charging periods.

[0299] Step S7, Parking-Charging-Departure Association Behavior Modeling: Identify the triple association of parking events, charging events, and departure events, calculate the association strength (three factors: arrival evidence, charging evidence, and departure evidence), and classify behavioral patterns (parking and charging immediately, unplugging and leaving immediately, charging at home overnight, charging at work, etc.).

[0300] Step S8, Low SOC Behavior Change Identification: Divide low SOC segments into normal segments, compare speed, distance, destination, and proportion of long-distance travel, calculate the low SOC behavior offset index, and determine whether there is conservative driving behavior due to low battery.

[0301] Step S9, Risk Event and Spatiotemporal Scene Coupling Analysis: Identify battery risk segments (voltage difference abnormality, temperature difference abnormality, high temperature, high power, high rate charging, low SOC high load), map risk events to location, time period, and scene category, calculate risk scene coupling score, and identify recurring risk points.

[0302] Step S10, Determining and Generating the Applicability of Scheduled Charging: Calculate the applicability score of scheduled charging based on the stability of arrival time, departure time, home charging availability, nighttime window, and charging habits. Generate a strategy benefit function based on off-peak electricity revenue, health friendliness, departure guarantee, and user disturbance cost, and select the optimal scheduled charging strategy.

[0303] Step S11, Evidence Chain Construction and Layered Label Output: For each inference conclusion, summarize the confidence levels of multiple pieces of evidence supporting the conclusion, calculate the total confidence level of the evidence chain, and categorize the conclusions into four types based on the confidence levels: factual statistics, strong inference, weak inference, and risk warning.

[0304] Step S12, Abnormal rollback and conflict correction: Detect inference conflicts (location conflicts, behavior conflicts, charging conflicts, etc.), calculate the conflict rollback loss function, and trigger rollback processing (degraded output, marked conflict, and suggested manual review) when the loss function exceeds the threshold.

[0305] Figure 5 This is an architecture diagram of an optional vehicle behavior analysis system based on fragment-level operational data according to an embodiment of this application, such as... Figure 5 As shown, the system includes the following modules:

[0306] The system comprises the following modules: Data Access Module: Accesses segment-level vehicle operation data and performs standardized processing; Credibility Assessment Module: Calculates and grades the credibility score for each segment; Dwell Point Analysis Module: Performs dwell point clustering and location inference; Path Pattern Module: Identifies fixed paths and commuting paths; Charging Behavior Module: Identifies charging events and extracts charging habit thresholds; Related Behavior Module: Models parking-charging-departure related behaviors; Low Battery Analysis Module: Calculates the low SOC behavior offset index; Risk Scenario Module: Calculates the risk scenario coupling score; Strategy Generation Module: Generates a reservation charging suitability score and recommendation strategy; Tag Output Module: Outputs hierarchical tags and evidence chains.

[0307] As can be seen from the above, the technical effects that the vehicle behavior analysis method based on fragment-level operation data can achieve include: (1) improving analysis accuracy: reducing misjudgments caused by low-quality slices through segment credibility layering; (2) enhancing system robustness: automatically downgrading or skipping previous low-confidence results through a multi-stage reasoning chain with credibility constraints, avoiding error accumulation and propagation; (3) improving interpretability: outputting evidence chains and hierarchical confidence levels, allowing users and engineers to trace the basis for judgment, facilitating manual review; each label has a clear evidence chain and confidence score, no longer an opaque black box judgment; (4) enhancing engineering application value: associating behavior analysis results with battery risk scenarios and scheduled charging strategies, which can be directly used for battery design optimization and charging function development; (5) adapting to fragment-level data: specifically designed for the characteristics of fragment-level data, not relying on complete trajectories, suitable for large-scale vehicle-side data upload scenarios; the data volume is only less than 1 / 100 of frame-by-frame data, significantly reducing transmission and storage costs; (6) supporting scheduled charging optimization: providing user scheduled charging adaptability judgment and personalized strategy recommendations, supporting grid peak shaving and valley filling and battery life optimization.

[0308] Example 2

[0309] This application embodiment also provides a device for determining a charging strategy based on vehicle behavior segments. It should be noted that the device for determining a charging strategy based on vehicle behavior segments in this application embodiment can be used to execute the method for determining a charging strategy based on vehicle behavior segments provided in this application embodiment. The following describes the device for determining a charging strategy based on vehicle behavior segments provided in this application embodiment.

[0310] According to an embodiment of this application, an apparatus for implementing the above-described method for determining a charging strategy based on vehicle behavior segments is also provided. Figure 6 This is a schematic diagram of an optional charging strategy determination device based on vehicle behavior segments according to an embodiment of this application, as shown below. Figure 6 As shown, the device includes: a credibility detection unit 601, a fragment removal unit 602, a first determination unit 603, a triplet generation unit 604, and a second determination unit 605.

[0311] Optionally, the credibility detection unit 601 is used to perform a weighted summation of the L segment scores of each standard segment in the standard segment dataset to obtain a credibility score for each standard segment, where L is a positive integer, and the L segment scores include a time continuity score, a physical consistency score, a field integrity score, and a geographical stability score. Each standard segment is used to record vehicle behavior data. The segment elimination unit 602 is used to eliminate standard segments in the standard segment dataset whose credibility scores are less than a preset credibility threshold to obtain a target segment dataset, where each target segment in the target segment dataset is one of the following: a charging segment, a driving segment, or a parking segment. The first determination unit 603 is used to determine the vehicle's destination location and charging location based on the target segment dataset. The dataset includes a target segment dataset, a commuting route list, a charging event list, charging control parameters, user driving behavior characteristics, and a risk event list, where the destination includes home and workplace locations. A triplet generation unit 604 generates X triplet sets based on the target segment dataset and the charging event list, where X is a positive integer. Each triplet records the correspondence between valid charging segments, preceding parking segments, and subsequent driving segments in the charging event list, and each triplet corresponds to a pattern label. A second determination unit 605 determines the vehicle's target charging strategy based on the destination, charging location list, commuting route list, charging event list, charging control parameters, user driving behavior characteristics, risk event list, the pattern labels corresponding to the X triplet sets, and preset electricity price data.

[0312] This application employs a multi-stage behavioral reasoning and credibility correction approach. First, it calculates credibility by weighted summation of scores for the temporal continuity, physical consistency, field completeness, and geographical stability of standard segments, and then removes low-credibility segments to obtain the target segment dataset. Next, based on this dataset, it generates X triples containing the correspondence between charging, driving, and parking segments. It then integrates multi-dimensional information including destination points, charging location lists, commuting route lists, charging event lists, charging control parameters, user driving behavior characteristics, risk event lists, and pattern labels, combined with preset electricity price data, to determine the vehicle's target charging strategy. This achieves the goal of full-process correlation modeling from data cleaning and behavioral pattern recognition to strategy generation, even in the context of quality defects in vehicle operation data and error propagation caused by independent processing of operation data by different inference modules. This improves the accuracy of operation data analysis, outputs interpretable vehicle charging strategies, and increases user adoption rates of recommended charging strategies, thus solving the technical problems of poor analysis results and inaccurate recommended charging strategies in existing technologies for new energy vehicle operation data.

[0313] In one optional embodiment, the first determining unit 603 includes: a first determining subunit, a generating subunit, a second determining subunit, a third determining subunit, and a fourth determining subunit.

[0314] Optionally, the first determining subunit is used to determine the destination point based on the time label table and the list of stopping points corresponding to the target segment dataset after determining the list of stopping points based on the parking segments in the target segment dataset, wherein the time label table includes the derived time label of each target segment; the generating subunit is used to generate a list of charging points based on the cluster center coordinates of each charging point cluster obtained by clustering the charging points of all charging segments in the target segment dataset; the second determining subunit is used to determine a list of commuting routes based on the driving segments and the destination point in the target segment dataset; the third determining subunit is used to determine a list of charging events and charging control parameters based on the charging segments in the target segment dataset; and the fourth determining subunit is used to determine the user's driving behavior characteristics based on the driving segments in the target segment dataset, and to determine a list of risk events based on the target segment dataset, the destination point, the list of charging points, and a preset risk field.

[0315] In one optional embodiment, the first determining subunit includes: a first determining module, a first clustering module, a first statistics module, a first filtering module, a first summarizing module, a second determining module, a third determining module, and a fourth determining module.

[0316] Optionally, the first determining module is used to select the endpoint location of each parking segment in the target segment dataset as a candidate dwell location, and calculate the kernel density weight of each parking segment with each candidate dwell location as the center; the first clustering module is used to perform spatial clustering of the candidate dwell locations of all parking segments based on the kernel density weight of each parking segment to obtain A dwell location clusters, where A is a positive integer, and each dwell location cluster includes at least two candidate dwell locations; the first statistics module is used to count the location cluster index of each dwell location cluster in the A dwell location clusters, where the location cluster index includes the cluster center coordinates, access frequency, average dwell time, and dwell time standard deviation; the first filtering module is used to perform cluster filtering on the A dwell location clusters based on the location cluster index of each dwell location cluster, where cluster filtering is used to remove dwell location clusters whose location cluster index is less than a preset index threshold; the first summarizing module is used to use the cluster center coordinates of the dwell location clusters that have passed the cluster filtering as dwell locations, and summarize all dwell locations. The system obtains a list of lodging locations; the second determination module is used to determine P location features for each lodging location based on the time label table corresponding to the target segment dataset and the location cluster index of the lodging location cluster to which each lodging location belongs in the list of lodging locations, where P is a positive integer, and the P location features include the proportion of nighttime stays, the proportion of weekday daytime stays, the normalized value of the average stay duration, the ratio of consecutive stay days to total stay days, and the proportion of times it is used as a departure point; the third determination module is used to determine the first location score and the second location score for each lodging location based on the P location features, where the first location score is used to represent the probability that each lodging location is a home location, and the second location score is used to represent the probability that each lodging location is a work location; the fourth determination module is used to select the lodging location with the largest first location score in the list of lodging locations as a home location, select the lodging location with the largest second location score in the list of lodging locations as a work location, and select the home location and the work location as destination points.

[0317] In one optional embodiment, the second determining subunit includes: a mapping module, a first partitioning module, a second statistics module, a fifth determining module, and a sixth determining module.

[0318] Optionally, the mapping module is used to map the starting point and ending point of each driving segment in the target segment dataset to a spatial grid of a preset scale, obtaining X spatial paths corresponding to X driving segments, where X is a positive integer; the first partitioning module is used to divide the X spatial paths into Y path sets, where Y is a positive integer less than or equal to X, and the starting points of any two spatial paths in each path set are located in the same spatial grid; the second statistics module is used to count the path indicators of each path set in the Y path sets, where the path indicators include frequency of occurrence, average distance, average duration, and average speed; the fifth determination module is used to determine the commuting score of each path set based on the path indicators of each path set; the sixth determination module is used to take any spatial path in each path set with a commuting score greater than or equal to a preset commuting threshold as a commuting path, and summarize the Y commuting paths corresponding to the Y path sets to obtain a commuting path list.

[0319] In one optional embodiment, the third determining subunit includes: a third statistical module, a seventh determining module, an eighth determining module, a second clustering module, a ninth determining module, a tenth determining module, and an eleventh determining module.

[0320] Optionally, the third statistical module is used to statistically analyze the charging indicators of each charging segment in the target segment dataset, wherein the charging indicators include the vehicle's state of charge, power information, and charging period; the seventh determination module is used to determine the charging confidence of each charging segment based on the charging indicators of each charging segment; the eighth determination module is used to identify charging segments with a charging confidence greater than or equal to a preset confidence threshold as valid charging segments, and to summarize all valid charging segments to obtain a charging event list; the second clustering module is used to cluster the charging start times corresponding to all valid charging segments in the charging event list to obtain the typical charging period of the vehicle, wherein each typical charging period is the cluster center period of the cluster of valid charging period clusters obtained by clustering; the ninth determination module is used to determine the typical starting charging charge value and typical ending charging charge value of the vehicle based on the median of the state of charge corresponding to all valid charging segments in the charging event list; the tenth determination module is used to determine the fast and slow charging ratio of the vehicle based on the power information corresponding to all valid charging segments in the charging event list; and the eleventh determination module is used to use the typical charging period, typical starting charging charge value, typical ending charging charge value, and fast and slow charging ratio of the vehicle as charging control parameters.

[0321] In one optional embodiment, the fourth determining subunit includes: a second dividing module, a twelfth determining module, a thirteenth determining module, a fourteenth determining module, a fifteenth determining module, and a risk event generation module.

[0322] Optionally, the second segmentation module is used to divide all driving segments in the target segment dataset into a first segment set and a second segment set based on a preset lower charge limit value, wherein the vehicle in each driving segment in the first segment set has a charge value less than the preset lower charge limit value at the start and / or end of driving, and the vehicle in each driving segment in the second segment set has a charge value greater than or equal to the preset lower charge limit value at both the start and end of driving; the twelfth determination module is used to determine the vehicle's driving behavior deviation index based on the driving indicators of the first segment set and the second segment set, and to determine the user's driving behavior characteristics based on the driving behavior deviation index, wherein the driving indicators include average speed, average distance, and driving destination (home / charging station). The module comprises the following components: the proportion of charging locations and the proportion of long-distance travel segments; the thirteenth determination module, which identifies target segments in the target segment dataset that hit a preset risk field as risk segments; the fourteenth determination module, which determines the associated scenarios for each risk segment based on the destination and charging location lists, wherein the associated scenarios are one of the following: home location, workplace location, or charging location; the fifteenth determination module, which determines the risk score for each risk segment based on the associated scenarios, wherein the risk scores include pressure difference risk score, temperature difference risk score, power risk score, and recurrence risk score; and the risk event generation module, which generates risk events based on risk segments with risk scores greater than or equal to a preset risk threshold and associated scenarios, and summarizes all risk events to obtain a risk event list.

[0323] In an optional embodiment, the triple generation unit 604 includes a fifth determining subunit and an associating subunit.

[0324] Optionally, the fifth determining subunit is used to determine the preceding parking segment and the following driving segment corresponding to each valid charging segment in the charging event list based on the target segment dataset, wherein the duration between the charging start time of each valid charging segment and the parking end start time of the corresponding preceding parking segment is less than a preset duration, and the duration between the charging end time of each valid charging segment and the driving start time of the corresponding following driving segment is less than a preset duration; the association subunit is used to associate each valid charging segment with the corresponding preceding parking segment and the following driving segment to obtain X triples, and to determine the pattern label corresponding to each triple based on the association strength and temporal characteristics of each triple, wherein the pattern label is one of the following: charging while parking, unplugging and leaving, charging overnight at home, and replenishing power at work.

[0325] In one optional embodiment, the second determining unit 605 includes: a sixth determining subunit, an acquisition subunit, a seventh determining subunit, and an eighth determining subunit.

[0326] Optionally, the sixth determining subunit is used to determine the vehicle's reservation charging suitability based on the destination, charging location list, charging event list, user driving behavior characteristics, and pattern labels corresponding to X triples; the obtaining subunit is used to obtain B candidate charging strategies when the vehicle's reservation charging suitability is greater than or equal to a preset suitability threshold, where B is a positive integer; the seventh determining subunit is used to determine the strategy benefit of each candidate charging strategy based on the commuting route list, charging control parameters, user driving behavior characteristics, risk event list, pattern labels corresponding to X triples, and preset electricity price data, where the strategy benefit is used to characterize off-peak electricity revenue, vehicle battery charge health, departure guarantee, and user disturbance cost; the eighth determining subunit is used to select the candidate charging strategy with the highest strategy benefit as the target charging strategy.

[0327] In one optional embodiment, the device for determining the charging strategy based on vehicle behavior segments further includes: a third determining unit and / or a conflict detection unit.

[0328] Optionally, the third determining unit is used to determine the confidence level of the evidence chain of the target charging strategy based on the evidence chain used in the determination process after determining the target charging strategy of the vehicle based on the destination, charging location list, commuting route list, charging event list, charging control parameters, user driving behavior characteristics, risk event list, pattern labels corresponding to X triples, and preset electricity price data. Furthermore, it determines the strategy output level of the target charging strategy based on the confidence level of the evidence chain. The evidence chain includes a confidence score, location score, commuting score, risk score, and charging confidence level. The conflict detection unit is used to perform conflict detection on the determination process of the target charging strategy. If there are conflicting operations in the determination process, it performs a rollback operation on the steps with conflicting operations and downgrades the strategy output level of the target charging strategy.

[0329] It should be noted that the units 601 to 605 mentioned above correspond to steps S101 to S105 in the method embodiment. The instances and application scenarios implemented by the units and the corresponding steps are the same, but are not limited to the content disclosed in the above embodiment.

[0330] Example 3

[0331] Embodiments of this application can also provide an electronic device. Figure 7 This is a structural block diagram of an electronic device according to an embodiment of this application, such as... Figure 7 As shown, the electronic device includes: one or more ( Figure 7 (Only one is shown) processor 702, memory 704, memory controller, and peripheral interface, wherein the peripheral interface is connected to the radio frequency module, audio module and display.

[0332] The memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the methods and devices in the embodiments of this application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, thereby realizing the above-mentioned method for determining the charging strategy based on vehicle behavior segments.

[0333] The memory may include high-speed random access memory (RAM), and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include memory remotely located relative to the processor, which can be connected to the terminal via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks (LANs), mobile communication networks, and combinations thereof.

[0334] The processor can access information and applications stored in the memory via a transmission device to perform the following steps: A weighted sum of L segment scores for each standard segment in the standard segment dataset is calculated to obtain a credibility score for each standard segment, where L is a positive integer, and the L segment scores include temporal continuity score, physical consistency score, field integrity score, and geographical stability score. Each standard segment is used to record vehicle behavior data. Standard segments with credibility scores below a preset credibility threshold are removed from the standard segment dataset to obtain a target segment dataset. Each target segment in the target segment dataset is one of the following: charging segment, driving segment, or parking segment. Based on the target segment dataset, the target segment of the vehicle is determined. The dataset includes a target segment dataset, a list of charging locations, a list of commuting routes, a list of charging events, charging control parameters, user driving behavior characteristics, and a list of risk events. Destination points include home and work locations. Based on the target segment dataset and the charging event list, X triples are generated, where X is a positive integer. Each triple records the correspondence between a valid charging segment, a preceding parking segment, and a subsequent driving segment in the charging event list, and each triple corresponds to a pattern label. The target charging strategy for the vehicle is determined based on the destination points, the charging location list, the commuting route list, the charging event list, the charging control parameters, user driving behavior characteristics, the risk event list, the pattern labels corresponding to the X triples, and preset electricity price data.

[0335] This application provides a vulnerability detection scheme for smart contracts. The scheme employs a combination of multi-stage behavioral reasoning and credibility correction. First, it calculates credibility by weighted summation of scores for the temporal continuity, physical consistency, field integrity, and geographical stability of standard segments, and then eliminates low-credibility segments to obtain a target segment dataset. Based on this dataset, it generates X triples containing the correspondence between charging, driving, and parking segments. It then integrates multi-dimensional information including destination, charging location list, commuting route list, charging event list, charging control parameters, user driving behavior characteristics, risk event list, and pattern labels, combined with preset electricity price data, to determine the vehicle's target charging strategy. This achieves the goal of full-process correlation modeling from data cleaning and behavioral pattern recognition to strategy generation, even in the context of quality defects in vehicle operation data and error propagation from independent processing of operation data by different reasoning modules. This improves the accuracy of operation data analysis, outputs interpretable vehicle charging strategies, and increases user adoption rates of recommended charging strategies, thus solving the technical problems of poor analysis results and inaccurate recommended charging strategies in existing technologies for new energy vehicle operation data.

[0336] Those skilled in the art will understand that Figure 7 The structure shown is for illustrative purposes only. Electronic devices can also be smartphones, tablets, PDAs, mobile internet devices, PADs, and other terminal devices. Figure 7 This does not limit the structure of the aforementioned electronic device. For example, electronic devices may also include components that are more... Figure 7 The more or fewer components shown (such as network interfaces, display devices, etc.), or having the same Figure 7 The different configurations shown.

[0337] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing the hardware related to the terminal device. The program can be stored in a computer-readable storage medium, which may include: flash drive, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0338] Example 4

[0339] Embodiments of this application may also provide a storage medium.

[0340] Optionally, in this embodiment of the application, the storage medium can be used to store the program code executed by the method for determining the charging strategy based on vehicle behavior segments provided in the above method embodiment.

[0341] Optionally, in this embodiment, the storage medium may be located in any computer terminal in a group of computer terminals in a computer network, or in any mobile terminal in a group of mobile terminals.

[0342] This application also provides a computer program product, which, when executed on a data processing device, is adapted to perform the steps of a method for determining a charging strategy based on vehicle behavior segments.

[0343] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0344] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0345] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0346] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0347] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0348] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0349] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for determining a charging strategy based on vehicle behavior segments, characterized in that, include: The credibility score of each standard segment is obtained by weighted summing of the L segment scores of each standard segment in the standard segment dataset, where L is a positive integer. The L segment scores include time continuity score, physical consistency score, field integrity score, and geographical stability score. Each standard segment is used to record vehicle behavior data. The standard segments with a confidence score lower than a preset confidence threshold in the standard segment dataset are removed to obtain the target segment dataset, wherein each target segment in the target segment dataset is one of the following: charging segment, driving segment, parking segment; Based on the target segment dataset, determine the vehicle's destination location, charging location list, commuting route list, charging event list, charging control parameters, user driving behavior characteristics, and risk event list, wherein the destination location includes home location and workplace location; X triples are generated based on the target segment dataset and the charging event list, where X is a positive integer. Each triple is used to record the correspondence between the valid charging segment, the preceding parking segment, and the following driving segment in the charging event list. Each triple corresponds to a pattern label. The target charging strategy for the vehicle is determined based on the destination, the list of charging locations, the list of commuting routes, the list of charging events, the charging control parameters, the user's driving behavior characteristics, the list of risk events, the pattern labels corresponding to the X triples, and the preset electricity price data.

2. The method for determining a charging strategy based on vehicle behavior segments according to claim 1, characterized in that, Based on the target segment dataset, the vehicle's destination location, charging location list, commuting route list, charging event list, charging control parameters, user driving behavior characteristics, and risk event list are determined, including: After determining the list of stopping locations based on the parking segments in the target segment dataset, the destination is determined based on the time tag table corresponding to the target segment dataset and the list of stopping locations, wherein the time tag table includes the derived time tag of each target segment; After clustering the charging locations of all charging segments in the target segment dataset, the charging location list is generated based on the cluster center coordinates of each charging location cluster obtained from the clustering. The commuter route list is determined based on the driving segments in the target segment dataset and the destination points; The charging event list and the charging control parameters are determined based on the charging segments in the target segment dataset. The user's driving behavior characteristics are determined based on driving segments in the target segment dataset, and the risk event list is determined based on the target segment dataset, the destination, the charging location list, and the preset risk field.

3. The method for determining a charging strategy based on vehicle behavior segments according to claim 2, characterized in that, After determining the list of stopping points based on parking segments in the target segment dataset, the destination point is determined based on the time-stamp table corresponding to the target segment dataset and the list of stopping points, including: The endpoint location of each parking segment in the target segment dataset is used as a candidate stopping point, and the kernel density weight of each parking segment is calculated with each candidate stopping point as the center. Based on the kernel density weight of each parking segment, spatial clustering is performed on the candidate parking locations of all parking segments to obtain A parking location clusters, where A is a positive integer, and each parking location cluster includes at least two candidate parking locations; Statistical analysis is performed on the location cluster indicators for each of the A location clusters, wherein the location cluster indicators include cluster center coordinates, access frequency, average stay duration, and standard deviation of stay duration; Cluster filtering is performed on the A clusters of the host locations based on the location cluster index of each host location cluster, wherein the cluster filtering is used to remove host location clusters whose location cluster index is less than a preset index threshold; The cluster center coordinates of the clusters of dwelling locations filtered by the cluster are used as dwelling locations. All dwelling locations are summarized to obtain the list of dwelling locations. Based on the time label table corresponding to the target segment dataset and the location cluster index of the location cluster to which each location belongs in the list of locations, P location features are determined for each location, where P is a positive integer. The P location features include the percentage of nighttime stays, the percentage of weekday daytime stays, the normalized value of the average stay duration, the ratio of consecutive stay days to total stay days, and the percentage of times the location is used as a departure point. Based on P location features of each residence location, a first location score and a second location score are determined for each residence location, wherein the first location score is used to characterize the probability that each residence location is the home location, and the second location score is used to characterize the probability that each residence location is the work location; The place with the highest first location rating in the list of places of residence is designated as the home location, the place with the highest second location rating in the list of places of residence is designated as the work location, and the home location and the work location are designated as the destination.

4. The method for determining a charging strategy based on vehicle behavior segments according to claim 2, characterized in that, The commuter route list is determined based on the driving segments in the target segment dataset and the destination points, including: The starting point and ending point of each driving segment in the target segment dataset are mapped to a spatial grid of a preset scale to obtain X spatial paths corresponding to X driving segments, where X is a positive integer; The X spatial paths are divided into Y path sets, where Y is a positive integer less than or equal to X. The starting points of any two spatial paths in each path set are in the same spatial grid. Statistical analysis is performed on the path metrics for each of the Y path sets, wherein the path metrics include frequency of occurrence, average distance, average duration, and average speed. A commuting score for each set of routes is determined based on the route metrics for each set of routes. Take any spatial path in each path set whose commuting score is greater than or equal to a preset commuting threshold as a commuting path, and summarize the Y commuting paths corresponding to the Y path sets to obtain the commuting path list.

5. The method for determining a charging strategy based on vehicle behavior segments according to claim 2, characterized in that, Determining the charging event list and the charging control parameters based on the charging segments in the target segment dataset includes: The charging metrics for each charging segment in the target segment dataset are statistically analyzed, wherein the charging metrics include the vehicle's state of charge, power information, and charging period. The charging confidence level of each charging segment is determined based on the charging index of each charging segment; Charging segments with a charging confidence level greater than or equal to a preset confidence threshold are considered valid charging segments. All valid charging segments are then aggregated to obtain the charging event list. Cluster the charging start times corresponding to all valid charging segments in the charging event list to obtain the typical charging periods of the vehicle, wherein each typical charging period is the cluster center period of the cluster of valid charging periods obtained by clustering. The typical starting charge value and typical ending charge value of the vehicle are determined based on the median of the state of charge corresponding to all valid charging segments in the charging event list. The fast and slow charging ratio of the vehicle is determined based on the power information corresponding to all valid charging segments in the charging event list. The typical charging period, typical starting charge value, typical ending charge value, and fast / slow charging ratio of the vehicle are used as the charging control parameters.

6. The method for determining a charging strategy based on vehicle behavior segments according to claim 2, characterized in that, The user's driving behavior characteristics are determined based on driving segments in the target segment dataset, and the risk event list is determined based on the target segment dataset, the destination, the charging location list, and preset risk fields, including: Based on a preset lower charge limit, all driving segments in the target segment dataset are divided into a first segment set and a second segment set. In the first segment set, the vehicle in each driving segment has a charge value less than the preset lower charge limit at the start and / or end of driving. In the second segment set, the vehicle in each driving segment has a charge value greater than or equal to the preset lower charge limit at both the start and end of driving. The driving behavior deviation index of the vehicle is determined based on the driving indicators of the first segment set and the driving indicators of the second segment set, and the user's driving behavior characteristics are determined based on the driving behavior deviation index. The driving indicators include average speed, average distance, percentage of driving destinations being home / charging locations, and percentage of long-distance travel segments. The target fragments in the target fragment dataset that hit the preset risk field are considered as risk fragments; Based on the list of destinations and charging locations, the associated scenarios for each risk segment are determined, wherein the associated scenarios are one of the following: home location, workplace, or charging location; A risk score is determined for each risk segment based on the associated scenarios of each risk segment, wherein the risk score includes a pressure difference risk score, a temperature difference risk score, a power risk score, and a recurrence risk score. Risk events are generated based on risk segments and associated scenarios whose risk scores are greater than or equal to a preset risk threshold. All risk events are then aggregated to obtain the risk event list.

7. The method for determining a charging strategy based on vehicle behavior segments according to claim 1, characterized in that, X triples are generated based on the target fragment dataset and the charging event list, including: Based on the target segment dataset, determine the preceding parking segment and the following driving segment corresponding to each valid charging segment in the charging event list, wherein the duration between the charging start time of each valid charging segment and the parking end start time of the corresponding preceding parking segment is less than a preset duration, and the duration between the charging end time of each valid charging segment and the driving start time of the corresponding following driving segment is less than the preset duration. Each effective charging segment is associated with its corresponding preceding parking segment and subsequent driving segment to obtain X triplet sets. Based on the association strength and temporal characteristics of each triplet set, a mode label corresponding to each triplet set is determined. The mode label is one of the following: charging while parking, unplugging and leaving, charging at home overnight, and charging at work.

8. The method for determining a charging strategy based on vehicle behavior segments according to claim 1, characterized in that, The target charging strategy for the vehicle is determined based on the destination, the list of charging locations, the list of commuting routes, the list of charging events, the charging control parameters, the user's driving behavior characteristics, the list of risk events, the pattern labels corresponding to the X triples, and preset electricity price data, including: The vehicle's scheduled charging suitability is determined based on the destination, the list of charging locations, the list of charging events, the user's driving behavior characteristics, and the pattern labels corresponding to the X triples. If the vehicle's scheduled charging compatibility is greater than or equal to a preset compatibility threshold, obtain B candidate charging strategies, where B is a positive integer. Based on the commuting route list, the charging control parameters, the user driving behavior characteristics, the risk event list, the pattern labels corresponding to the X triples, and the preset electricity price data, the strategy benefit of each candidate charging strategy is determined, wherein the strategy benefit is used to characterize off-peak electricity benefit, vehicle battery charge health, departure guarantee and user disturbance cost. The candidate charging strategy that yields the highest benefit is selected as the target charging strategy.

9. The method for determining a charging strategy based on vehicle behavior segments according to claim 1, characterized in that, After determining the target charging strategy for the vehicle based on the destination, the charging location list, the commuting route list, the charging event list, the charging control parameters, the user driving behavior characteristics, the risk event list, the pattern labels corresponding to the X triples, and the preset electricity price data, the method for determining the charging strategy based on vehicle behavior segments further includes: The confidence level of the evidence chain of the target charging strategy is determined based on the evidence chain used in the determination process, and the strategy output level of the target charging strategy is determined based on the confidence level of the evidence chain. The evidence chain includes confidence score, location score, commuting score, risk score, and charging confidence level. And / or, perform conflict detection on the determination process of the target charging strategy, and if there is a conflict operation in the determination process, perform a rollback operation on the step with the conflict operation, and downgrade the strategy output level of the target charging strategy.

10. A device for determining a charging strategy based on vehicle behavior segments, characterized in that, include: The credibility detection unit is used to perform a weighted summation of the L segment scores of each standard segment in the standard segment dataset to obtain the credibility score of each standard segment, where L is a positive integer, and the L segment scores include time continuity score, physical consistency score, field integrity score and geographical stability score, and each standard segment is used to record vehicle behavior data; The segment elimination unit is used to eliminate standard segments in the standard segment dataset whose confidence scores are less than a preset confidence threshold, to obtain a target segment dataset, wherein each target segment in the target segment dataset is one of the following: charging segment, driving segment, or parking segment; The first determining unit is used to determine the vehicle's destination location, charging location list, commuting route list, charging event list, charging control parameters, user driving behavior characteristics, and risk event list based on the target segment dataset, wherein the destination location includes home location and workplace location; The triplet generation unit is used to generate X triplets based on the target segment dataset and the charging event list, where X is a positive integer. Each triplet is used to record the correspondence between the valid charging segment, the preceding parking segment, and the following driving segment in the charging event list. Each triplet corresponds to a pattern label. The second determining unit is used to determine the target charging strategy for the vehicle based on the destination, the charging location list, the commuting route list, the charging event list, the charging control parameters, the user driving behavior characteristics, the risk event list, the pattern labels corresponding to the X triples, and the preset electricity price data.

Images