Vehicle-mounted data recompression storage method and device based on timetable learning

Abstract

The invention relates to a vehicle-mounted data recompression storage method and device based on timetable learning, and the method comprises the steps: predicting a currently applicable train working time template through a preset period model in combination with current time information when a train is awakened; during train operation, generating a low-compression-rate log, and extracting a key operation state set at the same time; after the train enters the dormant state, a CUSUM algorithm is combined with a DTW algorithm, the matching degree of the actual train operation mode and the predicted working time template in the past time is compared based on the key operation state set, and if the matching degree is larger than a preset template matching threshold value, the train is in the dormant state; if yes, estimating the subsequent sleep duration based on the working time template, selecting whether to use a recompression mode to decompress the low-compression-rate log and then recompress the low-compression-rate log or not based on an estimation result, and compressing the low-compression-rate log, otherwise, maintaining a log recording mode. Compared with the prior art, the method solves the problem that a traditional method cannot dynamically evaluate resource availability.

Description

Technical Field

[0001] This invention relates to the field of rail transit, and in particular to a method and device for recompressing and storing onboard data based on timetable learning. Background Technology

[0002] With the rapid development of the rail transit industry, ensuring operational efficiency and safety is of paramount importance. The onboard data logging unit (DLU) is primarily responsible for recording and storing key information data from the onboard control equipment of subway trains, such as position and speed, in real time. This data is used to provide reliable support for accurate diagnosis and problem solving in the event of a malfunction in the vehicle system.

[0003] DLU (Data Log Execution Unit) is an embedded device with high reliability requirements. It operates in a complex environment, needs to work continuously for extended periods, and is limited by size and weight, resulting in low processor performance and limited memory space. To ensure the real-time recording and processing of large amounts of log data during vehicle operation and to prevent loss or incompleteness, existing DLUs employ algorithms such as FastLZ and Deflate. These algorithms use smaller time windows and dictionary sizes, sacrificing compression ratio to reduce memory and CPU time consumption, and complete data compression, encoding, and storage operations in a short time.

[0004] In terms of data compression algorithms, algorithms such as LZMA and LZW achieve better compression ratios through complex matching and entropy coding techniques. However, compression and decompression consume a lot of processor computing resources. The DLU processor has limited performance, and real-time use will cause excessive system load, affecting other functions and even leading to data loss.

[0005] Chinese patent CN113190519B discloses a method for compressing train operation logs based on railway operation scenario templates. After the current operation scenario ends, the method constructs the ID of the current train operation log and the IDs of each node. Following a pre-defined operation scenario template, it integrates the log ID, the category of the current operation scenario, and the structure information of all nodes to obtain the current train operation log. Then, based on pre-set settings, the current train operation log is either compressed or uncompressed before being persisted. While this method can adaptively compress different railway operation scenarios, it fails to effectively distinguish the differences in resource states between train operation and sleep periods. The compression strategy lacks time-segmented adaptation, failing to balance the core contradiction between real-time performance and storage efficiency. Furthermore, if the terminal experiences a sudden service interruption during the scene log compression process, the completed compression progress will be lost. The next compression attempt will require rereading the original log to perform the entire compression process, wasting computing power and potentially leading to data redundancy or loss due to the original log being overwritten.

[0006] Overall, existing algorithms either have low compression rates, failing to address the storage space issue of DLU, or high computational complexity, making them unsuitable for real-time DLU execution. Currently, there is a need for a time-based learning method that can balance the core conflict between real-time performance and storage efficiency. Summary of the Invention

[0007] The purpose of this invention is to overcome the shortcomings of the prior art by providing a method and device for recompressing and storing vehicle data based on schedule learning. This method can intelligently predict whether a hibernation window can be used to perform high-load compression tasks, thus solving the problem that traditional methods cannot dynamically assess resource availability.

[0008] The objective of this invention can be achieved through the following technical solutions: A method for recompressing and storing in-vehicle data based on timetable learning, the method comprising: When the train is woken up, the current applicable train working time template is predicted by using a preset periodic model combined with the current time information. During train operation, a low-compression log is generated using a real-time log recording mode. At the same time, a log data frame is sampled and collected at preset time intervals, and key operating status fields are extracted and temporarily stored to obtain a set of key operating statuses. After the train enters a hibernation state, the CUSUM algorithm combined with the DTW algorithm is used to compare the matching degree between the actual train operation mode in the past time and the predicted working time template based on the key operating state set. If the matching degree is greater than the preset template matching threshold, the subsequent hibernation duration is estimated based on the working time template, and the decision is made on whether to use the recompression mode to decompress the low compression rate log and then recompress it based on the estimation result; otherwise, the log recording mode is maintained. In the recompression mode, if compression is interrupted, a breakpoint flag is written to the header of the compressed file to record the current compression progress, and the compression task continues from the breakpoint the next time the recompression mode is entered.

[0009] Furthermore, the pre-setting process of the periodic model includes: Collect multi-dimensional operational data accumulated from similar historical trains, generate a typical working time template based on the multi-dimensional operational data, and initialize the working time template configuration; Regularly collect onboard log data from similar trains, and clean and label the onboard log data based on instructions, operation plans and operational events issued by the ATS; Based on the vehicle log data, offline construction and optimization of multi-scenario working time templates are carried out, and periodic feature parameters of each working time template are extracted. The multi-scenario working time templates include weekday working time templates, weekend working time templates, holiday working time templates, and maintenance working time templates. Based on the periodic feature parameters, model parameters are set, including periodic priority weights, template matching thresholds, and periodic offset correction parameters. The model parameters, multi-scenario working time templates, and periodic feature parameters are encapsulated into a periodic model file that can be executed by the vehicle-mounted data recording unit, thus completing the preset of the periodic model.

[0010] Furthermore, the process of generating the typical work time template includes: Collect multi-dimensional operational data accumulated from similar historical trains and clean the data. The multi-dimensional operational data includes train onboard log data, ATS-issued operation plans / sleep wake-up commands, and actual operational event records. The K-means clustering algorithm is used to perform offline analysis on the cleaned multi-dimensional operational data, classify the scenarios according to the similarity of operational behaviors, and generate corresponding work time templates for each scenario.

[0011] Furthermore, the extraction process of the periodic feature parameters of each working time template includes: Fourier transform is used to perform frequency domain analysis on the multi-dimensional operational data corresponding to each working time template, identify the dominant frequency components in each working time template, and determine the boundary conditions of each working time template cycle. By combining the dominant frequency components and boundary conditions, periodic characteristic parameters are obtained.

[0012] Furthermore, the key operating status fields include train position, speed, and equipment operating status.

[0013] Furthermore, when the matching degree is greater than a preset threshold, the subsequent sleep duration is estimated again by the periodic prediction model after a preset delay time, and the recompression mode is selected based on the estimation result to decompress the low compression rate log and then recompress it.

[0014] Furthermore, the compression process of the recompression mode includes: Read the low-compression log in chronological order and perform a bitwise XOR operation on the same fields in adjacent log data frames; The LZMA high compression ratio algorithm is used to recompress the data after the bitwise XOR operation to generate and store high compression ratio logs. Delete the original low-compression log data corresponding to the high-compression log.

[0015] Furthermore, the process of estimating the subsequent sleep duration based on the working time template and selecting whether to use the recompression mode to decompress and recompress the low compression rate log based on the estimation result includes: Historical sleep duration data is retrieved from the aforementioned working time template as the estimated baseline value; Read the actual hibernation start time and the actual operating deviation of the day, correct the estimated base value, and obtain the final estimated hibernation duration; If the final estimated sleep duration is greater than the minimum effective compression duration threshold, then the recompression mode is started to compress the log; otherwise, the log recording mode is maintained. The minimum effective compression time threshold is the shortest time required to complete the recompression of a single low-compression log.

[0016] Furthermore, the process of comparing the matching degree between the actual train operation mode and the predicted working time template over a past period based on the key operating state set includes: The train operation data in the key operating status set and the predicted working time template are standardized and preprocessed to obtain the actual operation dataset and the template dataset. The deviation between the values ​​in the actual running dataset and the template dataset is calculated based on the CUSUM algorithm, and the total deviation is accumulated to obtain the overall deviation index. The overall deviation index is compared with a preset deviation threshold. If the overall deviation index is greater than the preset deviation threshold, the deviation trend is determined to be uncontrollable, the matching fails, and the log recording mode is maintained; otherwise, the deviation trend is determined to be controllable, and the comparison continues. The DTW algorithm is used to calculate the time series similarity between the actual running dataset and the template dataset, and the result is used as the matching score output.

[0017] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the time-based learning-based vehicle data recompression storage method as described above.

[0018] Compared with the prior art, the beneficial effects of the present invention include: 1. This invention performs low-compression logging during train operation to ensure real-time recording and no loss of critical data. During train hibernation, the data is recompressed using XOR differential encoding and LZMA high compression ratio algorithm, balancing the real-time nature of onboard data recording with long-term storage efficiency, significantly reducing long-term storage occupation, extending the lifespan of the onboard data recording unit's storage medium, and reducing data cleanup frequency. The breakpoint continuation compression mechanism of this invention adds a breakpoint flag segment to the header of incomplete high-pressure log files, recording the offset and timestamp of the data that has been compressed, ensuring that compression tasks across multiple hibernation cycles can be executed continuously, avoiding the waste of computing power and data redundancy caused by repeated compression. At the same time, the CUSUM and DTW dual-algorithm matching and verification ensures that recompression is only initiated when the operation mode is stable and the hibernation time is sufficient, avoiding invalid compression occupying hibernation resources.

[0019] 2. In this invention, the subsequent sleep duration is estimated by a periodic prediction model after a preset delay time, and the recompression mode is selected based on the prediction results to decompress the low compression rate log and then recompress it. This ensures that the train enters a stable sleep state completely and avoids the equipment status fluctuations in the early stage of sleep affecting the compression process.

[0020] 3. This invention uses K-means clustering to mine operational scenario patterns in historical data and generate standardized working time templates; it also uses Fourier transform to extract periodic features and constructs a periodic prediction model to achieve prediction based on patterns. Attached Figure Description

[0021] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a schematic diagram of the overall compressed storage system in this invention; Figure 3 This is a flowchart of the working time template learning and generation process in this invention; Figure 4 This is a flowchart of the log recompression process in this invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0023] Example 1 This embodiment discloses a method for recompressing and storing in-vehicle data based on timetable learning. This method is based on the in-vehicle data log unit (DLU), and the method is as follows: Figure 1 As shown, the specific steps are as follows: Step S1: When the train is woken up, the applicable train working time template is predicted by using a preset cycle model combined with the current time information.

[0024] The pre-set process for the periodic model includes: Collect multi-dimensional operational data accumulated from similar historical trains, generate typical working time templates based on the multi-dimensional operational data, and initialize the working time template configuration; Regularly collect onboard log data from similar trains, and clean and label the onboard log data based on instructions, operating plans and operational events issued by ATS (Automatic Train Monitoring System); Based on vehicle log data, offline construction and optimization of multi-scenario working time templates were carried out, and periodic feature parameters of each working time template were extracted. The multi-scenario working time templates include weekday working time templates, weekend working time templates, holiday working time templates, and maintenance working time templates. Based on the periodic feature parameters, model parameters are set, including periodic priority weights, template matching thresholds, and periodic offset correction parameters. The model parameters, multi-scenario working time templates, and periodic feature parameters are encapsulated into a periodic model file that can be executed by the vehicle-mounted data recording unit, thus completing the preset of the periodic model.

[0025] The process of generating a typical work time template includes: Collect and clean multi-dimensional operational data accumulated from similar historical trains. The multi-dimensional operational data includes train onboard log data, ATS-issued operation plans / sleep wake-up commands, and actual operational event records. The K-means clustering algorithm is used to perform offline analysis on the cleaned multi-dimensional operational data, classify the scenarios according to the similarity of operational behaviors, and generate corresponding work time templates for each scenario.

[0026] The process of extracting the periodic feature parameters of each working time template includes: Fourier transform is used to perform frequency domain analysis on the multi-dimensional operational data corresponding to each working time template, identify the dominant frequency components in each working time template, and determine the boundary conditions of each working time template cycle. By combining the dominant frequency components and boundary conditions, periodic characteristic parameters are obtained.

[0027] Step S2: During train operation, a low-compression log is generated using a real-time log recording mode. At the same time, a log data frame is sampled and collected at preset time intervals, and key operating status fields are extracted and temporarily stored to obtain a set of key operating statuses.

[0028] Key operational status fields include train position, speed, and equipment operating status.

[0029] Step S3: After the train enters the hibernation state, the CUSUM algorithm combined with the DTW algorithm is used to compare the matching degree between the actual train operation mode and the predicted working time template in the past time based on the key operating state set. If the matching degree is greater than the preset template matching threshold, the subsequent hibernation duration is estimated based on the working time template. Based on the estimation result, it is selected whether to use the recompression mode to decompress the low compression rate log and then recompress it. Otherwise, the log recording mode is maintained.

[0030] The recompression mode is a background task that starts during train hibernation and when the system load is extremely low. It utilizes idle computing resources to execute the highly complex LZMA compression algorithm to improve overall storage efficiency.

[0031] If compression is interrupted during the recompression mode, a breakpoint marker is written to the header of the compressed file to record the current compression progress. The compression task will resume from the breakpoint the next time the recompression mode is entered.

[0032] When the matching degree is greater than a preset threshold, the subsequent sleep duration is estimated again by the periodic prediction model after a preset delay time, and the recompression mode is selected based on the prediction result to decompress the low compression rate log and then recompress it.

[0033] Template matching judgment uses the CUSUM (Cumulative Sum Control Chart) algorithm to detect the trend of running deviation, and combines it with the DTW (Dynamic Time Warping) algorithm to calculate the time series similarity between the current running trajectory and the template.

[0034] The process of comparing the actual train operation mode with the predicted working time template over a past period based on the key operating state set includes: Standardize and preprocess the train operation data in the key operating status set and the predicted working time template to obtain the actual operation dataset and the template dataset; The deviation between the values ​​in the actual running dataset and the template dataset is calculated based on the CUSUM algorithm, and the total deviation is accumulated to obtain the overall deviation index. Compare the overall deviation index with the preset deviation threshold. If the overall deviation index is greater than the preset deviation threshold, the deviation trend is determined to be uncontrollable, the matching fails, and the log recording mode is maintained; otherwise, the deviation trend is determined to be controllable, and the comparison continues. The DTW algorithm is used to calculate the time series similarity between the actual running dataset and the template dataset, and the result is used as the matching score output.

[0035] The compression process in recompression mode includes: Read low-compression logs in chronological order and perform bitwise XOR operations on the same fields in adjacent log data frames. The LZMA high compression ratio algorithm is used to recompress the data after the bitwise XOR operation to generate and store high compression ratio logs. Delete the original low-compression log data corresponding to the high-compression log.

[0036] The process of estimating subsequent sleep duration based on a working time template and selecting whether to use a recompression mode to decompress and recompress the low-compression logs based on the estimation results includes: Historical sleep duration data is retrieved from the working time template as the estimated base value; Read the actual hibernation start time and the actual operating deviation of the day, correct the estimated base value, and obtain the final estimated hibernation duration; If the estimated sleep duration is greater than the minimum effective compression duration threshold, then the recompression mode will be started to compress the log; otherwise, the logging mode will be maintained. The minimum effective compression time threshold is the shortest time required to complete the recompression of a single low-compression log.

[0037] Example 2 This embodiment, based on Embodiment 1 above, discloses a specific implementation example of a timetable-based method for recompressing and storing vehicle data.

[0038] This embodiment is based on, as follows Figure 2 The image shows a heavily compressed storage system.

[0039] Figure 2 The diagram shows the main components of a heavy-duty compression storage system, including trackside equipment fixed at ground stations and vehicle-mounted equipment that moves with the vehicle. The two types of equipment communicate via a wireless network.

[0040] The Automatic Train Supervision (ATS) system is responsible for monitoring and managing the train operation status of the entire line, including information such as train position, speed, and operation plan. It sends sleep / wake commands to the vehicles when they are in sleep or wake-up mode.

[0041] The Data Storage Unit (DSU) is part of the ground-based data center. The DSU periodically downloads data log files from the onboard data logging unit to support offline learning of the schedule. After data processing is complete, the DSU can also distribute configurations to the onboard data logging unit.

[0042] The Carborne Controller (CC) is the core control unit of the train, which performs automatic operation control functions, including acceleration, deceleration, and stopping, and maintains communication with the ATS to receive instructions and send train status information.

[0043] The Depot Wake-up Unit (DWU) is primarily responsible for managing the train's sleep and wake-up times in the depot or at specific stations. It ensures the train can safely sleep during non-operational hours and accurately wake up before operation. After sleep mode, the DWU continues to operate via the onboard battery. The onboard data recording unit described in this invention is a component of the DWU. It is responsible for recording onboard logs during train operation and recompressing the data after the train goes into sleep mode.

[0044] The specific implementation process of the timetable-based learning-based in-vehicle data recompression and storage method is as follows: like Figure 3 As shown, the entire line's historical vehicle logs are downloaded through the data storage unit to obtain relevant information such as ATS commands, wake-up / sleep times, and operating plans. After manual annotation of information such as weekdays, holidays, and local special events, K-means clustering analysis is performed on the data, outputting various timetable template classifications and boundary thresholds for each type. Fourier transform is used to convert the daily sleep / wake-up time series to the frequency domain, and dominant frequencies, such as 7-day cycles and 30-day cycles, are extracted based on actual conditions. Finally, the timetable templates, matching thresholds, and cycle parameters are combined to form configuration data, encapsulated into a cycle model, and distributed to the onboard data recording units of each train.

[0045] Each time the train is woken up, the onboard data recording unit predicts the applicable working time template for the day based on the periodic model in the configuration. During train operation, the onboard data recording unit generates low-pressure logs using the conventional real-time log recording mode, samples and collects a log data frame every 10 minutes, extracts key operating status fields and temporarily stores them in the memory buffer.

[0046] like Figure 4 As shown, when the onboard data recording unit detects a sleep state, it determines the process based on the sampled data and the CUSUM p-value and DTW value of the predicted working time template. If the data matches the working time template and the remaining sleep time is sufficient, log recompression begins after 10 minutes. During recompression, the onboard data recording unit decompresses low-pressure logs generated by algorithms such as FastLZ and Deflate, performs inter-frame XOR operations, performs LZMA high-level compression, and writes the logs to a temporary file. Once a low-pressure log is completely written to the high-pressure log, it is deleted. If train wake-up is detected, the recompression process is immediately stopped, and a breakpoint marker segment is inserted into the temporary file to record the current compression progress.

[0047] The next time the conditions are met to enter recompression mode, the compression task will continue from the breakpoint until all logs have completed the high compression ratio conversion or are interrupted again; if the compression conditions are not met, the normal log recording mode will be maintained and the recompression process will not be started.

[0048] This method enables the vehicle-mounted data recording unit to intelligently predict whether a hibernation window can be used to perform high-load compression tasks, solving the problem that traditional methods cannot dynamically assess resource availability.

[0049] Example 3 Based on Embodiment 1, this embodiment provides an electronic device, including: one or more processors and a memory, wherein the memory stores one or more programs, the one or more programs including instructions for executing the aforementioned on-board data recompression and storage method based on schedule learning.

[0050] At the hardware level, the electronic device includes a processor, internal bus, network interface, memory, and non-volatile memory, and may also include other hardware required for business operations. The processor reads the corresponding computer program from the non-volatile memory into memory and then runs it to implement the above-mentioned time-based learning-based vehicle data recompression and storage method. Of course, in addition to the software implementation, this invention does not exclude other implementation methods, such as logic devices or a combination of hardware and software, etc. That is to say, the execution subject of the following processing flow is not limited to individual logic units, but can also be hardware or logic devices.

[0051] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0052] Computer-readable media include both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0053] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for recompressing and storing in-vehicle data based on schedule learning, characterized in that, The method includes: When the train is woken up, the current applicable train working time template is predicted by using a preset periodic model combined with the current time information. During train operation, a low-compression log is generated using a real-time log recording mode. At the same time, a log data frame is sampled and collected at preset time intervals, and key operating status fields are extracted and temporarily stored to obtain a set of key operating statuses. After the train enters a hibernation state, the CUSUM algorithm combined with the DTW algorithm is used to compare the matching degree between the actual train operation mode in the past time and the predicted working time template based on the key operating state set. If the matching degree is greater than the preset template matching threshold, the subsequent hibernation duration is estimated based on the working time template, and the decision is made on whether to use the recompression mode to decompress the low compression rate log and then recompress it based on the estimation result; otherwise, the log recording mode is maintained. In the recompression mode, if compression is interrupted, a breakpoint flag is written to the header of the compressed file to record the current compression progress, and the compression task continues from the breakpoint the next time the recompression mode is entered.

2. The method for recompressing and storing vehicle data based on timetable learning according to claim 1, characterized in that, The pre-setting process of the periodic model includes: Collect multi-dimensional operational data accumulated from similar historical trains, generate a typical working time template based on the multi-dimensional operational data, and initialize the working time template configuration; Regularly collect onboard log data from similar trains, and clean and label the onboard log data based on instructions, operation plans and operational events issued by the ATS; Based on the vehicle log data, offline construction and optimization of multi-scenario working time templates are carried out, and periodic feature parameters of each working time template are extracted. The multi-scenario working time templates include weekday working time templates, weekend working time templates, holiday working time templates, and maintenance working time templates. Based on the periodic feature parameters, model parameters are set, including periodic priority weights, template matching thresholds, and periodic offset correction parameters. The model parameters, multi-scenario working time templates, and periodic feature parameters are encapsulated into a periodic model file that can be executed by the vehicle-mounted data recording unit, thus completing the preset of the periodic model.

3. The method for recompressing and storing vehicle data based on timetable learning according to claim 2, characterized in that, The process of generating the typical work time template includes: Collect multi-dimensional operational data accumulated from similar historical trains and clean the data. The multi-dimensional operational data includes train onboard log data, ATS-issued operation plans / sleep wake-up commands, and actual operational event records. The K-means clustering algorithm is used to perform offline analysis on the cleaned multi-dimensional operational data, classify the scenarios according to the similarity of operational behaviors, and generate corresponding work time templates for each scenario.

4. The method for recompressing and storing vehicle data based on timetable learning according to claim 2, characterized in that, The extraction process of the periodic feature parameters of each working time template includes: Fourier transform is used to perform frequency domain analysis on the multi-dimensional operational data corresponding to each working time template, identify the dominant frequency components in each working time template, and determine the boundary conditions of each working time template period. By combining the dominant frequency components and boundary conditions, periodic characteristic parameters are obtained.

5. The method for recompressing and storing vehicle data based on timetable learning according to claim 1, characterized in that, The key operational status fields include train position, speed, and equipment operating status.

6. The method for recompressing and storing vehicle data based on timetable learning according to claim 1, characterized in that, When the matching degree is greater than a preset threshold, the subsequent sleep duration is estimated again by the periodic prediction model after a preset delay time, and the recompression mode is selected based on the prediction result to decompress the low compression rate log and then recompress it.

7. The method for recompressing and storing vehicle data based on timetable learning according to claim 1, characterized in that, The compression process of the recompression mode includes: Read the low-compression log in chronological order and perform a bitwise XOR operation on the same fields in adjacent log data frames; The LZMA high compression ratio algorithm is used to recompress the data after the bitwise XOR operation to generate and store high compression ratio logs. Delete the original low-compression log data corresponding to the high-compression log.

8. The method for recompressing and storing vehicle data based on timetable learning according to claim 1, characterized in that, The process of estimating subsequent sleep duration based on the working time template and selecting whether to use recompression mode to decompress and recompress the low compression rate logs based on the estimation results includes: Historical sleep duration data is retrieved from the aforementioned working time template as the estimated baseline value; Read the actual hibernation start time and the actual operating deviation of the day, correct the estimated base value, and obtain the final estimated hibernation duration; If the final estimated sleep duration is greater than the minimum effective compression duration threshold, then the recompression mode is started to compress the log; otherwise, the log recording mode is maintained. The minimum effective compression time threshold is the shortest time required to complete the recompression of a single low-compression log.

9. The method for recompressing and storing vehicle data based on timetable learning according to claim 1, characterized in that, The process of comparing the actual train operation mode with the predicted working time template over a past period based on the key operating state set includes: The train operation data in the key operating status set and the predicted working time template are standardized and preprocessed to obtain the actual operation dataset and the template dataset. The deviation between the values ​​in the actual running dataset and the template dataset is calculated based on the CUSUM algorithm, and the total deviation is accumulated to obtain the overall deviation index. The overall deviation index is compared with a preset deviation threshold. If the overall deviation index is greater than the preset deviation threshold, the deviation trend is determined to be uncontrollable, the matching fails, and the log recording mode is maintained; otherwise, the deviation trend is determined to be controllable, and the comparison continues. The DTW algorithm is used to calculate the time series similarity between the actual running dataset and the template dataset, and the result is used as the matching score output.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the on-board data recompression and storage method based on time schedule learning as described in any one of claims 1-9.

Images