A data processing method and device, electronic equipment and storage medium

Abstract

The application provides a data processing method and device, electronic equipment and storage medium, comprising: obtaining a root directory of original real vehicle running data through a graphical control interface; traversing each file in the root directory to obtain the original real vehicle running data; preprocessing the original real vehicle running data to obtain a training sample set; extracting feature parameters from each group of sample data in the training sample set; and generating an evaluation table of the training sample set based on the feature parameters corresponding to each group of samples. The technical solution provided by the application can obtain high-quality and high-value training samples.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a data processing method, apparatus, electronic device, and storage medium. Background Technology

[0002] Offline reinforcement learning is a key technology for addressing the high risk of online exploration in rule-based energy management strategies and the poor accuracy of simulation models in learning-based energy management strategies. By utilizing large-scale historical data collected during actual vehicle operation, a standardized Markov Decision Process (MDP) training dataset is constructed, consisting of a sequence structure of "state-action-reward-next state-termination flag".

[0003] However, in practical engineering applications, raw data collected from actual vehicles is difficult to use directly for offline reinforcement learning training. Specifically, the data sources exported from the CAN bus are complex, resulting in multiple signal sources for the same signal; the sampling frequencies of various signals are inconsistent, making it difficult to align multi-source data in the time dimension; the raw data generally contains a large amount of non-steady-state noise (such as torque fluctuations) and non-numerical logic (such as clutch status and gear position), which cannot be directly used for model calculation; the actual vehicle operation data itself does not contain the reward signal necessary for offline reinforcement learning training, and cannot directly form a standard dataset, and the data fragments usually do not have clear operating condition semantic information, resulting in a large number of invalid samples participating in learning during the training process, thereby reducing model training efficiency and affecting policy optimization results. Summary of the Invention

[0004] In view of this, embodiments of this application provide a data processing method, apparatus, electronic device, and storage medium that can obtain high-quality, high-value standardized training samples, thereby accelerating the convergence of offline reinforcement learning models and improving their robustness in energy management decisions under complex operating conditions.

[0005] This application mainly includes the following aspects: In a first aspect, embodiments of this application provide a data processing method, the processing method comprising: Obtain the root directory of the original real vehicle operation data through the graphical control interface; Traverse each file in the root directory to obtain the original vehicle operation data; The raw real-vehicle operation data is preprocessed to obtain a training sample set; Extract feature parameters from each set of sample data in the training sample set; An evaluation table for the training sample set is generated based on the feature parameters corresponding to each group of samples.

[0006] Furthermore, the preprocessing of the original real-vehicle operation data to obtain the training sample set includes: The negative values ​​in the engine torque signal in the original real vehicle operation data are corrected to zero, and the corrected original real vehicle operation data is determined as the first real vehicle operation data. Based on the preset signal quality priority rules, the single source signal with the highest confidence is selected from the multi-source signals of the first real vehicle operation data, and the selected first real vehicle operation data is determined as the second real vehicle operation data. Each signal in the second real vehicle operation data is interpolated according to a preset sampling frequency, and the interpolated second real vehicle operation data is determined as the third real vehicle operation data; Multiple third-vehicle operation data segments contained in each file are spliced ​​together in chronological order of collection time, and the spliced ​​third-vehicle operation data is determined as the fourth-vehicle operation data. The third real vehicle operation data of the sub-files in each file are spliced ​​together according to the collection time order, and the spliced ​​third real vehicle operation data is determined as the fourth real vehicle operation data; Based on the fourth real-vehicle operation data, a training sample set is constructed.

[0007] Furthermore, the construction of the training sample set based on the fourth real-vehicle operation data includes: Based on the vehicle speed, engine torque, and engine speed at each data collection moment in the fourth real vehicle operation data, the required power of the real vehicle at each data collection moment is determined. Based on the fuel consumption, SOC offset, and impact penalty at engine start-up moment of the actual vehicle at each data collection time, the reward value of the actual vehicle at each data collection time is determined. The clutch status signal in the fourth vehicle operation data is mapped to a binary value, and the mapped fourth vehicle operation data is determined as the fifth vehicle operation data. The fifth real-vehicle operation data, the required power, the fuel consumption, and the reward value are used to construct a training sample set.

[0008] Furthermore, the evaluation table for generating the training sample set based on the feature parameters corresponding to each group of samples includes: The feature parameters corresponding to each group of sample data are normalized respectively; Based on the feature parameters corresponding to each group of sample data after normalization, the comprehensive score of each group of samples is determined. Based on the comprehensive score of all group samples, all group samples are sorted. An evaluation table for the training sample set is generated based on all sorted group samples.

[0009] Furthermore, the processing method also includes: The feature parameters corresponding to each group of sample data are displayed on a visualization dashboard.

[0010] Secondly, embodiments of this application also provide a data processing apparatus, the processing apparatus comprising: The acquisition module is used to obtain the root directory of the original real vehicle operation data through a graphical control interface; The traversal module is used to traverse each file in the root directory to obtain the original real vehicle operation data; The preprocessing module is used to preprocess the raw real vehicle operation data to obtain a training sample set; The extraction module is used to extract feature parameters from each group of sample data in the training sample set; The generation module is used to generate an evaluation table for the training sample set based on the feature parameters corresponding to each group of samples.

[0011] Furthermore, the preprocessing module is specifically used for: The negative values ​​in the engine torque signal in the original real vehicle operation data are corrected to zero, and the corrected original real vehicle operation data is determined as the first real vehicle operation data. Based on the preset signal quality priority rules, the single source signal with the highest confidence is selected from the multi-source signals of the first real vehicle operation data, and the selected first real vehicle operation data is determined as the second real vehicle operation data. Each signal in the second real vehicle operation data is interpolated according to a preset sampling frequency, and the interpolated second real vehicle operation data is determined as the third real vehicle operation data; Multiple third-vehicle operation data segments contained in each file are spliced ​​together in chronological order of collection time, and the spliced ​​third-vehicle operation data is determined as the fourth-vehicle operation data. The third real vehicle operation data of the sub-files in each file are spliced ​​together according to the collection time order, and the spliced ​​third real vehicle operation data is determined as the fourth real vehicle operation data; Based on the fourth real-vehicle operation data, a training sample set is constructed.

[0012] Furthermore, when the preprocessing module is used to construct a training sample set based on the fourth real-vehicle operation data, it is specifically used for: Based on the vehicle speed, engine torque, and engine speed at each data collection moment in the fourth real vehicle operation data, the required power of the real vehicle at each data collection moment is determined. Based on the fuel consumption, SOC offset, and impact penalty at engine start-up moment of the actual vehicle at each data collection time, the reward value of the actual vehicle at each data collection time is determined. The clutch status signal in the fourth vehicle operation data is mapped to a binary value, and the mapped fourth vehicle operation data is determined as the fifth vehicle operation data. The fifth real-vehicle operation data, the required power, the fuel consumption, and the reward value are used to construct a training sample set.

[0013] Thirdly, embodiments of this application also provide an electronic device, including: a processor, a memory, and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor communicates with the memory through the bus, and the machine-readable instructions are executed by the processor to perform the steps of the data processing method described in the first aspect or any possible implementation of the first aspect.

[0014] Fourthly, embodiments of this application also provide a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the steps of the data processing method described in the first aspect or any possible implementation of the first aspect.

[0015] This application provides a data processing method, apparatus, electronic device, and storage medium, which obtains the root directory of the original real vehicle operation data through a graphical control interface; traverses each file in the root directory to obtain the original real vehicle operation data; preprocesses the original real vehicle operation data to obtain a training sample set; extracts feature parameters from each group of sample data in the training sample set; and generates an evaluation table for the training sample set based on the feature parameters corresponding to each group of samples.

[0016] In this way, high-quality, high-value training samples can be obtained.

[0017] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A flowchart of a data processing method provided in an embodiment of this application is shown; Figure 2 The graphical control interface provided in the embodiments of this application is shown; Figure 3 This application illustrates a visual dashboard provided in an embodiment of the present application; Figure 4 This invention provides a schematic diagram of the structure of a data processing method apparatus according to an embodiment of the present application. Figure 5 A schematic diagram of the structure of an electronic device provided in an embodiment of this application is shown. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the drawings in this application are for illustrative and descriptive purposes only and are not intended to limit the scope of protection of this application. Furthermore, it should be understood that the schematic drawings are not drawn to scale. The flowcharts used in this application illustrate operations implemented according to some embodiments of this application. It should be understood that the operations in the flowcharts may not be implemented in sequence, and steps without logical contextual relationships may be reversed or implemented simultaneously. In addition, those skilled in the art, guided by the content of this application, may add one or more other operations to the flowcharts, or remove one or more operations from the flowcharts.

[0021] Furthermore, the described embodiments are merely some, not all, of the embodiments of this application. The components of the embodiments of this application described and illustrated herein can typically be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0022] The methods, apparatus, electronic devices, or computer-readable storage media described in this application can be applied to any scenario that requires data processing. This application does not limit the specific application scenario, and any scheme that uses the data processing methods and apparatus provided in this application is within the protection scope of this application.

[0023] It is worth noting that offline reinforcement learning is a key technology for addressing the high risks of online exploration in rule-based energy management strategies and the poor accuracy of simulation models in learning-based energy management strategies. By utilizing large-scale historical data collected during real-vehicle operation, a standardized Markov Decision Process (MDP) training dataset is constructed, consisting of a sequence structure of "state-action-reward-next-moment state-termination flag". However, in practical engineering applications, the raw data collected from real vehicles is difficult to directly use for offline reinforcement learning training. Specifically, the data sources exported from the CAN bus are complex, resulting in multiple sources for the same signal; the sampling frequencies of various signals are inconsistent, making it difficult to align multi-source data in the time dimension; the raw data generally contains a large amount of non-steady-state noise (such as torque fluctuations) and non-numerical logic (such as clutch state and gear position), which cannot be directly used for model calculation; the real-vehicle operation data itself does not contain the reward signal necessary for offline reinforcement learning training, making it impossible to directly form a standard dataset; and data fragments usually lack clear operating condition semantic information, resulting in a large number of invalid samples participating in the learning process, thereby reducing model training efficiency and affecting strategy optimization performance.

[0024] To address the aforementioned issues, this application proposes a data processing method, apparatus, electronic device, and storage medium that can obtain high-quality, high-value standardized training samples, thereby accelerating the convergence of offline reinforcement learning models and improving their robustness in energy management decisions under complex operating conditions.

[0025] To facilitate understanding of this application, the technical solutions provided in this application will be described in detail below with reference to specific embodiments.

[0026] Please see Figure 1 , Figure 1 This is a flowchart illustrating a data processing method provided in an embodiment of this application.

[0027] like Figure 1 As shown in the figure, the data processing method provided in this application embodiment includes the following steps: Step S101: Obtain the root directory of the original actual vehicle operation data through the graphical control interface.

[0028] In this embodiment of the application, the graphical control interface refers to a visual operation panel used for human-computer interaction, such as... Figure 2As shown, the graphical control interface includes a data path input control, allowing users to specify the root directory path for storing raw vehicle operating data. For example, the raw vehicle operating data can originate from the CAN bus logs collected by a plug-in hybrid electric vehicle under suburban conditions. This data is extracted from the original BIF (Bus Interface File) file into scattered .mat format files. The raw vehicle operating data includes signals such as engine torque, battery SOC (State of Charge), vehicle speed, and clutch speed. The graphical control interface receives the user-inputted root directory path.

[0029] Step S102: Traverse each file in the root directory to obtain the original vehicle operation data.

[0030] In this embodiment, a recursive traversal operation is performed on the acquired root directory. Since the CAN bus data collected from the actual vehicle is scattered and stored in multiple folders after being exported, forming a large number of fragmented .mat format files, the recursive traversal automatically enters each folder under the root directory, searches and collects all the original actual vehicle operation data that conforms to the fragmented format one by one.

[0031] Step S103: Preprocess the original real vehicle operation data to obtain a training sample set.

[0032] Here, by preprocessing the raw real-vehicle operation data, the fragmented logs are reconstructed into a data chain consisting of "state—action—reward—next moment state—termination flag," thereby constructing a training dataset suitable for offline reinforcement learning. Through a graphical control interface, the background automatically executes the complete transformation process from raw data logs to the training dataset.

[0033] Regarding step S103, as an example in specific implementation, it may include the following steps: Step S1031: Correct the negative value in the engine torque signal in the original vehicle operation data to zero, and determine the corrected original vehicle operation data as the first vehicle operation data.

[0034] In this embodiment, the step performs engine torque correction based on physical constraints. The engine's output torque under normal driving conditions (i.e., non-reverse towing) is non-negative, but the vehicle's sensors may introduce negative signals due to zero-point drift. Here, abnormally negative values ​​of the engine torque signal in the original vehicle operating data are identified, and these abnormally negative values ​​are forcibly set to zero to eliminate physical negative values. This ensures that the corrected engine torque sequence matches the defined domain of the engine characteristic curve chart. The entire original vehicle operating data after correction is then determined as the first vehicle operating data.

[0035] Step S1032: Based on the preset signal quality priority rules, select the single source signal with the highest confidence from the multi-source signals of the first real vehicle operation data, and determine the selected first real vehicle operation data as the second real vehicle operation data.

[0036] In this embodiment, this step performs multi-source signal selection. Since multiple signals from different controllers (such as ECU, VCU, BMS) exist simultaneously on the vehicle's CAN bus, the signal with the highest confidence level is automatically selected as the valid signal source based on a preset signal quality priority rule. As an example, the preset signal quality priority rule can be set as follows: sum the multiple source signals; if the result is zero, randomly retain one source signal; if the result is non-zero, retain the non-zero source signal to eliminate irrelevant interference.

[0037] Step S1033: Interpolate each signal in the second actual vehicle operation data according to a preset sampling frequency, and determine the interpolated second actual vehicle operation data as the third actual vehicle operation data.

[0038] In this embodiment, this step performs asynchronous signal interpolation alignment to address the data misalignment problem caused by inconsistent sampling frequencies of different controllers. In actual engineering, the sampling frequency of the Battery Management System (BMS) and the Engine Control Unit (ECU) may not match; for example, one may be 10 Hz and the other 100 Hz, resulting in misalignment of state variables at the same timestamp on the time axis. This step first extracts a global time reference, and then uses a linear interpolation algorithm to calculate the time series of each signal in the second real vehicle operation data, uniformly interpolating it to a preset sampling frequency, so that all signals correspond point-by-point on a unified time axis. As an example, the preset sampling frequency can be 100 Hz.

[0039] Step S1034: The multiple third vehicle operation data segments contained in each file are spliced ​​together in the order of collection time, and the spliced ​​third vehicle operation data is determined as the fourth vehicle operation data.

[0040] In this embodiment, this step performs vertical splicing, aiming to reconstruct fragmented file content into a continuous long-time sequence. Since real-vehicle operation data is often stored in numerous fragments, multiple third real-vehicle operation data fragments contained in each file are sequentially spliced ​​vertically using vectorization according to their respective acquisition times, integrating the originally discrete fragments into a continuous, complete sequence in time. Here, vertical vectorization splicing refers to a merging operation where the first and last segments are joined together in the vector dimension. The resulting continuous long-time sequence set is determined as the fourth real-vehicle operation data.

[0041] Step S1035: Based on the fourth real vehicle operation data, construct a training sample set.

[0042] Regarding step S1035, as an example in specific implementation, it may include the following steps: Step S10351: Based on the vehicle speed, corrected engine torque, and engine speed at each acquisition time in the fourth real vehicle operation data, determine the required power of the real vehicle at each acquisition time.

[0043] In this embodiment, the required power is calculated based on the longitudinal dynamics model. The required power is determined using the vehicle speed signal, engine torque signal, and engine speed signal from the fourth real vehicle operation data. The required power can be calculated using formula (1).

[0044] (1).

[0045] in, For engine torque, Where 9550 is the engine speed, and 9550 is the unit conversion constant. This represents the required power. The required power sequence serves as an important physical constraint for offline reinforcement learning environment observations.

[0046] Step S10352: Based on the fuel consumption, SOC offset, and impact penalty at engine start-up moment of the actual vehicle at each data collection moment, determine the reward value of the actual vehicle at each data collection moment.

[0047] In this embodiment, an engine fuel map table corresponding to the actual vehicle model is introduced. This map table stores the fuel consumption of the engine under different combinations of engine torque and engine speed. Based on the engine torque and engine speed at each data collection time, the engine fuel map table is queried to obtain the fuel consumption at each data collection time.

[0048] In this embodiment, the reward value is calculated using a hierarchical step-wise reward function, serving as the reward label for offline reinforcement learning training. Specifically, the reward value calculation comprises three components: the first is an energy consumption penalty, which applies a negative incentive to fuel consumption, with higher fuel consumption resulting in a heavier penalty; the second is a SOC penalty, which penalizes the difference between the actual SOC and the target SOC (SOC offset), triggering a step-wise survival penalty when the SOC falls below a preset survival threshold, with the penalty intensity increasing progressively with the offset; the third is an impact penalty, which monitors the engine's start-up transient events (start-up from standstill to running) and applies a fixed penalty value to each start-up transient to suppress frequent engine start-stops. The sum of these three penalty items yields the reward value corresponding to each acquisition moment. ,in, As a reward value, As an energy consumption penalty item, This is a SOC penalty item. This is an impact penalty. For example, when SOC drops from 30% to 29.9%, a negative reward is given; when SOC falls below 25.8%, a tiered survival penalty is triggered; and when the moment of engine start-up is detected, an additional fixed penalty is applied.

[0049] Step S10353: Map the clutch status signal in the fourth vehicle operation data to a binary value, and determine the mapped fourth vehicle operation data as the fifth vehicle operation data.

[0050] In this embodiment, the clutch status signal is a logical status field in the original vehicle operation data, such as control mode fields like "Locked" and "Unlocked". It is mapped to a binary value according to a preset mapping rule, that is, two different values ​​(e.g., 1 and 0) are used to represent the clutch engagement and disengagement states respectively, thus completing the numerical and tensor encapsulation of the signal.

[0051] Step S10354: Construct a training sample set using the fifth real vehicle operation data, the required power, and the reward value.

[0052] In this embodiment of the application, the system encapsulates the fifth real vehicle operation data, the demand power sequence (the set of demand power at each collection time), and the reward value sequence (the set of reward values ​​at each collection time) obtained after the above preprocessing into a standardized offline experience sample file with a fixed suffix, which serves as the basic unit of the training sample set.

[0053] Step S104: Extract feature parameters from each group of sample data in the training sample set.

[0054] In this embodiment, the folder containing the standardized offline experience sample files is used as the input path. The system automatically recursively traverses each folder under the path, retrieves all folders that conform to the preset naming rules (e.g., *_S7_Final.mat), and completes the batch loading of multiple sets of samples (each folder corresponds to one set of samples), providing a unified data foundation for subsequent feature extraction.

[0055] Here, the feature parameters include two main dimensions: coverage features and value features. Coverage features characterize the completeness of the sample data's coverage of the vehicle's driving conditions, including vehicle speed coverage, power demand dispersion, and the effective engine operating percentage. Specifically, vehicle speed coverage assesses the sample's ability to cover different vehicle speed conditions; power demand dispersion assesses the sample's ability to cover different power load conditions; and the effective engine operating percentage assesses the sample's ability to cover engine intervention conditions. Value features characterize the contribution of the sample to the offline reinforcement learning model training, including driving intensity, battery SOC fluctuation range, and power mode switching frequency, represented by acceleration statistics. Driving intensity quantifies driving dynamics using the mean and quantiles of acceleration; battery SOC fluctuation range refers to the difference between the maximum and minimum battery SOC values ​​in the sample, quantifying the battery's charge and discharge boundary characteristics; and the number of power mode switching frequencies refers to the cumulative number of engine start and stop states, quantifying the frequency of power system switching. Through the above feature extraction, the original driving time-series signal is transformed into quantifiable and comparable multi-dimensional feature indicators, achieving standardized representation of the sample data.

[0056] Step S105: Generate an evaluation table for the training sample set based on the feature parameters corresponding to each group of samples.

[0057] Regarding step S105, as an example in specific implementation, it may include the following steps: Step S1051: Normalize the feature parameters corresponding to each group of sample data.

[0058] In this embodiment of the application, in order to eliminate the numerical differences between the feature parameters due to different dimensions, a minimum-maximum linear normalization process is performed on all extracted feature parameters, and each feature parameter is uniformly linearly mapped to the range of [0,1] to achieve the unification of dimensions.

[0059] Step S1052: Based on the feature parameters corresponding to each group of sample data after normalization, determine the comprehensive score of each group of samples.

[0060] In this embodiment, weight coefficients are pre-set for each feature parameter according to engineering application requirements, with higher weights assigned to parameters such as vehicle speed coverage, demand power dispersion, and driving intensity. A weighted sum is calculated for each normalized feature parameter and its corresponding weight coefficient to obtain a comprehensive score for each sample group. A higher comprehensive score indicates greater completeness of the working condition coverage and higher training value for that sample.

[0061] Step S1053: Sort all group samples based on the comprehensive score of all group samples.

[0062] In this embodiment, based on the comprehensive score calculated from all groups of samples, all samples are sorted in descending order, so that the sample with the highest score, i.e. the greatest training value, is placed at the front, and the sample with the lowest score is placed at the back. This can accurately extract real vehicle operation data that is of high value for training offline reinforcement learning models, thereby improving the adaptability and decision robustness of the energy management strategy under boundary conditions.

[0063] Step S1054: Based on all sorted group samples, generate an evaluation table for the training sample set.

[0064] In this embodiment, the evaluation table refers to a structured data table containing key information such as sample name, various characteristic parameters, and comprehensive score. Users can intuitively view the quantitative evaluation results of each sample through the table. The evaluation table is saved in CSV format to the root directory.

[0065] In this embodiment of the application, the processing method further includes: displaying the feature parameters corresponding to each group of sample data on a visualization dashboard.

[0066] Here, a multi-sample comparison visualization dashboard is automatically generated, such as... Figure 3 The diagram shows a comparison of multiple samples across six features. The visualization dashboard displays comparisons of vehicle speed distribution, power demand distribution, engine speed distribution, driving intensity distribution, battery SOC fluctuation, and power mode switching frequency. The dashboard visually demonstrates the differences in operating conditions between different samples, providing an intuitive basis for manual verification of sample quality.

[0067] The beneficial effects of the technical solution of this application are as follows: (1) High engineering efficiency: By integrating the processing logic of data cleaning, signal alignment, splicing and encapsulation, and designing a graphical control interface, the threshold for use is significantly reduced, the speed of batch data preprocessing is greatly improved, and the time and manpower cost of manual data cleaning are effectively reduced. (2) Good data quality: By introducing the vehicle longitudinal dynamics formula constraint and the linear interpolation alignment algorithm based on the global time base, the logical conflicts such as sensor numerical fluctuations and multi-source signal time sequence misalignment that are common in the data collected from real vehicles are solved, providing a high-quality, time-synchronized training dataset for subsequent offline reinforcement learning model training. (3) Fast model training: By constructing a dual-dimensional statistical feature system of coverage and value, and performing normalized weighted scoring and sorting screening on the cleaned samples, it is possible to automatically identify and remove low-value samples with low energy management strategy participation depth and low working condition information content, thereby reducing the number of convergence rounds required for model training and improving training efficiency.

[0068] This application provides a data processing method that can obtain high-quality, high-value standardized training samples, thereby accelerating the convergence of offline reinforcement learning models and improving their robustness in energy management decisions under complex operating conditions.

[0069] Based on the same application concept, this application also provides a data processing device corresponding to the data processing method provided in the above embodiments. Since the principle of the device in this application to solve the problem is similar to the data processing method in the above embodiments of this application, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be described again.

[0070] Please see Figure 4 , Figure 4 This is a schematic diagram of the structure of a data processing device provided in an embodiment of this application.

[0071] like Figure 4 As shown in the illustration, the data processing apparatus 410 provided in this application embodiment includes: The acquisition module 411 is used to acquire the root directory of the original real vehicle operation data through the graphical control interface; Traversal module 412 is used to traverse each file in the root directory to obtain the original real vehicle operation data; Preprocessing module 413 is used to preprocess the original real vehicle operation data to obtain a training sample set; Extraction module 414 is used to extract feature parameters from each group of sample data in the training sample set; The generation module 415 is used to generate an evaluation table for the training sample set based on the feature parameters corresponding to each group of samples.

[0072] Furthermore, the preprocessing module 413 is specifically used for: The negative values ​​in the engine torque signal in the original real vehicle operation data are corrected to zero, and the corrected original real vehicle operation data is determined as the first real vehicle operation data. Based on the preset signal quality priority rules, the single source signal with the highest confidence is selected from the multi-source signals of the first real vehicle operation data, and the selected first real vehicle operation data is determined as the second real vehicle operation data. Each signal in the second real vehicle operation data is interpolated according to a preset sampling frequency, and the interpolated second real vehicle operation data is determined as the third real vehicle operation data; Multiple third-vehicle operation data segments contained in each file are spliced ​​together in chronological order of collection time, and the spliced ​​third-vehicle operation data is determined as the fourth-vehicle operation data. The third real vehicle operation data of the sub-files in each file are spliced ​​together according to the collection time order, and the spliced ​​third real vehicle operation data is determined as the fourth real vehicle operation data; Based on the fourth real-vehicle operation data, a training sample set is constructed.

[0073] Furthermore, when the preprocessing module 413 is used to construct a training sample set based on the fourth real-vehicle operation data, it is also specifically used for: Based on the vehicle speed, engine torque, and engine speed at each data collection moment in the fourth real vehicle operation data, the required power of the real vehicle at each data collection moment is determined. Based on the fuel consumption, SOC offset, and impact penalty at engine start-up moment of the actual vehicle at each data collection time, the reward value of the actual vehicle at each data collection time is determined. The clutch status signal in the fourth vehicle operation data is mapped to a binary value, and the mapped fourth vehicle operation data is determined as the fifth vehicle operation data. The fifth real-vehicle operation data, the required power, the fuel consumption, and the reward value are used to construct a training sample set.

[0074] Furthermore, the generation module 415 is specifically used for: The feature parameters corresponding to each group of sample data are normalized respectively; Based on the feature parameters corresponding to each group of sample data after normalization, the comprehensive score of each group of samples is determined. Based on the comprehensive score of all group samples, all group samples are sorted. An evaluation table for the training sample set is generated based on all sorted group samples.

[0075] Furthermore, the processing device 410 also includes: The display module is used to show the feature parameters corresponding to each group of sample data on a visualization dashboard.

[0076] This application provides a data processing apparatus that can obtain high-quality, high-value standardized training samples, thereby accelerating the convergence of offline reinforcement learning models and improving their robustness in energy management decisions under complex operating conditions.

[0077] Please see Figure 5 , Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0078] like Figure 5 As shown, the electronic device 500 includes a processor 510, a memory 520, and a bus 530.

[0079] The memory 520 stores machine-readable instructions executable by the processor 510. When the electronic device 500 is running, the processor 510 and the memory 520 communicate via the bus 530. When the machine-readable instructions are executed by the processor 510, they can perform the operations described above. Figure 1 The steps of the data processing method in the method embodiment shown are described in detail in the method embodiment, and will not be repeated here.

[0080] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, can perform the above-described actions. Figure 1 The steps of the data processing method in the method embodiment shown are described in detail in the method embodiment, and will not be repeated here.

[0081] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and devices described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. In the several embodiments provided in this application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division; in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Another point is that the displayed or discussed mutual coupling or direct coupling or communication connection may be through some communication interfaces; the indirect coupling or communication connection of devices or units may be electrical, mechanical, or other forms.

[0082] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0083] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0084] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0085] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A data processing method, characterized in that, The processing method includes: Obtain the root directory of the original real vehicle operation data through the graphical control interface; Traverse each file in the root directory to obtain the original actual vehicle operation data; The raw real-vehicle operation data is preprocessed to obtain a training sample set; Extract feature parameters from each set of sample data in the training sample set; An evaluation table for the training sample set is generated based on the feature parameters corresponding to each group of samples.

2. The data processing method according to claim 1, characterized in that, The preprocessing of the original real-vehicle operation data to obtain the training sample set includes: The negative values ​​in the engine torque signal in the original real vehicle operation data are corrected to zero, and the corrected original real vehicle operation data is determined as the first real vehicle operation data. Based on the preset signal quality priority rules, the single source signal with the highest confidence is selected from the multi-source signals of the first real vehicle operation data, and the selected first real vehicle operation data is determined as the second real vehicle operation data. Each signal in the second real vehicle operation data is interpolated according to a preset sampling frequency, and the interpolated second real vehicle operation data is determined as the third real vehicle operation data; Multiple third-vehicle operation data segments contained in each file are spliced ​​together in chronological order of collection time, and the spliced ​​third-vehicle operation data is determined as the fourth-vehicle operation data. The third real vehicle operation data of the sub-files in each file are spliced ​​together according to the collection time order, and the spliced ​​third real vehicle operation data is determined as the fourth real vehicle operation data; Based on the fourth real-vehicle operation data, a training sample set is constructed.

3. The data processing method according to claim 2, characterized in that, The construction of the training sample set based on the fourth real-vehicle operation data includes: Based on the vehicle speed, engine torque, and engine speed at each data collection moment in the fourth real vehicle operation data, the required power of the real vehicle at each data collection moment is determined. Based on the fuel consumption, SOC offset, and impact penalty at engine start-up moment of the actual vehicle at each data collection time, the reward value of the actual vehicle at each data collection time is determined. The clutch status signal in the fourth vehicle operation data is mapped to a binary value, and the mapped fourth vehicle operation data is determined as the fifth vehicle operation data. The fifth real-vehicle operation data, the required power, the fuel consumption, and the reward value are used to construct a training sample set.

4. The data processing method according to claim 1, characterized in that, The evaluation table for generating the training sample set based on the feature parameters corresponding to each group of samples includes: The feature parameters corresponding to each group of sample data are normalized respectively; Based on the feature parameters corresponding to each group of sample data after normalization, the comprehensive score of each group of samples is determined. Based on the comprehensive score of all group samples, all group samples are sorted. An evaluation table for the training sample set is generated based on all sorted group samples.

5. The data processing method according to claim 1, characterized in that, The processing method further includes: The feature parameters corresponding to each group of sample data are displayed on a visualization dashboard.

6. A data processing apparatus, characterized in that, The processing device includes: The acquisition module is used to obtain the root directory of the original real vehicle operation data through a graphical control interface; The traversal module is used to traverse each file in the root directory to obtain the original real vehicle operation data; The preprocessing module is used to preprocess the raw real vehicle operation data to obtain a training sample set; The extraction module is used to extract feature parameters from each group of sample data in the training sample set; The generation module is used to generate an evaluation table for the training sample set based on the feature parameters corresponding to each group of samples.

7. The data processing apparatus according to claim 6, characterized in that, The preprocessing module is specifically used for: The negative values ​​in the engine torque signal in the original real vehicle operation data are corrected to zero, and the corrected original real vehicle operation data is determined as the first real vehicle operation data. Based on the preset signal quality priority rules, the single source signal with the highest confidence is selected from the multi-source signals of the first real vehicle operation data, and the selected first real vehicle operation data is determined as the second real vehicle operation data. Each signal in the second real vehicle operation data is interpolated according to a preset sampling frequency, and the interpolated second real vehicle operation data is determined as the third real vehicle operation data; Multiple third-vehicle operation data segments contained in each file are spliced ​​together in chronological order of collection time, and the spliced ​​third-vehicle operation data is determined as the fourth-vehicle operation data. The third real vehicle operation data of the sub-files in each file are spliced ​​together according to the collection time order, and the spliced ​​third real vehicle operation data is determined as the fourth real vehicle operation data; Based on the fourth real-vehicle operation data, a training sample set is constructed.

8. The data processing apparatus according to claim 7, characterized in that, When the preprocessing module is used to construct a training sample set based on the fourth real-vehicle operation data, it is also specifically used for: Based on the vehicle speed, engine torque, and engine speed at each data collection moment in the fourth real vehicle operation data, the required power of the real vehicle at each data collection moment is determined. Based on the fuel consumption, SOC offset, and impact penalty at engine start-up moment of the actual vehicle at each data collection time, the reward value of the actual vehicle at each data collection time is determined. The clutch status signal in the fourth vehicle operation data is mapped to a binary value, and the mapped fourth vehicle operation data is determined as the fifth vehicle operation data. The fifth real-vehicle operation data, the required power, the fuel consumption, and the reward value are used to construct a training sample set.

9. An electronic device, characterized in that, include: The device includes a processor, a memory, and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor communicates with the memory via the bus, and the machine-readable instructions are executed by the processor to perform the steps of the data processing method as described in any one of claims 1 to 5.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, performs the steps of the data processing method as described in any one of claims 1 to 5.

Images