Task adaptive allocation method for load balancing in multiple industrial networks

Abstract

In the context of multi-industrial network, the load balancing between network layers and components is crucial to the efficiency of industrial system to complete tasks. This patent proposes an adaptive task allocation method between components to solve the problem of load balancing in multi-industrial network. In this method, if the average load of the component and its adjacent components in the layer is less than the average load of the layer, inter-layer allocation is performed to achieve load balancing between network layers; if inter-layer allocation is not needed and the load of the component exceeds the average load of the layer, intra-layer task allocation is performed to achieve load balancing between components in the same layer. The industrial system applying this task allocation method can avoid the situation of system task completion timeout caused by uneven network layer load as much as possible. Compared with traditional task allocation methods, this method has better adaptability in the context of multi-industrial network.

Description

Technical Field

[0001] This invention relates to the construction of a load balancing method based on task allocation in a multi-industry network scenario. Specifically, in this method, the system adaptively allocates tasks between components in the same layer or between components in different layers according to the load conditions of different components within and between layers in the multi-industry network and the network layer in which the components are located. The allocated tasks will select nodes with lighter loads in the network layer to achieve the goal of optimizing the overall system's task execution time. Background Technology

[0002] With the prevalence of multi-network scenarios, the scheduling and coordination of intelligent components in industrial systems are affected by multiple coupling relationships. Existing industrial system models only consider single network connections between components, leading to information silos and a lack of collaborative relationships between different industrial systems connected by business coupling across different networks. Furthermore, in industrial systems, component capabilities change according to factors such as the operating environment and load at different times. For example, in a power grid system, the power generation capacity of wind turbines changes in real time due to varying wind conditions. A decrease in component capability can significantly increase the time required to execute tasks, resulting in an imbalance of load among components in the industrial system.

[0003] When an industrial system experiences load imbalance, tasks on overloaded components need to be reallocated to optimize overall system task completion time. In multi-layer industrial network scenarios, task reallocation can be divided into task reallocation between components at the same layer and task reallocation between components at different layers. To complete the reallocated task as quickly as possible, components with lighter loads should be selected. However, due to the heterogeneity of components across different network layers in a multi-layer industrial network, the cost of task execution varies for components in each layer. Reallocating tasks between components at different layers may cause the total cost of system task execution to exceed its limit, thus restricting cross-layer task reallocation. (Example follows.) Figure 1 As shown, load balancing within a layer and load balancing between layers can affect each other: when there is an imbalance in load between layers, cross-layer task allocation may cause other network layers that are already in a load-balanced state to become unbalanced again; conversely, when there is an imbalance in load within a layer, due to limitations such as topology and distance between components, cross-layer task allocation may also be chosen, as in the example. Figure 2 The overloaded network layer containing the component selected across layers will exacerbate the load imbalance between network layers. Therefore, load balancing within and between layers in a multi-layer industrial network system has a crucial impact on optimizing the overall industrial system's task completion time.

[0004] Generally, load balancing in industrial systems aims to balance the tasks undertaken by various components within the industrial production network, minimizing the total time required for the entire system to complete all tasks and achieving optimal industrial production efficiency. However, in the context of a single industrial network, the components are homogeneous and driven by the same factors. When a sudden event occurs, the capabilities of components at the same layer may change similarly simultaneously, potentially causing a significant drop in the system's load capacity and severely impacting its performance. Therefore, we aim to achieve load balancing across the entire industrial system through cross-layer and cross-system task allocation. This way, even if components in a certain network layer fail to function properly, other network layers can share the workload, ensuring the stable operation of the entire industrial system. Summary of the Invention

[0005] Technical Problem: The purpose of this invention is to propose an adaptive task allocation method for load balancing in multi-layer industrial networks. This allocation method involves two types of target components: components within the same layer and components across different layers. Tasks allocated in the system can autonomously choose to be allocated within the same layer or across layers based on the load conditions of the components and the layer they belong to. This method solves the problem that the mutual influence between load balancing within and between network layers in multi-layer industrial network scenarios leads to an excessively large solution strategy space and makes it difficult to obtain an approximate global optimal solution.

[0006] Technical Solution: In an industrial system with a multi-layered network, each industrial component undertakes a certain number of industrial tasks. During task execution, the component's capacity may decrease, leading to increased task execution time and some components becoming overloaded. To improve the efficiency of the industrial system in completing tasks and reduce the total completion time, some tasks on these components need to be allocated to other components at the same layer or different layers according to certain rules. The inter-layer relationships, intra-layer relationships, and inter-layer resource constraints among components in the entire system are crucial for task allocation. The main technical solution of this task allocation method is as follows: Overloaded components can choose one of two allocation strategies: intra-layer allocation or cross-layer allocation. Initially, tasks in the system are evenly distributed among components and network layers according to their execution capabilities. After a period of time, the system's... A component in the layer The ability to perform tasks If changes occur, The value decreases, by comparing components. Components Layer and component Determine the component's load status within the layer containing adjacent components. Is the component overloaded? If a component is under heavy load, a task allocation strategy needs to be selected. When choosing an allocation strategy, each component needs to consider not only the load of its adjacent components at the same and across layers, but also the overall load of the layer containing these components. If the average load of the layer containing these components is less than the average load of the layer containing the component requiring task allocation, then inter-layer allocation is needed; if inter-layer allocation is not needed and the component's own load exceeds the average load of its layer, then intra-layer task allocation is needed. See example. Figure 2 In the diagram, the component represented by the red node issues a request to allocate tasks. At this time, the components around it are all under heavy load, and the average load of its neighboring nodes is greater than the average load of its network layer. Therefore, a cross-layer allocation strategy needs to be selected.

[0007] Each component has a limited task processing capacity. The selected component, in addition to executing the reassigned task, must also perform its original tasks. Furthermore, the cost of task execution varies between different layers, and cross-layer task reassignment may increase the cost. Therefore, selecting the appropriate component to complete the reassignment task is crucial when the total cost of task execution is limited. When a component is requested to perform task reassignment, it first checks the sum of its own load and the sum of the capabilities of its adjacent components (including those at the same layer and those across layers), and records the result of dividing the two as the component's value. Neighborhood network layer utilization rate Then, examine the load and capacity of all components in the same layer as the component, and calculate the component's load and capacity. Floor utilization rate And compare and The value; if The value is less than Then, without exceeding the task execution cost limit, the component will select an adjacent component from other network layers using the optimal strategy to execute the task, until... Greater than or equal to ;if The value is already greater than Then compare the current utilization of the components. and The size, if Greater than Then, the optimal strategy is used to select adjacent components at the same level for task allocation until... Less than or equal to Once the above process is complete, the entire system is in a state of load balancing between components and between network layers. Beneficial effects

[0008] (1) The adaptive task allocation method for balancing the load between various industrial subsystems can select the task allocation strategy according to the load status between each layer of the whole system. This can solve the problem that the allocation result of the previous allocation method that only considered a single network fell into local optima and could not balance the load between various subsystems.

[0009] (2) Reduce the total time for the system to complete tasks. Because the tasks undertaken by the system components are constantly being dynamically adjusted, the load on the entire system components is continuously optimized and eventually reaches a balanced state, enabling the entire industrial system to complete all tasks in a shorter time.

[0010] (3) Enhancing the stability and safety of industrial systems In industrial systems, the overload of some components caused by load imbalance can lead to a decrease in the stability and safety of the components, which may result in system failure due to component instability; the adaptive task allocation method minimizes the possibility of component overload and ensures good stability and safety of industrial systems. Attached Figure Description

[0011] Figure 1 This diagram illustrates how load imbalance between layers affects load balancing within a layer in a multi-layer industrial network, using the power generation network in a circular industrial park as an example.

[0012] Figure 2 This diagram illustrates how load imbalance within a layer affects load balancing between layers in a multi-layer industrial network, using the power generation network in a circular industrial park as an example.

[0013] Figure 3 This is a schematic diagram of the main principle of the method of the present invention.

[0014] exist Figure 1 , Figure 2 The circular nodes in the diagram all perform the same task, representing different components in different layers of the industrial system. Blue nodes represent nodes with normal load, yellow nodes represent nodes with heavy load, and red nodes represent nodes undergoing task allocation. The red dashed line indicates the direction of task allocation. Detailed Implementation

[0015] (1) In the initial stage, the system senses the load conditions of multiple networks, components, and inter- and intra-layer loads. The multiple network sensing process identifies which components can allocate tasks, the component sensing process analyzes the tasks and capabilities of each component, and the inter- and intra-layer load sensing process detects whether there is an imbalance in load within or between layers in the industrial system. In the initial stage, we describe these four sensing objects and sensing processes in sequence: Multi-network sensing process: In the initial stage, it is first necessary to analyze the specific multi-industry network scenario. Multi-industry network Multiple industrial networks Composition, each industrial network Composed of isomorphic industrial components Composition. In industrial systems under the context of multiple industrial networks, there are two types of connection relationships: one is the connection relationship between components at the same layer, and the other is the connection relationship between components at different layers. Components that are connected to each other can allocate tasks. There is a real physical connection between components at the same layer, and they can pass tasks to each other and return the execution results of the tasks; while there is no real connection between components at different layers, and this virtual connection represents the business relationship between components. And because there is no real connection between components at different layers, there is no feedback process in the task allocation between these components, and the task completion results are directly provided by the adjacent components across layers. And because the cost of executing tasks is different for each layer of components, the task allocation between components at different layers in (3) will be limited by the total cost of executing tasks. Combining the same-layer connection and different-layer connection in multiple industrial networks, Multi-layer industrial networks Ultimately described as , No. Layered network representation , Representing the A collection of components in a layer. Representing the Connections between components in a layer Indicates the first The unit cost of a task performed by a layer network component. Inter-layer connections between components can be represented by an adjacency matrix. To indicate, representing the first Layer components and Inter-layer connections between layer components.

[0016] Component awareness process: After analyzing the network structure, each component needs to be analyzed in detail. Network layer Any component It can be represented by a triple: ,in Represents the first in the network layer Layer Components The set of tasks to be executed Representative task set Size, components The capacity size is expressed as Component capabilities It changes with the changes in external factors in real industrial production, and is expressed as Each task in the task set This can be represented by a pair: ,in Indicates task Size, Indicates task On which component is execution currently pending? Set of tasks to be executed .

[0017] Intra-layer load balancing perception process: After a detailed analysis of the components, the following definition is given for measuring load balancing among components in the same layer. Load balancing within the same layer is achieved through the average utilization of the components. Compared with the average utilization of the network layer The difference measurement, for each component in each layer of the system. Record its own utilization rate . The smaller the value, the more components that need to be reassigned for the task. The more balanced the load is among the network layer components, the better.

[0018] Inter-layer load balancing perception process: After defining load balancing within the same layer, the final definition for measuring load balancing between network layers is given. Inter-layer load balancing is achieved through the average utilization of components at each layer. Measurement, Component and the network layer where the components of adjacent domains reside. The average utilization rate (hereinafter referred to as the neighborhood network layer utilization rate) is expressed as: Components The utilization rate of the network layer in which it resides (hereinafter referred to as the network layer utilization rate). Represented as: . The smaller the value, the more components need to be allocated for the task. The more balanced the load is between the network layer where the component is located and the neighboring network layers, the better.

[0019] At the end of the initial phase, a decision needs to be made regarding whether to perform task allocation between components. If the components... Changes in capabilities may necessitate task reassignment. The goal of task reassignment for system components is to maximize the utilization of that layer of components. and the utilization rate of each component in this layer Utilize the neighboring network layer as close as possible to the component .when Greater than or The value is less than , indicating component Task reassignment is required.

[0020] When component When deciding to allocate tasks, you need to select which component will be used. On the waiting task Dispatch to adjacent components Above; such allocation strategies are represented as ,but The overall allocation strategy space can be represented as: .

[0021] As can be seen from this expression, directly searching the entire policy search space is extremely large, making it very difficult to find the optimal solution directly within it. Therefore, in this patent, the entire search policy space is allocated to the task components according to the request. utilization rate , Neighborhood network layer utilization and network layer utilization It is divided into intra-layer task allocation strategy (2) and cross-layer task allocation strategy (3).

[0022] For example Figure 1 For example, suppose the calorific value of the waste input to the generator unit represented by the red component in the current waste-to-energy incineration network is too low. This will reduce the power generation capacity of this generator unit, causing it to be unable to complete its power generation tasks on time. Therefore, this generator unit needs to request task allocation from its other components to reduce the tasks it is responsible for. The system can provide components with tasks. The network layers to be allocated include Network Layer 2 (biomass biogas power generation network), Network Layer 3 (wind power generation network), and its own Network Layer 1 (waste incineration power generation network). If the currently requesting task allocates the red component... and the average utilization rate of the layers containing adjacent nodes in each layer. Higher than the average utilization of all components in network layer 1 If so, the task will still be performed by a component in network layer 1; otherwise, other layer components will be needed to complete the task.

[0023] (2) Intra-layer allocation stage: when the component Neighborhood network layer utilization The value is already greater than its network layer utilization. This indicates a relative balance between the layers of the system. However, distributing tasks to components in other layers is more likely to exacerbate the load imbalance between layers. Therefore, components... Choose to allocate tasks within the component layer. When the component... When choosing to perform task allocation between components at the same level, it is only possible to allocate tasks between components. On the waiting task Dispatch to adjacent components on the same floor Above; such allocation strategies are represented as ,but The allocation strategy space at this point can be represented as: .

[0024] Same-level task allocation strategy The revenue can be defined as: ;in This indicates the component before task allocation. The standard deviation of utilization within the network layer in which it is located. This indicates the components after task allocation. The standard deviation of utilization within the network layer, and the standard deviation of component utilization within the layer are expressed as follows: .

[0025] The smaller the standard deviation within a layer, the more balanced the load is among the components within that layer, and after task allocation... The smaller the ratio, the greater the contribution of task allocation to load balancing within the layer, and the better the effect of achieving load balancing within the layer. In this case, the optimal task allocation strategy is the one that maximizes allocation benefits; therefore, the component... The strategy that maximizes profit will be adopted. Perform task allocation to ensure that each allocation achieves load balancing within the network layer as much as possible.

[0026] (3) Inter-layer allocation stage: when components Neighborhood network layer utilization The value is already less than the network layer utilization. This indicates that there is a load imbalance between different layers of the system, and tasks can no longer be distributed to other components in the same layer. Therefore, the components... Select other layer components for cross-layer task allocation. When the component When choosing to perform cross-layer task dispatch, it can only be done in the component. On the waiting task Dispatch to adjacent components across layers Above; such allocation strategies are represented as ,but The allocation strategy space at this point can be represented as: .

[0027] Cross-level task allocation strategy The return can also be defined as: ; Indicates before task allocation The average utilization rate of each component in the middle layer is compared to the whole The standard deviation of component utilization Indicates after task allocation The average utilization rate of each component in the middle layer is compared to the whole The standard deviation of component utilization. The standard deviation of neighborhood network layer utilization is expressed as: .

[0028] The smaller the standard deviation, the better. The more balanced the load between layers, the better the task allocation will be. The smaller the ratio, the greater the contribution of task allocation to inter-layer load balancing, and the better the effect of achieving inter-layer load balancing. In this case, the optimal strategy for inter-layer task allocation is the strategy that maximizes allocation benefits. However, since inter-layer allocation assigns tasks to heterogeneous components, the cost of executing these tasks differs from the cost of executing tasks in the original components. This may cause the total cost of completing the tasks to exceed the maximum limit. Therefore, the strategy... It is also subject to the upper limit of the total cost for the entire system to complete all tasks, as expressed below: .

[0029] in The upper limit of the cost for executing tasks across the entire system. Final component. The strategy will be based on maximizing profits while not exceeding the upper limit of the total cost of completing the task. Perform task allocation to ensure that each allocation achieves load balancing between neighboring network layers of the components as much as possible.

Claims

1. A method for adaptive task allocation for load balancing in a multi-industry network, characterized in that, The allocation method includes two types of target components: components at the same layer and components at different layers. The tasks being allocated in the system can choose to be allocated at the same layer or across layers based on the load of the component and the load of the layer in which it is located. Multiple industrial networks Multiple industrial networks Composition, each industrial network Composed of isomorphic industrial components In industrial systems within a multi-layer industrial network context, two types of connections exist: connections between components at the same layer and connections between components at different layers. Components connected to each other can allocate tasks. Components at the same layer have a tangible physical connection, allowing them to exchange tasks and receive feedback on execution results. However, there is no physical connection between components at different layers; this virtual connection represents the business relationship between components. Because there is no physical connection between components at different layers, task allocation between them lacks a feedback process; the results are directly provided by adjacent components across layers. Furthermore, since the cost of task execution varies for each layer, task allocation between components at different layers is limited by the total cost of task execution. In summary, considering both same-layer and different-layer connections in a multi-layer industrial network, [the following is a possible interpretation, but the context is unclear]. Multi-layer industrial networks Represented as , No. Layered network representation , Representing the A collection of components in a layer. Representing the Connections between components in a layer Indicates the first The unit cost of a task performed by a layer network component; the inter-layer connections between components are represented by an adjacency matrix. To indicate, representing the first Layer components and Inter-layer connections between layer components; Network layer Any component Represented by a triple: ,in Represents the first in the network layer Layer Components The set of tasks to be executed Representative task set Size, components The capacity size is expressed as Component capabilities It changes with the changes in external factors in real industrial production, and is expressed as Each task in the task set Represented by a tuple: ,in Indicates task Size, Indicates task On which component is execution pending? Set of tasks to be executed ; The component that allocates all search strategy space to tasks based on requests. utilization rate , Average utilization of the layer where the neighboring component is located and Average utilization of components in the layer It is divided into intra-layer task allocation strategy and cross-layer task allocation strategy; When component Average utilization of the layer where the neighboring component is located The value is already greater than the average utilization rate of the components in its layer. This indicates a relative balance between the layers of the system. However, distributing tasks to components in other layers is more likely to exacerbate the load imbalance between layers. Therefore, components... Choose to allocate tasks within the component layer, when the component When choosing to perform task allocation between components at the same level, it is only possible to allocate tasks between components. On the waiting task Dispatch to adjacent components on the same floor Above; such a allocation strategy is represented as ,but The allocation strategy space at this time is represented as: ; Same-level task allocation strategy The revenue is defined as: ;in This indicates the component before task allocation. The standard deviation of utilization within the network layer in which it is located. This indicates the components after task allocation. The standard deviation of utilization within the network layer, where the standard deviation of component utilization within the layer is expressed as: ; The smaller the standard deviation within a layer, the more balanced the load is among the components within that layer, and after task allocation... The smaller the ratio, the greater the contribution of task allocation to load balancing within the layer, and the better the effect of achieving load balancing within the layer. In this case, the optimal task allocation strategy is the strategy that maximizes allocation benefits. Will follow the strategy that maximizes profits Perform task allocation to ensure that each allocation achieves load balancing within the network layer as much as possible.

2. The adaptive task allocation method for load balancing in a multi-industry network according to claim 1, characterized in that: Load balancing between the same layer is achieved through component utilization. Average utilization rate of components in the same layer The difference measurement, for each component in each layer of the system. utilization rate , The smaller the value, the more components that need to be reassigned for the task. The more balanced the load among components within the same layer, the better the load balancing between layers: After defining load balancing within the same layer, the definition for measuring load balancing between network layers is finally given. Inter-layer load balancing is achieved by the average utilization of components in each layer. Measurement, Component The average utilization of the layer containing the neighboring components is expressed as: Components Average utilization of components in the layer Represented as: , The smaller the value, the more components need to be allocated for the task. Components of the layer and its adjacent domains The more balanced the load is on the layer.

3. The adaptive task allocation method for load balancing in a multi-industry network according to claim 2, characterized in that: When component Average utilization of the layer where the neighboring component is located The value is already lower than the average utilization rate of the components in the same layer. This indicates that there is a load imbalance between different layers of the system, and tasks can no longer be distributed to other components in the same layer. Therefore, the components... Select other layer components for cross-layer task allocation, when the component When choosing to perform cross-layer task dispatch, it can only be done in the component. On the waiting task Dispatch to adjacent components across layers Above; such allocation strategies are represented as ,but The allocation strategy space at this point is represented as follows: ; Cross-level task allocation strategy The revenue is defined as: ; Indicates before task allocation The average utilization rate of each component in the middle layer is compared to the whole The standard deviation of component utilization Indicates after task allocation The average utilization rate of each component in the middle layer is compared to the whole The standard deviation of component utilization, where the standard deviation of neighborhood network layer utilization is expressed as: ; Representation Component Components of adjacent domains in the same layer, The smaller the standard deviation, the better. The more balanced the load between layers, the better the task allocation will be. The smaller the ratio, the greater the contribution of task allocation to load balancing between layers, and the better the effect of achieving load balancing between layers. In this case, the optimal strategy for inter-layer task allocation is the strategy that maximizes allocation benefits. However, since inter-layer allocation allocates tasks to heterogeneous components, the cost of executing the task differs from the cost of executing the task in the original component. This may cause the total cost of completing the task to exceed the maximum limit. Therefore, the strategy... It is also subject to the upper limit of the total cost for the entire system to complete all tasks, as expressed below: ; in The upper limit of the cost for the entire system to execute tasks, and the final component The strategy will be based on maximizing profits while not exceeding the upper limit of the total cost of completing the task. Perform task allocation to ensure that each allocation achieves load balancing between neighboring network layers of the components as much as possible.

Images