Power van monitoring control method and device, vehicle-mounted terminal and network equipment

Abstract

The invention provides a power van monitoring control method and device, a vehicle-mounted terminal and network equipment, and belongs to the technical field of vehicles special for power production. The method comprises the following steps: acquiring a sensor data set and a parameter validity index matrix; based on the sensor data set and the parameter validity index matrix, determining parameter validity values of each piece of sensor data to all tasks; determining effective sensor data of each task in the set task sequence based on the parameter validity values of all the sensor data, and executing task calculation based on the effective sensor data of each task in the set task sequence; performing individual learning based on the effective sensor data of each task to determine a correction factor and a control function of each task; and sending individual learning results of all tasks to the network equipment. According to the invention, the defect that the switching of the special electric power production vehicle cannot be safely, reliably, flexibly, efficiently and automatically controlled when the existing special electric power vehicle meets the functional requirements of various electric power production vehicles is overcome.

Description

Technical Field

[0001] The present invention relates to the technical field of special vehicles for power production, and particularly to a monitoring and control method for a power vehicle, a monitoring and control device for a power vehicle, an in-vehicle terminal, a network device, a machine-readable storage medium, and a computer program product. Background Art

[0002] As an emergency backup power supply, a power vehicle (or a special vehicle for power production) can provide power support in case of a main power failure or an emergency, ensuring the continuous operation of critical facilities and services. The existing special power vehicles have relatively single functions, and most of them are designed for specific scenarios, such as vehicle controllers, generator controllers, vehicle position monitoring and communication units, etc. They have poor scalability and do not meet the functional requirements of various power production vehicles.

[0003] That is, when the existing special power vehicles face the functional requirements of various power production vehicles, they cannot achieve safe, reliable, flexible, and efficient automatic control of the power production special vehicle for connection and disconnection. Summary of the Invention

[0004] The purpose of the embodiments of the present invention is to provide a monitoring and control method, device, in-vehicle terminal, and network device for a power vehicle, so as to solve the defect that when the existing special power vehicles face the functional requirements of various power production vehicles, they cannot achieve safe, reliable, flexible, and efficient automatic control of the power production special vehicle for connection and disconnection.

[0005] To achieve the above purpose, the embodiments of the present invention provide a monitoring and control method for a power vehicle, which is applied to an in-vehicle terminal of the power vehicle. The method includes:

[0006] Obtain a sensor data set and a parameter validity index matrix; the sensor data set includes sensor data of multiple sensors; the parameter validity index matrix includes parameter validity indexes of all sensor data; the parameter validity index characterizes the correlation degree between each sensor data and the corresponding task control accuracy;

[0007] Determine the parameter validity value of each sensor data for all tasks based on the sensor data set and the parameter validity index matrix;

[0008] Determine the valid sensor data of each task in the set task sequence based on the parameter validity values of all sensor data, and perform task calculation based on the valid sensor data of each task in the set task sequence;

[0009] Perform individual learning based on the valid sensor data of each task to determine the correction factor and control function of each task;

[0010] Send the individual learning results of all tasks to the network device; the individual learning results include the parameter effectiveness index matrix, the correction factor of each task, and the control function.

[0011] Optionally, the parameter effectiveness index of each sensor data in the parameter effectiveness index matrix is calculated through the following steps:

[0012] Determine the probability that the sensor data of the first sensor belongs to the first task;

[0013] Calculate the information entropy based on the probability;

[0014] Calculate the conditional information entropy corresponding to the sensor data of the first sensor based on the information entropy;

[0015] Calculate the information gain based on the conditional information entropy and the information entropy;

[0016] Perform gain normalization processing on the information gain to obtain the parameter effectiveness index of the sensor data of the first sensor;

[0017] Wherein, the sensor data of the first sensor is the sensor data of any one sensor in the sensor data set; the first task is any one task in the set task sequence.

[0018] Optionally, determining the effective sensor data of each task in the set task sequence based on the parameter effectiveness values of all sensor data includes:

[0019] Determine the sensor data with a parameter effectiveness value greater than or equal to the set threshold in each task as the effective sensor data.

[0020] Optionally, after determining the effective sensor data of each task in the set task sequence based on the parameter effectiveness values of all sensor data, and performing task calculations based on the effective sensor data of each task in the set task sequence, it further includes:

[0021] In the case of triggering the core task, set the processing priority of the core task to the highest priority among all tasks in the set task sequence for priority processing; the core task represents an emergency situation that occurs during the normal use of the power vehicle.

[0022] Optionally, performing individual learning based on the effective sensor data of each task to determine the correction factor and control function of each task includes:

[0023] Repeat the following steps until the correction factor of the first task reaches the set condition, and output the corrected factor and control function after iteration is completed:

[0024] Input the valid sensor data of the first task into the first task to obtain the output result of the first task;

[0025] Calculate the correction factor of the first task based on the output result of the first task;

[0026] Calculate the control function of the first task based on the valid sensor data of previous samplings of the first task; the control function is calculated based on a cubic spline interpolation function;

[0027] Obtain an updated sensor data set including the updated sensor data of the multiple sensors;

[0028] Determine the updated parameter validity value of each sensor data for all tasks based on the product of the updated sensor data set and the parameter validity index matrix;

[0029] Determine the updated valid sensor data of the first task based on the updated parameter validity values of all sensor data for iterative calculation of the correction factor and control function of the first task;

[0030] Wherein, the first task is any one task in the set task sequence.

[0031] Optionally, the calculating the correction factor of the first task based on the output result of the first task includes:

[0032] Determine the mean points of all positive class outputs and the mean points of all negative class outputs based on the output result of the first task;

[0033] Based on the mean points of all positive class outputs and the mean points of all negative class outputs, determine a first hyperplane passing through the mean points of all positive class outputs and a second hyperplane passing through the mean points of all negative class outputs;

[0034] Calculate the average value of the distances between all positive class outputs and the first hyperplane as the first average value, calculate the average value of the distances between all negative class outputs and the second hyperplane as the second average value, calculate the maximum value of the distances between all positive class outputs and the first hyperplane as the first maximum value, and calculate the maximum value of the distances between all negative class outputs and the second hyperplane as the second maximum value;

[0035] Calculate the correction factor of the first task based on the first average value, the second average value, the first maximum value and the second maximum value.

[0036] Optionally, the calculating the correction factor of the first task based on the first average value, the second average value, the first maximum value and the second maximum value is calculated by the following formula:

[0037]

[0038] Among them, ε represents the correction factor of the first task, l + represents the first average value, l - represents the second average value, L + represents the first maximum value, L - represents the second maximum value.

[0039] Optionally, calculating the control function of the first task based on the valid sensor data of previous samplings of the first task includes:

[0040] Calculating the mean and standard deviation of the valid sensor data of previous samplings based on the valid sensor data of previous samplings of the first task;

[0041] Calculating the weight coefficient of the control function based on the mean and the standard deviation;

[0042] Constructing a cubic spline interpolation function based on the difference between the valid sensor data of the first task in two adjacent samplings, the calculated value of the previous sampling of the control function of the first task, and the weight coefficient of the control function;

[0043] Solving the cubic spline interpolation function to obtain the control function of the first task.

[0044] On the other hand, an embodiment of the present invention further provides a monitoring and control method for a power supply vehicle, which is applied to a network device, and the method includes:

[0045] Receiving the individual learning results of all tasks sent by multiple vehicle-mounted terminals; the individual learning results include a parameter validity index matrix, a correction factor and a control function for each task; the parameter validity index matrix includes the parameter validity indexes of all sensor data obtained by the vehicle-mounted terminals; the parameter validity index characterizes the correlation degree between each sensor data and the control accuracy of the corresponding task;

[0046] Performing ensemble learning based on the individual learning results of all tasks of the multiple vehicle-mounted terminals to obtain an ensemble learning result;

[0047] Sending the ensemble learning result to all vehicle-mounted terminals so that each vehicle-mounted terminal replaces the individual learning result based on the ensemble learning result.

[0048] Optionally, performing ensemble learning based on the individual learning results of all tasks of the multiple vehicle-mounted terminals to obtain an ensemble learning result includes:

[0049] Perform feature splicing on the individual learning results of all tasks of multiple vehicle-mounted terminals to obtain aggregated features;

[0050] Perform feature extraction on the aggregated features through a feature extraction network to obtain the common features of the individual learning results of all tasks of multiple vehicle-mounted terminals as the ensemble learning results.

[0051] On the other hand, an embodiment of the present invention further provides a power vehicle monitoring and control device, including:

[0052] An acquisition module, configured to acquire a sensor data set and a parameter validity index matrix; the sensor data set includes sensor data of multiple sensors; the parameter validity index matrix includes parameter validity indexes of all sensor data; the parameter validity index characterizes the correlation degree between each sensor data and the corresponding task control accuracy;

[0053] A determination module, configured to determine the parameter validity value of each sensor data for all tasks based on the sensor data set and the parameter validity index matrix;

[0054] A classification execution module, configured to determine the valid sensor data of each task in a set task sequence based on the parameter validity values of all sensor data, and perform task calculations based on the valid sensor data of each task in the set task sequence;

[0055] An individual learning module, configured to perform individual learning based on the valid sensor data of each task to determine the correction factor and control function of each task;

[0056] A sending module, configured to send the individual learning results of all tasks to a network device; the individual learning results include the parameter validity index matrix, the correction factor and control function of each task.

[0057] Optionally, the parameter validity index of each sensor data in the parameter validity index matrix is calculated through the following steps:

[0058] Determine the probability that the sensor data of the first sensor belongs to the first task;

[0059] Calculate the information entropy based on the probability;

[0060] Calculate the conditional information entropy corresponding to the sensor data of the first sensor based on the information entropy;

[0061] Calculate the information gain based on the conditional information entropy and the information entropy;

[0062] Perform gain normalization processing on the information gain to obtain the parameter validity index of the sensor data of the first sensor;

[0063] Among them, the sensor data of the first sensor is the sensor data of any one sensor in the sensor data set; the first task is any one task in the set task sequence.

[0064] Optionally, determining the valid sensor data of each task in the set task sequence based on the parameter validity values of all sensor data includes:

[0065] Determining the sensor data with the parameter validity value greater than or equal to the set threshold in each task as the valid sensor data.

[0066] Optionally, the device further includes:

[0067] A priority processing module, configured to, when triggering a core task, set the processing priority of the core task to the highest priority among all tasks in the set task sequence for priority processing; the core task represents an emergency situation that occurs during the normal use of the power vehicle.

[0068] Optionally, performing individual learning based on the valid sensor data of each task to determine the correction factor and control function of each task includes:

[0069] Repeatedly execute the following steps until the correction factor of the first task reaches the set condition, and output the correction factor and control function after iteration:

[0070] Input the valid sensor data of the first task into the first task to obtain the output result of the first task;

[0071] Calculate the correction factor of the first task based on the output result of the first task;

[0072] Calculate the control function of the first task based on the valid sensor data of previous samplings of the first task; the control function is calculated based on a cubic spline interpolation function;

[0073] Obtain an updated sensor data set including the updated sensor data of the multiple sensors;

[0074] Determine the updated parameter validity values of each sensor data for all tasks based on the product of the updated sensor data set and the parameter validity index matrix;

[0075] Determine the updated valid sensor data of the first task based on the updated parameter validity values of all sensor data for iterative calculation of the correction factor and control function of the first task;

[0076] Among them, the first task is any one task in the set task sequence.

[0077] Optionally, calculating the correction factor of the first task based on the output result of the first task includes:

[0078] Determining the mean point of all positive-class outputs and the mean point of all negative-class outputs based on the output result of the first task;

[0079] Determining a first hyperplane passing through the mean point of all positive-class outputs and a second hyperplane passing through the mean point of all negative-class outputs based on the mean point of all positive-class outputs and the mean point of all negative-class outputs;

[0080] Calculating the average value of the distances between all positive-class outputs and the first hyperplane as a first average value, calculating the average value of the distances between all negative-class outputs and the second hyperplane as a second average value, calculating the maximum value of the distances between all positive-class outputs and the first hyperplane as a first maximum value, and calculating the maximum value of the distances between all negative-class outputs and the second hyperplane as a second maximum value;

[0081] Calculating the correction factor of the first task based on the first average value, the second average value, the first maximum value, and the second maximum value.

[0082] Optionally, calculating the correction factor of the first task based on the first average value, the second average value, the first maximum value, and the second maximum value is calculated through the following formula:

[0083]

[0084] where ε represents the correction factor of the first task, l + represents the first average value, l - represents the second average value, L + represents the first maximum value, L - represents the second maximum value.

[0085] Optionally, calculating the control function of the first task based on the effective sensor data of previous samplings of the first task includes:

[0086] Calculating the mean and standard deviation of the effective sensor data of previous samplings based on the effective sensor data of previous samplings of the first task;

[0087] Calculating the weight coefficient of the control function based on the mean and the standard deviation;

[0088] Constructing a cubic spline interpolation function based on the difference between the effective sensor data of the first task in two adjacent samplings, the calculated value of the previous sampling of the control function of the first task, and the weight coefficient of the control function.

[0089] Solve the cubic spline interpolation function to obtain the control function of the first task.

[0090] On the other hand, an embodiment of the present invention further provides a power vehicle monitoring and control device, including:

[0091] A receiving module, configured to receive the individual learning results of all tasks sent by multiple vehicle-mounted terminals; the individual learning results include a parameter validity index matrix, a correction factor for each task, and a control function; the parameter validity index matrix includes the parameter validity indexes of all sensor data acquired by the vehicle-mounted terminals; the parameter validity index characterizes the correlation degree between each sensor data and the control accuracy of the corresponding task;

[0092] An integrated learning module, configured to perform integrated learning based on the individual learning results of all tasks of the multiple vehicle-mounted terminals to obtain an integrated learning result;

[0093] A sending module, configured to send the integrated learning result to all vehicle-mounted terminals, so that each vehicle-mounted terminal replaces the individual learning result based on the integrated learning result.

[0094] Optionally, the performing integrated learning based on the individual learning results of all tasks of the multiple vehicle-mounted terminals to obtain an integrated learning result includes:

[0095] Perform feature stitching on the individual learning results of all tasks of multiple vehicle-mounted terminals to obtain aggregated features;

[0096] Extract features from the aggregated features through a feature extraction network, and obtain the common features of the individual learning results of all tasks of the multiple vehicle-mounted terminals as the integrated learning result.

[0097] On the other hand, an embodiment of the present invention further provides a vehicle-mounted terminal, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the above-mentioned power vehicle monitoring and control method is implemented.

[0098] On the other hand, an embodiment of the present invention further provides a network device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the above-mentioned power vehicle monitoring and control method is implemented.

[0099] On the other hand, the present invention further provides a machine-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the above-mentioned power vehicle monitoring and control method is implemented.

[0100] On the other hand, the present invention also provides a computer program product, including a computer program which, when executed by a processor, implements the above-mentioned power vehicle monitoring and control method.

[0101] Through the above technical solutions, the data classification method based on the parameter validity index matrix in the embodiments of the present invention quickly classifies and inputs the corresponding data; performs task calculations based on the valid sensor data of each task in the set task sequence to achieve task priority resource coordination and meet the response requirements of real-time tasks; and through the update of the correction factor and the control function, combined with the iteration of individual learning and ensemble learning, realizes end-to-end decision control. Therefore, the embodiments of the present invention realize intelligent control from data acquisition and classification, task real-time scheduling, and efficient end-to-end processing, and can automatically control the power-on and power-off of special vehicles for power production safely, reliably, flexibly and efficiently.

[0102] Other features and advantages of the embodiments of the present invention will be described in detail in the subsequent specific implementation part. BRIEF DESCRIPTION OF THE DRAWINGS

[0103] The drawings are used to provide a further understanding of the embodiments of the present invention, and constitute a part of the specification, and are used together with the following specific implementation to explain the embodiments of the present invention, but do not constitute a limitation to the embodiments of the present invention. In the drawings:

[0104] Figure 1 is one of the flow schematic diagrams of the power vehicle monitoring and control method provided by the present invention;

[0105] Figure 2 is another flow schematic diagram of the power vehicle monitoring and control method provided by the present invention;

[0106] Figure 3 is yet another flow schematic diagram of the power vehicle monitoring and control method provided by the present invention;

[0107] Figure 4 is still another flow schematic diagram of the power vehicle monitoring and control method provided by the present invention;

[0108] Figure 5 is another flow schematic diagram of the power vehicle monitoring and control method provided by the present invention;

[0109] Figure 6 is one of the structural schematic diagrams of the power vehicle monitoring and control device provided by the present invention;

[0110] Figure 7 is another structural schematic diagram of the power vehicle monitoring and control device provided by the present invention;

[0111] Figure 8 is one of the structural schematic diagrams of the on-vehicle terminal provided by the present invention;

[0112] Figure 9 This is the second schematic structural diagram of the vehicle-mounted terminal provided by the present invention. Specific embodiments

[0113] The following will describe in detail the specific embodiments of the embodiments of the present invention with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only used to illustrate and explain the embodiments of the present invention, and are not used to limit the embodiments of the present invention.

[0114] Method embodiments

[0115] Please refer to Figure 1 , an embodiment of the present invention provides a power vehicle monitoring and control method, which is applied to a vehicle-mounted terminal of a power vehicle. The method includes:

[0116] Step 110, obtain a sensor data set and a parameter validity index matrix.

[0117] The sensor data set includes sensor data of multiple sensors. Among them, the sensor data of multiple sensors can be sensor data collected for the environmental state of the power vehicle, the vehicle state, and the power production unit. Among them, the sensor data collected for the environmental state of the power vehicle includes temperature data, humidity data, air pressure data, etc. of the environment where the power vehicle is located. The sensor data collected for the vehicle state includes real-time vehicle condition information (such as fuel consumption data, battery voltage data, intake pipe temperature data, current vehicle speed data, engine water temperature data, and engine speed data, etc.), and vehicle driving monitoring information (vehicle position information, area vehicle inspection, real-time monitoring and tracking, trajectory query, trajectory playback, alarm information reminder view, photo data, and video monitoring data, etc.). The sensor data collected for the power production unit includes power parameters, equipment status information, and alarm information, etc. Power parameters refer to parameters such as voltage, current, and power collected in real time. Equipment status information refers to the switch status and online / offline status of the equipment, which helps to realize remote control and fault warning of the equipment. Alarm information means that when a fault or abnormal situation occurs in the power production unit, the sensor will trigger an alarm signal, including information such as alarm type, alarm time, and alarm location, so as to take measures in time for processing.

[0118] The parameter validity index matrix includes parameter validity indexes of all sensor data; the parameter validity index characterizes the correlation degree of each sensor data with the corresponding task control accuracy. Please refer to Figure 2 , an embodiment of the present invention uses the parameter validity index matrix to perform validity analysis on the sensor data of multiple sensors, and then the classification result of the sensor data of multiple sensors can be obtained, so as to realize the rapid preprocessing operation of the data.

[0119] Since the characteristics of the input sensor data contribute differently to the task control accuracy, a parameter effectiveness index is introduced to quantify the correlation between each sensor data and the task control accuracy, so as to solve the influence of the characteristics of multi-input or redundant sensor data on the classification complexity, realize the rapid preprocessing of sensor data, and reduce the computing pressure of the vehicle-mounted terminal. The parameter effectiveness index matrix Q: For the sensor data collected by the nth sensor, there is one or more corresponding service functions, which are used to characterize the influence degree of a certain sensor data on the control output accuracy. If the influence ability of a certain sensor data on the control accuracy is smaller, the value of the parameter effectiveness index is smaller.

[0120] In one embodiment, the parameter effectiveness index of each sensor data in the parameter effectiveness index matrix is calculated through the following steps:

[0121] Step 11: Determine the probability that the sensor data of the first sensor belongs to the first task.

[0122] Step 12: Calculate the information entropy based on the probability.

[0123] Step 13: Calculate the conditional information entropy corresponding to the sensor data of the first sensor based on the information entropy.

[0124] Step 14: Calculate the information gain based on the conditional information entropy and the information entropy.

[0125] Step 15: Perform gain normalization processing on the information gain to obtain the parameter effectiveness index of the sensor data of the first sensor. Wherein, the sensor data of the first sensor is the sensor data of any one sensor in the sensor data set; the first task is any one task in the set task sequence.

[0126] Specifically, the sensor data streams collected by multiple sensors are a typical multi-classification problem, which can be abstracted as a sensor data set C, C = {C1, C2…C n}, where n is the number of sensors.

[0127] Construct a sensor data classification model C→R, where R represents the task scenario of the power supply vehicle.

[0128] S(C n →R m ) = S nm , which represents the input of the nth sensor data to the mth task of the power supply vehicle.

[0129] |S 1m | + |S 2m | + … |S nm | = |S m |; Formula (1).

[0130] |S m | is the total number of states to which all sensor data are mapped to task m, where m is the number of scenarios.

[0131] The parameter validity index is used to measure the correlation degree between each sensor data and the corresponding task control accuracy. For any sensor data C n (i.e., the sensor data of the first sensor) belonging to task R m (i.e., the first task) is:

[0132] P nm = |S nm | / |S m |; Formula (2).

[0133] Among them, the probability that each sensor data belongs to the corresponding task can be set in advance.

[0134] The information entropy constructed for the sensor data is:

[0135]

[0136] The conditional information entropy value corresponding to the sensor data C n is;

[0137]

[0138] The information gain of the sensor data C n is:

[0139] Gain = Mesg(S nm ) - Mesg(S m ); Formula (5).

[0140] The validity index obtained by normalizing the information gain is:

[0141]

[0142] k is the number of times the task calls or samples the sensor data.

[0143] Performing the above calculations on each sensor data can obtain the parameter validity index matrix Q as shown below:

[0144]

[0145] Step 130: Determine the parameter validity value of each sensor data for all tasks based on the sensor data set and the parameter validity index matrix.

[0146] In the embodiment of the present invention, based on the product R of the sensor data set C and the parameter validity index matrix Qm The parameter effectiveness value of each sensor data for all tasks can be determined. If the influence ability of a certain sensor data on the control accuracy is smaller, the calculated parameter effectiveness value is smaller, and it should be excluded as a redundant feature or an irrelevant feature; on the contrary, if the influence ability of a certain sensor data on the control accuracy is smaller, the calculated parameter effectiveness value is larger, and it should be retained.

[0147] Step 150: Determine the effective sensor data for each task in the set task sequence based on the parameter effectiveness values of all sensor data, and perform task calculations based on the effective sensor data for each task in the set task sequence.

[0148] In the embodiment of the present invention, the effective or ineffective sensor data for each task can be determined through the comparison result between the parameter effectiveness value of all sensor data and a set threshold.

[0149] In one embodiment, the determining the effective sensor data for each task in the set task sequence based on the parameter effectiveness values of all sensor data includes: determining the sensor data with a parameter effectiveness value greater than or equal to the set threshold in each task as the effective sensor data.

[0150] For example, the product R of the sensor data set C and the parameter effectiveness index matrix Q m includes the parameter effectiveness value of each sensor data. If the parameter effectiveness value of the sensor data is greater than or equal to the set threshold (for example, 0.3), then at this time the sensor data is effective sensor data for the task; otherwise, if the parameter effectiveness value of the sensor data is less than the set threshold of 0.3, then at this time the sensor data is ineffective sensor data for the task. Thus, in the embodiment of the present invention, through the parameter effectiveness index matrix Q, the screening of multiple sensor data for each task in the set task sequence is realized, and each task only screens out the sensor data with a parameter effectiveness value greater than or equal to the set threshold as the effective sensor data. Thus, the ineffective sensor data with a parameter effectiveness value less than the set threshold is removed. For example, if the current task in the set task sequence is to collect the environmental state of the power vehicle. At this time, the parameter effectiveness value of the vehicle real-time vehicle condition information in the sensor data set is less than the set threshold, which indicates that the vehicle real-time vehicle condition information belongs to ineffective sensor data and should be discarded.

[0151] It should be noted that, please refer to Figure 2 , the set task sequence includes tasks 1 to task m arranged and processed in chronological order. For example, the set task sequence includes tasks for collecting environmental state, tasks for collecting and controlling vehicle real-time vehicle condition information or vehicle driving monitoring information, and tasks for collecting power production units, and the collection and monitoring are carried out according to the set task sequence.

[0152] The tasks of the power supply vehicle are divided into multiple subtasks according to scenarios. According to the relevance between the subtasks, task embedding encoding is performed in the order of each subtask to obtain a task embedding sequence (i.e., a set task sequence). The task embedding sequence is input into the sequence-to-sequence model to generate a task loading and unloading strategy, and an end-to-end control model from input perception to output control is established from the input end to the output end. Multiple subtasks such as scenario environment perception, target recognition, and target decision-making can be unified into an intelligent model to complete the unification from cognition to control decision-making.

[0153] In other aspects of the embodiments of the present invention, after step 150, determining the valid sensor data of each task in the set task sequence based on the parameter validity values of all sensor data, and performing task calculations based on the valid sensor data of each task in the set task sequence, further includes:

[0154] Step 160, in the case of triggering a core task, set the processing priority of the core task to the highest priority among all tasks in the set task sequence for priority processing.

[0155] Wherein the core task represents an emergency situation that occurs during the normal use of the power supply vehicle. Please refer to Figure 3 , for example, the engine oil temperature of the power supply vehicle is too high, the oil level is too low, a fault code appears, etc. When the above emergency situation occurs, set the processing priority of the core task to the highest priority among all tasks in the set task sequence (or task scheduling sequence) for priority processing. For example, alarms for too high engine oil temperature, too low oil level, and fault code appear through voice or text, etc. Thus, while ensuring real-time and orderly scheduling of tasks, it can also timely handle emergency situations that occur during the normal use of the power supply vehicle to avoid affecting normal operations.

[0156] Step 170, perform individual learning based on the valid sensor data of each task to determine the correction factor and control function of each task.

[0157] Through the previous data classification and task arrangement, the valid sensor data is quickly input into the intelligent control model of the embodiments of the present invention to obtain a control output. The intelligent control model of the embodiments of the present invention includes: an intelligent control algorithm, and realizes shared update through online learning and offline learning (or individual learning and integrated learning).

[0158] In one embodiment, the performing individual learning based on the valid sensor data of each task to determine the correction factor and control function of each task includes:

[0159] Repeat the following steps until the correction factor of the first task reaches the set condition, and output the correction factor and control function after the iteration is completed:

[0160] Step 171: Input the valid sensor data of the first task into the first task to obtain the output result of the first task. Herein, the first task is any one task in the set task sequence.

[0161] Step 172: Calculate the correction factor of the first task based on the output result of the first task.

[0162] In one embodiment, step 172: Calculate the correction factor of the first task based on the output result of the first task, including:

[0163] Step 1721: Determine the mean point of all positive class outputs and the mean point of all negative class outputs based on the output result of the first task.

[0164] In the embodiment of the present invention, the mean point of all positive class outputs in the output result CR of the sensor data set C → task R of the first task is denoted as 0 + and the mean point of all negative class outputs is denoted as 0 - .

[0165] Step 1722: Determine a first hyperplane passing through the mean point of all positive class outputs and a second hyperplane passing through the mean point of all negative class outputs based on the mean point of all positive class outputs and the mean point of all negative class outputs.

[0166] Specifically, in the embodiment of the present invention, it is determined that the plane with φ = 0 + -0 - as the normal vector and passing through the mean point of all positive class outputs is the first hyperplane, and it is determined that the plane with φ = 0 + -0 - as the normal vector and passing through the mean point of all negative class outputs is the second hyperplane. It is represented by formula (8) as:

[0167]

[0168] Step 1723: Calculate the average value of the distances between all positive class outputs and the first hyperplane as the first average value, calculate the average value of the distances between all negative class outputs and the second hyperplane as the second average value, calculate the maximum value of the distances between all positive class outputs and the first hyperplane as the first maximum value, and calculate the maximum value of the distances between all negative class outputs and the second hyperplane as the second maximum value.

[0169] In the embodiment of the present invention, the average value of the distances between all positive class outputs and the first hyperplane is calculated as the first average value l + , and the average value of the distances between all negative class outputs and the second hyperplane is calculated as the second average value l - . Take the first maximum value L+ = max{l +}, the second maximum value L - = max{l -} respectively represent the maximum value of the distances between all positive-class outputs and the first hyperplane and the maximum value of the distances between all negative-class outputs and the second hyperplane.

[0170] Step 1724, calculate the correction factor for the first task based on the first average value, the second average value, the first maximum value, and the second maximum value.

[0171] In one embodiment, the calculating the correction factor for the first task based on the first average value, the second average value, the first maximum value, and the second maximum value is calculated by the following formula:

[0172]

[0173] where ε represents the correction factor for the first task, l + represents the first average value, l - represents the second average value, L + represents the first maximum value, L - represents the second maximum value.

[0174] Step 173, calculate the control function for the first task based on the valid sensor data of previous samplings of the first task; the control function is calculated based on a cubic spline interpolation function.

[0175] In the case where the correction factor is calculated, it is judged whether it meets the requirements according to the size of the correction factor. Specifically, in one embodiment, after the task R inputs the sensor data C, the obtained control result is updated by the following formula:

[0176]

[0177] where CR represents the output result of the sensor data set C → task R, or it can be understood as the control result obtained by the task R after inputting the sensor data C. k represents the number of times the task calls or samples the sensor data. When the correction factor is less than 0.6 or greater than or equal to 0.6 and less than 0.9, it does not meet the requirements at this time, calculate the control function, and obtain the updated sensor data set for iterative calculation of the correction factor and control function of the first task.

[0178] where step 173, calculate the control function for the first task based on the valid sensor data of previous samplings of the first task, includes:

[0179] Step 1731: Calculate the mean and standard deviation of the valid sensor data for each sampling of the first task based on the valid sensor data for each sampling of the first task.

[0180] The mean m of the valid sensor data for each sampling k is calculated by the following formula:

[0181]

[0182] where Cn refers to the nth sensor data, and k represents the number of times the task calls or samples the sensor data.

[0183] The standard deviation h of the valid sensor data for each sampling k is calculated by the following formula:

[0184]

[0185] where Cn refers to the nth sensor data, and k represents the number of times the task calls or samples the sensor data.

[0186] Step 1732: Calculate the weight coefficient of the control function based on the mean and the standard deviation.

[0187] Step 1733: Construct a cubic spline interpolation function based on the difference between the valid sensor data of the first task for two adjacent samplings, the calculated value of the previous sampling of the control function of the first task, and the weight coefficient of the control function.

[0188] Step 1734: Solve the cubic spline interpolation function to obtain the control function of the first task.

[0189] In one embodiment, the control function of the first task calculated based on the cubic spline interpolation function is defined as:

[0190] g k = a k + b k (Cn k - Cn k-1 ) + c k (Cn k - Cn k-1 ) 2 + d k (Cn k - Cn k-1 ) 3 ; Formula (13).

[0191]

[0192] where g k represents the control function of the first task, gk-1 The previous sampling calculation value of the control function representing the first task, b k , c k and d k represent the weight coefficients of the control function, b k are calculated through the mean value m k and the standard deviation h k c is calculated. k d is calculated through the mean value m k d is calculated. k Cn is calculated through the mean value m k and the standard deviation h k Cn is calculated. k -Cn k-1 represents the difference between the effective sensor data of the first task in two adjacent samplings, and k represents the number of times the task calls the sensor data or the number of samplings. Substitute the previous sampling calculation value g of the control function of the first task k-1 , the weight coefficients b k , c k and d k , and the difference Cn k -Cn k-1 of the effective sensor data of the first task in two adjacent samplings into formula (13), solve formula (13), and obtain the control function of the first task.

[0193] Step 174: Obtain an updated sensor data set including the updated sensor data of the multiple sensors.

[0194] Step 175: Determine the updated parameter effectiveness value of each sensor data for all tasks based on the product of the updated sensor data set and the parameter effectiveness index matrix.

[0195] Step 176: Determine the updated effective sensor data of the first task based on the updated parameter effectiveness values of all sensor data for iterative calculation of the correction factor and control function of the first task.

[0196] When the correction factor ε is less than 0.6 or greater than or equal to 0.6 and less than 0.9, it does not meet the requirements at this time. Calculate the control function, and obtain the updated sensor data set, and calculate the updated effective sensor data of the first task again, so as to perform iterative calculation of the correction factor and control function of the first task again until the correction factor reaches the set condition (ε is greater than or equal to 0.9), and the individual learning result is obtained at this time. The individual learning result includes the parameter effectiveness index matrix, the correction factor and control function of each task.

[0197] Step 190: Send the individual learning results of all tasks to the network device; the individual learning results include the parameter validity index matrix, the correction factor for each task, and the control function.

[0198] The in-vehicle terminal sends the individual learning results of all tasks to the network device so that the network device can perform integrated learning. The individual learning results include the parameter validity index matrix, the correction factor for each task, and the control function.

[0199] The data classification method based on the parameter validity index matrix in the embodiments of the present invention enables fast classification input of corresponding data; performs task calculations based on the valid sensor data of each task in the set task sequence to achieve task priority resource coordination and meet the response requirements of real-time tasks; and through the update of the correction factor and the control function, combined with the iteration of individual learning and integrated learning, realizes end-to-end decision control. Therefore, the embodiments of the present invention realize intelligent control from data acquisition and classification, task real-time scheduling, and efficient end-to-end processing, and can automatically control the power production special vehicle to be put into and withdrawn from service safely, reliably, flexibly, and efficiently.

[0200] The embodiments of the present invention propose a power vehicle monitoring and control method, which realizes the perception and upload of key data such as the intelligent operation and monitoring and maintenance of the power vehicle through the set in-vehicle terminal, sensors and other sensing devices, and interacts with the command system in real time to meet the high-efficiency command requirements of emergency power supply support tasks. Specifically, it has the following application scenarios:

[0201] 1. Online monitoring: Omnidirectional, fully intelligent, and accurate monitoring

[0202] Through the background system, functions such as querying the location information of the fleet and vehicles, checking vehicles in a region, real-time monitoring and tracking, trajectory query, trajectory playback, viewing alarm information reminders, taking pictures, and video monitoring are provided to monitor the vehicle information in real time and facilitate management.

[0203] 2. Scheduling management: Automatic scheduling, cost saving, and time saving

[0204] It mainly includes functions of automatic scheduling, manual scheduling, and status query. Among them, automatic scheduling combines the vehicle task situation and the vehicle position range, automatically calculates and matches, and automatically sends scheduling task information to it; manual scheduling provides vehicle checking in a region, SMS scheduling, and voice scheduling. A variety of comprehensive information scheduling saves more costs and time for enterprises.

[0205] 3. Report management: Comprehensive and rich business information content

[0206] The TCU provides functions of multi-condition query of reports and export of reports. With the help of the report function, the business information content is comprehensively enriched and the vehicle information is more abundant.

[0207] The reports mainly include alarm statistics (SOS alarm, out-of-bounds alarm, overspeed alarm, power-off alarm), driving reports (start / stop statistics, historical tracks), fuel consumption reports (daily fuel consumption, refueling statistics), mileage statistics, temperature statistics, short message statistics, and image statistics.

[0208] 4. OBD information: Keep track of the vehicle's health status at any time and anywhere

[0209] Provide real-time query functions for the vehicle's real-time condition information (fuel consumption information, battery voltage, intake pipe temperature, current vehicle speed, engine water temperature, engine speed), fault records, maintenance management, and driving reports.

[0210] On the other hand, please refer to Figure 4 , the embodiment of the present invention also provides a monitoring and control method for a power supply vehicle, which is applied to a network device. The method includes:

[0211] Step 210: Receive the individual learning results of all tasks sent by multiple in-vehicle terminals.

[0212] The individual learning results include a parameter effectiveness index matrix Q, a correction factor ε for each task, and a control function g k ; the parameter effectiveness index matrix includes the parameter effectiveness indexes of all sensor data obtained by the in-vehicle terminal; the parameter effectiveness index characterizes the correlation degree between each sensor data and the control accuracy of the corresponding task. The network device in the embodiment of the present invention can be a cloud server platform.

[0213] Step 220: Perform ensemble learning based on the individual learning results of all tasks of the multiple in-vehicle terminals to obtain an ensemble learning result.

[0214] The cloud server platform performs ensemble learning (such as Bagging, Boosting, and Stacking, etc.) based on the individual learning results of all tasks of the multiple in-vehicle terminals to obtain an ensemble learning result. Please refer to Figure 5 , in one embodiment, Step 220: Perform ensemble learning based on the individual learning results of all tasks of the multiple in-vehicle terminals to obtain an ensemble learning result, including: Step 221: Perform feature splicing on the individual learning results of all tasks of the multiple in-vehicle terminals to obtain aggregated features; Step 222: Perform feature extraction on the aggregated features through a feature extraction network to obtain the common features of the individual learning results of all tasks of the multiple in-vehicle terminals as the ensemble learning result.

[0215] The cloud server platform performs feature splicing on the individual learning results of all tasks of multiple vehicle-mounted terminals. Specifically, any one of horizontal splicing, vertical splicing, or overlay splicing can be selected to perform feature splicing on the individual learning results of all tasks of multiple vehicle-mounted terminals to obtain aggregated features. The aggregated features are then subjected to feature extraction through a feature extraction network such as a convolutional neural network, a recurrent neural network, or an autoencoder to obtain the common features of the individual learning results of all tasks of multiple vehicle-mounted terminals as the integrated learning result.

[0216] Step 230: Send the integrated learning result to all vehicle-mounted terminals so that each vehicle-mounted terminal replaces the individual learning result based on the integrated learning result.

[0217] The cloud server platform sends the integrated learning result to all vehicle-mounted terminals so that each vehicle-mounted terminal replaces the individual learning result based on the integrated learning result.

[0218] The vehicle-mounted terminal in the embodiment of the present invention acquires sensor data and sensor data labels, constructs a parameter validity index matrix Q through a local parameter validity index, performs end-to-end mapping control analysis on the sensor data, performs control calculation output through a correction factor and a control function, and updates the Q matrix, the correction factor ε, and the control function g k in combination with the output result, and sends the individual result to the cloud server. The cloud server performs feature splicing on the individual results of multiple vehicle-mounted terminals to obtain aggregated features, extracts shared features (i.e., common features) through large model aggregated feature extraction, and sends the shared features to each vehicle-mounted terminal for storage and update.

[0219] Please refer to Figure 2 and Figure 3 , the data classification method based on parameter index validity in the embodiment of the present invention quickly realizes the classification input of corresponding sensor data. The embodiment of the present invention also proposes a task priority resource coordination method to meet the response requirements of real-time tasks. In addition, the embodiment of the present invention also proposes an intelligent control algorithm for integrated learning to realize end-to-end decision control. The embodiment of the present invention realizes intelligent control from data acquisition and classification, task real-time scheduling, and efficient end-to-end processing, and can automatically control the power-on and power-off of power production special vehicles safely, reliably, flexibly and efficiently, reduce the complexity of operation, and meet the high-efficiency command requirements of emergency supply support tasks.

[0220] Device Embodiment

[0221] Please refer to Figure 6 , on the other hand, the embodiment of the present invention also provides a power vehicle monitoring and control device, including:

[0222] An acquisition module 601, configured to acquire a set of sensor data and a matrix of parameter validity indicators; the set of sensor data includes sensor data of multiple sensors; the matrix of parameter validity indicators includes parameter validity indicators of all sensor data; the parameter validity indicator characterizes the correlation degree between each sensor data and the corresponding task control accuracy;

[0223] A determination module 602, configured to determine the parameter validity value of each sensor data for all tasks based on the set of sensor data and the matrix of parameter validity indicators;

[0224] A classification execution module 603, configured to determine the valid sensor data of each task in a set task sequence based on the parameter validity values of all sensor data, and perform task calculations based on the valid sensor data of each task in the set task sequence;

[0225] An individual learning module 604, configured to perform individual learning based on the valid sensor data of each task to determine the correction factor and control function of each task;

[0226] A sending module 605, configured to send the individual learning results of all tasks to a network device; the individual learning results include the matrix of parameter validity indicators, the correction factor and control function of each task.

[0227] Optionally, the parameter validity indicator of each sensor data in the matrix of parameter validity indicators is calculated through the following steps:

[0228] Determine the probability that the sensor data of the first sensor belongs to the first task;

[0229] Calculate the information entropy based on the probability;

[0230] Calculate the conditional information entropy corresponding to the sensor data of the first sensor based on the information entropy;

[0231] Calculate the information gain based on the conditional information entropy and the information entropy;

[0232] Perform gain normalization processing on the information gain to obtain the parameter validity indicator of the sensor data of the first sensor;

[0233] Wherein, the sensor data of the first sensor is the sensor data of any one sensor in the set of sensor data; the first task is any one task in the set task sequence.

[0234] Optionally, the determining the valid sensor data of each task in the set task sequence based on the parameter validity values of all sensor data includes:

[0235] Determine the sensor data with the parameter validity value greater than or equal to the set threshold in each task as the valid sensor data.

[0236] Optionally, the device further includes:

[0237] A priority processing module, configured to, when triggering a core task, set the processing priority of the core task to the highest priority among all tasks in the set task sequence for priority processing; the core task represents an emergency situation that occurs during the normal use of the power vehicle.

[0238] Optionally, the individual learning based on the valid sensor data of each task to determine the correction factor and control function of each task includes:

[0239] Repeat the following steps until the correction factor of the first task reaches the set condition, and output the correction factor and control function when the iteration is completed:

[0240] Input the valid sensor data of the first task into the first task to obtain the output result of the first task;

[0241] Calculate the correction factor of the first task based on the output result of the first task;

[0242] Calculate the control function of the first task based on the valid sensor data of previous samplings of the first task; the control function is calculated based on a cubic spline interpolation function;

[0243] Obtain an updated sensor data set including the updated sensor data of the multiple sensors;

[0244] Determine the updated parameter validity value of each sensor data for all tasks based on the product of the updated sensor data set and the parameter validity index matrix;

[0245] Determine the updated valid sensor data of the first task based on the updated parameter validity values of all sensor data for iterative calculation of the correction factor and control function of the first task;

[0246] Wherein, the first task is any task in the set task sequence.

[0247] Optionally, the calculating the correction factor of the first task based on the output result of the first task includes:

[0248] Determine the mean points of all positive class outputs and the mean points of all negative class outputs based on the output result of the first task;

[0249] Determine a first hyperplane passing through the mean point of all positive class outputs and a second hyperplane passing through the mean point of all negative class outputs based on the mean point of all positive class outputs and the mean point of all negative class outputs;

[0250] Calculate the average of the distances between all positive class outputs and the first hyperplane as the first average value, calculate the average of the distances between all negative class outputs and the second hyperplane as the second average value, calculate the maximum value of the distances between all positive class outputs and the first hyperplane as the first maximum value, and calculate the maximum value of the distances between all negative class outputs and the second hyperplane as the second maximum value;

[0251] Calculate the correction factor of the first task based on the first average value, the second average value, the first maximum value, and the second maximum value.

[0252] Optionally, the calculating the correction factor of the first task based on the first average value, the second average value, the first maximum value, and the second maximum value is calculated by the following formula:

[0253]

[0254] where ε represents the correction factor of the first task, l + represents the first average value, l - represents the second average value, L + represents the first maximum value, L - represents the second maximum value.

[0255] Optionally, the calculating the control function of the first task based on the valid sensor data of previous samplings of the first task includes:

[0256] Calculate the mean and standard deviation of the valid sensor data of previous samplings of the first task based on the valid sensor data of previous samplings of the first task;

[0257] Calculate the weight coefficient of the control function based on the mean and the standard deviation;

[0258] Construct a cubic spline interpolation function based on the difference between the valid sensor data of the first task in two adjacent samplings, the calculated value of the previous sampling of the control function of the first task, and the weight coefficient of the control function;

[0259] Solve the cubic spline interpolation function to obtain the control function of the first task.

[0260] The power vehicle monitoring and control device includes a processor and a memory. The above-mentioned acquisition module 601, determination module 602, classification and execution module 603, individual learning module 604, and sending module 605 are all stored in the memory as program units, and the processor executes the above program units stored in the memory to implement corresponding functions.

[0261] The processor contains a kernel, and the kernel retrieves the corresponding program units from the memory. One or more kernels can be set.

[0262] The memory may include non-permanent memory in a computer-readable medium, in the form of random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. The memory includes at least one storage chip.

[0263] On the other hand, please refer to Figure 7 , the embodiment of the present invention also provides a power vehicle monitoring and control device, including:

[0264] A receiving module 701, configured to receive the individual learning results of all tasks sent by multiple vehicle-mounted terminals; the individual learning results include a parameter validity index matrix, a correction factor for each task, and a control function; the parameter validity index matrix includes the parameter validity indexes of all sensor data obtained by the vehicle-mounted terminals; the parameter validity index characterizes the correlation degree between each sensor data and the control accuracy of the corresponding task;

[0265] An integrated learning module 702, configured to perform integrated learning based on the individual learning results of all tasks of the multiple vehicle-mounted terminals to obtain an integrated learning result;

[0266] A sending module 703, configured to send the integrated learning result to all vehicle-mounted terminals, so that each vehicle-mounted terminal replaces the individual learning result based on the integrated learning result.

[0267] Optionally, the performing integrated learning based on the individual learning results of all tasks of the multiple vehicle-mounted terminals to obtain an integrated learning result includes:

[0268] Performing feature splicing on the individual learning results of all tasks of multiple vehicle-mounted terminals to obtain an aggregated feature;

[0269] Performing feature extraction on the aggregated feature through a feature extraction network to obtain the common feature of the individual learning results of all tasks of multiple vehicle-mounted terminals as the integrated learning result.

[0270] The power vehicle monitoring and control device includes a processor and a memory. The above-mentioned receiving module 701, integrated learning module 702, sending module 703, etc. are all stored in the memory as program units, and the processor executes the above program units stored in the memory to implement corresponding functions.

[0271] The processor contains a kernel, and the kernel retrieves the corresponding program units from the memory. One or more kernels can be set.

[0272] The memory may include non-permanent memory in a computer-readable medium, in the form of random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash memory (flash RAM). The memory includes at least one storage chip.

[0273] On the other hand, an embodiment of the present invention further provides a vehicle-mounted terminal, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the above-mentioned power vehicle monitoring and control method is implemented. Figure 8 The schematic diagram of the entity structure of a vehicle-mounted terminal is exemplified, as Figure 8 shown, the electronic device may include: a processor 810, a communication interface 820, a memory 830, and a communication bus 840. Among them, the processor 810, the communication interface 820, and the memory 830 complete mutual communication through the communication bus 840. The processor 810 can call the logical instructions in the memory 830 to execute the power vehicle monitoring and control method, and the method includes: obtaining a sensor data set and a parameter validity index matrix; the sensor data set includes sensor data of multiple sensors; the parameter validity index matrix includes parameter validity indexes of all sensor data; the parameter validity index characterizes the correlation degree between each sensor data and the corresponding task control accuracy; determining the parameter validity value of each sensor data for all tasks based on the sensor data set and the parameter validity index matrix; determining the valid sensor data of each task in the set task sequence based on the parameter validity values of all sensor data, and performing task calculations based on the valid sensor data of each task in the set task sequence; performing individual learning based on the valid sensor data of each task to determine the correction factor and control function of each task; sending the individual learning results of all tasks to the network device; the individual learning results include the parameter validity index matrix, the correction factor and control function of each task.

[0274] In addition, when the logical instructions in the above-mentioned memory 830 are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, 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 for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0275] In other embodiments, please refer to Figure 9 , a vehicle-mounted terminal (or vehicle-mounted intelligent terminal), including a main control storage module, a data acquisition and distribution module, a task scheduling module, a control algorithm module, and a communication interface module.

[0276] The main control storage module collects a set of sensor data through the communication interface module. The main control storage module calls the data acquisition and distribution module to obtain the parameter validity index matrix Q. The main control storage module calls the task scheduling module to obtain the set task sequence. The main control storage module calls the control algorithm module to calculate the output result, and analyzes the correction factor ε to realize the optimization and update of the result, and performs the update of the control function g k .

[0277] Among them, the communication interface module includes a CAN interface, a multi-channel RS-485 communication interface, an Ethernet interface, a tele-signaling and remote control interface, a wireless public / private network remote communication interface, a Bluetooth wireless interface, and a Beidou communication interface.

[0278] The vehicle-mounted terminal according to the embodiment of the present invention realizes intelligent control from data acquisition and classification, task real-time scheduling, and efficient end-to-end processing, and can automatically control the power-on and power-off of the power vehicle safely, reliably, flexibly, and efficiently, reducing the complexity of operation. By integrating the perception, scheduling, and control modules through an integrated model, the boundaries between the three are eliminated, making them an integrated whole, which can be more rapid and find more suitable coping strategies for the scenario, meeting the high-efficiency command requirements of emergency supply support tasks.

[0279] On the other hand, the embodiment of the present invention also provides a network device (not shown), including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, it implements the above-mentioned power vehicle monitoring and control method.

[0280] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a machine-readable storage medium. When the computer program is executed by a processor, the computer can execute the monitoring and control method for a power vehicle, and the method includes: obtaining a set of sensor data and a parameter validity index matrix; the set of sensor data includes sensor data of multiple sensors; the parameter validity index matrix includes parameter validity indexes of all sensor data; the parameter validity index represents the correlation degree between each sensor data and the corresponding task control accuracy; determining the parameter validity value of each sensor data for all tasks based on the set of sensor data and the parameter validity index matrix; determining the valid sensor data of each task in a set task sequence based on the parameter validity values of all sensor data, and performing task calculations based on the valid sensor data of each task in the set task sequence; performing individual learning based on the valid sensor data of each task to determine the correction factor and control function of each task; sending the individual learning results of all tasks to a network device; the individual learning results include the parameter validity index matrix, the correction factor and control function of each task. Or,

[0281] Receiving the individual learning results of all tasks sent by multiple vehicle-mounted terminals; the individual learning results include a parameter validity index matrix, a correction factor and a control function of each task; the parameter validity index matrix includes parameter validity indexes of all sensor data obtained by the vehicle-mounted terminals; the parameter validity index represents the correlation degree between each sensor data and the corresponding task control accuracy; performing integrated learning based on the individual learning results of all tasks of the multiple vehicle-mounted terminals to obtain an integrated learning result; sending the integrated learning result to all vehicle-mounted terminals, so that each vehicle-mounted terminal replaces the individual learning result based on the integrated learning result.

[0282] In another aspect, the present invention further provides a machine-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it is configured to execute a power vehicle monitoring and control method, which includes: obtaining a sensor data set and a parameter validity index matrix; the sensor data set includes sensor data of multiple sensors; the parameter validity index matrix includes parameter validity indexes of all sensor data; the parameter validity index characterizes the correlation degree between each sensor data and the corresponding task control accuracy; determining the parameter validity value of each sensor data for all tasks based on the sensor data set and the parameter validity index matrix; determining the valid sensor data of each task in a set task sequence based on the parameter validity values of all sensor data, and performing task calculations based on the valid sensor data of each task in the set task sequence; performing individual learning based on the valid sensor data of each task to determine the correction factor and control function of each task; sending the individual learning results of all tasks to a network device; the individual learning results include the parameter validity index matrix, the correction factor and control function of each task. Or,

[0283] Receiving individual learning results of all tasks sent by multiple vehicle-mounted terminals; the individual learning results include a parameter validity index matrix, a correction factor and control function of each task; the parameter validity index matrix includes parameter validity indexes of all sensor data obtained by the vehicle-mounted terminals; the parameter validity index characterizes the correlation degree between each sensor data and the corresponding task control accuracy; performing integrated learning based on the individual learning results of all tasks of the multiple vehicle-mounted terminals to obtain an integrated learning result; sending the integrated learning result to all vehicle-mounted terminals, so that each vehicle-mounted terminal replaces the individual learning result based on the integrated learning result.

[0284] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. A person of ordinary skill in the art can understand and implement it without creative labor.

[0285] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the essence of the above technical solution or the part that contributes to the prior art can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0286] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A power supply vehicle monitoring and control method, characterized in that: The vehicle-mounted terminal applied to a power supply vehicle comprises: Acquire a sensor data set and a parameter validity index matrix; the sensor data set includes sensor data of multiple sensors; the parameter validity index matrix includes parameter validity indexes of all sensor data; the parameter validity index represents the correlation between each sensor data and the corresponding task control accuracy; Determine a parameter validity value of each sensor data for all tasks based on the sensor data set and the parameter validity indicator matrix; determining valid sensor data for each task in the set task sequence based on parameter validity values ​​of all sensor data, and performing task calculation based on the valid sensor data for each task in the set task sequence; Perform individual learning based on the valid sensor data for each task to determine the correction factor and control function for each task; The individual learning results of all tasks are sent to the network device; the individual learning results include the parameter effectiveness indicator matrix, the correction factor and the control function of each task.

2. The power supply vehicle monitoring and control method according to claim 1, characterized in that: The parameter validity index of each sensor data in the parameter validity index matrix is ​​calculated by the following steps: determining a probability that sensor data from a first sensor belongs to a first task; Calculating information entropy based on the probability; Calculating conditional information entropy corresponding to sensor data of the first sensor based on the information entropy; Calculate information gain based on the conditional information entropy and the information entropy; Performing gain normalization processing on the information gain to obtain a parameter validity index of the sensor data of the first sensor; The sensor data of the first sensor is sensor data of any sensor in a sensor data set; and the first task is any task in a set task sequence.

3. The power supply vehicle monitoring and control method according to claim 1, characterized in that: The determining of valid sensor data for each task in the task sequence based on parameter validity values ​​of all sensor data includes: The sensor data whose parameter validity value in each task is greater than or equal to the set threshold is determined as valid sensor data.

4. The power supply vehicle monitoring and control method according to claim 3, characterized in that: After determining the valid sensor data of each task in the set task sequence based on the parameter validity values ​​of all sensor data, and performing task calculation based on the valid sensor data of each task in the set task sequence, the method further includes: When a core task is triggered, the processing priority of the core task is set to the highest priority of all tasks in the set task sequence for priority processing; the core task represents an emergency situation that occurs during normal use of the power supply vehicle.

5. The power supply vehicle monitoring and control method according to claim 1, characterized in that: The individual learning is performed based on the effective sensor data of each task to determine the correction factor and control function of each task, including: Repeat the following steps until the correction factor of the first task reaches the set condition, and output the correction factor and control function after iteration: Inputting valid sensor data of the first task into the first task to obtain an output result of the first task; Calculating a correction factor for the first task based on an output result of the first task; Calculating a control function of the first task based on the valid sensor data sampled for each time of the first task; the control function is calculated based on a cubic spline interpolation function; obtaining an updated sensor data set including updated sensor data of the plurality of sensors; Determining an updated parameter validity value of each sensor data for all tasks based on the product of the updated sensor data set and the parameter validity indicator matrix; determining updated valid sensor data of the first task based on updated parameter validity values ​​of all sensor data, so as to perform iterative calculation of correction factors and control functions of the first task; The first task is any task in a set task sequence.

6. The power supply vehicle monitoring and control method according to claim 5, characterized in that: The calculating the correction factor of the first task based on the output result of the first task includes: Determine the mean point of all positive class outputs and the mean point of all negative class outputs based on the output result of the first task; Based on the mean point of all the positive class outputs and the mean point of all the negative class outputs, determine a first hyperplane passing through the mean point of all the positive class outputs and a second hyperplane passing through the mean point of all the negative class outputs; Calculate the average value of the distances between all positive class outputs and the first hyperplane as a first average value, calculate the average value of the distances between all negative class outputs and the second hyperplane as a second average value, calculate the maximum value of the distances between all positive class outputs and the first hyperplane as a first maximum value, and calculate the maximum value of the distances between all negative class outputs and the second hyperplane as a second maximum value; A correction factor for the first task is calculated based on the first average value, the second average value, the first maximum value, and the second maximum value.

7. The power supply vehicle monitoring and control method according to claim 6, characterized in that: The correction factor of the first task is calculated based on the first average value, the second average value, the first maximum value, and the second maximum value, and is calculated by the following formula: Where ε represents the correction factor of the first task, l + represents the first average value, l - represents the second average value, L + represents the first maximum value, L - represents the second maximum value.

8. The power supply vehicle monitoring and control method according to claim 5, characterized in that: The calculating the control function of the first task based on the valid sensor data sampled for the first task comprises: Calculate the mean and standard deviation of the valid sensor data sampled previously based on the valid sensor data sampled previously for the first task; Calculating a weight coefficient of the control function based on the mean and the standard deviation; Constructing a cubic spline interpolation function based on a difference in effective sensor data of the first task between two adjacent samples, a previous sampled calculated value of a control function of the first task, and a weight coefficient of the control function; The cubic spline interpolation function is solved to obtain a control function of the first task.

9. A power supply vehicle monitoring and control method, characterized in that: Applied to a network device, the method comprises: Receiving individual learning results of all tasks sent by multiple vehicle terminals; the individual learning results include a parameter effectiveness index matrix, a correction factor of each task and a control function; the parameter effectiveness index matrix includes parameter effectiveness indexes of all sensor data acquired by the vehicle terminal; the parameter effectiveness index represents the correlation between each sensor data and the control accuracy of the corresponding task; Performing integrated learning based on the individual learning results of all tasks of the multiple vehicle-mounted terminals to obtain an integrated learning result; The integrated learning result is sent to all vehicle-mounted terminals, so that each of the vehicle-mounted terminals replaces the individual learning result based on the integrated learning result.

10. The power supply vehicle monitoring and control method according to claim 9, characterized in that: The integrated learning is performed based on the individual learning results of all tasks of the multiple vehicle-mounted terminals to obtain an integrated learning result, including: Perform feature concatenation on the individual learning results of all tasks of multiple vehicle terminals to obtain aggregate features; The aggregated features are extracted through a feature extraction network to obtain common features of individual learning results of all tasks of multiple vehicle-mounted terminals as the integrated learning results.

11. A power supply vehicle monitoring and control device, characterized in that: include: An acquisition module, used to acquire a sensor data set and a parameter validity indicator matrix; the sensor data set includes sensor data of multiple sensors; The parameter validity index matrix includes parameter validity indexes of all sensor data; the parameter validity index represents the correlation between each sensor data and the corresponding task control accuracy; A determination module, configured to determine a parameter validity value of each sensor data for all tasks based on the sensor data set and the parameter validity indicator matrix; A classification execution module, used for determining valid sensor data of each task in a set task sequence based on parameter validity values ​​of all sensor data, and performing task calculation based on the valid sensor data of each task in the set task sequence; An individual learning module, used for performing individual learning based on effective sensor data of each task to determine a correction factor and a control function for each task; A sending module, used to send individual learning results of all tasks to network devices; The individual learning results include the parameter effectiveness indicator matrix, the correction factor of each task and the control function.

12. The power supply vehicle monitoring and control device according to claim 11, characterized in that: The parameter validity index of each sensor data in the parameter validity index matrix is ​​calculated by the following steps: determining a probability that sensor data from a first sensor belongs to a first task; Calculating information entropy based on the probability; Calculating conditional information entropy corresponding to sensor data of the first sensor based on the information entropy; Calculate information gain based on the conditional information entropy and the information entropy; Performing gain normalization processing on the information gain to obtain a parameter validity index of the sensor data of the first sensor; The sensor data of the first sensor is sensor data of any sensor in a sensor data set; and the first task is any task in a set task sequence.

13. The power supply vehicle monitoring and control device according to claim 11, characterized in that: The determining of valid sensor data for each task in the task sequence based on parameter validity values ​​of all sensor data includes: The sensor data whose parameter validity value in each task is greater than or equal to the set threshold is determined as valid sensor data.

14. The power supply vehicle monitoring and control device according to claim 11, characterized in that: The device also includes: The priority processing module is used to set the processing priority of the core task to the highest priority of all tasks in the set task sequence for priority processing when the core task is triggered; the core task represents an emergency situation that occurs during the normal use of the power supply vehicle.

15. The power supply vehicle monitoring and control device according to claim 11, characterized in that: The individual learning is performed based on the effective sensor data of each task to determine the correction factor and control function of each task, including: Repeat the following steps until the correction factor of the first task reaches the set condition, and output the correction factor and control function after iteration: Inputting valid sensor data of the first task into the first task to obtain an output result of the first task; Calculating a correction factor for the first task based on an output result of the first task; Calculating a control function of the first task based on the valid sensor data sampled for each time of the first task; the control function is calculated based on a cubic spline interpolation function; obtaining an updated sensor data set including updated sensor data of the plurality of sensors; Determining an updated parameter validity value of each sensor data for all tasks based on the product of the updated sensor data set and the parameter validity indicator matrix; determining updated valid sensor data of the first task based on updated parameter validity values ​​of all sensor data, so as to perform iterative calculation of correction factors and control functions of the first task; The first task is any task in a set task sequence.

16. The power supply vehicle monitoring and control device according to claim 15, characterized in that: The calculating the correction factor of the first task based on the output result of the first task includes: Determine the mean point of all positive class outputs and the mean point of all negative class outputs based on the output result of the first task; Based on the mean point of all the positive class outputs and the mean point of all the negative class outputs, determine a first hyperplane passing through the mean point of all the positive class outputs and a second hyperplane passing through the mean point of all the negative class outputs; Calculate the average value of the distances between all positive class outputs and the first hyperplane as a first average value, calculate the average value of the distances between all negative class outputs and the second hyperplane as a second average value, calculate the maximum value of the distances between all positive class outputs and the first hyperplane as a first maximum value, and calculate the maximum value of the distances between all negative class outputs and the second hyperplane as a second maximum value; A correction factor for the first task is calculated based on the first average value, the second average value, the first maximum value, and the second maximum value.

17. The power supply vehicle monitoring and control device according to claim 16, characterized in that: The correction factor of the first task is calculated based on the first average value, the second average value, the first maximum value, and the second maximum value, and is calculated by the following formula: Where ε represents the correction factor of the first task, l + represents the first average value, l - represents the second average value, L + represents the first maximum value, L - represents the second maximum value.

18. The power supply vehicle monitoring and control device according to claim 15, characterized in that: The calculating the control function of the first task based on the valid sensor data sampled for the first task comprises: Calculate the mean and standard deviation of the valid sensor data sampled previously based on the valid sensor data sampled previously for the first task; Calculating a weight coefficient of the control function based on the mean and the standard deviation; Constructing a cubic spline interpolation function based on a difference in effective sensor data of the first task between two adjacent samples, a previous sampled calculated value of a control function of the first task, and a weight coefficient of the control function; The cubic spline interpolation function is solved to obtain a control function of the first task.

19. A power supply vehicle monitoring and control device, characterized in that: include: A receiving module, used for receiving individual learning results of all tasks sent by multiple vehicle terminals; The individual learning results include a parameter validity index matrix, a correction factor for each task, and a control function; the parameter validity index matrix includes parameter validity indexes of all sensor data acquired by the vehicle-mounted terminal; the parameter validity index represents the correlation between each sensor data and the control accuracy of the corresponding task; An integrated learning module, used for performing integrated learning based on the individual learning results of all tasks of the plurality of vehicle-mounted terminals to obtain an integrated learning result; The sending module is used to send the integrated learning result to all vehicle-mounted terminals, so that each of the vehicle-mounted terminals replaces the individual learning result based on the integrated learning result.

20. The power supply vehicle monitoring and control device according to claim 19, characterized in that: The integrated learning is performed based on the individual learning results of all tasks of the multiple vehicle-mounted terminals to obtain an integrated learning result, including: Perform feature concatenation on the individual learning results of all tasks of multiple vehicle terminals to obtain aggregate features; The aggregated features are extracted through a feature extraction network to obtain common features of individual learning results of all tasks of multiple vehicle-mounted terminals as the integrated learning results.

21. A vehicle-mounted terminal, 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 program, the power supply vehicle monitoring and control method according to any one of claims 1 to 8 is implemented.

22. A network 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 program, the power supply vehicle monitoring and control method described in any one of claims 9 to 10 is implemented.

23. A machine-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the power supply vehicle monitoring and control method described in any one of claims 1 to 8 is implemented, or the power supply vehicle monitoring and control method described in any one of claims 9 to 10 is implemented.

24. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the power supply vehicle monitoring and control method described in any one of claims 1 to 8 is implemented, or the power supply vehicle monitoring and control method described in any one of claims 9 to 10 is implemented.

Images