CN115840640A - Operation acceleration system and method for power edge intelligent algorithm - Google Patents

Abstract

The invention discloses an operation acceleration system and method for an intelligent algorithm of a power edge, which comprises the following steps: determining target operation data to be calculated, and splitting the target operation data into a plurality of data groups according to different operation nodes; determining scheduling levels corresponding to the operation submodules respectively according to the historical resource loss and the calculation amount required for executing the corresponding operation tasks and performing data scheduling distribution; and monitoring each operation submodule, and carrying out operation acceleration according to a preset acceleration calculation rule when the operation time and the resource consumption of the corresponding target operation submodule exceed preset thresholds. The operation acceleration system and method for the power edge intelligent algorithm ensure that each operation task can be allocated to a proper operation sub-module to be effectively implemented, improve the operation efficiency, accelerate the operation when the operation timeout is determined, improve the capability and efficiency of the system for processing complex operation and reduce the power consumption overhead.

Description

A computing acceleration system and method for power edge intelligent algorithm

technical field

The invention relates to the technical field of operation acceleration, in particular to an operation acceleration system and method for power edge intelligent algorithms.

Background technique

With the entry of the information age, the application of power edge intelligent algorithms in power systems has become more and more common. The running algorithm model is accelerated to avoid reducing the execution efficiency of the operation.

At present, the calculation acceleration methods adopted include introducing calculation acceleration components to perform complex calculations. Although the above methods can improve the ability and efficiency of processing complex calculations to a certain extent, for the large amount of calculation data and how to effectively schedule computing resources, existing The method does not propose an effective solution, and there is a problem of poor operation effect.

Contents of the invention

The purpose of this section is to outline some aspects of embodiments of the invention and briefly describe some preferred embodiments. Some simplifications or omissions may be made in this section, as well as in the abstract and titles of this application, to avoid obscuring the purpose of this section, the abstract and titles, and such simplifications or omissions should not be used to limit the scope of the invention.

In view of the above problems, the present invention has been proposed.

Therefore, the technical problem to be solved by the present invention is: the existing electric power edge computing method has the problem of poor operation effect on large amount of computing data and complex computing, and the problem of how to effectively schedule the optimization of computing resources.

In order to solve the above-mentioned technical problems, the present invention provides the following technical solutions: an operation acceleration method for power edge intelligent algorithms, including:

determining the target operation data to be calculated, and splitting the target operation data into multiple data packets according to different computing nodes;

According to the amount of historical resource consumption and the amount of calculation required to perform corresponding calculation tasks, determine the corresponding scheduling level of each calculation sub-module and perform data scheduling and allocation;

Each operation sub-module is monitored, and when it is determined that the operation time and resource consumption of the corresponding target operation sub-module exceed a preset threshold, the operation is accelerated according to a preset acceleration calculation rule.

As a preferred solution of the calculation acceleration method for power edge intelligent algorithms in the present invention, wherein: the determination of the target operation data to be calculated includes: obtaining the initial operation data, and according to the preset traversal method, the acquired During the traversal process, process the traversed initial operation data according to the preset data processing method to determine the target operation data.

As a preferred solution of the calculation acceleration method for power edge intelligent algorithms in the present invention, wherein: the traversal methods include three traversal methods: recursive, non-recursive and hierarchical; At least one of an invalid data filtering method for invalid data of keywords, an information completion method for filling missing data, and a noise data processing method is set.

As a preferred solution of the computing acceleration method for power edge intelligent algorithms described in the present invention, wherein: each of the data packets corresponds to the corresponding computing tasks covered by the corresponding computing nodes, and includes executing the computing tasks The required operational data.

As a preferred solution of the calculation acceleration method for the power edge intelligent algorithm described in the present invention, wherein: the calculation node means that the complete calculation stage corresponding to the target operation data is composed of multiple calculation nodes, and each calculation node There is a sequence of execution among them.

As a preferred solution of the calculation acceleration method for power edge intelligent algorithms described in the present invention, wherein: the data scheduling allocation includes: processing the data packets to obtain corresponding local calculation results; according to the order of the calculation nodes The execution sequence is to integrate the partial operation results output by the corresponding operation sub-modules to obtain the overall operation results corresponding to the complete operation stage.

As a preferred solution of the computing acceleration system for power edge intelligent algorithms described in the present invention, wherein: the data acquisition module is used to determine the target operation data to be calculated, and according to different computing nodes, the target operation The data is split into multiple data packets;

A parallel computing module, including a plurality of operation sub-modules and a main control module connected to each of the operation sub-modules, wherein: the main control module is used to perform data scheduling and distribution and monitor each operation sub-module to determine whether operation is required accelerate.

As a preferred scheme of the computing acceleration system for power edge intelligent algorithms described in the present invention, wherein: the computing sub-module is used to process the data packets allocated via the main control module to obtain the corresponding local operation result;

The main control module is also used to integrate the local operation results output by the corresponding operation sub-modules according to the execution sequence of the operation nodes, so as to obtain the overall operation results corresponding to the complete operation stage.

As a preferred solution of the computing acceleration system for power edge intelligent algorithms described in the present invention, wherein: the computing accelerating system further includes a performance optimization module, which is used to optimize the The execution data synchronously generated in the parallel computing task is segmented to obtain sub-execution data blocks corresponding to each sub-task.

As a preferred solution of the computing acceleration system for power edge intelligent algorithms described in the present invention, wherein: the performance optimization module is also used to respectively determine the performance consumption generated when performing parallel processing operations on corresponding sub-execution data blocks Information; according to the performance consumption information, perform fault operation analysis, and when it is determined that there is a hidden danger of fault operation, feed back the corresponding performance optimization strategy to the main control end, so as to trigger the main control end to perform system optimization according to the received performance optimization strategy .

Beneficial effects of the present invention: the computing acceleration system and method for power edge intelligent algorithms provided by the present invention analyze the parallel computing tasks and identify the scheduling level corresponding to each computing sub-module before data scheduling and allocation , so as to ensure that each computing task can be assigned to a suitable computing sub-module for effective implementation, which improves the computing efficiency. In addition, the main control module monitors the calculation time and resource consumption generated by each of the calculation sub-modules when performing calculation tasks, and when it is determined that there is a calculation timeout, the calculation is accelerated according to the preset acceleration calculation rule , improve the ability and efficiency of the system to process complex operations, and reduce power consumption.

Description of drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the accompanying drawings that need to be used in the description of the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For Those of ordinary skill in the art can also obtain other drawings based on these drawings without any creative effort. in:

Fig. 1 is a method flowchart of a computing acceleration system and method for power edge intelligent algorithms provided by the first embodiment of the present invention;

FIG. 2 is a schematic diagram of the system structure of a computing acceleration system and method for power edge intelligent algorithms provided by the second embodiment of the present invention;

Fig. 3 is a comparison graph of 48h load forecast results and historical real data of a computing acceleration system and method for power edge intelligent algorithms provided by the third embodiment of the present invention.

Detailed ways

In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation modes of the present invention will be described in detail below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of them. Example. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative efforts shall fall within the protection scope of the present invention.

In the following description, a lot of specific details are set forth in order to fully understand the present invention, but the present invention can also be implemented in other ways different from those described here, and those skilled in the art can do it without departing from the meaning of the present invention. By analogy, the present invention is therefore not limited to the specific examples disclosed below.

Second, "one embodiment" or "an embodiment" referred to herein refers to a specific feature, structure or characteristic that may be included in at least one implementation of the present invention. "In one embodiment" appearing in different places in this specification does not all refer to the same embodiment, nor is it a separate or selective embodiment that is mutually exclusive with other embodiments.

The present invention is described in detail in conjunction with schematic diagrams. When describing the embodiments of the present invention in detail, for the convenience of explanation, the cross-sectional view showing the device structure will not be partially enlarged according to the general scale, and the schematic diagram is only an example, which should not limit the present invention. scope of protection. In addition, the three-dimensional space dimensions of length, width and depth should be included in actual production.

At the same time, in the description of the present invention, it should be noted that the orientation or positional relationship indicated by "upper, lower, inner and outer" in the terms is based on the orientation or positional relationship shown in the accompanying drawings, and is only for the convenience of describing the present invention. The invention and the simplified description do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operate in a specific orientation, and thus should not be construed as limiting the present invention. In addition, the terms "first, second or third" are used for descriptive purposes only, and should not be construed as indicating or implying relative importance.

Unless otherwise specified and limited in the present invention, the term "installation, connection, connection" should be understood in a broad sense, for example: it can be a fixed connection, a detachable connection or an integrated connection; it can also be a mechanical connection, an electrical connection or a direct connection. A connection can also be an indirect connection through an intermediary, or it can be an internal communication between two elements. Those of ordinary skill in the art can understand the specific meanings of the above terms in the present invention in specific situations.

Example 1

Referring to Fig. 1, it is an embodiment of the present invention, which provides a calculation acceleration method for power edge intelligent algorithm, including:

S1: Determine the target operation data to be calculated, and split the target operation data into multiple data packets according to different computing nodes;

Furthermore, each data group corresponds to a computing task covered by a corresponding computing node, and includes computing data required to execute the computing task.

Furthermore, determine the target operation data to be calculated, including:

The data acquisition module acquires initial operation data, and traverses the acquired initial operation data according to a preset traversal method. The traversal modes include three traversal modes: recursive, non-recursive and hierarchical.

During the traversal process, the data acquisition module processes the traversed initial operation data according to the preset data processing method to determine the target operation data, wherein the data processing method includes removing invalid data containing preset keywords. At least one of an invalid data filtering method of data, an information complementing method of supplementing missing data, and a noise data processing method.

It should be noted that when constructing the traversal rules, the data acquisition module can traverse from the data acquired at the beginning to the data acquired at the last time point in sequence based on the order of data acquisition time. Or according to the data type and the degree of correlation between the data, for example: perform unified traversal for multiple target data with a high degree of correlation; for the same type of data, perform unified traversal. or other data characteristics to determine the traversal rule, which is not limited in this embodiment of the present application.

The data acquisition module can store the acquired initial operation data in the storage addresses corresponding to different storage identifiers in the preset data table, and then traverse multiple pieces of data stored in the data table according to the storage identifiers.

Furthermore, the complete operation stage corresponding to the target operation data is composed of multiple operation nodes, and each operation node has a sequence of execution.

S2: According to the amount of historical resource consumption and the amount of calculation required to perform corresponding calculation tasks, determine the corresponding scheduling level of each calculation sub-module and perform data scheduling and allocation;

It should be noted that by analyzing parallel computing tasks and identifying the scheduling level corresponding to each computing sub-module before data scheduling and allocation, it is ensured that each computing task can be assigned to a suitable computing sub-module for execution. Effective implementation improves operational efficiency.

S3: Monitor each operation sub-module, and when it is determined that the operation time and resource consumption of the corresponding target operation sub-module exceed a preset threshold, perform operation acceleration according to a preset acceleration calculation rule;

Furthermore, the operation sub-module processes the data packets allocated via the main control module to obtain corresponding local operation results.

It should be noted that before processing the assigned data group, the operation sub-module can first query the historical operation records of the entire system to see whether the same operation has been done in the past, and if so, determine the corresponding operation according to the historical operation records. Local calculation results, in this way, can improve calculation efficiency and resource utilization while avoiding repeated calculation operations.

It should also be noted that the local operation results calculated by the operation sub-module will be uniformly cached in the preset storage space, so as to facilitate the subsequent integrated call and the search of historical operation records.

Furthermore, in terms of increasing the computing speed, for example, it aims at the problem of low data processing efficiency caused by a large amount of computing and long running time. The operation sub-module can also segment the obtained data group based on the preset operation key field; the data segmentation method can be horizontal segmentation or vertical segmentation. Afterwards, calculations are performed on each data segment included in the segmented data, so as to improve data processing efficiency.

Furthermore, the main control module integrates the local operation results output by the corresponding operation sub-modules according to the execution sequence of the operation nodes, so as to obtain the overall operation results corresponding to the complete operation stage.

It should be noted that the main control module monitors the calculation time and resource consumption generated by each of the calculation sub-modules when performing calculation tasks, and when it is determined that there is a calculation timeout, the calculation is performed according to the preset accelerated calculation rules. Computing acceleration, improving the system's ability and efficiency in processing complex operations, and reducing power consumption. The main control module can also optimize the operation speed of each operation sub-module by increasing the main frequency and adjusting the allocation of computing resources in a timely manner, so as to improve the working efficiency of the system.

It should be noted that through the analysis of parallel computing tasks, it is possible to know in a timely manner whether each computing sub-module has a hidden danger of faulty operation and whether it can support the current computing processing, and optimize performance in time to avoid the occurrence of outages, so as to ensure that each computing The tasks can be effectively implemented, and the computing efficiency is improved.

Example 2

Referring to FIG. 2 , an embodiment of the present invention provides an operation acceleration system for intelligent algorithms at the power edge, including: a data acquisition module 100 and a parallel computing module 200 .

Furthermore, the data acquisition module 100 is configured to determine the target operation data to be calculated, and split the target operation data into multiple data packets according to different computing nodes, wherein each data packet corresponds to The computing tasks covered in the corresponding computing nodes include the computing data needed to execute the computing tasks; the data acquisition module is also used to obtain initial operating data, and perform each acquired Item initial operation data to traverse.

The parallel computing module 200 includes a plurality of computing sub-modules 201 and a main control module 202 connected to each of the computing sub-modules.

Wherein, the main control module 202 is used to determine the scheduling level corresponding to each of the operation sub-modules 201 according to the amount of historical resource consumption and the amount of calculation required to perform the corresponding calculation task, and perform data scheduling and allocation according to the scheduling level; The main control module 202 is also used to monitor the computing time and resource consumption generated by each computing sub-module 201 when performing computing tasks, and determine the computing time and resource consumption of the corresponding target computing sub-module 201. When the preset threshold is exceeded, the calculation is accelerated according to the preset acceleration calculation rules.

Furthermore, the system also includes a performance optimization module 300. When the performance optimization module 300 determines that there are multiple subtasks in the parallel computing task, it performs a segmentation operation on the execution data synchronously generated in the parallel computing task to obtain the subtasks The corresponding sub-execution data blocks respectively.

The performance optimization module 300 respectively determines performance consumption information generated when performing parallel processing operations on corresponding sub-execution data blocks. Wherein, the performance consumption information includes cache information, occupancy of computing resources, for example, occupancy of CPU, occupancy of content resources, consumption of hard disk, and the like.

Furthermore, according to the performance consumption information, the failure operation analysis is performed, and when it is determined that there is a hidden danger of failure operation, the corresponding performance optimization strategy is fed back to the main control terminal, so as to trigger the main control terminal to perform the operation according to the received performance optimization strategy System Optimization.

The performance optimization module 300 can also combine the resource consumption trend within a certain period of time to determine whether each operation sub-module has excess performance, and if so, can request the main control module 202 to reduce the corresponding computing resource configuration. And when it is determined that the performance of the corresponding operation sub-module is not enough to support the current operation processing, the main control module may be requested to allocate more computing resources to the operation sub-module, so as to avoid the occurrence of operation interruption.

Example 3

The following is an embodiment of the present invention, which provides a computing acceleration system and method for power edge intelligent algorithms. In order to verify the beneficial effects of the present invention, a scientific demonstration is carried out through simulation and comparison experiments.

The data load value of a charging station in Guiyang City in December 2020 was selected as the data source of this experiment for simulation processing, and combined with historical weather conditions, the 48h load forecast was performed using power edge computing.

For the same execution task, the calculation acceleration system and method of the power edge intelligent algorithm used in this simulation experiment are compared with the traditional calculation acceleration system and method, and the experimental results are shown in the following table:

traditional method This method time required to perform tasks 0.637s 0.475s efficiency 47.5% 81.5% power consumption 7.2W 1.2W

It can be seen that the operation acceleration system adopting the power edge intelligent algorithm of the present invention greatly improves the acceleration effect, doubles the operation time, improves the efficiency by more than 70%, and greatly reduces the power consumption.

As can be seen in Figure 3, the curve trend obtained by comparing the 48h load prediction results obtained by this method with the historical real data is roughly the same, which proves the accuracy and effectiveness of this method.

It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention without limitation, although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be carried out Modifications or equivalent replacements without departing from the spirit and scope of the technical solution of the present invention shall be covered by the claims of the present invention.

Claims (10)

1. An operation acceleration method for a power edge intelligent algorithm is characterized by comprising the following steps:

determining target operation data to be calculated, and splitting the target operation data into a plurality of data groups according to different operation nodes;

determining scheduling levels corresponding to the operation submodules respectively according to the historical resource loss and the calculation amount required for executing the corresponding operation tasks and performing data scheduling distribution;

and monitoring each operation submodule, and carrying out operation acceleration according to a preset acceleration calculation rule when the operation time and the resource consumption of the corresponding target operation submodule exceed preset thresholds.

2. The operation acceleration method for the power edge intelligence algorithm according to claim 1, characterized in that: the determining target operation data of a calculation to be performed comprises: acquiring initial operation data, and traversing each item of acquired initial operation data according to a preset traversal mode; and in the traversing process, processing the traversed initial operation data according to a preset data processing mode to determine target operation data.

3. The operation acceleration method for the power edge intelligence algorithm according to claim 2, characterized in that: the traversal modes comprise a recursive mode, a non-recursive mode and a hierarchical traversal mode; the data processing mode comprises at least one of an invalid data filtering mode for eliminating invalid data containing preset keywords, an information complementing mode for complementing missing data and a noise data processing mode.

4. The operation acceleration method for the power edge intelligence algorithm according to claim 1, characterized in that: each data packet corresponds to the operation task correspondingly covered in the corresponding operation node and comprises operation data required to be used for executing the operation task.

5. An operation acceleration method for a power edge intelligence algorithm according to claim 1 or 4, characterized by: the operation nodes mean that the complete operation stage corresponding to the target operation data is composed of a plurality of operation nodes, and the operation nodes have a sequential execution sequence.

6. The operation acceleration method for the power edge intelligence algorithm according to claim 1, characterized in that: the data scheduling assignment includes: processing the data packet to obtain a corresponding local operation result; and integrating the local operation results output by the corresponding operation sub-modules according to the sequential execution sequence of the operation nodes to obtain an overall operation result corresponding to the complete operation stage.

7. An operation acceleration system for a power edge intelligence algorithm, comprising:

the data acquisition module (100) is used for determining target operation data to be calculated, and dividing the target operation data into a plurality of data groups according to different operation nodes;

a parallel computing module (200) comprising a plurality of operational sub-modules (201) and a master control module (202) connected to each of said operational sub-modules, wherein: the main control module (202) is used for carrying out data scheduling distribution and monitoring each operation submodule (201) to judge whether operation acceleration is needed.

8. The operation acceleration system for a power edge intelligence algorithm of claim 7, characterized by: the operation submodule is used for processing the data packets distributed by the main control module to obtain corresponding local operation results;

the main control module is further configured to integrate the local operation results output by the corresponding operation sub-modules according to the execution sequence of the operation nodes, so as to obtain an overall operation result corresponding to the complete operation stage.

9. The operation acceleration system for a power edge intelligence algorithm of claim 7, characterized by: the operation acceleration system further comprises a performance optimization module (300) which is used for performing segmentation operation on the execution data synchronously generated in the parallel operation task when the parallel operation task is determined to have a plurality of subtasks so as to obtain the sub execution data blocks respectively corresponding to the subtasks.

10. The system for operational acceleration of a power edge intelligence algorithm of claim 9, wherein: the performance optimization module (300) is further configured to determine performance consumption information generated when performing parallel processing operations on corresponding sub-execution data blocks, respectively; and performing fault operation analysis according to the performance consumption information, and feeding back a corresponding performance optimization strategy to the main control end when determining that the fault operation hidden danger exists, so as to trigger the main control end to perform system optimization according to the received performance optimization strategy.

Images