Industrial load management and control method based on cloud edge cooperation technology

Abstract

The invention relates to the technical field of industrial load management, and discloses an industrial load management and control method based on a cloud-side cooperation technology, and the method comprises the steps: 1, collecting historical load data, environment data and equipment operation state data through the deployment of a load sensor, environment monitoring equipment and equipment state monitoring equipment, and carrying out the collection of the historical load data, environment data and equipment operation state data; the historical load data is used for representing load change conditions of the system at multiple time points, the environment data comprises external temperature and electricity price information which are acquired in real time, the equipment operation state data comprises operation state parameters of equipment, and data fusion is carried out by combining the historical load data, the environment data and the equipment operation state data. A dynamic weighting mechanism is introduced in a data fusion stage, weights are adaptively adjusted according to the importance of historical load data, environmental data and equipment state data, accurate description of contribution of different features to fusion features is realized, and the effect of remarkably reducing load prediction errors is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial load management, and specifically to an industrial load control method based on cloud-edge collaborative technology. Background Art

[0002] The demand for intelligence and efficiency in the industrial field is increasing day by day, and industrial load control has become a key technology for achieving energy conservation and emission reduction and optimizing energy distribution. However, the current industrial load management methods have many defects when facing dynamic and complex industrial environments.

[0003] On the one hand, traditional industrial load forecasting methods mainly rely on historical load data, ignoring the significant impact of environmental data and equipment operating status on load changes, resulting in poor adaptability and accuracy of forecasting models in complex scenarios. At the same time, existing multimodal data fusion technology fails to effectively measure the importance of different data sources, cannot give full play to the complementary characteristics of multimodal data, and has limitations in the characterization of load characteristics.

[0004] On the other hand, the load control process usually assumes that the relationship between devices is fixed, which cannot cope with the complex collaboration requirements brought about by the dynamic changes in loads in industrial systems. If the load distribution strategy between industrial equipment lacks flexible adjustment of real-time collaboration relationships, it is easy to cause uneven load distribution and waste of resources, reducing the overall operating efficiency of the system.

[0005] In addition, although the current load control method based on cloud-edge collaboration has made some progress in computing resource allocation and communication efficiency, the overhead problem of cloud-edge collaborative communication is particularly prominent in high-frequency data interaction scenarios. Since the cloud needs to frequently interact with edge devices, redundant communication causes a significant increase in system energy consumption, restricting the efficient operation of the system.

[0006] In summary, the shortcomings of the existing technology in load forecasting, load regulation and cloud-edge collaborative communication seriously restrict the performance of the industrial load control system. Therefore, those skilled in the art provide an industrial load control method based on cloud-edge collaborative technology to solve the above problems. Summary of the invention

[0007] In view of the shortcomings of the prior art, the present invention provides an industrial load management and control method based on cloud-edge collaborative technology to solve the problems raised in the above background technology.

[0008] To achieve the above objectives, the present invention is implemented through the following technical solutions: an industrial load control method based on cloud-edge collaborative technology, comprising: Step 1: By deploying load sensors, environmental monitoring equipment and equipment status monitoring equipment, historical load data, environmental data and equipment operation status data are collected. The historical load data is used to represent the load changes of the system at multiple time points. The environmental data includes real-time external temperature and electricity price information. The equipment operation status data includes equipment operation status parameters. The historical load data, environmental data and equipment operation status data are combined to perform data fusion and generate a unified fusion feature as the input of load forecasting. Step 2: Input the fusion features generated in step 1 into the load forecasting model, process the input features through the encoding module of the model, extract the potential load association pattern, and then generate the load forecast value at the future time point according to the potential pattern through the decoding module. The forecasting model is constructed based on the correlation between the historical change trend of the load, environmental data and equipment status; Step 3: Combine the load forecast value generated in step 2 with the industrial equipment network structure to establish a dynamic graph model that reflects the load transmission relationship between devices. The dynamic graph model includes a set of device nodes, connection relationships between nodes, and an edge weight matrix that reflects the load transmission strength between devices. The dynamic graph is updated in real time through the load forecast value and the association relationship between devices. Step 4: Based on the dynamic graph model established in step 3, define the local optimization goals of the device nodes, build the load control costs between the nodes according to the collaborative relationship between the device nodes, coordinate the load distribution of each node through the global optimization framework, and calculate the optimization results in real time by the edge device and feed back the processing results to the cloud; Step 5. Combined with the load distribution results optimized in step 4, a game model including device nodes is constructed in the cloud. As participants in the game model, the device nodes dynamically adjust the load control strategy according to the combined influence of short-term load forecast error, long-term trend deviation and communication overhead, and achieve a global equilibrium state through the game model. Finally, the control strategy parameters of cloud computing are sent to edge devices for execution to complete the global coordination and dynamic optimization of the load.

[0009] Preferably, the data fusion in step 1 is accomplished by the following weighting mechanism: , in, For the i-th type of input data, is the number of feature categories, is the input feature vector of the i-th class, Calculated by the following formula: , in, is the weight coefficient of each type of data, is the attention score of the i-th feature, is the number of feature categories, is an exponential function, A normalization term for the exponential value of the feature attention score.

[0010] Preferably, the load forecasting model in step 2 generates a load forecast value at a future time point through the following process: The encoder extracts the potential correlation pattern of the fused features: , in, is a potential latent variable, is the feature mean, is the characteristic standard deviation, is random noise; The decoder generates predicted load values ​​from latent variables ,in, is the current time point, is the prediction time interval.

[0011] Preferably, the dynamic graph model established in step 3 The following relations are satisfied: , in, is the edge weight between node i and node j at time t, and is the load value of node i and node j, Indicates the difference in node load.

[0012] Preferably, the local optimization objective function in step 4 is defined as: , in, is the actual load of node i, is the predicted load of node i, represents the square of the load forecast error of node i, represents the load forecast error of node i, is the optimization variable of node i.

[0013] Preferably, the load regulation cost function between the nodes is defined as: , in, represents the collaboration cost function between nodes i and j, is the load distribution value of node i, is the load distribution value of node j, is the penalty coefficient for load regulation between nodes, is the square of the difference in load distribution values ​​between nodes.

[0014] Preferably, the global optimization framework is solved by the following objective function: , in, To optimize the goal, is the load distribution value of device node i, is a collection of device nodes. is the edge set between device nodes, represents the load forecast error of node i, is the optimization variable of node i, represents the collaboration cost function between nodes i and j, is the optimization variable of node j.

[0015] Preferably, the game model in step 5 is constructed by the following utility function: , in, is the utility function of device node i, is the short-term load forecast error, is the long-term load trend error, is the communication overhead, , and is the weight parameter.

[0016] Preferably, the utility function is optimized by the following equilibrium conditions: , , , in, , and is the weight parameter, is the utility function of device node i, is the weight coefficient of the utility function for short-term error The partial derivative of is the weight coefficient of the utility function for short-term error The partial derivative of is the weight coefficient of the utility function for short-term error The partial derivative of .

[0017] Preferably, the equilibrium strategy parameters generated by the cloud through game optimization are periodically sent to the edge device. The edge device adjusts the load control strategy in real time according to the received strategy parameters combined with local data, and uploads the load control results to the cloud for updating the next round of strategy parameters.

[0018] The present invention provides an industrial load control method based on cloud-edge collaboration technology. It has the following beneficial effects: 1. The present invention introduces a dynamic weighting mechanism in the data fusion stage, adaptively adjusts the weights according to the importance of historical load data, environmental data and equipment status data, and accurately characterizes the contribution of different features to the fusion feature, thereby significantly reducing the load forecast error.

[0019] 2. The present invention constructs a dynamic graph model based on load forecasting results, updates the load transmission relationship between device nodes in real time, and realizes flexible adjustment of the load cooperation relationship between devices, thereby improving the load control accuracy and enhancing the dynamic response capability of the system.

[0020] 3. The present invention comprehensively considers the communication overhead in game optimization, dynamically adjusts the communication frequency and data transmission volume between the cloud and edge devices, achieves the goal of reducing cloud-edge communication redundancy, and obtains the effects of reducing system operation energy consumption and improving collaborative efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0022] In order to make the technical personnel in the technical field understand the scheme of the present invention, the technical scheme in the embodiment of the present invention will be clearly and completely described below in combination with the drawings in the embodiment of the present invention. Obviously, the described embodiment is a partial embodiment of the present invention, not a complete embodiment. Based on the embodiment of the present invention, other embodiments obtained by ordinary technicians in the field without creative work should fall within the scope of protection of the present invention.

[0023] The present invention is described in detail below in conjunction with the accompanying drawings: Example

[0024] Please refer to the attached Figure 1 , an embodiment of the present invention provides an industrial load control method based on cloud-edge collaboration technology, including: Step 1: Deploy load sensors, environmental monitoring equipment, and equipment status monitoring equipment to collect historical load data, environmental data, and equipment operation status data. The historical load data is used to represent the load changes of the system at multiple time points. The environmental data includes real-time external temperature and electricity price information. The equipment operation status data includes the equipment operation status parameters. The historical load data, environmental data, and equipment operation status data are combined to perform data fusion and generate a unified fusion feature as the input for load forecasting. Step 2: Input the fusion features generated in step 1 into the load forecasting model, process the input features through the encoding module of the model, extract the potential load association pattern, and then generate the load forecast value at the future time point according to the potential pattern through the decoding module. The forecasting model is constructed based on the correlation between the historical change trend of the load, environmental data and equipment status; Step 3: Combine the load forecast value generated in step 2 with the industrial equipment network structure to establish a dynamic graph model that reflects the load transmission relationship between devices. The dynamic graph model includes a set of device nodes, the connection relationship between nodes, and an edge weight matrix that reflects the load transmission intensity between devices. The dynamic graph is updated in real time through the load forecast value and the relationship between devices. Step 4: Based on the dynamic graph model established in step 3, define the local optimization goals of the device nodes, build the load control costs between the nodes according to the collaborative relationship between the device nodes, coordinate the load distribution of each node through the global optimization framework, and calculate the optimization results in real time by the edge device and feed back the processing results to the cloud; Step 5. Combined with the load distribution results optimized in step 4, a game model including device nodes is constructed in the cloud. As participants in the game model, the device nodes dynamically adjust the load control strategy according to the combined influence of short-term load forecast error, long-term trend deviation and communication overhead, and achieve a global equilibrium state through the game model. Finally, the control strategy parameters of cloud computing are sent to edge devices for execution to complete the global coordination and dynamic optimization of the load.

[0025] Benefits of step 1: By deploying load sensors, environmental monitoring equipment, and equipment status monitoring equipment, we can comprehensively collect historical load, environmental data, and equipment operating status data to provide rich data support for load forecasting. By combining multimodal data for fusion processing and dynamically adjusting the weights of different data sources, we can effectively improve the data expression ability and the accuracy of load feature description, and solve the problem of the limitation of a single data source in traditional load forecasting.

[0026] Benefits of step 2: By introducing the load forecasting model, using the encoding module to extract potential load correlation patterns, and using the decoding module to generate load forecast values ​​at future time points, the nonlinear change trend of the load can be effectively captured. The model is built based on the correlation between historical loads, environmental data and equipment status, which significantly improves the accuracy of load forecasting.

[0027] Benefits of step 3: Combine the load forecast value and the equipment network structure to establish a dynamically changing equipment node relationship diagram. By updating the connection relationship and load transmission strength of the equipment nodes in real time, the dynamic diagram can flexibly adapt to the load transmission requirements and collaboration changes between equipment, solve the problem that the traditional static topology model cannot reflect dynamic load changes, and improve the real-time and flexibility of load regulation.

[0028] Benefits of step 4: By defining the local optimization goals of device nodes and the load control costs between nodes, the global optimization framework is used to coordinate the load distribution of each node to achieve distributed management of load control. The optimization results are calculated in real time by edge devices and fed back to the cloud, reducing excessive reliance on cloud computing, improving the system's adaptive capabilities, and solving the problem of high computing latency in centralized load control.

[0029] Benefits of step 5: By building a game model, the impact of short-term load forecast error, long-term trend deviation and communication overhead is comprehensively considered, and the load control strategy is dynamically adjusted to achieve a global equilibrium state. The cloud sends the equilibrium strategy parameters to the edge device for execution to ensure the multi-objective balance of the control target, while reducing communication redundancy and system operation energy consumption, solving the problem that the load control strategy in the traditional method cannot take into account global coordination and communication efficiency.

[0030] The data fusion in step 1 is accomplished by the following weighting mechanism: , in, For the i-th type of input data, is the number of feature categories, is the input feature vector of the i-th class, Calculated by the following formula: , in, is the weight coefficient of each type of data, is the attention score of the i-th feature, is the number of feature categories, is an exponential function, A normalization term for the exponential value of the feature attention score.

[0031] By introducing a dynamic weighting mechanism, different types of input data can be adaptively adjusted according to their importance in specific scenarios. The dynamic weighting method solves the problem of assigning fixed weights to data in traditional feature fusion, enhances the flexibility and accuracy of feature expression, and improves the effect of load forecasting.

[0032] The attention mechanism dynamically assigns weights based on the attention scores of various types of data, ensuring that the key characteristics of multimodal data are preserved in different industrial operating environments. For example, when the environment changes significantly, the weight of environmental data is automatically increased, and when the equipment status changes dramatically, the weight of equipment status data is increased, effectively adapting to the dynamic changes in load characteristics.

[0033] By calculating the weight coefficient through exponential function and normalization, the weighted values ​​of different feature categories can be quickly obtained, avoiding the high computational overhead caused by complex feature fusion methods. This method achieves efficient fusion of multimodal data at a low computational cost and is suitable for the real-time prediction needs of large-scale industrial load systems.

[0034] The load forecasting model in step 2 generates the load forecast value at a future time point through the following process: The encoder extracts the potential correlation pattern of the fused features: , in, is a potential latent variable, is the feature mean, is the characteristic standard deviation, is random noise; The decoder generates predicted load values ​​from latent variables ,in, is the current time point, is the prediction time interval.

[0035] By introducing an encoder to extract the potential correlation mode of fusion features, the complex nonlinear relationship in industrial load data, especially the implicit correlation between multimodal data, can be effectively captured. The generation process of latent variables integrates the global and local characteristics of multimodal data, solving the problem that traditional methods cannot predict complex load changes, and significantly improving the accuracy of load prediction.

[0036] The generation process of latent variables ensures the adaptability of the model to random fluctuations and dynamic changes of load data through dynamic adjustment of feature mean and feature standard deviation, as well as the introduction of random noise. The decoder generates future load forecast values ​​from latent variables, which can cope with sudden changes in industrial load scenarios and improve the robustness and reliability of the model.

[0037] The structural design of extracting feature latent patterns through encoders and decoding to generate load forecast values ​​enables the model to adapt to different industrial scenarios and various load change patterns. The generation of latent variables provides a unified feature representation for the model, avoids overfitting problems, and improves the versatility and generalization ability of the model in different data sets and operating scenarios.

[0038] The decoder generates predicted load values ​​from latent variables in an efficient and stable manner, avoiding complex calculations on multimodal data and reducing the consumption of computing resources. In the scenario of industrial real-time load forecasting, it can meet the efficiency requirements of high-frequency and large-scale forecasting.

[0039] The dynamic graph model created in step 3 The following relations are satisfied: , in, is the edge weight between node i and node j at time t, and is the load value of node i and node j, Indicates the difference in node load.

[0040] By defining the edge weights As a function of the load difference between nodes i and j, the dynamic graph model can reflect the load transmission intensity between nodes in real time. The dynamic adjustment mechanism solves the problem that the traditional static model cannot adapt to the load changes of industrial equipment, making the collaborative relationship between equipment flexible.

[0041] Edge Weight The calculation of is closely related to the difference in node load, which enables the model to dynamically capture the imbalance of load distribution between devices and optimize the load control strategy in a timely manner. By updating the relationship between nodes in real time, accurate load control is ensured to meet the needs of complex industrial scenarios.

[0042] By dynamically adjusting the edge weights, the connection weights of the load differences between nodes are reduced, thereby reducing unnecessary load transmission cooperation. This mechanism can optimize the load regulation cost between nodes and improve the overall operating efficiency of the industrial load regulation system.

[0043] The dynamic graph model is updated in real time based on load forecast values ​​and equipment association relationships. It can quickly adapt to scenarios where equipment operating status and load characteristics change, improving the model's adaptability and robustness, especially in industrial environments with load fluctuations.

[0044] The local optimization objective function in step 4 is defined as: , in, is the actual load of node i, is the predicted load of node i, represents the square of the load forecast error of node i, represents the load forecast error of node i, is the optimization variable of node i.

[0045] The load regulation cost function between nodes is defined as: , in, represents the collaboration cost function between nodes i and j, is the load distribution value of node i, is the load distribution value of node j, is the penalty coefficient for load regulation between nodes, is the square of the difference in load distribution values ​​between nodes.

[0046] The global optimization framework is solved by the following objective function: , in, To optimize the goal, is the load distribution value of device node i, is a collection of device nodes. is the edge set between device nodes, represents the load forecast error of node i, is the optimization variable of node i, represents the collaboration cost function between nodes i and j, is the optimization variable of node j.

[0047] By defining the local optimization objective function Taking the error between the actual load and the predicted load of each equipment node as the optimization target can ensure the accuracy of the load forecasting results, solve the problem that the error of a single node cannot be effectively controlled in traditional load forecasting, and improve the accuracy of load distribution.

[0048] Collaboration cost function between nodes By penalizing the square of the node load distribution difference, the load transmission intensity between nodes can be dynamically balanced, the system instability caused by load differences can be reduced, the load collaboration cost between nodes can be effectively optimized, and the overall operation efficiency of the system can be improved.

[0049] The global optimization framework uses the objective function The solution can ensure that the local load error of each node is minimized, optimize the global load distribution through the collaborative relationship between nodes, achieve global coordination of load forecasting and regulation, and resolve the conflict between a single node and the overall system goals in the industrial load system.

[0050] The results of local optimization are calculated in real time by edge devices, reducing the computing pressure and data transmission delay of the cloud. The synergy between edge devices and the cloud can timely adjust the load control strategy according to dynamic load changes, enhancing the real-time and adaptive capabilities of the industrial load control system.

[0051] The game model in step 5 is constructed by the following utility function: , in, is the utility function of device node i, is the short-term load forecast error, is the long-term load trend error, is the communication overhead, , and is the weight parameter.

[0052] The utility function is optimized by the following equilibrium condition: , , , in, , and is the weight parameter, is the utility function of device node i, is the weight coefficient of the utility function for short-term error The partial derivative of is the weight coefficient of the utility function for short-term error The partial derivative of is the weight coefficient of the utility function for short-term error The partial derivative of .

[0053] The equilibrium strategy parameters generated by game optimization in the cloud are periodically sent to the edge devices. The edge devices adjust the load control strategy in real time based on the received strategy parameters and local data, and upload the load control results to the cloud for updating the next round of strategy parameters.

[0054] Through the utility function in the game model The construction of , comprehensively considers the impact of short-term load forecasting error, long-term trend error and communication overhead. Using weight parameters , and The dynamic adjustment can effectively balance the conflicts among multiple objectives and solve the problem that the traditional single-objective optimization method cannot meet the global objective requirements at the same time.

[0055] By optimizing the equilibrium condition of the utility function , , , solve the balancing strategy of each device node to maximize the benefit of each node, while the system as a whole reaches the global optimal state, significantly improve the balance of the load control strategy, and ensure the stability of system operation.

[0056] In game optimization, the utility function incorporates communication overhead , by adjusting the communication weight The optimized adjustment can dynamically control the data transmission frequency and communication volume between the cloud and edge devices, effectively reduce redundant communications, reduce system energy consumption, and ensure the real-time performance of the control strategy.

[0057] The balancing strategy parameters generated in the cloud are periodically sent to the edge devices. The edge devices adjust the load control strategy in real time based on the received strategy parameters and local data, and feed the control results back to the cloud for updating. This method ensures that the load control strategy can be continuously optimized according to the dynamic changes of the industrial system and enhances the system's adaptive ability.

[0058] By comprehensively considering short-term errors, long-term trends and communication overhead, the game model can optimize the resource allocation of the entire system, significantly reduce the operating costs of the industrial system and improve economic benefits.

[0059] Although 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 the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An industrial load control method based on cloud-edge collaborative technology, characterized in that: include: Step 1: By deploying load sensors, environmental monitoring equipment and equipment status monitoring equipment, historical load data, environmental data and equipment operation status data are collected. The historical load data is used to represent the load changes of the system at multiple time points. The environmental data includes real-time external temperature and electricity price information. The equipment operation status data includes equipment operation status parameters. The historical load data, environmental data and equipment operation status data are combined to perform data fusion and generate a unified fusion feature as the input of load forecasting. Step 2: Input the fusion features generated in step 1 into the load forecasting model, process the input features through the encoding module of the model, extract the potential load association pattern, and then generate the load forecast value at the future time point according to the potential pattern through the decoding module. The forecasting model is constructed based on the correlation between the historical change trend of the load, environmental data and equipment status; Step 3: Combine the load forecast value generated in step 2 with the industrial equipment network structure to establish a dynamic graph model that reflects the load transmission relationship between devices. The dynamic graph model includes a set of device nodes, connection relationships between nodes, and an edge weight matrix that reflects the load transmission strength between devices. The dynamic graph is updated in real time through the load forecast value and the association relationship between devices. Step 4: Based on the dynamic graph model established in step 3, define the local optimization goals of the device nodes, build the load control costs between the nodes according to the collaborative relationship between the device nodes, coordinate the load distribution of each node through the global optimization framework, and calculate the optimization results in real time by the edge device and feed back the processing results to the cloud; Step 5. Combined with the load distribution results optimized in step 4, a game model including device nodes is constructed in the cloud. As participants in the game model, the device nodes dynamically adjust the load control strategy according to the combined influence of short-term load forecast error, long-term trend deviation and communication overhead, and achieve a global equilibrium state through the game model. Finally, the control strategy parameters of cloud computing are sent to edge devices for execution to complete the global coordination and dynamic optimization of the load.

2. According to claim 1, an industrial load control method based on cloud-edge collaborative technology is characterized in that: The data fusion in step 1 is accomplished through the following weighting mechanism: ， in, For the i-th type of input data, is the number of feature categories, is the input feature vector of the i-th class, Calculated by the following formula: ， in, is the weight coefficient of each type of data, is the attention score of the i-th feature, is the number of feature categories, is an exponential function, A normalization term for the exponential value of the feature attention score.

3. According to claim 1, the industrial load control method based on cloud-edge collaborative technology is characterized in that: The load forecasting model in step 2 generates the load forecast value at a future time point through the following process: The encoder extracts the potential correlation pattern of the fused features: ， in, is a potential latent variable, is the feature mean, is the characteristic standard deviation, is random noise; The decoder generates predicted load values ​​from latent variables ,in, is the current time point, is the prediction time interval.

4. According to claim 1, the industrial load control method based on cloud-edge collaborative technology is characterized in that: The dynamic graph model established in step 3 The following relations are satisfied: ， in, is the edge weight between node i and node j at time t, and is the load value of node i and node j, Indicates the difference in node load.

5. According to claim 4, the industrial load control method based on cloud-edge collaborative technology is characterized in that: The local optimization objective function in step 4 is defined as: ， in, is the actual load of node i, is the predicted load of node i, represents the square of the load forecast error of node i, represents the load forecast error of node i, is the optimization variable of node i.

6. The industrial load control method based on cloud-edge collaborative technology according to claim 5 is characterized in that: The load regulation cost function between the nodes is defined as: ， in, represents the collaboration cost function between nodes i and j, is the load distribution value of node i, is the load distribution value of node j, is the penalty coefficient for load regulation between nodes, is the square of the difference in load distribution values ​​between nodes.

7. The industrial load control method based on cloud-edge collaborative technology according to claim 6 is characterized in that: The global optimization framework is solved by the following objective function: ， in, To optimize the goal, is the load distribution value of device node i, is a collection of device nodes. is the edge set between device nodes, represents the load forecast error of node i, is the optimization variable of node i, represents the collaboration cost function between nodes i and j, is the optimization variable of node j.

8. The industrial load control method based on cloud-edge collaborative technology according to claim 1 is characterized in that: The game model in step 5 is constructed by the following utility function: ， in, is the utility function of device node i, is the short-term load forecast error, is the long-term load trend error, is the communication overhead, , and is the weight parameter.

9. The industrial load control method based on cloud-edge collaborative technology according to claim 8 is characterized in that: The utility function is optimized by the following equilibrium condition: ， ， ， in, , and is the weight parameter, is the utility function of device node i, is the weight coefficient of the utility function for short-term error The partial derivative of is the weight coefficient of the utility function for short-term error The partial derivative of is the weight coefficient of the utility function for short-term error The partial derivative of .

10. The industrial load control method based on cloud-edge collaborative technology according to claim 9 is characterized in that: The equilibrium strategy parameters generated by the cloud through game optimization are periodically sent to the edge device. The edge device adjusts the load control strategy in real time according to the received strategy parameters combined with local data, and uploads the load control results to the cloud for updating the next round of strategy parameters.

Images