Converter station scene computing power arrangement algorithm equipment

Abstract

The invention discloses computing power arrangement algorithm equipment for a converter station scene, and relates to the technical field of computing power arrangement algorithms. Comprising a service layer, a control layer, a resource layer and an arrangement management layer, the service layer is internally provided with resource information processing and task management, the control layer is internally provided with resource information collection, resource allocation and network connection scheduling, and the resource layer is internally provided with computing resources, storage resources, network resources and spectrum resources. For highly heterogeneous and high-frequency and short-time change time sequence characteristics of computing power requirements in a converter station, a distribution function of sequential arrival of business objects is estimated by counting arrival probabilities of different types of sequential business objects, and the time sequence computing power requirements of a system and proportions of various business objects are estimated according to the distribution function. And according to characteristics of transmission topology and computing power distribution, computing power resources are dynamically distributed and optimized by using a computing power slicing strategy based on deep reinforcement learning.

Description

Technical Field

[0001] This invention relates to the field of computing power orchestration algorithm technology, specifically to a computing power orchestration algorithm device for a converter station scenario. Background Technology

[0002] A compute network (CNN) is an emerging computing model designed to improve computational efficiency and performance by leveraging massively parallel computing resources. A CNN typically consists of multiple compute nodes, each containing one or more processors and memory. These nodes are connected via a high-speed network to form a compute graph, where various computational tasks are executed. The core idea of ​​a CNN is to connect distributed compute nodes. A CNN can dynamically and promptly perceive user needs and multi-dimensional resources such as applications, network, computing power, and storage. It collectively allocates compute and network resources, coordinates and schedules computational tasks, enabling applications to access geographically distributed compute resources on demand. A CNN is a network capable of intelligently perceiving, allocating, and scheduling compute resources, providing optimal user experience and utilization of compute and network resources based on the computational needs of each node.

[0003] The computational resource slicing strategy based on deep reinforcement learning (DRL) is a method that uses deep reinforcement learning algorithms to dynamically allocate and optimize computational resources. In this strategy, the DRL algorithm acts as a decision-maker, learning how to allocate computational, storage, and spectrum resources across different tasks to achieve the highest efficiency and performance.

[0004] To address the highly heterogeneous and high-frequency, short-term changing timing characteristics of computing power demand in converter stations, this paper estimates the distribution function of sequential arrival of service objects by statistically analyzing the arrival probabilities of different types of sequential service objects. Based on this distribution function, the system's timing computing power demand and the proportion of each type of service object are estimated. According to the characteristics of transmission topology and computing power distribution, deep learning methods are used to predict the timing computing power demand space within the construction scenario, and reinforcement learning is used to converge the optimal global computing power orchestration combination scheme. Summary of the Invention

[0005] This invention provides a computing power orchestration algorithm device for converter station scenarios to solve the problems in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a computing power orchestration algorithm device for a converter station scenario, comprising a service layer, a control layer, a resource layer, and an orchestration management layer. The service layer is internally configured with resource information processing and task management. The control layer is internally configured with resource information collection, resource allocation, and network connection scheduling. The resource layer is internally configured with computing resources, storage resources, network resources, and spectrum resources. The orchestration management layer is internally configured with computing power network orchestration, computing power modeling, and computing power OAM.

[0007] The resource allocation unit is internally equipped with a computing power router. A thermal grease layer is fixedly installed at the bottom of the computing power router. A heat-conducting plate is fixedly installed at the bottom of the thermal grease layer. A heat dissipation fin is fixedly installed at the bottom of the heat-conducting plate. One end of the heat dissipation fin penetrates through the top of the water-cooled box and extends into the interior of the water-cooled box. Coolant is installed inside the water-cooled box. A cooling plate is fixedly installed at the bottom of the water-cooled box. A cooling rod is fixedly installed at the top of the cooling plate. One end of the cooling rod penetrates through the bottom of the water-cooled box and extends into the interior of the water-cooled box. Support plates are fixedly installed on both sides of the top of the water-cooled box.

[0008] Motors are installed on both sides of the bottom of the water-cooled box. The output shaft of the motor is fixedly connected to a rotating shaft. One end of the rotating shaft passes through the bottom of the water-cooled box and extends into the interior of the water-cooled box. Stirring blades are fixedly installed on both sides of the rotating shaft. An air-cooled box is fixedly installed at the bottom of the water-cooled box. A cooling fan is installed inside the air-cooled box. Heat dissipation vents are opened on both sides of the air-cooled box. Dustproof nets are fixedly installed inside the heat dissipation vents. Mounting shells are fixedly installed on both sides of the water-cooled box.

[0009] Furthermore, the top of the support plate is provided with a threaded hole, and a screw is threadedly connected to the inner wall of the threaded hole. One end of the screw is rotatably connected to a pressure plate.

[0010] Furthermore, a rubber pad is fixedly provided at the bottom of the clamping plate, and the bottom of the rubber pad is in contact with the top of the computing power router.

[0011] Furthermore, a first electric push rod is fixedly installed inside the mounting housing. A brush plate is movably inserted into the telescopic end of the first electric push rod, and a groove adapted to the telescopic end of the first electric push rod is provided on the top of the brush plate.

[0012] Furthermore, a slot is provided on one side of the telescopic end of the first electric push rod, and a bolt is movably provided on the inner wall of the slot. A threaded hole adapted to the bolt is provided on one side of the brush plate, and the connection between the bolt and the brush plate is a threaded connection.

[0013] Furthermore, the connection between the first electric push rod and the bolt is a movable plug-in connection.

[0014] Furthermore, the interior of the air-cooled box is equipped with a dust collection hood, a first air pipe is fixedly installed at the bottom of the dust collection hood, a vacuum cleaner is fixedly installed at one end of the first air pipe, and a second air pipe is fixedly installed on one side of the vacuum cleaner.

[0015] Furthermore, a second electric push rod is fixedly installed on both sides of the bottom of the dust hood.

[0016] Furthermore, a temperature sensor is fixedly installed on one side of the bottom of the water-cooled box, and a PLC controller is installed on the other side of the bottom of the water-cooled box.

[0017] Furthermore, the connection between the screw and the support plate is a threaded connection.

[0018] Compared with existing technologies, this invention provides a computing power orchestration algorithm device for converter station scenarios, which has the following beneficial effects:

[0019] 1. The computing power orchestration algorithm equipment in this converter station scenario, taking into account the highly heterogeneous and high-frequency, short-term changing time-series characteristics of computing power demand in the converter station, estimates the distribution function of sequential arrival of service objects by statistically analyzing the arrival probability of different types of sequential service objects. Based on this distribution function, it estimates the system's time-series computing power demand and the proportion of each type of service object. Based on the characteristics of transmission topology and computing power distribution, it uses a computing power slicing strategy based on deep reinforcement learning to dynamically allocate and optimize computing power resources.

[0020] 2. The computing power orchestration algorithm equipment in this converter station scenario uses a thermal grease layer, heat-conducting plate, heat dissipation fins, water-cooled box, and coolant. The thermal grease layer, heat-conducting plate, and heat dissipation fins can conduct heat to the computing power router, transferring the heat of the computing power router to the water-cooled box. The coolant in the water-cooled box can absorb the heat of the computing power router, reduce the temperature of the computing power router, and ensure the stable operation of the computing power router, preventing high temperature conditions from affecting the stability and efficiency of the computing power router's resource allocation.

[0021] 3. The computing power orchestration algorithm equipment in this converter station scenario can cool the coolant in the water-cooled box by setting up a cooling chip, so that the coolant in the water-cooled box can always maintain a low temperature, thereby maintaining a good heat absorption effect of the coolant.

[0022] 4. The computing power orchestration algorithm equipment in this converter station scenario uses a motor shaft and a stirring blade. The motor drives the shaft and stirring blade to rotate, and the rotation of the stirring blade stirs the coolant in the water-cooled tank, keeping it in a flowing state, thereby achieving uniform cooling of the coolant.

[0023] 5. The computing power orchestration algorithm equipment in this converter station scenario is equipped with cooling fans, heat dissipation vents, and dust filters. The cooling fans can provide air cooling to the hot end of the cooling chip, allowing the heat from the hot end of the cooling chip to be quickly discharged from the heat dissipation vents, ensuring the normal operation of the cooling chip and preventing overheating. The dust filters inside the heat dissipation vents can intercept most of the dust in the external environment. Attached Figure Description

[0024] Figure 1 This is a schematic diagram of the structure of the present invention;

[0025] Figure 2 This is a schematic diagram of the computing power router structure of the present invention;

[0026] Figure 3 This is a schematic diagram of the internal structure of the computing power router of the present invention;

[0027] Figure 4 This is a schematic diagram of the water-cooled box structure of the present invention;

[0028] Figure 5 This is a schematic diagram of the internal structure of the mounting shell of the present invention;

[0029] Figure 6 This is a schematic diagram of the dust cover structure of the present invention.

[0030] In the diagram: 1. Service Layer; 101. Resource Information Processing; 102. Task Management; 2. Control Layer; 201. Resource Information Collection; 202. Resource Allocation; 203. Network Connection Scheduling; 204. Computing Power Router; 205. Thermal Grease Layer; 206. Heat Dissipation Plate; 207. Heat Sink; 208. Support Plate; 209. Screw; 210. Pressure Plate; 211. Rubber Pad; 3. Resource Layer; 301. Computing Resources; 302. Storage Resources; 303. Network Resources; 304. Spectrum Resources; 4. Orchestration and Management Layer; 401. Computing Power Network Orchestration; 02. Computational Modeling; 403. Computational OAM; 5. Water-cooled Box; 501. Coolant; 502. Cooling Chip; 503. Cooling Rod; 504. Temperature Sensor; 505. PLC Controller; 6. Motor; 601. Shaft; 602. Stirring Blade; 7. Air-cooled Box; 701. Cooling Fan; 702. Heat Dissipation Vent; 703. Dustproof Net; 8. Mounting Housing; 801. First Electric Actuator; 802. Brush Plate; 803. Bolt; 9. Dust Collection Hood; 901. First Air Pipe; 902. Vacuum Cleaner; 903. Second Air Pipe; 904. Second Electric Actuator. Detailed Implementation

[0031] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0032] Please see Figure 1-6 This invention discloses a computing power orchestration algorithm device for a converter station scenario, comprising a service layer 1, a control layer 2, a resource layer 3, and an orchestration management layer 4. The service layer 1 internally includes resource information processing 101 and task management 102. The control layer 2 internally includes resource information collection 201, resource allocation 202, and network connection scheduling 203. The resource layer 3 internally includes computing resources 301, storage resources 302, network resources 303, and spectrum resources 304. The orchestration management layer 4 internally includes computing power network orchestration 401, computing power modeling 402, and computing power OAM 403.

[0033] The resource allocation 202 is internally equipped with a computing power router 204. A thermal grease layer 205 is fixedly installed at the bottom of the computing power router 204. A heat-conducting plate 206 is fixedly installed at the bottom of the thermal grease layer 205. A heat dissipation fin 207 is fixedly installed at the bottom of the heat-conducting plate 206. One end of the heat dissipation fin 207 penetrates the top of the water-cooled box 5 and extends into the interior of the water-cooled box 5. Coolant 501 is installed inside the water-cooled box 5. A cooling chip 502 is fixedly installed at the bottom of the water-cooled box 5. A cooling rod 503 is fixedly installed at the top of the cooling chip 502. One end of the cooling rod 503 penetrates the bottom of the water-cooled box 5 and extends into the interior of the water-cooled box 5. Support plates 208 are fixedly installed on both sides of the top of the water-cooled box 5.

[0034] Motors 6 are installed on both sides of the bottom of the water-cooled box 5. The output shaft of the motor 6 is fixedly connected to a rotating shaft 601. One end of the rotating shaft 601 passes through the bottom of the water-cooled box 5 and extends into the interior of the water-cooled box 5. Stirring blades 602 are fixedly installed on both sides of the rotating shaft 601. An air-cooled box 7 is fixedly installed at the bottom of the water-cooled box 5. A cooling fan 701 is installed inside the air-cooled box 7. Heat dissipation vents 702 are opened on both sides of the air-cooled box 7. A dustproof net 703 is fixedly installed inside the heat dissipation vents 702. Mounting shells 8 are fixedly installed on both sides of the water-cooled box 5.

[0035] Specifically, the top of the support plate 208 is provided with a threaded hole, and the inner wall of the threaded hole is threaded with a screw 209, one end of which is rotatably connected to a pressure plate 210.

[0036] In this embodiment, by setting a screw 209, the clamping plate 210 can be moved vertically by turning the screw 209. When the clamping plate 210 moves down, the computing power router 204 can be clamped and installed, which facilitates the subsequent disassembly and maintenance of the computing power router 204.

[0037] Specifically, a rubber pad 211 is fixedly provided at the bottom of the clamping plate 210, and the bottom of the rubber pad 211 is in contact with the top of the computing power router 204.

[0038] In this embodiment, by setting a rubber pad 211, the friction between the pressing plate 210 and the computing router 204 can be enhanced, thereby further enhancing the pressing effect.

[0039] Specifically, a first electric push rod 801 is fixedly installed inside the mounting shell 8. A brush plate 802 is movably inserted into the telescopic end of the first electric push rod 801. A groove adapted to the telescopic end of the first electric push rod 801 is provided on the top of the brush plate 802.

[0040] In this embodiment, by setting a first electric push rod 801 and a brush plate 802, the first electric push rod 801 drives the brush plate 802 to move vertically. The moving brush plate 802 can clean the surface of the dustproof net 703, preventing the dustproof net 703 from becoming clogged and hindering the discharge of heat.

[0041] Specifically, a slot is provided on one side of the telescopic end of the first electric push rod 801, and a bolt 803 is movably provided on the inner wall of the slot. A threaded hole adapted to the bolt 803 is provided on one side of the brush plate 802, and the connection between the bolt 803 and the brush plate 802 is a threaded connection.

[0042] In this embodiment, by setting a bolt 803, the bolt 803 is moved by turning the bolt 803. When one end of the bolt 803 is moved out of the slot on the first electric push rod 801, the restriction on the brush plate 802 can be released, and then the brush plate 802 can be disassembled, thereby realizing the function of the brush plate 802 being detachable.

[0043] Specifically, the connection between the first electric push rod 801 and the bolt 803 is a movable plug-in connection.

[0044] In this embodiment, by setting a bolt 803, the bolt 803 is moved by turning the bolt 803. When one end of the bolt 803 is inserted into the slot on the first electric push rod 801, the brush plate 802 can be restricted.

[0045] Specifically, the air-cooled box 7 is equipped with a dust collection hood 9 inside, a first air pipe 901 is fixedly installed at the bottom of the dust collection hood 9, a vacuum cleaner 902 is fixedly installed at one end of the first air pipe 901, and a second air pipe 903 is fixedly installed on one side of the vacuum cleaner 902.

[0046] In this embodiment, by setting up a dust suction hood 9, a vacuum cleaner 902, a first air pipe 901 and a second air pipe 903, the vacuum cleaner 902 can generate negative pressure and suck away the dust accumulated on the surface of the cooling chip 502 through the dust suction hood 9, so as to prevent the dust from affecting the heat dissipation of the cooling chip 502.

[0047] Specifically, a second electric push rod 904 is fixedly installed on both sides of the bottom of the dust hood 9.

[0048] In this embodiment, by setting a second electric push rod 904, when it is necessary to vacuum clean the cooling chip 502, the vacuum cover 9 can be moved to a position close to the cooling chip 502.

[0049] Specifically, a temperature sensor 504 is fixedly installed on one side of the bottom of the water-cooled box 5, and a PLC controller 505 is installed on the other side of the bottom of the water-cooled box 5.

[0050] In this embodiment, by setting a temperature sensor 504 and a PLC controller 505, the temperature sensor 504 can detect the temperature of the coolant 501 in real time. Once the temperature of the coolant 501 rises to a preset threshold, it will send a signal to the PLC controller 505, and then the PLC controller 505 will start the cooling chip 502.

[0051] Specifically, the connection between the screw 209 and the support plate 208 is a threaded connection.

[0052] In this embodiment, by setting a screw 209, the clamping plate 210 can be moved vertically by turning the screw 209. When the clamping plate 210 moves upward, the clamping on the computing router 204 can be released.

[0053] When using it, 1. Business object definition: The computing power network proposed in this invention takes the integration of communication, sensing and computing as its core support. Communication, sensing and computing complement each other. It connects and coordinates multi-level computing power resources and communication resources of cloud, edge and terminal through virtualization. Through real-time sensing, it realizes on-demand scheduling and efficient sharing of computing power services, and flexibly allocates computing power resources for various services in converter stations, such as converter station equipment monitoring, personnel positioning and behavior monitoring.

[0054] 2. The computing network architecture design distributes computing power, storage, and other resource information of service nodes through the network control layer (including centralized controllers, distributed routing protocols, etc.). Combined with network information and the needs of upper-layer applications (such as personnel location and behavior monitoring at converter stations, equipment status monitoring, etc.), it provides optimal distribution, association, and allocation of computing, storage, and network resources, thereby achieving optimal configuration and utilization of the entire network resources. The computing network architecture is as follows: Figure 1 As shown.

[0055] 3. Regarding computing power measurement, consensus needs to be reached based on standards and specifications. A unified descriptive language should be established to quantify heterogeneous computing power resources and diverse business needs, and measurable standard units should be assigned to computing power resources. Different measurement methods can be used depending on the application scenario. In the integrated sensing and computing scenario of converter stations, coarse-grained units such as virtual machines and containers can be used for measurement.

[0056] 4. Computing power awareness: Computing power awareness is the awareness of all types of resource information, including business needs. The computing power network proposed in this invention can proactively provide resource demand information in a business demand-driven manner, or it can use the ubiquitous sensing capability of the network to sense the usage status of network computing power resources. At the same time, based on historical prior information, supervised learning is used to predict resource demand and resource consumption, thereby realizing resource allocation.

[0057] 5. Computing power routing: Computing power routing distributes network computing power resource information to obtain a computing power resource view centered on converter station services. This provides a clear understanding of the distribution and usage of various computing power resources, laying the foundation for computing power orchestration.

[0058] 6. Establishing Timing and Resource Constraints: The integrated communication, sensing, and computing scenario in converter stations also includes various services such as personnel positioning and behavior monitoring, and equipment status monitoring. Priorities need to be set for different service objects, and the potential temporal dependencies between these objects need to be analyzed to ensure the order of data processing. Based on the high-frequency, short-term changes in the arrival of service objects in the converter station, a dynamic adjustment mechanism should be established to address the periodic fluctuations and sudden requests in computing power demand within the integrated communication, sensing, and computing scenario. Simultaneously, the communication, sensing, and computing needs of the service objects should be considered to ensure timely and flexible resource allocation and stable and efficient data transmission and storage.

[0059] 7. Construction of distribution function: Collect arrival data of different types of business objects, estimate their arrival probability and consider priority and time series characteristics, and construct the arrival distribution function of business objects based on arrival probability and business constraints.

[0060] 8. Resource requirement standardization: Based on the constructed distribution function and the computing resources required by the business, requirements are standardized, including task type, task constraints, priority, timing characteristics, bandwidth resource requirements, frequency resource requirements, and environmental security requirements. Requirements standardization provides a clear description of the business objects in the converter station's integrated sensing and computing scenario, which helps optimize resource allocation and task scheduling, and improves overall efficiency.

[0061] 9. Requirements parsing and mapping: This process parses and maps standardized business resource requirements, extracting specific computational requirements, resource requirements, and constraints. The parsed intent requests and constraints are then mapped to specific resources and operations to generate strategies and plans.

[0062] 10. Definition of Request Northbound Interface: The function of the Request Northbound Interface (RNBI) is to analyze the business requirements in the integrated computing and communication scenario of converter stations, map the requirements into a standard computing power configuration model, and play a role in the conversion between resource requirements and specific computing power parameters.

[0063] 11. Demand Strategy Definition: The demand strategy layer possesses management, control, and decision-making capabilities. It accepts standardized computing resource demand requests from RNB1 and processes them within the current system. Subsequently, a demand-based management and orchestration system is adopted to achieve unified scheduling and slicing configuration of integrated computing resources.

[0064] 12. The DRL-based slicing strategy model incorporates the joint allocation of computing power resources into the service demand slicing management of the converter station's integrated sensing and computing scenario. Defining the environment, actions, and rewards applicable to the integrated sensing and computing scenario of the converter station is a core element of Deep Reinforcement Learning (DRL), which is explained in detail below.

[0065] 13. Environment Settings: The environment settings define the performance status and resource usage allocated to the current slice in the converter station's integrated sensing and computing scenario. State S = [C S F S M S This indicates the usage of computing resources, communication bandwidth resources, and storage resources.

[0066] 14. Action Settings: The computing power slicing strategy for service requirements includes computing resource slicing, communication bandwidth resource slicing, and storage resource slicing. These fine-grained strategies will be combined into a set of scalable slices, directly affecting various performance indicators of the converter station's integrated communication, sensing, and computing network. In this paper, the slicing action involves a combination of three fine-grained strategies, represented as: A i =[F i Ci M i ].

[0067] 15. Reward Setting: The reward function is the feedback calculated after executing the slicing policy. The agent effectively adjusts the policy based on the reward feedback. The action objective of the optimal slicing policy is to find the policy with the highest slice reward value that satisfies the slice requirement constraints during the slice configuration process. If the computing power slice meets the service requirements, positive feedback is obtained; otherwise, there is no feedback. In the DRL-based computing power slicing model, in order to obtain the maximum reward, the computing power coordinator needs to continuously try different actions in the integrated sensory computing environment to learn policy π, while updating parameters in the direction of maximizing reward.

[0068]

[0069] 16. When cooling the computing power router 204, the heat conduction plate 206 and the heat dissipation fins 207 can conduct heat to the computing power router 204, transferring the heat of the computing power router 204 to the water cooling box 5. The coolant 501 in the water cooling box 5 can absorb the heat of the computing power router 204 and reduce the temperature of the computing power router 204.

[0070] When cooling the coolant 501 in the water-cooled box 5, the temperature sensor 504 can detect the temperature of the coolant 501 in real time. Once the temperature of the coolant 501 rises to the preset threshold, it will send a signal to the PLC controller 505. Then the PLC controller 505 will start the cooling chip 502. The cooling chip 502 can cool the coolant 501 in the water-cooled box 5, so that the coolant 501 in the water-cooled box 5 always maintains a low temperature.

[0071] In summary, the computing power orchestration algorithm device for this converter station scenario addresses the highly heterogeneous and high-frequency, short-term changing timing characteristics of computing power demand in converter stations. It estimates the distribution function of sequential arrival of different types of sequential service objects by statistically analyzing their arrival probabilities. Based on this distribution function, it estimates the system's timing computing power demand and the proportion of each type of service object. Furthermore, based on the characteristics of the transmission topology and computing power distribution, it dynamically allocates and optimizes computing power resources using a deep reinforcement learning-based computing power slicing strategy. By setting up a silicone grease layer 205, a heat-conducting plate 206, heat dissipation fins 207, a water-cooled box 5, and coolant 501, the silicone grease layer 205, heat-conducting plate 206, and heat dissipation fins can conduct heat to the computing power router 204, transferring its heat to the water-cooled box 5. The coolant 501 in the water-cooled box 5 absorbs the heat from the computing power router 204, reducing its temperature and ensuring stable operation, preventing overheating. The dynamic state affects the stability and efficiency of resource allocation in the computing power router 204. By setting up a cooling chip 502, the coolant 501 in the water-cooled box 5 can be cooled down, keeping the coolant 501 in the water-cooled box 5 at a low temperature, thus maintaining a good heat absorption effect of the coolant 501. By setting up a motor 6 shaft 601 and a stirring blade 602, the motor 6 drives the shaft 601 and the stirring blade 602 to rotate. The rotation of the stirring blade 602 will agitate the coolant 501 in the water-cooled box 5, keeping it in a flowing state, thereby achieving uniform cooling of the coolant 501. By setting up a cooling fan 701, a heat dissipation port 702 and a dust filter 703, the cooling fan 701 can provide air cooling for the hot end of the cooling chip 502, allowing the heat from the hot end of the cooling chip 502 to be quickly discharged from the heat dissipation port 702, ensuring the normal operation of the cooling chip 502 and preventing overheating. The dust filter 703 inside the heat dissipation port 702 can intercept most of the dust in the external environment.

[0072] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A computing power orchestration algorithm device for a converter station scenario, comprising a service layer (1), a control layer (2), a resource layer (3), and an orchestration management layer (4), characterized in that: The service layer (1) is internally configured with resource information processing (101) and task management (102); the control layer (2) is internally configured with resource information collection (201), resource allocation (202) and network connection scheduling (203); the resource layer (3) is internally configured with computing resources (301), storage resources (302), network resources (303) and spectrum resources (304); and the orchestration management layer (4) is internally configured with computing power network orchestration (401), computing power modeling (402) and computing power OAM (403). The resource allocation (202) is internally equipped with a computing power router (204). A thermal grease layer (205) is fixedly installed at the bottom of the computing power router (204). A heat-conducting plate (206) is fixedly installed at the bottom of the thermal grease layer (205). A heat dissipation fin (207) is fixedly installed at the bottom of the heat-conducting plate (206). One end of the heat dissipation fin (207) penetrates the top of the water-cooled box (5) and extends into the interior of the water-cooled box (5). Coolant (501) is installed inside the water-cooled box (5). A cooling chip (502) is fixedly installed at the bottom of the water-cooled box (5). A cooling rod (503) is fixedly installed at the top of the cooling chip (502). One end of the cooling rod (503) penetrates the bottom of the water-cooled box (5) and extends into the interior of the water-cooled box (5). Support plates (208) are fixedly installed on both sides of the top of the water-cooled box (5). Motors (6) are provided on both sides of the bottom of the water-cooled box (5). The output shaft of the motor (6) is fixedly connected to a rotating shaft (601). One end of the rotating shaft (601) passes through the bottom of the water-cooled box (5) and extends into the interior of the water-cooled box (5). Stirring blades (602) are fixedly provided on both sides of the rotating shaft (601). An air-cooled box (7) is fixedly provided at the bottom of the water-cooled box (5). A cooling fan (701) is provided inside the air-cooled box (7). A heat dissipation vent (702) is provided on both sides of the air-cooled box (7). A dustproof net (703) is fixedly provided inside the heat dissipation vent (702). An installation shell (8) is fixedly provided on both sides of the water-cooled box (5).

2. The computing power orchestration algorithm device for a converter station scenario according to claim 1, characterized in that: The top of the support plate (208) is provided with a threaded hole, and a screw (209) is threadedly connected to the inner wall of the threaded hole. One end of the screw (209) is rotatably connected to a pressure plate (210).

3. The computing power orchestration algorithm device for a converter station scenario according to claim 2, characterized in that: A rubber pad (211) is fixedly provided at the bottom of the clamping plate (210), and the bottom of the rubber pad (211) is in contact with the top of the computing power router (204).

4. The computing power orchestration algorithm device for a converter station scenario according to claim 1, characterized in that: The mounting housing (8) is fixedly provided with a first electric push rod (801). A brush plate (802) is movably inserted into the telescopic end of the first electric push rod (801). The top of the brush plate (802) is provided with a groove that matches the telescopic end of the first electric push rod (801).

5. The computing power orchestration algorithm device for a converter station scenario according to claim 4, characterized in that: The first electric push rod (801) has a slot on one side of its telescopic end, and a bolt (803) is movably installed on the inner wall of the slot. The brush plate (802) has a threaded hole on one side that matches the bolt (803). The connection between the bolt (803) and the brush plate (802) is a threaded connection.

6. The computing power orchestration algorithm device for a converter station scenario according to claim 5, characterized in that: The connection between the first electric push rod (801) and the bolt (803) is a movable plug-in connection.

7. The computing power orchestration algorithm device for a converter station scenario according to claim 1, characterized in that: The air-cooled box (7) is equipped with a dust collection hood (9). A first air pipe (901) is fixedly installed at the bottom of the dust collection hood (9). A vacuum cleaner (902) is fixedly installed at one end of the first air pipe (901). A second air pipe (903) is fixedly installed on one side of the vacuum cleaner (902).

8. The computing power orchestration algorithm device for a converter station scenario according to claim 7, characterized in that: The bottom of the dust hood (9) is fixedly provided with a second electric push rod (904) on both sides.

9. The computing power orchestration algorithm device for a converter station scenario according to claim 1, characterized in that: A temperature sensor (504) is fixedly installed on one side of the bottom of the water-cooled box (5), and a PLC controller (505) is installed on the other side of the bottom of the water-cooled box (5).

10. The computing power orchestration algorithm device for a converter station scenario according to claim 3, characterized in that: The connection between the screw (209) and the support plate (208) is a threaded connection.

Images