WO2023241388A1 - 模型训练方法及装置、补能意图识别方法及装置、设备、介质 - Google Patents

模型训练方法及装置、补能意图识别方法及装置、设备、介质 Download PDF

Info

Publication number
WO2023241388A1
WO2023241388A1 PCT/CN2023/098229 CN2023098229W WO2023241388A1 WO 2023241388 A1 WO2023241388 A1 WO 2023241388A1 CN 2023098229 W CN2023098229 W CN 2023098229W WO 2023241388 A1 WO2023241388 A1 WO 2023241388A1
Authority
WO
WIPO (PCT)
Prior art keywords
data
energy
energy replenishment
intention
replenishment
Prior art date
Application number
PCT/CN2023/098229
Other languages
English (en)
French (fr)
Inventor
袁鲁峰
付振
王明月
刘拼拼
贾振坤
Original Assignee
中国第一汽车股份有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 中国第一汽车股份有限公司 filed Critical 中国第一汽车股份有限公司
Publication of WO2023241388A1 publication Critical patent/WO2023241388A1/zh

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/907—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
    • G06F16/909—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using geographical or spatial information, e.g. location
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/40—Business processes related to the transportation industry
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T10/00—Road transport of goods or passengers
    • Y02T10/60—Other road transportation technologies with climate change mitigation effect
    • Y02T10/70—Energy storage systems for electromobility, e.g. batteries

Definitions

  • This application relates to the field of intelligent transportation technology, such as model training methods and device replenishment intention identification methods and devices, equipment, and media.
  • This application provides model training methods and device energy replenishment intention identification methods and devices, equipment, and media to solve defects in related technologies and improve the accuracy of predictions.
  • this application provides a model training method, which includes:
  • the portrait data of the sample vehicle is constructed
  • the energy replenishment intention recognition model is trained.
  • this application also provides a method for identifying energy supplementation intentions, which includes:
  • the energy-replenishing intention recognition model is trained through the above-mentioned model training method.
  • this application also provides a model training device, which includes:
  • the sample data extraction module is configured to extract sample energy replenishment reference data from sample vehicle data
  • the behavior data determination module is configured to determine the energy replenishment behavior data based on the sample energy replenishment reference data
  • the portrait data construction module is configured to construct portrait data of the sample vehicle based on energy replenishment behavior data
  • the model training module is set to train the energy replenishment intention recognition model based on the energy replenishment behavior data and portrait data.
  • this application also provides an energy supplement intention recognition device, which includes:
  • the current data acquisition module is set to obtain the current energy replenishment reference data from the current vehicle data
  • the energy replenishment intention recognition module is configured to input the current energy replenishment reference data into the energy replenishment intention recognition model to obtain the energy replenishment intention recognition result at the current moment;
  • the energy-replenishing intention recognition model is trained through the model training device.
  • this application also provides an electronic device, which includes:
  • processors one or more processors
  • memory configured to store one or more programs
  • the one or more processors When one or more programs are executed by one or more processors, the one or more processors implement the above-mentioned model training method or energizing intention identification method.
  • this application example also provides a computer-readable storage medium on which a computer program is stored, wherein when the program is executed by a processor, the above-mentioned model training method or energy-enhancing intention identification method is implemented.
  • Figure 1 is a flow chart of a model training method provided in Embodiment 1 of the present application.
  • FIG. 2 is a schematic diagram of the energy replenishment intention data selection principle provided by Embodiment 1 of the present application;
  • Figure 3 is a flow chart of a model training method provided in Embodiment 2 of the present application.
  • Figure 4 is a flow chart of a model training method provided in Embodiment 3 of the present application.
  • Figure 5 is a statistical distribution diagram of refueling time intervals provided in Embodiment 3 of the present application.
  • Figure 6 is a statistical distribution diagram of a single refueling amount provided in Embodiment 3 of the present application.
  • Figure 7 is a flow chart of a method for identifying energy replenishment intentions provided in Embodiment 4 of the present application.
  • Figure 8 is a schematic diagram of the training and application scenarios of a supplementary intention recognition model provided in Embodiment 5 of the present application;
  • Figure 9 is a system framework diagram for energy replenishment intention recognition provided in Embodiment 5 of the present application.
  • Figure 10 is a schematic structural diagram of a model training device provided in Embodiment 6 of the present application.
  • FIG 11 is a schematic structural diagram of an energy replenishment intention recognition device provided in Embodiment 7 of the present application.
  • FIG. 12 is a schematic structural diagram of an electronic device provided in Embodiment 8 of the present application.
  • Figure 1 is a flow chart of a model training method provided in Embodiment 1 of the present application. This embodiment can be applied to the situation of vehicle energy replenishment intention model training.
  • the method can be executed by a model training device.
  • the model training device can use Implemented in the form of hardware and/or software, the model training device can be integrated into an electronic device with model training.
  • the method includes:
  • Sample vehicles can be vehicles that provide data for model training.
  • the number of sample vehicles in this embodiment is usually multiple.
  • the sample vehicle data includes: vehicle terminal log data and vehicle sensor data.
  • the vehicle-mounted terminal log data can be hidden log data generated by vehicle-machine user behavior.
  • the vehicle-mounted terminal log data may include: record data of user navigation searches or screen clicks, etc.
  • Vehicle terminal log data can also be called color data.
  • Car console can be the abbreviation of in-vehicle infotainment products installed in vehicles. Car console can functionally realize information communication between people and vehicles, vehicles and the outside world, or vehicles and vehicles.
  • Vehicle sensor data may be data generated by multiple types of sensors of the vehicle.
  • the vehicle sensor data may include data such as vehicle speed or remaining fuel level of the vehicle.
  • Vehicle sensor data can also be called gray data.
  • preliminary data preparation can be made for determining data related to vehicle energy replenishment, determining vehicle energy replenishment behavior, and identifying vehicle energy replenishment intentions.
  • the sample energy replenishment reference data may be data related to energy replenishment in the sample vehicle data.
  • Sample energy replenishment reference data may include but is not limited to: engine status data, motor status data, vehicle speed data, vehicle energy data, vehicle mileage data, vehicle information data (such as vehicle identification code), time data and location data (such as longitude and latitude information) ) and so on.
  • the engine status data can be It is the engine starting or stalling status data
  • the motor status data can be the motor is not powered on or the motor is powered on
  • the vehicle speed data can include the current vehicle speed and vehicle speed changes
  • the vehicle energy data can be the vehicle's remaining energy and remaining energy percentage and other data
  • vehicle mileage data can be the total mileage of the vehicle or the number of miles that the vehicle can still travel with remaining energy.
  • the sample energy replenishment reference data in this embodiment includes both data when energy is replenished and data when energy is not replenished.
  • Extracting the sample energy replenishment reference data from the sample vehicle data may be a process of initially screening out data related to energy replenishment from all data of the sample vehicle.
  • the relevant script code can be called to read the fields of the sample energy replenishment reference data from the database to extract the sample energy replenishment reference data.
  • the sample vehicle data was initially screened, the data irrelevant to energy replenishment was removed, and the data related to energy replenishment was preliminarily screened out, which facilitated the subsequent rapid and accurate determination of replenishment. Preliminary preparations were made for the behavioral data.
  • the energy replenishment behavior data may be data corresponding to the time of energy replenishment behavior filtered out from the sample energy replenishment reference data.
  • the energy-replenishing behavior data can be a set of sample energy-replenishing reference data combined with the corresponding moments of the energy-replenishing behavior; the energy-replenishing behavior data can be matched one-to-one by determining the corresponding moments of the energy-replenishing behavior.
  • the refueling behavior data may include: “Vehicle Identifier (ID)", “Engine Status”, “Remaining Gasoline Amount”, “Remaining Gasoline Amount Percentage”, “Kilometers that the vehicle can still travel with the remaining gasoline amount” “Number”, “Total Mileage”, “Current Speed”, “Season of the Day”, “Month of the Day”, “Week of the Day”, “Time Period of the Day”, “Hour of the Day”, “Day of the Day” “minute range” and “whether it is a special holiday or public holiday” and other data.
  • the energy replenishment behavior data is determined based on the sample energy replenishment reference data, including: directly locating the energy replenishment moment based on the sample energy replenishment reference data, and then determining the energy replenishment behavior data.
  • the energy replenishment time may be a time when the vehicle performs energy replenishment behavior. For example, it may be a time when the vehicle performs refueling behavior or charging behavior.
  • the energy replenishment moments during and before and after the data change can be extracted, and then the energy replenishment behavior data corresponding to the energy replenishment moments can be determined. For example, when the sample vehicle replenishes energy, the vehicle's remaining energy, remaining energy percentage, and the mileage that the vehicle can still travel with the remaining energy in the sample energy replenishment reference data will increase. This can be determined by changes in the sample energy replenishment reference data. It is always at the energy replenishment moment, and then the energy replenishment behavior data corresponding to the energy replenishment moment is determined.
  • position and vehicle energy data changes corresponding to the energy replenishment time, and then determine the energy replenishment behavior data corresponding to the energy replenishment time.
  • the energy replenishment behavior data is determined based on the sample energy replenishment reference data, including: extracting the energy replenishment intention data related to the energy replenishment intention from the sample energy replenishment reference data; from the energy replenishment intention data, Extract energy-replenishing behavior data related to energy-replenishing behavior.
  • the energy replenishment intention data may be the sample energy replenishment reference data corresponding to the moment containing the energy replenishment intention.
  • the energy-replenishing intention data may correspond to at least one moment containing energy-replenishing intention.
  • FIG. 2 is a schematic diagram of the energy-replenishing intention data selection principle provided in Embodiment 1 of the present application. As shown in Figure 2, the energy replenishment behavior is set to "refuel", the horizontal axis represents time, and the circular dot represents the time node when the user has the intention to refuel.
  • the energy replenishment intention data may be the sample energy replenishment reference data corresponding between the "intention t" time and the “refueling t” time, that is, the positive sample data.
  • the energy replenishment intention data related to the energy replenishment intention can be extracted by judging the moment when the energy replenishment intention is generated or the operation when the energy replenishment intention is generated, and judging the point at which the energy replenishment intention is generated, thereby determining The process of generating energy-replenishing intention data from energy-replenishing intention to energy-replenishing behavior.
  • the moment when the intention to replenish energy is judged can be the midpoint value in the process of two replenishment behaviors, or it can be the value at other time in the process of two replenishment behaviors.
  • the operation when the intention to replenish energy is generated may be related to going to a gas station or a battery swap station.
  • Extracting energy-replenishing behavior data related to energy-replenishing behavior from the energy-replenishing intention data may be a process of determining whether the vehicle is in energy-replenishing behavior based on changes in the data of the sample vehicle. For example, when the remaining fuel capacity of the fuel vehicle increases and the fuel vehicle is in a stalled state, it can be considered that the fuel vehicle has replenished energy.
  • the sample energy replenishment reference data screening is realized; at the same time, the user's energy replenishment intention is converted into an obtainable
  • the energy replenishment intention data allows the user's energy replenishment intention to be visualized.
  • the portrait data can be generated through the behavioral data of the sample vehicle to generate label data describing the sample vehicle.
  • the portrait data may include at least one of the following: parking location, energy replenishment location, driving route and energy data. wait.
  • the portrait data may include: "distance from the last gas station” and “distance from frequently visited gas stations” and other data.
  • portrait data is constructed based on the energy replenishment behavior data, including: determining the energy replenishment location based on the energy replenishment behavior data; and constructing portrait data of the sample vehicle based on the energy replenishment location.
  • Energy replenishment locations can include: gas stations, charging stations, charging piles or battery swap stations, etc.
  • Determining the energy replenishment location based on the energy replenishment behavior data may involve filtering out the longitude and latitude coordinates of the energy replenishment location from the energy replenishment behavior data. For example, based on the refueling and charging behavior discovery algorithm, the latitude and longitude coordinates of each refueling and charging of the sample vehicle can be identified from the energy replenishment behavior data.
  • Constructing the portrait data of the sample vehicle based on the energy replenishment location may be based on the longitude and latitude coordinates of the energy replenishment location, and determining the energy replenishment location that the sample vehicle frequently visits and/or the last energy replenishment location that the sample vehicle visited from the energy replenishment behavior data.
  • the longitude and latitude coordinates of refueling and charging use the "reverse geographical analysis" service to obtain the gas stations or charging stations near the current coordinate point, and store the results in the database to obtain the location of each refueling and charging of the sample vehicle. and name.
  • the energy-replenishing location that appears most often is the "frequently visited gas station"; the energy-replenishing location closest to the current moment is the "last gas station visited”.
  • the sample vehicle's frequent energy replenishment location or the last energy replenishment location visited is clarified, and the distance replenishment distance of the sample vehicle can be determined based on the energy replenishment location.
  • the distance between the energy positions makes the auxiliary judgment of the energy replenishment intention of the sample vehicle more targeted.
  • the energy replenishment behavior data of the sample vehicle is more specific, making the identification data of the energy replenishment intention of the sample vehicle more comprehensive.
  • the energy replenishment intention recognition model may be a model that performs the energy replenishment intention recognition task based on energy replenishment behavior data and portrait data. Input the energy-replenishing behavior data and portrait data into the energy-replenishing intention recognition model. The model can give the result of whether the sample vehicle has the intention of replenishing energy.
  • the energy replenishment behavior data and portrait data in this embodiment correspond to the data of multiple sample vehicles.
  • a relationship between the vehicle ID and the vehicle's energy replenishment behavior data and portrait data can be established. Mapping relationship, then the corresponding energy replenishment behavior data and portrait data of each vehicle ID can be input into the energy replenishment intention recognition model as a set of training data to train the model.
  • the process of training the energy-replenishing intention recognition model can be: input the energy-replenishing behavior data and portrait data into the energy-replenishing intention recognition model.
  • the energy-replenishing intention recognition model can predict whether the set of data contains energy-replenishing intentions through the intention prediction algorithm, and Compare the prediction results of the energy replenishment intention identification model with the actual energy replenishment intention results of the sample vehicles, calculate the loss function, and then use the parameter configuration of the mediation model based on the loss function to perform multiple iterations of the energy replenishment intention identification model according to the above scheme.
  • the trained energy-replenishing intention recognition model can be obtained.
  • the intent prediction algorithm can be Linear Support Vector Classification (LinearSVC), Logistic Regression (LR), Decision Tree (DecisionTree, DT), Random Forest (RandomForest, RF) or Gradient Boosting Decision Tree (Gradient Boosting Decision) Tree, GBDT), etc. You can choose the GBDT algorithm.
  • LinearSVC Linear Support Vector Classification
  • LR Logistic Regression
  • DecisionTree Decision Tree
  • DT Random Forest
  • Gradient Boosting Decision Tree Gradient Boosting Decision Tree
  • GBDT Gradient Boosting Decision
  • the energy replenishment behavior data and portrait data can also be processed. For example, it can be based on feature engineering processing, including feature extraction, aggregation, formatting, etc., to improve the significance of model feature recognition.
  • the energy replenishment intention recognition model is finally used to evaluate the final energy replenishment intention recognition model using the test set. If the evaluation passes, the training process of the energy replenishment intention recognition model is completed.
  • the technical solution of the embodiment of the present application extracts sample energy replenishment reference data from sample vehicle data; determines energy replenishment behavior data based on the sample energy replenishment reference data; constructs portrait data of the sample vehicle based on the energy replenishment behavior data; Behavioral data and portrait data are used to train the replenishment intention recognition model.
  • the technical solution of the embodiment of the present application adopts a two-stage data screening operation, that is, preliminary screening of data related to energy replenishment, and secondary screening of energy replenishment behavior data corresponding to the energy replenishment moment, which improves the accuracy and accuracy of vehicle energy replenishment behavior data extraction.
  • Efficiency The portrait data of the sample vehicle and the energy replenishment behavior data are used as model training data to make the model training data more comprehensive, thereby improving the accuracy and generalization of the energy replenishment intention recognition model training.
  • FIG 3 is a flow chart of a model training method provided in Embodiment 2 of the present application. Based on the above embodiment, this embodiment will extract energy replenishment intention data related to energy replenishment intention from the sample energy replenishment reference data. optimization. As shown in Figure 3, the method includes:
  • Determining the energy replenishment time based on the vehicle sensor data in the sample energy replenishment reference data may include identifying the energy replenishment time with energy replenishment behavior through the vehicle sensor data in the sample energy replenishment reference data. For example, the refueling moment of the vehicle's refueling/charging behavior can be identified through data such as "remaining fuel level/power" and "current speed" in the vehicle sensor data.
  • the vehicle The sensor data corresponds to the energy replenishment time.
  • the vehicle sensor data corresponding to that time can be found.
  • the energy replenishment intention period may be a time period in which there is an energy replenishment intention. It can be the time period between generating the intention to replenish energy and executing the replenishment behavior.
  • the model training method determines the energy replenishment moment, it also includes:
  • the average driving time is determined based on at least two energy replenishing moments; the reference driving time is determined based on the average driving time and the preset interval percentage.
  • the average driving time may be the average of the time intervals between adjacent energy replenishment moments. Different sample vehicles have different energy replenishment frequencies, and the average driving time can be different. For example, the replenishment energies are set to t 1 , t 2 , t 3 , and t 4 , and the time intervals between adjacent replenishment moments are (t 2 -t 1 ), (t 3 –t 2 ), (t 4 –t 3 ), the average driving time is the average of the above time intervals.
  • the preset interval percentage may be the percentage of the driving time interval from when the energy replenishment intention is generated until the vehicle starts to perform the energy replenishment behavior.
  • the preset interval percentage may be a value based on empirical assumptions.
  • the reference driving time may be the driving time interval from when the intention to replenish energy is generated until the vehicle starts to perform energy replenishment behavior.
  • the reference driving time can be an empirically assumed value or a fixed time interval.
  • the reference driving time can also be a value determined by a preset interval percentage and the average driving time. For example, the average driving time is set to t, the preset interval percentage is ⁇ , and the reference driving time is t* ⁇ .
  • the average driving time of different sample vehicles can be counted, making the average driving time data more accurate.
  • the reference driving time data can also be made more accurate.
  • the reference driving time can be obtained through statistics based on the data of different sample vehicles, which can make the obtained reference driving time data more targeted.
  • Determining the energy replenishment intention period can be by setting the energy replenishment moment as the time point when the energy replenishment behavior is performed, and setting the time point corresponding to the reference driving time from the energy replenishment moment back to the energy replenishment intention as the time point when the energy replenishment intention is generated.
  • Two The time period between time points is the energy replenishment intention period.
  • the energy replenishment time is set to T
  • the reference driving time is ⁇ T
  • the time when the energy replenishment intention is generated is (T- ⁇ T)
  • the energy replenishment intention period is the time from (T- ⁇ T) time to T time. part.
  • the energy replenishment intention time period is specified to facilitate data extraction.
  • the sample energy replenishment reference data contains vehicle information data and time data. Through the energy replenishment intention period, based on the vehicle information data and time data, the sample energy replenishment reference data of the corresponding vehicle can be found.
  • This energy replenishment reference data can be used as energy replenishment intention data.
  • the sample energy replenishment data of a single sample vehicle at a single time can be used as a data set, and the data set includes the sample energy replenishment reference data of the vehicle at that time.
  • the energy replenishment intention data is extracted, which provides a guarantee for the subsequent rapid and accurate extraction of energy replenishment behavior data.
  • the technical solution of the embodiment of the present application extracts sample energy replenishment reference data from sample vehicle data; determines the energy replenishment time based on the vehicle sensor data in the sample energy replenishment reference data; determines the energy replenishment time based on the energy replenishment time and the reference driving time Intention period; extract sample energy replenishment reference data located in the energy replenishment intention period as energy replenishment intention data; extract energy replenishment behavior data related to energy replenishment behavior from the energy replenishment intention data; construct a sample vehicle based on the energy replenishment behavior data portrait data; train the energy replenishment intention recognition model based on the energy replenishment behavior data and portrait data.
  • the energy replenishment intention period is confirmed in time, and the energy replenishment intention data is confirmed and extracted through the energy replenishment intention period, which improves the accuracy of energy replenishment intention identification.
  • Another way of extracting energy-replenishing intention data related to energy-replenishing intention from the sample energy-replenishing reference data may include the following steps:
  • Step A Screen the energy replenishment intention keywords from the vehicle terminal log data in the sample energy replenishment reference data.
  • the energy-replenishing intention keyword may be keyword information related to the energy-replenishing behavior. For example, it can be information about energy replenishment positions. For example, the address of the battery swap station or the name of the battery swap station, etc. It can also be the general information of energy replenishment positions. For example, battery swap stations, charging piles or gas stations, etc.
  • Screening the energy replenishment intention keywords from the vehicle terminal log data in the sample energy replenishment reference data may be a process of screening out keywords related to the energy replenishment intention from the search records in the vehicle terminal log data. For example, you can check the historical search records in the log data of the vehicle terminal, and check the keywords corresponding to "gas station", "battery swap station", the address of the gas station or the address of the battery swap station, which can be the key to screening energy replenishment intentions. word.
  • Step B Screen the associated sensor data of the energy replenishment intention keywords from the vehicle sensor data in the sample energy replenishment reference data.
  • the associated sensing data may be related to the vehicle sensor data containing the keyword of energy replenishment intention.
  • the energy replenishment intention keyword is set to "gas station".
  • the vehicle sensor data includes the location data of the gas station and the number of other vehicle sensors corresponding to the location data, such as engine status data, vehicle speed data, and vehicle energy. Data, vehicle mileage data, vehicle information data, time data, etc. are all associated sensing data.
  • Step C From the vehicle sensor data in the sample energy replenishment reference data, filtering the associated sensor data of the energy replenishment intention keyword can be through the field of the energy replenishment intention keyword, and determining the field containing the keyword in the vehicle sensor data. By determining the vehicle sensor data corresponding to the data, the associated sensing data is confirmed.
  • the energy-replenishing intention data is determined.
  • the advantage of this embodiment is that by associating the energy replenishment intention keywords in the vehicle terminal log data with the vehicle sensor data, the moment when the energy replenishment intention is generated can be determined more accurately, thereby improving the accuracy of determining the energy replenishment intention.
  • FIG 4 is a flow chart of a model training method provided in Embodiment 3 of the present application. Based on the above embodiment, this embodiment will extract energy-replenishing behavior data related to energy-replenishing behavior from the energy-replenishing intention data and optimize it. . As shown in Figure 4, the method includes:
  • S430 Traverse the energy replenishment intention data at each moment in sequence, and determine whether the push condition is satisfied at that moment. If the push condition is satisfied at this moment, S440 is executed. If the push condition is not satisfied at this moment, S430 is returned to execution.
  • the condition for pushing onto the stack may be that the energy replenishment intention data satisfies the conditions for the energy replenishment behavior.
  • the push conditions can be engine status data, vehicle speed data, and event types; when the vehicle is an electric vehicle, the push conditions can be motor status data, vehicle speed data, and event types.
  • the event type can be "vehicle stalled” or “vehicle started”; the event type can also be “vehicle powered off” or “vehicle powered on”.
  • the stacking conditions can be that the vehicle speed is 0, the engine is turned off, and the event If the event type is "vehicle stalled", this scenario may be to stop and prepare to replenish energy; the stacking condition may also be that the vehicle speed is 0, the engine is started and the event type is "vehicle start", this scenario may be to drive out of the gas station after replenishing energy. .
  • This embodiment can sequentially determine whether the energy replenishment intention data at each moment meets the stacking condition. If it satisfies the stacking condition, perform the subsequent operation of S440. If the energy replenishment intention data for the next moment does not meet, continue to perform this step.
  • the number of data items in the stack can be the number of energy replenishment intention data items that have been pushed into the stack.
  • the number of data items in the stack does not exceed the preset number. For example, the default number of items is set to 2. If after the energy replenishment intention data is pushed into the stack, the number of data items in the stack is 1, then return to the operation of S430 and continue to extract the energy replenishment intention data that meets the conditions for being pushed into the stack. . If after the replenishment intention data is pushed into the stack, the number of data items in the stack is 2, it will no longer be pushed into the stack and the subsequent operation of S450 will be performed.
  • Energy growth events are events in which vehicle energy data increases. It can be increased by the remaining fuel volume or the remaining electric power.
  • determining whether there is an energy growth event based on the energy replenishment intention data in the stack may include: determining the energy replenishment increase amount and/or energy replenishment time interval based on the energy replenishment intention data in the stack; The energy replenishment increase amount and/or the energy replenishment time interval determine whether there is an energy increase event.
  • the energy replenishment increment may be the change in vehicle energy obtained by comparing the energy replenishment intention data in the stack. For example, it may be an increase in the remaining fuel amount or an increase in the remaining electric power.
  • the energy replenishment time interval may be the time interval data obtained by comparing the energy replenishment intention data in the stack. For example, if there are two pieces of data in the stack, the replenishment time interval can be the time interval between the two pieces of data.
  • Table 1 below is a historical data statistics table of energy replenishment intention data in the stack.
  • the time from "2020-12-01 18:16:49" to the time "2020-12-01 18:19:19” corresponds to refueling event 1.
  • Refueling event 1 is a normal refueling event.
  • the current fuel volume Stable changes.
  • the time "2020-12-01 18:19:59” corresponds to refueling event 2.
  • Refueling event 2 is an abnormal refueling event, and the current oil volume fluctuates briefly. Limited by the accuracy of the sensor, the refueling data sequence has brief fluctuations. In order to ensure the accuracy of the data, the refueling data needs to be cleaned.
  • the energy replenishment increase amount and/or the energy replenishment time interval are determined based on the energy replenishment intention data in the stack; based on the energy replenishment increase amount and/or the energy replenishment time interval, it is determined whether the energy replenishment increase amount is Reach the growth amount threshold and/or whether the energy replenishment time interval reaches the energy replenishment time interval threshold, and then determine whether there is an energy growth event, thereby realizing the cleaning of energy replenishment intention data.
  • the growth threshold is the minimum set value for energy supplement growth.
  • the energy replenishment time interval threshold is the minimum setting value of the energy replenishment time interval.
  • the replenishment time interval threshold can be used to smooth data by setting a fluctuation tolerance time window.
  • the selection method of the energy replenishment time interval threshold is as follows:
  • FIG. 5 is a statistical distribution diagram of refueling time intervals provided in Embodiment 3 of the present application.
  • the energy replenishment time interval is set as the refueling time interval.
  • the horizontal axis is the time interval between two refuelings.
  • the vertical axis is the cumulative proportion of normal data.
  • the energy replenishment time interval threshold is selected as follows:
  • the threshold value of the time interval between two refuelings is selected as 300s.
  • the data processing process is as follows:
  • FIG. 6 is a statistical distribution diagram of a single refueling amount provided in Embodiment 3 of the present application.
  • the energy replenishment growth amount is set to the refueling amount of the refueling event
  • the horizontal axis is the refueling amount of each refueling event
  • the vertical axis is the frequency of the refueling amount.
  • the effective refueling volume threshold is selected to be >2L.
  • the energy replenishment intention data in the stack is used as the energy replenishment behavior data by extracting the data as the energy replenishment behavior data.
  • Clearing the energy replenishment intention data in the stack may be a process of clearing the energy replenishment intention data in the stack so that the energy replenishment intention data in the stack becomes 0.
  • the energy replenishment intention data in the stack As energy replenishment behavior data, and clearing the energy replenishment intention data in the stack, the energy replenishment behavior data is determined, and the extraction of the energy replenishment behavior data is completed. At the same time, the energy replenishment intention data in the stack is cleared.
  • the energy replenishment intention data provides conditions for the subsequent judgment of the energy replenishment intention data, realizing the circular judgment of the data.
  • the technical solution of the embodiment of the present application extracts sample energy replenishment reference data from sample vehicle data; extracts energy replenishment intention data related to energy replenishment intention from the sample energy replenishment reference data; and sequentially traverses the energy replenishment intention data at each moment. , determine whether the push condition is met at this moment; if it is satisfied, then after pushing the replenishment intention data at that moment into the stack, it will be judged whether the number of data items in the stack meets the preset number; if it is satisfied, then according to the replenishment intention data in the stack data to determine whether there is an energy growth event; if so, use the energy replenishment intention data in the stack as energy replenishment behavior data, and clear the energy replenishment intention data in the stack; construct the portrait data of the sample vehicle based on the energy replenishment behavior data; Based on the energy replenishment behavior data and portrait data, the energy replenishment intention recognition model is trained.
  • the energy replenishment intention data were gradually filtered to achieve accurate judgment on the energy replenishment behavior data.
  • the energy replenishment intention recognition model is trained.
  • Figure 7 is a flow chart of a method for identifying energy replenishment intentions provided in Embodiment 4 of the present application. This embodiment can be applied to the situation of vehicle replenishment intention identification.
  • the method can be executed by an energy replenishment intention identification device.
  • the intention recognition device can be implemented in the form of hardware and/or software.
  • the power-supply intention recognition device can be integrated into an electronic device configured with a power-supplement intention recognition model.
  • the electronic device can be a vehicle-mounted terminal.
  • the method includes:
  • the current vehicle data can be all real-time data generated by the current vehicle at the current moment.
  • the current vehicle data includes: vehicle terminal log data and vehicle sensor data.
  • the vehicle terminal log data and vehicle sensor data are mapped one-to-one to prepare for the energy replenishment intention identification process.
  • the current energy replenishment reference data may be all associated data related to energy replenishment in the current vehicle data.
  • Obtaining the current energy replenishment reference data from the current vehicle data is similar to the process of extracting the sample energy replenishment reference data from the sample vehicle data introduced in the above embodiment, and will not be described in detail here.
  • the energy-replenishing intention recognition model can be trained through the model training method of any embodiment of the present application.
  • the energy replenishment intention identification model is used to analyze the input energy replenishment reference data and provide the identification result of whether there is an energy replenishment intention at the current moment.
  • the identification result of the energy replenishment intention may be the result of determining whether the current vehicle has the energy replenishment intention. For example, there may or may not be a supplementary intention.
  • the data can be processed before being input into the energizing intent recognition model. For example, it can be based on feature engineering processing, including feature extraction, aggregation, formatting, etc., to improve the significance of model feature recognition.
  • the problem of being unable to predict the vehicle's energy replenishment intention in advance is solved, and the current vehicle's energy replenishment intention can be identified and predicted through the current energy replenishment reference data.
  • the energy replenishment intention identification method obtains the current energy replenishment reference data from the current vehicle data, it also includes:
  • the historical energy replenishment reference data may be all associated data related to energy replenishment in the current vehicle's historical data.
  • the historical energy replenishment reference data is the historical energy replenishment experience value of the vehicle, which records the relevant data of the vehicle's historical energy replenishment record.
  • the supplementary reference data may be historical energy replenishment reference data through the common correlation between the historical energy replenishment reference data and the current energy replenishment reference data, or it may directly use the historical energy replenishment reference data of the current vehicle as the energy replenishment reference data.
  • the current energy replenishment reference data is input into the energy replenishment intention recognition model to obtain the energy replenishment intention recognition results at the current moment, including:
  • the current energy replenishment reference data and the supplementary reference data may be input into the energy replenishment intention identification model as a whole data group to obtain the current energy replenishment intention identification model. Complementary intention identification results.
  • the current energy replenishment reference data is supplemented from historical data, providing more sufficient data support for subsequent predictions.
  • the current energy replenishment reference data and supplementary reference data are input into the energy replenishment intention identification model to obtain the current moment.
  • the identification results of energy replenishment intentions and the input data are more comprehensive, ensuring the accuracy of predictions.
  • the energy replenishment intention identification method inputs the current energy replenishment reference data into the energy replenishment intention identification model, and obtains the energy replenishment intention identification results at the current moment, including:
  • the portrait data of the current vehicle may be generated in advance through the historical behavior data of the current vehicle to generate label data describing the current vehicle.
  • the portrait data of the current vehicle and the current energy replenishment reference data can be input into the energy replenishment intention recognition model as a data group, and the model determines whether there is an energy replenishment intention at the current moment through calculation.
  • the input data of the energy replenishment intention recognition model is more comprehensive, ensuring the accuracy of energy replenishment intention recognition.
  • the technical solution of the embodiment of the present application obtains the current energy replenishment reference data from the current vehicle data; inputs the current energy replenishment reference data into the energy replenishment intention identification model to obtain the energy replenishment intention identification result at the current moment.
  • the energy replenishment intention recognition model By applying the energy replenishment intention recognition model, the function of directly predicting and identifying energy replenishment intentions through the current energy replenishment reference data is realized, which improves the efficiency and accuracy of energy replenishment intention identification.
  • Figure 8 is a schematic diagram of the training and application scenarios of a supplementary intention recognition model provided in Embodiment 5 of the present application.
  • This embodiment can implement the model training method and energy replenishment intention identification method provided in the above embodiments of this application.
  • model training is performed first, and algorithm evaluation/optimization is repeatedly performed to obtain an intent recognition model. Then the model is applied to realize the identification process of energy replenishment intention.
  • the process of model training includes: source data collection, energy replenishment intention sample selection, portrait extraction, training data set construction, modeling and algorithm evaluation/optimization.
  • Source data collection is equivalent to the process of extracting sample energy reference data.
  • the selection of energy-replenishing intention samples is equivalent to extracting the energy-replenishing intention data related to the energy-replenishing intention.
  • Portrait extraction is equivalent to constructing portrait data.
  • training data set is equivalent to extracting energy-replenishing behavior data and constructing portrait data.
  • Modeling can be building an energy-replenishing intention recognition model, which is equivalent to training an energy-replenishing intention recognition model.
  • the energy-replenishing intention identification process includes: real-time data sources, portrait extraction, feature set construction, models and intentions.
  • the energy-replenishing intention identification process is the process of identifying energy-replenishing intentions using the energy-replenishing intention recognition model.
  • the real-time data source is equivalent to obtaining the current energy replenishment reference data from the current vehicle data.
  • the portrait data is equivalent to obtaining the portrait data of the current vehicle.
  • the construction of the feature set is equivalent to obtaining the current energy supplement reference data and the portrait data of the current vehicle.
  • the model is equivalent to inputting the current vehicle's portrait data and current energy replenishment reference data into the energy replenishment intention recognition model.
  • FIG 9 is a system framework diagram for energy replenishment intention recognition provided in Embodiment 5 of the present application.
  • the system framework diagram can be integrated into the vehicle terminal.
  • This embodiment can implement the energy replenishment intention identification method provided in the above embodiments of this application.
  • the system framework consists of the basic layer of data access and storage, the real-time/offline computing engine layer, the technology implementation layer and the backward service application layer from bottom to top.
  • the data source can be vehicle terminal log data and vehicle sensor data collected in real time.
  • Data sources are transmitted through a data bus.
  • the data bus may be a distributed publish/subscribe based messaging system.
  • Data storage can be implemented through a database, and can be implemented through multiple databases.
  • the type of database may include: a database management system based on a distributed file system, an open source non-relational distributed database management system, or a relational database management system, etc.
  • Real-time computing can be implemented based on the distributed processing engine (Flink) for streaming data and batch data.
  • Vehicle terminal log data and vehicle sensor data can be calculated in real time through the real-time calculation engine, and the calculated data can be stored in the database.
  • Offline computing can calculate the vehicle terminal log data and vehicle sensor data through the offline computing engine, and calculate the vehicle's portrait data, which can be stored in the database.
  • Feature engineering is used to process data before inputting it into the model, including feature extraction, aggregation and formatting, to help the model improve the significance of feature recognition.
  • Energy replenishment identification can be used to obtain current energy replenishment reference data from current vehicle data.
  • Intention prediction can be based on vehicle terminal log data and sensor data, and through the energy replenishment intention recognition model, the energy replenishment intention can be predicted and the energy replenishment intention prediction result can be output.
  • Corresponding services can also be provided for supplementary image recognition. For example, activity recommendations can be made, and suitable preferential activities can be recommended for the purpose of replenishing energy. Points of interest (POI) can also be recommended to provide personalized demand recommendations based on geographical information, public service sites and information about buildings such as bus stations or service sites that can provide services. Refueling reminders can also be carried out. When it is determined that the user has the intention to replenish energy, the refueling reminder can be carried out through the vehicle machine. Among them, the refueling reminder can include: sound, pop-up window or image information reminder, etc.
  • Figure 10 is a schematic structural diagram of a model training device provided in Embodiment 6 of the present application.
  • This device can implement the model training method provided in the above embodiments of this application.
  • This embodiment can be applied to the situation of vehicle replenishment intention model training.
  • the device can be implemented by software and/or hardware, and can be integrated into an electronic device with model training. As shown in Figure 10, the device includes:
  • the sample data extraction module 1010 is configured to extract the sample energy replenishment reference data from the sample vehicle data; the behavior data determination module 1020 is configured to determine the energy replenishment behavior data based on the sample energy replenishment reference data; the portrait data construction module 1030 is configured to determine the energy replenishment behavior data based on the sample energy replenishment reference data.
  • the energy replenishment behavior data is used to construct portrait data of the sample vehicle; the model training module 1040 is configured to train the energy replenishment intention recognition model based on the energy replenishment behavior data and portrait data.
  • the technical solution of the embodiment of the present application extracts sample energy replenishment reference data from sample vehicle data; determines energy replenishment behavior data based on the sample energy replenishment reference data; constructs portrait data of the sample vehicle based on the energy replenishment behavior data; Behavioral data and portrait data are used to train the replenishment intention recognition model.
  • the technical solution of the embodiment of the present application adopts a two-stage data screening operation, that is, preliminary screening of data related to energy replenishment, and secondary screening of energy replenishment behavior data corresponding to the energy replenishment moment, which improves the accuracy and accuracy of vehicle energy replenishment behavior data extraction.
  • Efficiency The portrait data of the sample vehicle and the energy replenishment behavior data are used as model training data to make the model training data more comprehensive, thereby improving the accuracy and generalization of the energy replenishment intention recognition model training.
  • the behavior data determination module 1020 includes:
  • the intention data extraction unit is configured to extract the energy replenishment intention data related to the energy replenishment intention from the sample energy replenishment reference data;
  • the behavior data extraction unit is configured to extract the energy replenishment intention data related to the energy replenishment behavior from the energy replenishment intention data. behavioral data.
  • the intent data extraction unit may include:
  • the keyword filtering subunit is set to filter energy replenishment intention keywords from the vehicle terminal log data in the sample energy replenishment reference data; the associated data filtering subunit is set to filter the energy replenishment intention keywords from the vehicle sensor data in the sample energy replenishment reference data.
  • the associated sensing data of the energy-replenishing intention keywords is screened; the intention data determination subunit is configured to determine the energy-replenishing intention data based on the energy-replenishing intention keywords and the associated sensing data.
  • the intent data extraction unit may also include:
  • the energy replenishment time determination subunit is set to determine the energy replenishment time based on the vehicle sensor data in the sample energy replenishment reference data; the intention period determination subunit is set to determine the energy replenishment intention period based on the energy replenishment time and the reference driving time; intention The data determination subunit is configured to extract sample energy replenishment reference data located in the energy replenishment intention period as energy replenishment intention data.
  • the energy replenishment time determination subunit further includes:
  • the average duration determination subunit is set so that if the energy replenishment time is at least two, it will be determined based on at least two replenishment moments. At the available time, the average driving time is determined; the reference duration determination subunit is set to determine the reference driving time based on the average driving time and the preset interval percentage.
  • the energy-replenishing behavior data extraction unit includes:
  • the condition judgment subunit is set to traverse the energy replenishment intention data at each moment in sequence to determine whether the stacking condition is met at that moment; the number judgment subunit is set to push the energy replenishment intention data at that moment into the stack if it is met. , determine whether the number of data items in the stack meets the preset number; the growth determination subunit is set to If it is satisfied, determine whether there is an energy growth event based on the energy replenishment intention data in the stack; the data determination subunit is set to If it exists , then the energy replenishment intention data in the stack will be used as the energy replenishment behavior data, and the energy replenishment intention data in the stack will be cleared.
  • the growth determination subunit may be set to:
  • the energy replenishment increase amount and/or the energy replenishment time interval is determined; based on the energy replenishment increase amount and/or the energy replenishment time interval, it is determined whether there is an energy increase event.
  • the portrait data building module includes:
  • the energy replenishment position determination unit is configured to determine the energy replenishment position based on the energy replenishment behavior data; the portrait data construction unit is configured to construct portrait data of the sample vehicle based on the energy replenishment location.
  • the sample vehicle data includes: vehicle terminal log data and vehicle sensor data.
  • the model training device provided by the embodiments of this application can execute the model training method provided by any embodiment of this application, and has functional modules and effects corresponding to the execution method.
  • FIG 11 is a schematic structural diagram of an energy replenishment intention recognition device provided in Embodiment 7 of the present application.
  • This device can implement the energy replenishment intention identification method provided by the above embodiments of the present application.
  • This embodiment can be applied to the situation of vehicle energy replenishment intention recognition.
  • the device can be implemented by software and/or hardware, and can be integrated into an electronic device equipped with an energy replenishment intention recognition model.
  • the electronic device can be a vehicle-mounted device. terminal. As shown in Figure 11, the device includes:
  • the current data acquisition module 1110 is configured to obtain the current energy replenishment reference data from the current vehicle data;
  • the energy replenishment intention identification module 1120 is configured to input the current energy replenishment reference data into the energy replenishment intention identification model to obtain the energy replenishment at the current moment.
  • Intention recognition results; among them, the energy-supplementing intention recognition model is trained by the model training device.
  • the technical solution of the embodiment of the present application obtains the current energy replenishment reference data from the current vehicle data; inputs the current energy replenishment reference data into the energy replenishment intention recognition model to obtain the energy replenishment intention awareness at the current moment. Don't get results.
  • the energy replenishment intention recognition model By applying the energy replenishment intention recognition model, the function of directly predicting and identifying energy replenishment intentions through the current energy replenishment reference data is realized, which improves the efficiency and accuracy of energy replenishment intention identification.
  • the current energy replenishment data acquisition module 1110 obtains the current energy replenishment reference data from the current vehicle data, it also includes:
  • the supplementary data acquisition module is configured to obtain the supplementary reference data of the current energy supplement reference data from the historical energy supplement reference data; accordingly, the energy supplement intention identification module 1120 includes:
  • the current data acquisition module 1110 includes:
  • the portrait data acquisition unit is configured to acquire the portrait data of the current vehicle;
  • the recognition result acquisition unit is configured to input the portrait data of the current vehicle and the current energy replenishment reference data into the energy replenishment intention recognition model to obtain the energy replenishment intention recognition at the current moment. result.
  • the current vehicle data includes: vehicle terminal log data and vehicle sensor data.
  • the energy replenishment intention identification device provided by the embodiments of the present application can execute the energy replenishment intention identification method provided by any embodiment of the present application, and has corresponding functional modules and effects of the execution method.
  • FIG. 12 is a schematic structural diagram of an electronic device provided in Embodiment 8 of the present application.
  • FIG. 12 shows a block diagram of an exemplary device suitable for implementing the implementation of the embodiment of the present application.
  • the device shown in Figure 12 is only an example and should not impose any restrictions on the functions and usage scope of the embodiments of the present application.
  • electronic device 1200 is embodied in the form of a general computing device.
  • the components of the electronic device 1200 may include, but are not limited to: one or more processors or processing units 1210, a system memory 1220, and a bus 1230 connecting different system components (including the system memory 1220 and the processing unit 1210).
  • Bus 1230 represents one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, a graphics accelerated port, a processor, or a local bus using any of a variety of bus structures.
  • these architectures include, but are not limited to, Industry Standard Architecture (ISA) bus, Micro Channel Architecture (MAC) bus, Enhanced ISA bus, Video Electronics Standards Association (Video Electronics Standards) Association, VESA) local bus and Peripheral Component Interconnect (PCI) bus.
  • Electronic device 1200 includes a variety of computer system readable media. These media can be any available media that can be accessed by electronic device 1200, including volatile and nonvolatile media, removable and non-removable media.
  • System memory 1220 may include computer system readable media in the form of volatile memory, such as random access memory (Random Access Memory, RAM) 1221 and/or cache memory (cache 1222).
  • Electronic device 1200 may include other removable/non-removable, volatile/non-volatile computer system storage media.
  • storage system 1223 may be configured to read and write to non-removable, non-volatile magnetic media (not shown in Figure 12, commonly referred to as a "hard drive”).
  • a disk drive configured to read and write to a removable non-volatile disk (e.g., a "floppy disk") and a removable non-volatile optical disk (e.g., a Compact Disc) may be provided.
  • each drive may be connected to bus 1230 through one or more data media interfaces.
  • the system memory 1220 may include at least one program product having a set of (eg, at least one) program modules configured to perform the functions of embodiments of the present application.
  • Program modules 1224 generally perform functions and/or methods in the embodiments described in the embodiments of this application.
  • Electronic device 1200 may also communicate with one or more external devices 1300 (e.g., keyboard, pointing device, display 1310, etc.), may also communicate with one or more devices that enable a user to interact with electronic device 1200, and/or with Any device (eg, network card, modem, etc.) that enables the electronic device 1200 to communicate with one or more other computing devices. This communication may occur through an input/output (I/O) interface 1240.
  • the electronic device 1200 can also communicate with one or more networks (such as a local area network (Local Area Network, LAN), a wide area network (Wide Area Network, WAN), and/or a public network, such as the Internet) through the network adapter 1250.
  • networks such as a local area network (Local Area Network, LAN), a wide area network (Wide Area Network, WAN), and/or a public network, such as the Internet
  • network adapter 1250 communicates with other modules of electronic device 1200 via bus 1230.
  • other hardware and/or software modules may be used in conjunction with electronic device 1200, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, redundant arrays of independent disks (Redundant Arrays of Independent Disks, RAID) systems, tape drives and data backup storage systems, etc.
  • the processing unit 1210 executes a variety of functional applications and data processing by running programs stored in the system memory 1220, such as implementing the model training method and energy supplement provided by the embodiments of the present application. Intent identification methods.
  • Embodiment 8 of the present application also provides a computer-readable storage medium on which a computer program (also known as computer executable instructions) is stored.
  • a computer program also known as computer executable instructions
  • the program When the program is executed by a processor, it is used to perform model training provided by the embodiment of the present application. Methods and empowering intention identification methods.
  • the computer storage medium in the embodiment of the present application may be any combination of one or more computer-readable media.
  • the computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium.
  • the computer-readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination thereof. Examples of computer-readable storage media (a non-exhaustive list) include: electrical connections having one or more conductors, portable computer disks, hard drives, RAM, ROM, Erasable Programmable Read-Only Memory, EPROM or flash memory), optical fiber, CD-ROM, optical storage device, magnetic storage device, or any suitable combination of the above.
  • a computer-readable storage medium may be any tangible medium that contains or stores a program for use by or in connection with an instruction execution system, apparatus, or device.
  • a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave carrying computer-readable program code therein. Such propagated data signals may take many forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above.
  • a computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device .
  • Program code embodied on a computer-readable medium can be transmitted using any appropriate medium, including but not limited to wireless, wire, optical cable, radio frequency (Radio Frequency, RF), etc., or any suitable combination of the above.
  • any appropriate medium including but not limited to wireless, wire, optical cable, radio frequency (Radio Frequency, RF), etc., or any suitable combination of the above.
  • Computer program code for performing operations of embodiments of the present application may be written in one or more programming languages, including object-oriented programming languages such as Java, Smalltalk, C++, and conventional A procedural programming language such as the "C" language or similar programming language.
  • the program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server.
  • the remote computer may be connected to the user's computer through any kind of network, including a LAN or WAN, or may be connected to an external computer (such as through the Internet using an Internet service provider).
  • the acquisition, storage and application of current vehicle data, current energy replenishment reference data, historical energy replenishment reference data, supplementary reference data, current vehicle portrait data, etc. are in compliance with relevant laws and regulations and do not violate public order and good customs.
  • the energizing intention recognition model in this embodiment is not targeted at a specific user and cannot reflect the personal information of a specific user.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Business, Economics & Management (AREA)
  • Databases & Information Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Economics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Library & Information Science (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • General Health & Medical Sciences (AREA)
  • Human Resources & Organizations (AREA)
  • Marketing (AREA)
  • Primary Health Care (AREA)
  • Strategic Management (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

本申请公开了模型训练方法及装置补能意图识别方法及装置、设备、介质。模型训练方法，包括：从样本车辆数据中提取样本补能参考数据；根据样本补能参考数据，确定补能行为数据；根据补能行为数据，构建样本车辆的画像数据；根据补能行为数据和画像数据，训练补能意图识别模型。

Description

模型训练方法及装置、补能意图识别方法及装置、设备、介质
本申请要求在2022年06月14日提交中国专利局、申请号为202210670153.1的中国专利申请的优先权，该申请的全部内容通过引用结合在本申请中。
技术领域
本申请涉及智能交通技术领域，例如涉及模型训练方法及装置补能意图识别方法及装置、设备、介质。
背景技术
随着智能交通技术的迅速发展，智能服务场景的细分和预测性主动服务的需求越来越高。例如，通过预测用户补能意图，来为用户提供补能提醒或补能服务推荐等服务。
相关技术通常根据用户当前车辆的燃油余量或电池电量，设定规则，当触发设定的阈值，即认为存在补能意图。但该方式预测准确性较低，亟需改进。
发明内容
本申请提供模型训练方法及装置补能意图识别方法及装置、设备、介质，用以解决相关技术中的缺陷，提高了预测的准确性。
第一方面，本申请提供了一种模型训练方法，其中，包括：
从样本车辆数据中提取样本补能参考数据；
根据样本补能参考数据，确定补能行为数据；
根据补能行为数据，构建样本车辆的画像数据；
根据补能行为数据和画像数据，训练补能意图识别模型。
第二方面，本申请还提供了一种补能意图识别方法，其中，包括：
从当前车辆数据中获取当前补能参考数据；
将当前补能参考数据输入到补能意图识别模型中，得到当前时刻的补能意图识别结果；
其中，补能意图识别模型通过上述的模型训练方法训练得到。
第三方面，本申请还提供了一种模型训练装置，其中，包括：
样本数据提取模块，设置为从样本车辆数据中提取样本补能参考数据；
行为数据确定模块，设置为根据样本补能参考数据，确定补能行为数据；
画像数据构建模块，设置为根据补能行为数据，构建样本车辆的画像数据；
模型训练模块，设置为根据补能行为数据和画像数据，训练补能意图识别模型。
第四方面，本申请还提供了一种补能意图识别装置，其中，包括：
当前数据获取模块，设置为从当前车辆数据中获取当前补能参考数据；
补能意图识别模块，设置为将当前补能参考数据输入到补能意图识别模型中，得到当前时刻的补能意图识别结果；
其中，补能意图识别模型通过模型训练装置训练得到。
第五方面，本申请还提供了一种电子设备，其中，包括：
一个或多个处理器；
存储器，设置为存储一个或多个程序；
当一个或多个程序被一个或多个处理器执行，使得一个或多个处理器实现上述的模型训练方法或补能意图识别方法。
第六方面，本申请例还提供了一种计算机可读存储介质，其上存储有计算机程序，其中，该程序被处理器执行时实现上述的模型训练方法或补能意图识别方法。
附图说明
图1是本申请实施例一提供的一种模型训练方法的流程图；
图2是本申请实施例一提供的一种补能意图数据选择原理示意图；
图3是本申请实施例二提供的一种模型训练方法的流程图；
图4是本申请实施例三提供的一种模型训练方法的流程图；
图5是本申请实施例三提供的一种加油时间间隔的统计分布图；
图6是本申请实施例三提供的一种单次加油量的统计分布图；
图7是本申请实施例四提供的一种补能意图识别方法的流程图；
图8是本申请实施例五提供的一种补能意图识别模型的训练及应用场景的示意图；
图9为本申请实施例五提供的一种补能意图识别的***框架图；
图10是本申请实施例六提供的一种模型训练装置的结构示意图；
图11是本申请实施例七提供的一种补能意图识别装置的结构示意图；
图12是本申请实施例八提供的一种电子设备的结构示意图。
具体实施方式
下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行描述，所描述的实施例仅仅是本申请一部分的实施例。
实施例一
图1为本申请实施例一提供的一种模型训练方法的流程图，本实施例可适用于车辆补能意图模型训练的情况，该方法可以由模型训练装置来执行，该模型训练装置可以采用硬件和/或软件的形式实现，该模型训练装置可集成于具有模型训练的电子设备中。
如图1所示，该方法包括：
S110、从样本车辆数据中提取样本补能参考数据。
样本车辆可以是为模型训练提供数据的车辆。为了保证模型训练的准确性，本实施例的样本车辆的数量通常为多个。
本申请任一实施例的模型训练的方法中，样本车辆数据包括：车载终端日志数据和车辆传感器数据。
车载终端日志数据可以是车机用户行为产生的埋点日志数据。示例性的，车载终端日志数据可以包括：用户导航搜索或点击屏幕等的记录数据。车载终端日志数据也可以称为彩色数据。车机可以是安装在车辆中的车载信息娱乐产品的简称，车机可以在功能上可以实现人与车、车与外界，或车与车的信息通讯。
车辆传感器数据可以是车辆的多类传感器产生的数据。示例性的，车辆传感器数据可以包括：车速或车辆剩余油量等数据。车辆传感器数据也可以称为灰色数据。
通过车载终端日志数据和车辆传感器数据，可以为确定车辆补能相关的数据、确定车辆补能行为以及车辆补能意图识别做数据的前期准备。
样本补能参考数据可以是样本车辆数据中与补能相关的数据。样本补能参考数据可以包括但不限于：发动机状态数据、电机状态数据、车辆速度数据、车辆能量数据、车辆里程数据、车辆信息数据(例如车辆识别代码)、时间数据和位置数据(如经纬度信息)等中的至少一个。其中，发动机状态数据可以 是发动机启动或熄火状态数据；电机状态数据可以是电机未上电或电机上电状态数据；车辆速度数据可以包括当前车速和车速变化等数据；车辆能量数据可以是车辆的剩余能量和剩余能量百分比等数据；车辆里程数据可以是车辆的总里程数或车辆剩余能量还可以行驶的里程数等数据。
本实施例的样本补能参考数据既包括补能时的数据，又包括未补能时的数据。
从样本车辆数据中提取样本补能参考数据可以是从样本车辆的全部数据中初步筛选出与补能相关的数据的过程。示例性的，可以调用相关脚本代码从数据库中读取样本补能参考数据的字段，实现提取样本补能参考数据。
通过从样本车辆数据中提取样本补能参考数据，对样本车辆数据进行了初步筛选，去掉了与补能无关的数据，初步筛选出了与补能相关的数据，为后续快速且精准的确定补能行为数据做了初步的准备。
S120、根据样本补能参考数据，确定补能行为数据。
补能行为数据可以是从样本补能参考数据中筛选出来的补能行为对应时刻的数据。补能行为数据可以是与补能行为对应时刻组合起来的一组样本补能参考数据；补能行为数据可以通过确定补能行为对应时刻，将补能行为数据进行一一对应。示例性的，补能行为数据可以包括：“车辆标识(Identifier，ID)”、“发动机状态”、“剩余汽油量”、“剩余汽油量百分比”’、“车辆剩余汽油量还能行驶的公里数”、“总里程”、“当前车速”、“当日所处季节”、“当日所处月份”、“当日所处星期”、“当日所处时段”、“当日所处小时”、“当日所处分钟区间”和“是否特殊节日或公休日”等数据。
在上述技术方案的基础上，根据样本补能参考数据，确定补能行为数据，包括：根据样本补能参考数据，直接定位补能时刻，进而确定补能行为数据。
补能时刻可以是车辆有补能行为的时刻，示例性的，可以是有加油行为或充电行为的时刻。
可以通过车辆能量数据变化，提取出数据变化过程中和数据变化前后的补能时刻，进而确定补能时刻对应的补能行为数据。示例性的，当样本车辆补能时，样本补能参考数据中的车辆的剩余能量、剩余能量百分比和车辆剩余能量还可以行驶的里程数会增长，可以通过样本补能参考数据的变化确定该时刻处于补能时刻，进而确定补能时刻对应的补能行为数据。
也可以是通过车辆位置信息，提取出位于车辆位于补能位置的数据所对应的补能时刻，进而确定补能时刻对应的补能行为数据。
也可以通过车辆能量数据变化和车辆位置信息，提取出位于车辆位于补能 位置且车辆能量数据变化所对应的补能时刻，进而确定补能时刻对应的补能行为数据。
在上述技术方案的基础上，根据样本补能参考数据，确定补能行为数据，包括：从样本补能参考数据中，提取与补能意图相关的补能意图数据；从补能意图数据中，提取与补能行为相关的补能行为数据。
补能意图数据可以是含有补能意图的时刻所对应的样本补能参考数据。补能意图数据可以对应至少一个含有补能意图的时刻。示例性的，图2为本申请实施例一提供的一种补能意图数据选择原理示意图。如图2所示，补能行为设定为“加油”，横轴代表时间，圆形点表示用户产生加油意图时的时间节点。假设在“意图t”时刻至“加油t”时刻之间，用户会一直保持着“加油”的意图，此区间内的样本补能参考数据为正样本；在“加油t-1”时刻至“意图t”时刻之间，用户会一直保持着“不加油”的意图，此区间内的样本补能参考数据为负样本。补能意图数据可以是在“意图t”时刻至“加油t”时刻之间所对应的样本补能参考数据，即正样本数据。
从样本补能参考数据中，提取与补能意图相关的补能意图数据，可以是通过判断补能意图产生的时刻或补能意图产生时的操作，判断出补能意图产生的点，从而确定补能意图产生到补能行为之间的补能意图数据的过程。判断补能意图产生的时刻，可以是两次补能行为过程中的中点值，也可以是两次补能行为过程中的其他时刻值。补能意图产生时的操作可以是前往加油站或者换电站的相关操作。确定了补能意图产生的点，就可以确认补能意图行为数据的区间，进而提取出了补能意图行为数据。
从补能意图数据中，提取与补能行为相关的补能行为数据，可以是通过样本车辆的数据发生变化判断车辆是否处于补能行为的过程。示例性的，当燃油车的剩余油量发生了增长，且燃油车处于熄火状态时，可以认为燃油车发生了补能行为。
在上述方案的基础上，通过从样本补能参考数据中，提取与补能意图相关的补能意图数据，实现了样本补能参考数据筛选；同时，将用户的补能意图转化为可以获取到的补能意图数据，让用户的补能意图可视化。通过从补能意图数据中，提取与补能行为相关的补能行为数据，对补能意图数据筛选，使所得到的补能行为数据更加准确。
S130、根据补能行为数据，构建样本车辆的画像数据。
画像数据可以是通过样本车辆的行为数据，产出描述样本车辆的标签数据。画像数据可以包含下述至少一项：停车位置、补能位置、行驶路线和能量数据 等。示例性的，画像数据可以包括：“距离上次加油站的距离”和“距离常去加油站的距离”等数据。
在上述技术方案的基础上，根据补能行为数据，构建画像数据，包括：根据补能行为数据，确定补能位置；根据补能位置，构建样本车辆的画像数据。
补能位置可以包括：加油站、充电站、充电桩或换电站等。
根据补能行为数据，确定补能位置，可以是从补能行为数据中筛选出补能位置的经纬度坐标。示例性的，可以基于加油充电行为发现算法，从补能行为数据中识别出样本车辆每次加油充电的经纬度坐标。
根据补能位置，构建样本车辆的画像数据，可以是根据补能位置的经纬度坐标，从补能行为数据中确定样本车辆常去的补能位置和/或样本车辆上次去的补能位置。示例性的，根据加油充电的经纬度坐标，使用“逆地理解析”服务获取当前坐标点附近的加油站或充电站，并将本次结果存入数据库，就可以获得样本车辆每次加油充电的地点和名称。其中，出现次数最多的补能位置，即为“常去的加油站”；距离当前时刻最近一次的补能位置，即为“上次去的加油站”。
通过确定补能位置，并根据补能位置，构建样本车辆的画像数据，明确了样本车辆的常去的补能位置或上次去的补能位置，并可以根据补能位置判断样本车辆距离补能位置的距离，使辅助判断样本车辆的补能意图更加有针对性。
通过构建样本车辆的画像数据，将样本车辆的补能行为数据更加具体化，使样本车辆的补能意图识别数据更加全面。
S140、根据补能行为数据和画像数据，训练补能意图识别模型。
补能意图识别模型可以是基于补能行为数据和画像数据数据，执行补能意图识别任务的模型。将补能行为数据和画像数据输入到补能意图识别模型中，该模型可以给出样本车辆是否有补能意图的结果。
本实施例的补能行为数据和画像数据对应的是多台样本车辆的数据，可以是针对每一台样本车辆的数据，建立该车辆ID与该车辆的补能行为数据和画像数据之间的映射关系，后续可以依次将每一车辆ID的对应补能行为数据和画像数据作为一组训练数据输入到的补能意图识别模型中，对模型进行训练。
训练补能意图识别模型的过程可以是：将补能行为数据和画像数据输入补能意图识别模型中，该补能意图识别模型即可通过意图预测算法预测该组数据是否含有补能意图，并将补能意图识别模型的预测结果和样本车辆的实际补能意图结果做对比，计算损失函数，然后基于该损失函数调解模型的参数配置，按照上述方案对补能意图识别模型进行多次迭代，即可得到训练好的补能意图识别模型。
意图预测算法可以是线性支持向量分类(Linear Support Vector Classification，LinearSVC)、逻辑回归(LogisticRegression，LR)、决策树(DecisionTree，DT)、随机森林(RandomForest，RF)或梯度提升决策树(Gradient Boosting Decision Tree，GBDT)等。可以选择GBDT算法。
在训练补能意图识别模型之前，还可以对补能行为数据和画像数据进行处理。示例性的，可以基于特征工程处理，包括特征提取、聚合和格式化等，用于提高模型特征识别的显著性。
示例性的，假定补能行为数据有39000条，首先对39000条数据进行随机打乱，随机数种子可以设定为42；然后可以根据“训练集：验证集：测试集”等于“7：1：2”的比例，选出训练集、验证集和测试集；然后用训练集训练补能意图识别模型，用验证集验证补能意图识别模型，再根据情况不断调整补能意图识别模型，得到调整后优化的补能意图识别模型，最后用测试集评估最终的补能意图识别模型，若评估通过，则完成了补能意图识别模型的训练过程。
本申请实施例的技术方案，通过从样本车辆数据中提取样本补能参考数据；根据样本补能参考数据，确定补能行为数据；根据补能行为数据，构建样本车辆的画像数据；根据补能行为数据和画像数据，训练补能意图识别模型。本申请实施例的技术方案采用两阶段的数据筛选操作，即初步筛选补能相关的数据，和二次筛选补能时刻对应的补能行为数据，提高了车辆补能行为数据提取的准确性和高效性。将样本车辆的画像数据与补能行为数据一并作为模型训练数据，使得模型训练数据更加全面，进而提高了补能意图识别模型训练的精准性和泛化性。
实施例二
图3为本申请实施例二提供的一种模型训练方法的流程图，本实施例在上述实施例的基础上，将从样本补能参考数据中，提取与补能意图相关的补能意图数据优化。如图3所示，该方法包括：
S310、从样本车辆数据中提取样本补能参考数据。
S320、根据样本补能参考数据中的车辆传感器数据，确定补能时刻。
根据样本补能参考数据中的车辆传感器数据，确定补能时刻，可以是通过样本补能参考数据中的车辆传感器数据，识别出有补能行为的补能时刻。示例性的，可以通过车辆传感器数据中的“剩余油量/电量”和“当前时速”等数据，识别出车辆加油/充电行为的补能时刻。
通过根据样本补能参考数据中的车辆传感器数据，确定补能时刻，将车辆 传感器数据和补能时刻对应起来，通过确认补能时刻，就可以对应找到该时刻对应的车辆传感器数据。
S330、根据补能时刻和参考行车时长，确定补能意图时段。
补能意图时段可以是有补能意图的时间段。可以是从产生补能意图至执行补能行为之间的时间段。
在上述技术方案的基础上，该模型训练方法确定补能时刻之后，还包括：
若补能时刻为至少两个，则根据至少两个补能时刻，确定平均行车时长；根据平均行车时长和预设间隔百分比，确定参考行车时长。
平均行车时长可以是相邻补能时刻之间的时间间隔的平均值。不同的样本车辆补能频率不同，平均行车时长可以不同。示例性的，设定补能为t1、t2、t3、t4，相邻补能时刻之间的时间间隔为(t2-t1)、(t3–t2)、(t4–t3)，平均行车时长为上述时间间隔的平均值。
预设间隔百分比可以是从产生补能意图直到车辆开始执补能行为的行车时间间隔的百分比。预设间隔百分比可以是基于经验假设的值。
参考行车时长可以是从产生补能意图直到车辆开始执补能行为的行车时间间隔。参考行车时长可以是经验假设的值，可以是一个固定的时间间隔。参考行车时长也可以是通过预设间隔百分比和平均行车时长进行确定的值。示例性的，设定平均行车时长为t，预设间隔百分比为φ，参考行车时长为t*φ。
通过预设间隔百分比和平均行车时长进行确定参考行车时长，可以将不同样本车辆的平均行车时长进行统计，使平均行车时长的数据更加准确，同时，也可以使参考行车时长的数据更准确。根据不同的样本车辆的数据进行统计得到参考行车时长，可以使所得到的参考行车时长数据更有针对性。
确定补能意图时段可以是通过将补能时刻设定为执行补能行为的时间点，从补能时刻回推至参考行车时长所对应的时刻设定为产生补能意图的时间点，两个时间点之间的时间段即为补能意图时段。示例性的，设定补能时刻为T，参考行车时长为ΔT，产生补能意图的时刻为(T-ΔT)，补能意图时段为从(T-ΔT)时刻至T时刻之间的时间段。
通过根据补能时刻和参考行车时长，确定补能意图时段，将补能意图的时间段具体化，便于数据的提取。
S340、提取位于补能意图时段的样本补能参考数据，作为补能意图数据。
样本补能参考数据中包含车辆信息数据和时间数据，通过补能意图时段，根据车辆信息数据和时间数据，可以找到对应的车辆的样本补能参考数据，样 本补能参考数据就可以作为补能意图数据。示例性的，可以将单台样本车辆的单个时刻的样本补能数据作为一条数据集，该数据集包含该时刻下该车辆的样本补能参考数据。可以通过确认车辆ID和车辆的对应时刻，就可以找到对应的样本补能参考数据。
通过提取位于补能意图时段的样本补能参考数据，作为补能意图数据，将补能意图数据的提取出来，为后续快速且精准的提取补能行为数据提供了保障。
S350、从补能意图数据中，提取与补能行为相关的补能行为数据。
S360、根据补能行为数据，构建样本车辆的画像数据。
S370、根据补能行为数据和画像数据，训练补能意图识别模型。
本申请实施例的技术方案，通过从样本车辆数据中提取样本补能参考数据；根据样本补能参考数据中的车辆传感器数据，确定补能时刻；根据补能时刻和参考行车时长，确定补能意图时段；提取位于补能意图时段的样本补能参考数据，作为补能意图数据；从补能意图数据中，提取与补能行为相关的补能行为数据；根据补能行为数据，构建样本车辆的画像数据；根据补能行为数据和画像数据，训练补能意图识别模型。通过补能时刻和参考行车时长，从时间上确认了补能意图时段，并通过补能意图时段实现了对补能意图数据的确认和提取，提高了补能意图识别的准确性。
在上述实施例的基础上，从样本补能参考数据中，提取与补能意图相关的补能意图数据的另一种方式可以包括如下几个步骤：
步骤A、从样本补能参考数据中的车载终端日志数据中，筛选补能意图关键词。
补能意图关键词可以是与补能行为有关的关键词信息。示例性的，可以是补能位置的信息。例如，换电站的地址或换电站的名称等。也可以是补能位置的统称信息。例如，换电站、充电桩或加油站等。
从样本补能参考数据中的车载终端日志数据中，筛选补能意图关键词，可以是从车载终端日志数据中的搜索记录中，筛选出和补能意图相关的关键词的过程。示例性的，可以查看车载终端日志数据中的历史搜索记录，查看“加油站”、“换电站”、加油站的地址或换电站的地址所对应的关键词，就可以是筛选补能意图关键词。
步骤B、从样本补能参考数据中的车辆传感器数据中，筛选补能意图关键词的关联传感数据。
关联传感数据可以是车辆传感器数据中包含补能意图关键词的数据的相关 数据。示例性的，设定补能意图关键词为“加油站”，车辆传感器数据中包含加油站的位置数据，该位置数据所对应的其他车辆传感器数，例如发动机状态数据、车辆速度数据、车辆能量数据、车辆里程数据、车辆信息数据和时间数据等，均为关联传感数据。
步骤C、从样本补能参考数据中的车辆传感器数据中，筛选补能意图关键词的关联传感数据可以是通过补能意图关键词的字段，在车辆传感器数据中确定含有该关键词的字段的数据，通过确定该数据所对应的车辆传感器数据，即确认关联传感数据。
根据补能意图关键词和关联传感数据，确定补能意图数据。
根据补能意图关键词和关联传感数据，确定补能意图数据，可以是通过补能意图关键词，确定车载终端日志数据中关键词所对应的车辆信息数据或时间数据等，再从关联传感数据中找出该车辆信息数据或该时间数据对应的关联传感数据，从而确定了补能意图产生时的补能意图数据，再结合补能行为，可以确定补能意图数据。
本实施例这样设置的好处是：将车辆终端日志数据中的补能意图关键词和车辆传感器数据关联起来，可以更加准确地确定补能意图产生的时刻，提高了补能意图确定的准确性。
实施例三
图4为本申请实施例三提供的一种模型训练方法的流程图，本实施例在上述实施例的基础上，将从补能意图数据中，提取与补能行为相关的补能行为数据优化。如图4所示，该方法包括：
S410、从样本车辆数据中提取样本补能参考数据。
S420、从样本补能参考数据中，提取与补能意图相关的补能意图数据。
S430、依次遍历每个时刻的补能意图数据，判断该时刻是否满足入栈条件，若该时刻满足入栈条件，则执行S440，若该时刻不满足入栈条件，则返回执行S430。
入栈条件可以是补能意图数据满足补能行为的条件。当车辆为燃油车的时候，入栈条件可以是发动机状态数据、车速数据和事件类型；当车辆为电动车的时候，入栈条件可以是电机状态数据、车速数据和事件类型。其中，事件类型可以为“车辆熄火”或“车辆启动”；事件类型也可以是“车辆下电”或“车辆上电”。示例性的，假定车辆为燃油车，入栈条件可以是车辆车速为0、发动机关闭且事 件类型为“车辆熄火”，该场景可能是停车准备补能；入栈条件也可以是车速为0、发动机启动且事件类型为“车辆启动”，该场景可能是补能结束后驶出加油站。
本实施例可以依次针对每一时刻的补能意图数据，判断该时刻是否满足入栈条件，若满足执行后续S440的操作，若不满足针对下一时刻的补能意图数据，继续执行本步骤。
S440、若该时刻满足入栈条件，则将该时刻的补能意图数据入栈后，判断栈内数据条数是否满足预设条数，若栈内数据条数满足预设条数，则执行S450，若栈内数据条数不满足预设条数，则返回执行S430。
栈内数据条数可以为已经入栈的补能意图数据的条数。栈内数据条数不超过预设条数。示例性的，设定预设条数为2条，若补能意图数据入栈后，栈内数据条数为1条，则返回执行S430的操作，继续提取符合入栈条件的补能意图数据。若补能意图数据入栈后，栈内数据条数为2条，则不再继续入栈，执行后续S450的操作。
S450、若栈内数据条数满足预设条数，则根据栈内的补能意图数据，确定是否存在能量增长事件，若存在能量增长事件，则执行S460，若不存在能量增长事件，则清空栈内的补能意图数据，并返回执行S430。
能量增长事件为车辆能量数据发生增长的事件。可以为剩余油量增长，也可以为剩余电量增长。
确定是否存在能量增长事件，对栈内的补能意图数据进行筛选，确认车辆状态数据发生变化的过程中是否存在补能行为，保证了补能行为获取的准确性。
在上述技术方案的基础上，根据栈内的补能意图数据，确定是否存在能量增长事件，可以包括：根据栈内的补能意图数据，确定补能增长量和/或补能时间间隔；根据补能增长量和/或补能时间间隔，确定是否存在能量增长事件。
补能增加量可以是对比栈内补能意图数据得到的车辆能量的变化量。示例性的，可以是剩余油量的增长量，也可以是剩余电量的增长量。
补能时间间隔可以是对比栈内补能意图数据得到的时间间隔数据。示例性的，若栈内有两条数据，补能时间间隔可以为两条数据之间的时间间隔。
示例性的，下表1为栈内补能意图数据的历史数据统计表。如下表1所示，从“2020-12-01 18:16:49”时刻到“2020-12-01 18:19:19”时刻对应加油事件1，加油事件1为正常加油事件，当前油量稳定变化。“2020-12-01 18:19:59”时刻对应加油事件2，加油事件2为异常加油事件，当前油量短暂波动。受限于传感器的精确度，加油数据序列有短暂的波动。为了保证数据的准确性，需要对加油数据进行清洗。
根据本申请实施例的技术方案，根据栈内的补能意图数据，确定补能增长量和/或补能时间间隔；根据补能增长量和/或补能时间间隔，判断补能增长量是否达到增长量阈值和/或补能时间间隔是否达到补能时间间隔阈值，进而判断是否存在能量增长事件，从而实现了对补能意图数据的清洗。
增长量阈值为补能增长量的最小设定值。
补能时间间隔阈值为补能时间间隔的最小设定值。
补能时间间隔阈值可以通过设定波动容忍时间窗，进行数据的平滑处理。补能时间间隔阈值的选取方式如下：
计算两次加油时间间隔，选取时间间隔为T1-Tn。
绘制累积分布占比图：判断累积占比达到90％以上的时间间隔。
选取该时间间隔为补能时间间隔阈值。
示例性的，图5为本申请实施例三提供的一种加油时间间隔的统计分布图。
如图5所示，设定补能时间间隔为加油时间间隔，横轴为两次加油时间间隔，纵轴为正常数据累计占比，补能时间间隔阈值的选取方式如下：
计算两次加油时间间隔，选取时间间隔为(10-600s)。
绘制累积分布占比图：当两次加油时间间隔为300s时，累积占比达到92％。
故选取两次加油时间间隔阈值为300s。
通过设定补能时间间隔阈值，去掉了短暂异常波动的补能意图数据，保障了补能意图参考数据的准确性。
增长量阈值的选取，数据处理过程如下：
计算补能行为数据的车辆能量数据，只保留车辆能量数据＞1；绘制车辆能量数据频数图：确定占比异常多的车辆能量数据最小值；该车辆能量数据为增长量阈值。
示例性的，图6为本申请实施例三提供的一种单次加油量的统计分布图。
如图6所示，设定补能增长量为加油事件的加油量，横轴为每次加油事件的加油量，纵轴为加油量的频数。在单次加油过程中单次加油量异常值的清洗过程，涉及单次加油量阈值的选取，数据处理过程如下：
计算每次加油事件的加油量，只保留加油量>1L。
绘制加油量频数图：当加油量为2L时，占比异常多。
故选取有效加油量阈值为>2L。
通过设定增长量阈值，避免了单次补能异常数据的干扰，保障了数据的准确性。
S460、若存在能量增长事件，则将栈内的补能意图数据作为补能行为数据，并清空栈内的补能意图数据。
本实施例将栈内的补能意图数据作为补能行为数据可以是将该数据提取出来，作为补能行为数据。
清空栈内的补能意图数据可以是将栈内的补能意图数据清空，使栈内补能意图数据变为0条的过程。
通过将栈内的补能意图数据作为补能行为数据，并清空栈内的补能意图数据，确定了补能行为数据，完成了对补能行为数据的提取，同时，清空栈内的 补能意图数据，为后续的补能意图数据的判断提供条件，实现了对数据的循环判断。
S470、根据补能行为数据，构建样本车辆的画像数据。
S480、根据补能行为数据和画像数据，训练补能意图识别模型。
本申请实施例的技术方案通过从样本车辆数据中提取样本补能参考数据；从样本补能参考数据中，提取与补能意图相关的补能意图数据；依次遍历每个时刻的补能意图数据，判断该时刻是否满足入栈条件；若满足，则将该时刻的补能意图数据入栈后，判断栈内数据条数是否满足预设条数；若满足，则根据栈内的补能意图数据，确定是否存在能量增长事件；若存在，则将栈内的补能意图数据作为补能行为数据，并清空栈内的补能意图数据；根据补能行为数据，构建样本车辆的画像数据；根据补能行为数据和画像数据，训练补能意图识别模型。将补能意图数据进行了逐步筛选，实现了对补能行为数据的准确判断。同时，通过清空栈内补能参考数据，实现了对数据的循环判断。
实施例四
图7为本申请实施例四提供的一种补能意图识别方法的流程图，本实施例可适用于车辆补能意图识别的情况，该方法可以由补能意图识别装置来执行，该补能意图识别装置可以采用硬件和/或软件的形式实现，该补能意图识别装置可集成于配置有补能意图识别模型的电子设备中，该电子设备可以是车载终端。
如图7所示，该方法包括：
S710、从当前车辆数据中获取当前补能参考数据。
当前车辆数据可以当前车辆在当前时刻产生的全部实时数据。
本申请任一实施例的补能意图识别的方法中，当前车辆数据包括：车载终端日志数据和车辆传感器数据。
通过车载终端日志数据和车辆传感器数据，将车载终端日志数据和车辆传感器数据一一对应，为补能意图识别过程做准备。
当前补能参考数据可以是当前车辆数据中与补能相关的全部关联数据。
从当前车辆数据中获取当前补能参考数据与上述实施例介绍的从样本车辆数据中提取样本补能参考数据过程类似，在此不进行赘述。
通过从当前车辆数据中获取当前补能参考数据，实现了对数据的初步筛选，去掉了和补能无关的数据。
S720、将当前补能参考数据输入到补能意图识别模型中，得到当前时刻的补能意图识别结果。
补能意图识别模型可以通过本申请任一实施例的模型训练方法训练得到。补能意图识别模型用于对输入的补能参考数据进行分析，给出当前时刻是否存在补能意图识别结果。
补能意图识别结果可以是判断出当前车辆是否有补能意图的结果。示例性的，可能有补能意图，也可能没有补能意图。
在输入补能意图识别模型中之前，可以对数据进行处理。示例性的，可以基于特征工程处理，包括特征提取、聚合和格式化等，用于提高模型特征识别的显著性。
通过补能意图识别模型，解决了无法***车辆的补能意图的问题，实现了通过当前补能参考数据对当前车辆的补能意图的识别预测。
在上述技术方案的基础上，补能意图识别方法从当前车辆数据中获取当前补能参考数据之后，还包括：
从历史补能参考数据中，获取当前补能参考数据的补充参考数据。
历史补能参考数据可以是当前车辆的历史数据中与补能相关的全部关联数据。历史补能参考数据是车辆的历史补能经验值，记录了车辆历史补能记录的相关数据。
补充参考数据可以是通过历史补能参考数据和当前补能参考数据共同的关联性的历史补能参考数据，也可以是直接把当前车辆的历史补能参考数据直接作为补能参考数据。
相应的，将当前补能参考数据输入到补能意图识别模型中，得到当前时刻的补能意图识别结果，包括：
将当前补能参考数据和补充参考数据输入到补能意图识别模型中，得到当前时刻的补能意图识别结果。
将当前补能参考数据和补充参考数据输入到补能意图识别模型中，可以是，将当前补能参考数据和补充参考数据作为一个数据组整体输入到补能意图识别模型中，得到当前时刻的补能意图识别结果。
通过获取补充参考数据，从历史数据上补充了当前补能参考数据，为后续预测提供了更加充足的数据支持，将当前补能参考数据和补充参考数据输入到补能意图识别模型，得到当前时刻的补能意图识别结果，输入的数据更加全面，确保了预测的准确性。
在上述技术方案的基础上，补能意图识别方法将当前补能参考数据输入到补能意图识别模型中，得到当前时刻的补能意图识别结果，包括：
获取当前车辆的画像数据；将当前车辆的画像数据和当前补能参考数据输入到补能意图识别模型中，得到当前时刻的补能意图识别结果。
当前车辆的画像数据可以是预先通过当前车辆的历史行为数据，产出描述当前车辆的标签数据。
可以是将当前车辆的画像数据和当前补能参考数据作为数据组输入到补能意图识别模型中，模型通过计算判断出当前时刻是否具有补能意图。
通过获取画像数据，使补能意图识别模型的输入数据更加全面，保证了补能意图识别的准确性。
本申请实施例的技术方案通过从当前车辆数据中获取当前补能参考数据；将当前补能参考数据输入到补能意图识别模型中，得到当前时刻的补能意图识别结果。通过应用补能意图识别模型，实现了通过当前补能参考数据直接进行补能意图预测识别的功能，提高了补能意图识别的高效性和准确性。
实施例五
图8为本申请实施例五提供的一种补能意图识别模型的训练及应用场景的示意图。本实施例可以实现本申请上述实施例所提供的模型训练方法和补能意图识别方法。如图8所示，先进行模型训练，并反复进行算法评估/优化，得到意图识别模型。然后再将模型进行应用，实现对补能意图的识别过程。
模型训练的过程包括：源数据采集、补能意图样本选取、画像提取、训练数据集构建、建模和算法评估/优。
源数据采集相当于提取样本补能参考数据的过程。
补能意图样本选取相当于提取与补能意图相关的补能意图数据。
画像提取相当于构建画像数据。
训练数据集构建相当于提取补能行为数据和构建画像数据。
建模可以是构建补能意图识别模型，相当于训练补能意图识别模型。
补能意图识别过程包括：实时数据源、画像提取、特征集构建、模型和意图，补能意图识别过程是使用补能意图识别模型进行补能意图识别的过程。
实时数据源相当于从当前车辆数据中获取当前补能参考数据。
画像数据相当于获取当前车辆的画像数据。
特征集构建相当于获取当前补能参考数据和当前车辆的画像数据。
模型相当于将当前车辆的画像数据和当前补能参考数据输入到补能意图识别模型中。
意图相当于得到当前时刻的补能意图识别结果。
图9为本申请实施例五提供的一种补能意图识别的***框架图。该***框架图可集成于车载终端中。本实施例可以实现本申请上述实施例所提供的补能意图识别方法。如图9所示，该***框架由底向上分别为数据接入和存储的基础层、实时/离线计算引擎层及技术实现层和后向服务应用层。
数据源可以是实时采集的车载终端日志数据和车辆传感器数据。数据源通过数据总线进行传输，示例性的，数据总线可以是一种分布式的基于发布/订阅的消息***。数据存储可以通过数据库实现，可以通过多个数据库实现。示例性的，数据库的类型可以包括：基于分布式文件***的数据库管理***、开源的非关系型分布式数据库管理***或关系型数据库管理***等。
实时计算可以基于针对流数据和批数据的分布式处理引擎(Flink)实现。车载终端日志数据和车辆传感器数据可以通过实时计算引擎实现实时计算，计算后的数据可以存储在数据库中。离线计算可以通过离线计算引擎计算车载终端日志数据和车辆传感器数据，计算出车辆的画像数据，画像数据可以存储在数据库中。
特征工程用于输入模型之前对数据进行处理，包括特征提取、聚合和格式化，帮助模型提高特征识别的显著性。补能识别可以用于从当前车辆数据中获取当前补能参考数据。意图预测可以基于车载终端日志数据和传感器数据，通过补能意图识别模型，对补能意图进行预测并输出补能意图预测结果。
还可以针对补意图识别提供相应的服务。示例性的，可以进行活动推荐，针对补能意图推荐合适的优惠活动。也可以进行补能兴趣点(pointofinterest，poi)推荐，基于地理信息、公共服务站点以及公交站等建筑或能够提供服务的服务站点的信息，提供个性化需求推荐。也可以进行加油提醒，当确定用户具有补能意图时，可以通过车机进行加油提醒。其中，加油提醒可以包括：声音、弹窗或图像信息提醒等。
通过构建补能意图识别的***框架，实现了补能意图识别模型的应用，并对补能意图识别服务进行扩展，可以实现更个性化的性能。
实施例六
图10为本申请实施例六提供的一种模型训练装置的结构示意图。该装置可以实现本申请上述实施例所提供的模型训练方法。本实施例可适用于车辆补能意图模型训练的情况，该装置可以由软件和/或硬件的方式来实现，并可集成于具有模型训练的电子设备中。如图10所示，该装置包括：
样本数据提取模块1010，设置为从样本车辆数据中提取样本补能参考数据；行为数据确定模块1020，设置为根据样本补能参考数据，确定补能行为数据；画像数据构建模块1030，设置为根据补能行为数据，构建样本车辆的画像数据；模型训练模块1040，设置为根据补能行为数据和画像数据，训练补能意图识别模型。
本申请实施例的技术方案，通过从样本车辆数据中提取样本补能参考数据；根据样本补能参考数据，确定补能行为数据；根据补能行为数据，构建样本车辆的画像数据；根据补能行为数据和画像数据，训练补能意图识别模型。本申请实施例的技术方案采用两阶段的数据筛选操作，即初步筛选补能相关的数据，和二次筛选补能时刻对应的补能行为数据，提高了车辆补能行为数据提取的准确性和高效性。将样本车辆的画像数据与补能行为数据一并作为模型训练数据，使得模型训练数据更加全面，进而提高了补能意图识别模型训练的精准性和泛化性。
一实施例中，行为数据确定模块1020包括：
意图数据提取单元，设置为从样本补能参考数据中，提取与补能意图相关的补能意图数据；行为数据提取单元，设置为从补能意图数据中，提取与补能行为相关的补能行为数据。
一实施例中，意图数据提取单元，可以包括：
关键词筛选子单元，设置为从样本补能参考数据中的车载终端日志数据中，筛选补能意图关键词；关联数据筛选子单元，设置为从样本补能参考数据中的车辆传感器数据中，筛选补能意图关键词的关联传感数据；意图数据确定子单元，设置为根据补能意图关键词和关联传感数据，确定补能意图数据。
一实施例中，意图数据提取单元，也可以包括：
补能时刻确定子单元，设置为根据样本补能参考数据中的车辆传感器数据，确定补能时刻；意图时段确定子单元，设置为根据补能时刻和参考行车时长，确定补能意图时段；意图数据确定子单元，设置为提取位于补能意图时段的样本补能参考数据，作为补能意图数据。
一实施例中，补能时刻确定子单元在确定补能时刻之后，还包括：
平均时长确定子单元，设置为若补能时刻为至少两个，则根据至少两个补 能时刻，确定平均行车时长；参考时长确定子单元，设置为根据平均行车时长和预设间隔百分比，确定参考行车时长。
一实施例中，补能行为数据提取单元包括：
条件判断子单元，设置为依次遍历每个时刻的补能意图数据，判断该时刻是否满足入栈条件；条数判断子单元，设置为若满足，则将该时刻的补能意图数据入栈后，判断栈内数据条数是否满足预设条数；增长确定子单元，设置为若满足，则根据栈内的补能意图数据，确定是否存在能量增长事件；数据确定子单元，设置为若存在，则将栈内的补能意图数据作为补能行为数据，并清空栈内的补能意图数据。
一实施例中，增长确定子单元可以设置为：
根据栈内的补能意图数据，确定补能增长量和/或补能时间间隔；根据补能增长量和/或补能时间间隔，确定是否存在能量增长事件。
一实施例中，画像数据构建模块包括：
补能位置确定单元，设置为根据补能行为数据，确定补能位置；画像数据构建单元，设置为根据补能位置，构建样本车辆的画像数据。
本申请实施例中任一模型训练的装置中，样本车辆数据包括：车载终端日志数据和车辆传感器数据。
本申请实施例所提供的模型训练装置可执行本申请任意实施例所提供的模型训练方法，具备执行方法相应的功能模块和效果。
实施例七
图11为本申请实施例七提供的一种补能意图识别装置的结构示意图。该装置可以实现本申请上述实施例所提供的补能意图识别方法。本实施例可适用于车辆补能意图识别的情况，该装置可以由软件和/或硬件的方式来实现，并可集成于配置有补能意图识别模型的电子设备中，该电子设备可以是车载终端。如图11所示，该装置包括：
当前数据获取模块1110，设置为从当前车辆数据中获取当前补能参考数据；补能意图识别模块1120，设置为将当前补能参考数据输入到补能意图识别模型中，得到当前时刻的补能意图识别结果；其中，补能意图识别模型通过模型训练装置训练得到。
本申请实施例的技术方案通过从当前车辆数据中获取当前补能参考数据；将当前补能参考数据输入到补能意图识别模型中，得到当前时刻的补能意图识 别结果。通过应用补能意图识别模型，实现了通过当前补能参考数据直接进行补能意图预测识别的功能，提高了补能意图识别的高效性和准确性。
相应的，当前补能数据获取模块1110从当前车辆数据中获取当前补能参考数据之后，还包括：
补充数据获取模块，设置为从历史补能参考数据中，获取当前补能参考数据的补充参考数据；相应的，补能意图识别模块1120包括：
将当前补能参考数据和补充参考数据输入到补能意图识别模型中，得到当前时刻的补能意图识别结果。
相应的，当前数据获取模块1110包括：
画像数据获取单元，设置为获取当前车辆的画像数据；识别结果获取单元，设置为将当前车辆的画像数据和当前补能参考数据输入到补能意图识别模型中，得到当前时刻的补能意图识别结果。
本申请实施例中任一补能意图识别的装置中，当前车辆数据包括：车载终端日志数据和车辆传感器数据。
本申请实施例所提供的补能意图识别装置可执行本申请任意实施例所提供的补能意图识别方法，具备执行方法相应的功能模块和效果。
实施例八
图12为本申请实施例八提供的一种电子设备的结构示意图，图12示出了适于用来实现本申请实施例实施方式的示例性设备的框图。图12显示的设备仅仅是一个示例，不应对本申请实施例的功能和使用范围带来任何限制。
如图12所示，电子设备1200以通用计算设备的形式表现。电子设备1200的组件可以包括但不限于：一个或者多个处理器或者处理单元1210，***存储器1220，连接不同***组件(包括***存储器1220和处理单元1210)的总线1230。
总线1230表示几类总线结构中的一种或多种，包括存储器总线或者存储器控制器，***总线，图形加速端口，处理器或者使用多种总线结构中的任意总线结构的局域总线。举例来说，这些体系结构包括但不限于工业标准体系结构(Industry Standard Architecture，ISA)总线，微通道体系结构(Micro Channel Architecture，MAC)总线，增强型ISA总线、视频电子标准协会(Video Electronics Standards Association，VESA)局域总线以及***组件互连(Peripheral Component Interconnect，PCI)总线。
电子设备1200包括多种计算机***可读介质。这些介质可以是任何能够被电子设备1200访问的可用介质，包括易失性和非易失性介质，可移动的和不可移动的介质。
***存储器1220可以包括易失性存储器形式的计算机***可读介质，例如随机存取存储器(Random Access Memory，RAM)1221和/或高速缓存存储器(高速缓存1222)。电子设备1200可以包括其它可移动/不可移动的、易失性/非易失性计算机***存储介质。仅作为举例，存储***1223可以设置为读写不可移动的、非易失性磁介质(图12未显示，通常称为“硬盘驱动器”)。尽管图12中未示出，可以提供设置为对可移动非易失性磁盘(例如“软盘”)读写的磁盘驱动器，以及对可移动非易失性光盘(例如光盘只读存储器(Compact Disc Read-Only Memory，CD-ROM)，数字视频光盘只读存储器(Digital Video Disk Read-Only Memory，DVD-ROM)或者其它光介质)读写的光盘驱动器。在这些情况下，每个驱动器可以通过一个或者多个数据介质接口与总线1230相连。***存储器1220可以包括至少一个程序产品，该程序产品具有一组(例如至少一个)程序模块，这些程序模块被配置以执行本申请实施例的功能。
具有一组(至少一个)程序模块1224的程序/实用工具1225，可以存储在例如***存储器1220中，这样的程序模块1224包括但不限于操作***、一个或者多个应用程序、其它程序模块以及程序数据，这些示例中的每一个或一种组合中可能包括网络环境的实现。程序模块1224通常执行本申请实施例所描述的实施例中的功能和/或方法。
电子设备1200也可以与一个或多个外部设备1300(例如键盘、指向设备、显示器1310等)通信，还可与一个或者多个使得用户能与该电子设备1200交互的设备通信，和/或与使得该电子设备1200能与一个或多个其它计算设备进行通信的任何设备(例如网卡，调制解调器等等)通信。这种通信可以通过输入/输出(Input/Output，I/O)接口1240进行。并且，电子设备1200还可以通过网络适配器1250与一个或者多个网络(例如局域网(Local Area Network，LAN)，广域网(Wide Area Network，WAN)和/或公共网络，例如因特网)通信。如图所示，网络适配器1250通过总线1230与电子设备1200的其它模块通信。应当明白，尽管图中未示出，可以结合电子设备1200使用其它硬件和/或软件模块，包括但不限于：微代码、设备驱动器、冗余处理单元、外部磁盘驱动阵列、独立磁盘冗余阵列(Redundant Arrays of Independent Disks，RAID)***、磁带驱动器以及数据备份存储***等。
处理单元1210通过运行存储在***存储器1220中的程序，从而执行多种功能应用以及数据处理，例如实现本申请实施例所提供的模型训练方法和补能 意图识别方法。
实施例八
本申请实施例八还提供一种计算机可读存储介质，其上存储有计算机程序(或称为计算机可执行指令)，该程序被处理器执行时用于执行本申请实施例所提供的模型训练方法和补能意图识别方法。
本申请实施例的计算机存储介质，可以采用一个或多个计算机可读的介质的任意组合。计算机可读介质可以是计算机可读信号介质或者计算机可读存储介质。计算机可读存储介质例如可以是但不限于电、磁、光、电磁、红外线、或半导体的***、装置或器件，或者任意以上的组合。计算机可读存储介质的例子(非穷举的列表)包括：具有一个或多个导线的电连接、便携式计算机磁盘、硬盘、RAM、ROM、可擦式可编程只读存储器(Erasable Programmable Read-Only Memory，EPROM或闪存)、光纤、CD-ROM、光存储器件、磁存储器件、或者上述的任意合适的组合。在本文件中，计算机可读存储介质可以是任何包含或存储程序的有形介质，该程序可以被指令执行***、装置或者器件使用或者与其结合使用。
计算机可读的信号介质可以包括在基带中或者作为载波一部分传播的数据信号，其中承载了计算机可读的程序代码。这种传播的数据信号可以采用多种形式，包括但不限于电磁信号、光信号或上述的任意合适的组合。计算机可读的信号介质还可以是计算机可读存储介质以外的任何计算机可读介质，该计算机可读介质可以发送、传播或者传输用于由指令执行***、装置或者器件使用或者与其结合使用的程序。
计算机可读介质上包含的程序代码可以用任何适当的介质传输，包括但不限于无线、电线、光缆、射频(Radio Frequency，RF)等等，或者上述的任意合适的组合。
可以以一种或多种程序设计语言或其组合来编写用于执行本申请实施例操作的计算机程序代码，程序设计语言包括面向对象的程序设计语言—诸如Java、Smalltalk、C++，还包括常规的过程式程序设计语言诸如“C”语言或类似的程序设计语言。程序代码可以完全地在用户计算机上执行、部分地在用户计算机上执行、作为一个独立的软件包执行、部分在用户计算机上部分在远程计算机上执行、或者完全在远程计算机或服务器上执行。在涉及远程计算机的情形中，远程计算机可以通过任意种类的网络包括LAN或WAN连接到用户计算机，或者，可以连接到外部计算机(例如利用因特网服务提供商来通过因特网连接)。
本公开的技术方案中，所涉及的样本车辆数据、样本补能参考数据、补能行为数据、样本车辆的画像数据、车载终端日志数据、车辆传感器数据、关联传感数据、补能意图数据、当前车辆数据、当前补能参考数据、历史补能参考数据、补充参考数据、当前车辆的画像数据等的获取，存储和应用等，均符合相关法律法规的规定，且不违背公序良俗。
本实施例中的补能意图识别模型并不是针对一特定用户的，并不能反映出一特定用户的个人信息。

Claims (17)

  1. 一种模型训练方法，包括：
    从样本车辆数据中提取样本补能参考数据；
    根据所述样本补能参考数据，确定补能行为数据；
    根据所述补能行为数据，构建所述样本车辆的画像数据；
    根据所述补能行为数据和所述画像数据，训练补能意图识别模型。
  2. 根据权利要求1所述的方法，其中，所述根据所述样本补能参考数据，确定补能行为数据，包括：
    从所述样本补能参考数据中，提取与补能意图相关的补能意图数据；
    从所述补能意图数据中，提取与补能行为相关的所述补能行为数据。
  3. 根据权利要求2所述的方法，其中，所述从所述样本补能参考数据中，提取与补能意图相关的补能意图数据，包括：
    从所述样本补能参考数据中的车载终端日志数据中，筛选补能意图关键词；
    从所述样本补能参考数据中的车辆传感器数据中，筛选所述补能意图关键词的关联传感数据；
    根据所述补能意图关键词和所述关联传感数据，确定所述补能意图数据。
  4. 根据权利要求2所述的方法，其中，所述从所述样本补能参考数据中，提取与补能意图相关的补能意图数据，包括：
    根据所述样本补能参考数据中的车辆传感器数据，确定补能时刻；
    根据所述补能时刻和参考行车时长，确定补能意图时段；
    提取位于所述补能意图时段的样本补能参考数据，作为所述补能意图数据。
  5. 根据权利要求4所述的方法，在所述确定补能时刻之后，还包括：
    在所述补能时刻为至少两个的情况下，根据所述至少两个补能时刻，确定平均行车时长；
    根据所述平均行车时长和预设间隔百分比，确定所述参考行车时长。
  6. 根据权利要求2所述的方法，其中，所述从所述补能意图数据中，提取与补能行为相关的所述补能行为数据，包括：
    依次遍历每个时刻的补能意图数据，判断所述时刻是否满足入栈条件；
    响应于所述时刻满足所述所述时刻，将所述时刻的补能意图数据入栈后，判断栈内数据条数是否满足预设条数；
    响应于所述栈内数据条数满足所述预设条数，根据栈内的补能意图数据，确定是否存在能量增长事件；
    作用于存在所述能量增长事件，将所述栈内的补能意图数据作为所述补能行为数据，并清空栈内的补能意图数据。
  7. 根据权利要求6所述的方法，其中，所述根据栈内的补能意图数据，确定是否存在能量增长事件，包括：
    根据所述栈内的补能意图数据，确定补能增长量和补能时间间隔中的至少之一；
    根据所述补能增长量和所述补能时间间隔中的至少之一，确定是否存在所述能量增长事件。
  8. 根据权利要求1所述的方法，其中，所述根据所述补能行为数据，构建所述样本车辆的画像数据，包括：
    根据所述补能行为数据，确定补能位置；
    根据所述补能位置，构建所述样本车辆的画像数据。
  9. 根据权利要求1-8中任一项所述的方法，其中，所述样本车辆数据包括：车载终端日志数据和车辆传感器数据。
  10. 一种补能意图识别方法，包括：
    从当前车辆数据中获取当前补能参考数据；
    将所述当前补能参考数据输入到补能意图识别模型中，得到当前时刻的补能意图识别结果；
    其中，所述补能意图识别模型通过权利要求1-9中任一项所述的方法训练得到。
  11. 根据权利要求10所述的方法，在所述从当前车辆数据中获取当前补能参考数据之后，还包括：
    从历史补能参考数据中，获取所述当前补能参考数据的补充参考数据；
    所述将所述当前补能参考数据输入到补能意图识别模型中，得到当前时刻的补能意图识别结果，包括：
    将所述当前补能参考数据和所述补充参考数据输入到所述补能意图识别模型中，得到当前时刻的补能意图识别结果。
  12. 根据权利要求10所述的方法，其中，所述将所述当前补能参考数据输 入到补能意图识别模型中，得到当前时刻的补能意图识别结果，包括：
    获取当前车辆的画像数据；
    将所述当前车辆的画像数据和所述当前补能参考数据输入到所述补能意图识别模型中，得到当前时刻的补能意图识别结果。
  13. 根据权利要求10-12中任一项所述的方法，其中，所述当前车辆数据包括：车载终端日志数据和车辆传感器数据。
  14. 一种模型训练装置，包括：
    样本数据提取模块，设置为从样本车辆数据中提取样本补能参考数据；
    行为数据确定模块，设置为根据所述样本补能参考数据，确定补能行为数据；
    画像数据构建模块，设置为根据所述补能行为数据，构建所述样本车辆的画像数据；
    模型训练模块，设置为根据所述补能行为数据和所述画像数据，训练补能意图识别模型。
  15. 一种补能意图识别装置，包括：
    当前数据获取模块，设置为从当前车辆数据中获取当前补能参考数据；
    补能意图识别模块，设置为将所述当前补能参考数据输入到补能意图识别模型中，得到当前时刻的补能意图识别结果；
    其中，所述补能意图识别模型通过权利要求1-9中任一项所述的方法训练得到。
  16. 一种电子设备，包括：
    至少一个处理器；
    存储器，设置为存储至少一个程序；
    当所述至少一个程序被所述至少一个处理器执行，使得所述至少一个处理器实现如权利要求1-9中任一项所述的模型训练方法，或权利要求10-13中任一项所述的补能意图识别方法。
  17. 一种计算机可读存储介质，存储有计算机程序，所述程序被处理器执行时实现如权利要求1-9中任一所述的模型训练方法，或权利要求10-13中任一所述的补能意图识别方法。
PCT/CN2023/098229 2022-06-14 2023-06-05 模型训练方法及装置、补能意图识别方法及装置、设备、介质 WO2023241388A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202210670153.1 2022-06-14
CN202210670153.1A CN114896482A (zh) 2022-06-14 2022-06-14 一种模型训练和补能意图识别方法、装置、设备及介质

Publications (1)

Publication Number Publication Date
WO2023241388A1 true WO2023241388A1 (zh) 2023-12-21

Family

ID=82728979

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2023/098229 WO2023241388A1 (zh) 2022-06-14 2023-06-05 模型训练方法及装置、补能意图识别方法及装置、设备、介质

Country Status (2)

Country Link
CN (1) CN114896482A (zh)
WO (1) WO2023241388A1 (zh)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114896482A (zh) * 2022-06-14 2022-08-12 中国第一汽车股份有限公司 一种模型训练和补能意图识别方法、装置、设备及介质

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20170228840A1 (en) * 2016-02-08 2017-08-10 Xerox Corporation Method and system for identifying locations for placement of replenishment stations for vehicles
CN111741261A (zh) * 2020-06-17 2020-10-02 上海蔚来汽车有限公司 车辆补能意图预测方法、装置以及车辆补能***
CN112307335A (zh) * 2020-10-29 2021-02-02 中国第一汽车股份有限公司 车辆服务信息的推送方法、装置、设备及车辆
CN113050656A (zh) * 2021-03-29 2021-06-29 恒大新能源汽车投资控股集团有限公司 车辆补能方法、车载终端、补能控制平台和移动终端
CN114896482A (zh) * 2022-06-14 2022-08-12 中国第一汽车股份有限公司 一种模型训练和补能意图识别方法、装置、设备及介质

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20170228840A1 (en) * 2016-02-08 2017-08-10 Xerox Corporation Method and system for identifying locations for placement of replenishment stations for vehicles
CN111741261A (zh) * 2020-06-17 2020-10-02 上海蔚来汽车有限公司 车辆补能意图预测方法、装置以及车辆补能***
CN112307335A (zh) * 2020-10-29 2021-02-02 中国第一汽车股份有限公司 车辆服务信息的推送方法、装置、设备及车辆
CN113050656A (zh) * 2021-03-29 2021-06-29 恒大新能源汽车投资控股集团有限公司 车辆补能方法、车载终端、补能控制平台和移动终端
CN114896482A (zh) * 2022-06-14 2022-08-12 中国第一汽车股份有限公司 一种模型训练和补能意图识别方法、装置、设备及介质

Also Published As

Publication number Publication date
CN114896482A (zh) 2022-08-12

Similar Documents

Publication Publication Date Title
CN112867628B (zh) 显示电动车辆的充电选项
WO2021135653A1 (zh) 一种识别车辆的异常停留的方法和***
JP6559792B2 (ja) オーダーペアリングのシステム及び方法
US9043151B2 (en) Large scale demand responsive transit framework
WO2016127918A1 (zh) 一种运力调度方法及***
CN109712392B (zh) 基于区块链的停车信息处理方法、装置、设备和存储介质
CN106875670B (zh) Spark平台下基于GPS数据的出租车调配方法
WO2022116662A1 (zh) 车辆数据处理方法、装置、计算机设备和存储介质
CN102163214B (zh) 一种数字地图生成装置及方法
CN102902752A (zh) 一种日志监控方法及***
WO2023241388A1 (zh) 模型训练方法及装置、补能意图识别方法及装置、设备、介质
US20210333119A1 (en) System and method for cloud computing-based vehicle configuration
CN112307335A (zh) 车辆服务信息的推送方法、装置、设备及车辆
CN113222057A (zh) 数据预测模型训练、数据预测方法、装置、设备及介质
US20150330798A1 (en) Grid-based arrival time prediction
WO2021029822A1 (en) System and method for dispatching vehicles technical field
CN115122977A (zh) 电动汽车充电站的检测、表征和呈现
WO2024032132A1 (zh) 净水量的校验方法、装置、设备和介质
CN117241212A (zh) 车辆常驻停车点的确定方法、装置、设备以及存储介质
CN112070377A (zh) 出行服务处理方法、装置、电子设备及存储介质
CN110704745A (zh) 一种车载终端的信息搜索方法和装置
CN114463029A (zh) 一种车辆加油日确定方法、装置、设备和相关***
CN114187775A (zh) 基于大数据的停车位推荐的方法、装置、设备及存储介质
Khan Real time predictive monitoring system for urban transport
CN114077858A (zh) 向量数据处理方法、装置、设备及存储介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 23822957

Country of ref document: EP

Kind code of ref document: A1