Multi-task resource optimization method applied to cable explosion-proof monitoring terminal

Abstract

A multi-task resource optimization method applied to a cable explosion-proof monitoring terminal belongs to the technical field of digital information transmission, and comprises the following steps: obtaining task information to be processed; obtaining a to-be-processed task type; sorting the to-be-processed tasks according to the to-be-processed task information and the to-be-processed task types; wherein the to-be-processed tasks are sorted according to the to-be-processed task information and the to-be-processed task types, and the step of sorting the to-be-processed tasks of at least one type through an optimization model objective function. According to the method and the device, the processing problem when a plurality of tasks such as early warning or data uploading are concurrently encountered by the FPGA is solved, the miniaturization of the cable explosion-proof monitoring terminal is facilitated, and a relatively good programming effect is achieved by utilizing a relatively small space.

Description

Technical Field

[0001] The present invention belongs to the technical field of digital information transmission, and particularly relates to a multi-task resource optimization method applied to a cable explosion-proof monitoring terminal. Background Technique

[0002] A cable explosion-proof monitoring terminal is a device used to monitor the operating state of a cable and prevent potential failures. It can timely detect abnormal situations and issue early warnings by real-time monitoring parameters such as the temperature, ambient humidity, and vibration state of the cable, thereby ensuring the safe operation of the cable.

[0003] For example, Chinese Patent No. CN109361181B discloses a high-voltage cable joint early warning explosion-proof box and an early warning monitoring method. The explosion-proof box body is arranged at the middle joint of the high-voltage cable. The sensor, controller, and in-box fire extinguishing device are all arranged inside the explosion-proof box body, and the out-box fire extinguishing device is arranged in the high-voltage cable trench outside the explosion-proof box body. It issues an early warning according to the monitoring data value being greater than the corresponding set temperature data and / or spatial displacement data threshold, and controls the in-box fire extinguishing device and / or the out-box fire extinguishing device to extinguish the fire of the explosion-proof box body through the controller.

[0004] However, due to the limited space in the cable well, the cable explosion-proof monitoring terminal arranged in the cable well generally needs to be attached to the surface of the cable joint, resulting in a relatively small internal space of the monitoring terminal, and higher-integration electronic components need to be used. Although some field-programmable gate arrays (FPGAs) have the advantage of high integration and can arrange more electronic components on a very small circuit board, the GPU soft core inside is generally single-core and has limited performance. It is difficult to process in the case of multiple early warning tasks or data upload tasks occurring concurrently.

[0005] Therefore, it is urgently necessary to develop a multi-task resource optimization method applied to a cable explosion-proof monitoring terminal to solve the problems in the prior art. Summary of the Invention

[0006] The purpose of the present invention is to provide a multi-task resource optimization method applied to a cable explosion-proof monitoring terminal. Based on the task allocation mechanism of the timing pipeline, multiple tasks are gradually divided into single tasks according to the time sequence, so as to realize the processing under the single-core hardware resources, and solve the processing problem of the FPGA when multiple early warning or data upload tasks occur concurrently as mentioned in the above background technique.

[0007] To solve the above technical problems, the specific technical solution of the present invention is as follows:

[0008] A multi-task resource optimization method applied to a cable explosion-proof monitoring terminal includes the following steps:

[0009] Obtain the information of the task to be processed;

[0010] Obtain the type of task to be processed;

[0011] Sort the tasks to be processed according to the information of the tasks to be processed and the type of tasks to be processed;

[0012] Among them, the sorting of the tasks to be processed according to the information of the tasks to be processed and the type of tasks to be processed includes:

[0013] Sort at least one type of task to be processed through an optimized model objective function;

[0014] The optimized model objective function is as follows:

[0015]

[0016] Among them, C(t) is the total cost of task processing by the computing unit at the t-th moment; e is the number of the internal PE of the computing unit, ranging from 1 to K, and K is the total number of computing units; is the computing cost generated when the internal PE of the computing unit processes the task; is the delay cost generated when the internal PE of the computing unit processes the task; is the communication cost generated when the internal PE of the computing unit processes the task.

[0017] Furthermore, the communication cost is obtained through the following formula:

[0018]

[0019] Among them, is the communication cost generated when the internal PE of the computing unit processes the task; w a (t) is the computing load of the a-th task at the t-th moment, q a is the computing load per unit data volume when processing this task; u a (t) is the data volume of the a-th task at the t-th moment.

[0020] Furthermore, the computing cost is obtained through the following formula:

[0021]

[0022] Among them, is the computing cost generated when the internal PE of the computing unit processes the task, is the cost required for occupying the unit computing resource of the computing unit; d ae (t) is the time required for the computing unit to process the a-th task; f ae (t) is the computing resource allocated by the computing unit to the buffer to which the a-th task belongs.

[0023] Further, the time required for the arithmetic unit to process the ath task is calculated by the following formula:

[0024]

[0025] where d ae (t) is the time required for the arithmetic unit to process the ath task; w a (t) is the computational load of the ath task at the t-th moment; f ae (t) is the computing resource allocated by the arithmetic unit to the buffer to which the ath task belongs; χ ae is the space consumed for data operation.

[0026] Further, the computing resource f ae (t) allocated by the arithmetic unit to the buffer to which the ath task belongs satisfies the following formula:

[0027]

[0028] where f ae (t) is the computing resource allocated by the arithmetic unit to the buffer to which the ath task belongs; χ ae is the space consumed for data operation of the ath task, and F e is the total computing resource of the arithmetic unit.

[0029] Further, the types of tasks to be processed include time-insensitive requests and time-sensitive requests.

[0030] Further, the delay cost calculation formula includes a first calculation formula as follows:

[0031]

[0032] where is the delay cost generated when the internal PE of the arithmetic unit processes the task, d ee′ is the time consumed for data upload, d ae (t) is the time required for the arithmetic unit to process the ath task, and χ ee′ is the space required for data upload and temporary storage.

[0033] Further, the time-sensitive request is attached with a special task i, and the delay cost calculation formula further includes a second calculation formula as follows:

[0034]

[0035] where is the delay cost generated when the internal PE of the arithmetic unit processes the task, 2d ie (t) is the total time required to process the special task i, and χ ieThe space required to process the special task i; χ ae The space required for the operation of the a-th task data.

[0036] Furthermore, the delay constraint model of the first calculation formula is as follows:

[0037] d ee′ +d ae χ ee′ ≤T m2 ;

[0038] The delay constraint model of the second calculation formula is as follows:

[0039] 2d ie +d ae χ ie χ ae ≤T m1 ;

[0040] The maximum delay constraint T max is as follows:

[0041] T max = max(T m1 , T m2 );

[0042] When the delay cost of the task is greater than the maximum delay constraint T max , the system automatically deletes the task.

[0043] A cable explosion-proof monitoring terminal, including a terminal main body, a memory, a processor disposed in the terminal main body, and a computer program stored in the memory. The processor executes the computer program to implement the steps of the method.

[0044] The present invention has the following advantages:

[0045] Based on the task allocation mechanism of the timing pipeline, this application can divide multiple tasks such as three physical fields plus partial discharge electromagnetic signals into single tasks according to the time sequence, so as to achieve processing under single-core hardware resources, solve the processing problem of FPGA when multiple early warnings or data upload tasks occur concurrently, and contribute to the miniaturization of the cable explosion-proof monitoring terminal, so as to achieve better programming effects with less space.

[0046] This application calculates the calculation cost, delay cost, and communication cost generated when the PE inside the operation unit processes tasks, with high accuracy, and can also more carefully consider the energy consumption generated by calculation resource communication information, etc., and has high reliability.

[0047] Other features and advantages of the present invention will be disclosed in detail in the following specific embodiments and drawings. Description of the Drawings

[0048] Figure 1 Schematic diagram of the process of the present invention Figure 1 ;

[0049] Figure 2 Schematic diagram of the process of the present invention Figure 2 。 Detailed implementation manners

[0050] In order to better understand the purpose, structure and function of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings.

[0051] A multi-task resource optimization method applied to a cable explosion-proof monitoring terminal, as Figure 1 shown, includes the following steps:

[0052] S1: Obtain the information of the task to be processed;

[0053] S2: Obtain the type of the task to be processed;

[0054] S3: Sort the task to be processed according to the information of the task to be processed and the type of the task to be processed.

[0055] The information of the task to be processed in S1 is the information required in the subsequent sorting process. In this embodiment, the change of the data volume processed by each PE inside the FPGA neural network operation unit can be measured in real time by means of register reading, and the fluctuation of the PE data volume satisfies the normal distribution and they are independent of each other, and the relative positions between PEs are known; the calculation load of the unit data volume of each data processing task is known, and the processing timing logic of the service is known; the time for the controller to allocate computing resources for data buffering inside the operation unit can be ignored, and the creation and destruction of the buffer can be completed in real time.

[0056] In this embodiment, the type of the task to be processed in S2 can be divided into time-sensitive requests TSDP and time-insensitive requests NTSDP; as Figure 2 shown, the time-sensitive requests TSDP include T1D3, T2D1, T3D2 and T4D1, and the time-insensitive requests NTSDP include D1, D2, D3 and D4. Among them, the priority of the time-sensitive requests TSDP is higher than that of the time-insensitive requests NTSDP. The time-sensitive requests TSDP and time-insensitive requests NTSDP can be classified according to the differences in data types, spatio-temporal characteristics, etc. of the three physical fields of the 10 kV cable operation state plus the partial discharge electromagnetic small signal in the power distribution network. For example, the types of tasks to be processed include alarm tasks and data upload tasks. The alarm tasks are classified as time-sensitive requests TSDP, and the data upload tasks are classified as time-insensitive requests NTSDP.

[0057] Sorting the tasks to be processed according to the task information to be processed and the type of the tasks to be processed in S3 includes the following steps:

[0058] Sort at least one type of tasks to be processed by optimizing the model objective function;

[0059] The optimized model objective function is as follows:

[0060]

[0061] Among them, C(t) is the total cost of task processing by the computing unit at the t-th moment; e is the number of the internal PE of the computing unit, ranging from 1 to K, and K is the total number of computing units; is the computing cost generated when the internal PE of the computing unit processes the task; is the delay cost generated when the internal PE of the computing unit processes the task; is the communication cost generated when the internal PE of the computing unit processes the task. In this embodiment, the t-th moment is the current moment, that is, the moment when the current multiple tasks to be processed are input into the FPGA.

[0062] The processing of data processing tasks requires occupying the computing resources and communication resources of the neural network computing unit and generating energy consumption. Since the computing loads of different types of tasks change with the amount of data processed, these changes have relatively large differences in the time dimension and the space dimension, resulting in changes in the computing resources and communication resources occupied during the data processing and transmission processes. In this application, the total cost of task processing by the computing unit at the t-th moment is used to sort the priorities of each task from small to large according to the value of the total cost, which helps to quickly process the tasks and save the energy consumption of processing.

[0063] Assume that when the neural network computing unit is operating in a steady state, the data processing tasks input from the front side are periodic, and the cycle of one trigger is from T0 to T7, T0 is the start time of the data processing task, and T7 is the end time of the task. The optimized model objective function in this embodiment can sort when multiple tasks appear at the start time, or can sort when one or more new tasks appear at the middle time. In this embodiment, when one or more new tasks appear at the middle time, all the ongoing tasks are sorted through the optimized model objective function.

[0064] In this embodiment, the communication cost is obtained through the following formula:

[0065]

[0066] Among them, w a (t) is the computing load of the a-th task in the PE numbered e at the t-th moment, representing the communication cost; q aThe computational load per unit data volume when processing this type of task; u a (t) is the data volume of the a-th task at the t-th moment.

[0067] Specifically, the communication cost is the computational load of the a-th task at the t-th moment, with the unit of ms, referring to the generation time of the task data frame. q a It can be calculated according to the type of the a-th task, which is prior art and will not be elaborated in this application. When an alarm task appears in the system, during the task reporting process, it is necessary to edit and pack the data frame according to the protocol. The data frame length is the data volume, with the unit of byte; the length of the data frame is specified by the communication protocol. For example, in the MODBUS communication protocol, the maximum length of a normal working data frame is 253 bytes, depending on the transmitted data volume.

[0068] The characteristic model expression of the a-th task is as follows:

[0069] θ a =[p a , w a (t), T a ;

[0070] Among them, θ a represents the characteristic model of the a-th task, p a represents the type of the a-th task; w a (t) is the communication cost of the a-th task at the t-th moment; T a is the delay constraint of the a-th task, which can be set according to the type of the task. For example, the delay constraint of an alarm task is generally 50 ms; the time of w a (t) cannot exceed the time of the delay constraint, otherwise there is a risk of task timeout. a ∈ [1, m], where m is the total number of tasks, and a is an integer from 1 to m.

[0071] The computational cost is obtained through the following formula:

[0072]

[0073] Among them, is the computational cost generated when the PE inside the arithmetic unit processes the task, is the cost required for occupying the unit computational resource of the arithmetic unit; d ae (t) is the time required for the arithmetic unit to process the a-th task; f ae (t) is the computational resource allocated by the arithmetic unit to the buffer to which the a-th task belongs.

[0074] The time required for the arithmetic unit to process the a-th task can be calculated through the following formula:

[0075]

[0076] Among them, d ae (t) is the time required for the operation unit to process the a-th task; w a (t) is the computing load of the a-th task at the t-th moment; f ae (t) is the computing resource allocated by the operation unit to the buffer to which the a-th task belongs; χ ae is the space consumed for data operation, which is determined by the data size. a ∈ [1, m], e ∈ [1, K].

[0077] Among them, the computing resource f ae (t) allocated by the operation unit to the buffer to which the a-th task belongs satisfies the following formula:

[0078]

[0079] Among them, f ae (t) is the computing resource allocated by the operation unit to the buffer to which the a-th task belongs; χ ae is the space consumed for data operation of the a-th task, F e is the total computing resource of the operation unit.

[0080] In this embodiment, the delay cost calculation formula of the time-insensitive request NTSDP is as follows:

[0081]

[0082] Among them, is the delay cost generated when the internal PE of the operation unit processes the task, d ee′ is the time consumed for data upload, d ae (t) is the time required for the operation unit to process the a-th task, χ ee′ is the space required for data upload and temporary storage;

[0083] The delay constraint model of NTSDP is as follows:

[0084] d ee′ +d ae χ ee′ ≤T m2 ; among them, T m2 is the delay constraint of NTSDP, which is set according to the requirements of tasks of the NTSDP type.

[0085] In this embodiment, the time-sensitive request TSDP is attached with a special task i. Specifically, the special task i is a handshake task and requires a round trip.

[0086] In this embodiment, the delay cost calculation formula of the time-sensitive request TSDP is as follows:

[0087]

[0088] Among them, is the delay cost generated when the PE inside the operation unit processes tasks, and d ie (t) is the time required to process the special task i. In this embodiment, since it is a round trip, it is twice the time, χ ie is the space consumed for handshake task data encapsulation; χ ae is the space consumed for the operation of the a-th task data.

[0089] The delay constraint model of TSDP is as follows:

[0090] 2d ie +d ae χ ie χ ae ≤T m1 ;

[0091] Among them, T m1 is the delay constraint of the TSDP task, which is set according to the requirements of the TSDP type task.

[0092] In this embodiment, it also includes the maximum delay constraint T max ;

[0093] T max = max(T m1 , T m2 );

[0094] The purpose of setting the maximum delay constraint is to exclude incorrect tasks. When the delay cost of a task is very long, it can be determined that this task is an incorrect task, such as an error message reported due to hardware influence, for example, an error message caused by a Bug. When the delay cost of the task is greater than the maximum delay constraint T max , the system automatically deletes the task.

[0095] Since many data of the cable operating state are sensitive to time, the processing task has a maximum delay tolerance, that is, the sum of the delays of each successive part and the maximum delay of each parallel part on the processing timing logic chain does not exceed the delay tolerance.

[0096] A cable explosion-proof monitoring terminal includes a terminal main body, a memory, a processor disposed in the terminal main body, and a computer program stored on the memory. The processor executes the computer program to implement the steps of the foregoing method.

[0097] Embodiment 2

[0098] The difference between this embodiment and the first embodiment is that the time-sensitive requests are sorted from the smallest to the largest according to the delay cost.

[0099] Embodiment Three

[0100] The difference between this embodiment and the first embodiment is that the time-insensitive requests are arranged from the smallest to the largest according to the communication cost.

[0101] In addition, it should be understood that although this specification is described according to the implementation manners, not every implementation manner only includes an independent technical solution. This narrative manner of the specification is only for clarity. Those skilled in the art should regard the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other implementation manners that can be understood by those skilled in the art.

Claims

1. A multi-task resource optimization method applied to a cable explosion-proof monitoring terminal, characterized in that: The steps include: Get information about pending tasks; Get the type of task to be processed; Sort the pending tasks according to the pending task information and pending task types; The step of sorting the tasks to be processed according to the information of the tasks to be processed and the types of the tasks to be processed includes: sorting at least one type of pending tasks by optimizing a model objective function; The optimization model objective function is as follows: Where C(t) is the total cost of the computing unit for processing the task at time t; e is the number of the PE inside the computing unit, ranging from 1 to K, and K is the total number of computing units; It is the computational cost generated when the PE inside the computing unit processes tasks; The delay cost incurred when the PE inside the computing unit processes tasks; It is the communication cost generated when the PE inside the computing unit processes tasks.

2. The multi-task resource optimization method applied to the cable explosion-proof monitoring terminal according to claim 1 is characterized in that: The communication cost is obtained by the following formula: in, The communication cost generated when the PE inside the computing unit processes tasks; w a (t) is the computational load of the ath task at time t, q a is the computational load per unit of data volume when processing tasks; u a (t) is the amount of data for the a-th task at the t-th moment.

3. The multi-task resource optimization method applied to the cable explosion-proof monitoring terminal according to claim 2 is characterized in that: The calculation cost is obtained by the following formula: in, It is the computational cost generated when the PE inside the computing unit processes tasks. The cost required to calculate the resources for each unit of computing unit; d ae (t) is the time required for the computing unit to process the ath task; f ae (t) is the computing resource allocated by the computing unit to the buffer belonging to the a-th task.

4. The multi-task resource optimization method applied to the cable explosion-proof monitoring terminal according to claim 3 is characterized in that: The time required for the computing unit to process the ath task is calculated by the following formula: Among them, d ae (t) is the time required for the computing unit to process the ath task; w a (t) is the computational load of the ath task at the tth time; f ae (t) is the computing resource allocated by the computing unit to the buffer of the ath task; x ae The space required for data operations.

5. The multi-task resource optimization method applied to the cable explosion-proof monitoring terminal according to claim 3 or 4, characterized in that: The computing unit allocates the computing resources f to the buffer belonging to the ath task ae (t) satisfies the following formula: Among them, f ae (t) is the computing resource allocated by the computing unit to the buffer belonging to the a-th task; ae is the space consumed by the a-th task data operation, F e is the total amount of computing resources of the computing unit.

6. The multi-task resource optimization method applied to the cable explosion-proof monitoring terminal according to any one of claims 1 to 4, characterized in that: The types of tasks to be processed include time-insensitive requests and time-sensitive requests.

7. The multi-task resource optimization method applied to the cable explosion-proof monitoring terminal according to claim 6 is characterized in that: The delay cost calculation formula includes a first calculation formula, as follows: in, is the delay cost generated when the PE inside the computing unit processes tasks, d ee′ The time consumed for data upload, d ae (t) is the time required for the computing unit to process the ath task, ee′ The space required for temporarily storing data uploads.

8. The multi-task resource optimization method applied to the cable explosion-proof monitoring terminal according to claim 7 is characterized in that: The time-sensitive request is accompanied by a special task i, and the delay cost calculation formula also includes a second calculation formula, as follows: in, is the delay cost generated when the PE inside the computing unit processes tasks, 2d ie (t) is the total time required to process a specific task i, χ ie The space required to process special task i; χ ae The space consumed by the a-th task data operation.

9. The multi-task resource optimization method applied to the cable explosion-proof monitoring terminal according to claim 8 is characterized in that: The delay constraint model of the first calculation formula is as follows: d ee′ +d ae x ee′ ≤T m2 ; The delay constraint model of the second calculation formula is as follows: 2d ie +d ae x ie x ae ≤T m1 ; Maximum delay constraint T max as follows: T max =max(T m1 ,T m2 ); When the delay cost of the task is greater than the maximum delay constraint T max The system automatically deletes the task.

10. A cable explosion-proof monitoring terminal, comprising a terminal body, a memory arranged in the terminal body, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 9.

Images