An electrical automation ecological Internet of Things system based on meta-learning multi-variable graph network

Abstract

The present invention discloses an electrical automation ecological Internet of Things system based on meta-learning polymorphic graph network, comprising an interface integration module, an ecological Internet of Things module, a polymorphic graph network module, an adaptive meta-learning module, an electrical automation module, and a real-time monitoring module. The interface integration module is used to coordinate and integrate system interfaces, the ecological Internet of Things module is used for Internet of Things data collection, storage and transmission, the polymorphic graph network module is used to process dynamic changes of Internet of Things nodes, the adaptive meta-learning module is used for transfer learning and adaptive adjustment between different electrical tasks, the electrical automation module is used for task processing and automatic control, and the real-time monitoring module is used to provide a user with a real-time monitoring interface. The present invention discloses an electrical automation ecological Internet of Things system based on meta-learning polymorphic graph network, proposes a dynamic adaptation algorithm based on polymorphic graph network to dynamically adapt to multiple nodes of the ecological Internet of Things, and proposes an adaptive task adjustment algorithm based on meta-learning to learn and adjust system tasks.

Description

Technical Field

[0001] The present invention relates to the fields of adaptive meta-learning, graph neural networks and ecological Internet of Things, and specifically to an electrical automation ecological Internet of Things system based on meta-learning variable graph networks. Background Art

[0002] Adaptive meta-learning technology is an adaptive machine learning method that aims to solve the learning and adjustment problems between different tasks in the electrical automation ecological Internet of Things system. This technology builds a meta-learning framework and focuses on task feature extraction, so that the system can better adapt to new tasks by learning from the experience of a series of tasks, and realize online learning and real-time adjustment. At the same time, it has the ability of transfer learning. Adaptive meta-learning technology makes full use of previously learned meta-knowledge to achieve rapid adaptation when solving new tasks. Through adaptive strategies, the system can automatically select appropriate learning and adjustment strategies according to the nature of the task and changes in the environment, thereby improving the overall performance of the system and providing strong support for the intelligence and adaptability of the electrical automation ecological Internet of Things system.

[0003] Graph neural network technology is a deep learning method that aims to solve the problem of processing complex graph structure data generated by nodes and edges in the ecological Internet of Things. By learning the dynamic relationship between nodes and edges in the ecological Internet of Things system, this technology can achieve the ability to dynamically adapt to multiple nodes in the ecological Internet of Things. The graph neural network can retain the topological structure information between nodes, effectively process node attributes and the relationship between nodes, and is suitable for dynamically changing graph structures, so that the system can better perceive and adapt to changes in the ecological environment, and handle fluctuations in sensor data and changes in equipment status. Combining graph neural network technology with the ecological Internet of Things, the system can more effectively handle the dynamic adaptability of multiple nodes in the ecological Internet of Things, so as to achieve intelligent perception and rapid response to changes in each node in the ecological Internet of Things.

[0004] However, the existing electrical automation ecological Internet of Things system based on meta-learning multi-variable graph network cannot achieve rapid adjustment when facing new tasks, lacks flexibility and adaptability in the face of different tasks and environments, and lacks the ability to effectively respond to the dynamic changes of electrical equipment and ecological Internet of Things nodes, which affects the system's accurate monitoring and analysis of the data of each node of the ecological Internet of Things, thereby limiting the real-time and reliability of the system. In addition, the system lacks the ability to process intelligent tasks and cannot fully understand and process the relationship between different tasks, thus affecting the overall intelligence of the system. Summary of the invention

[0005] The purpose of the present invention is to provide an electrical automation ecological Internet of Things system based on meta-learning polymorphic graph networks, so as to solve the problems proposed in the above background technology that the existing electrical automation ecological Internet of Things system based on meta-learning polymorphic graph networks cannot be quickly adjusted when facing new tasks, lacks flexibility and adaptability in the face of different tasks and environments, and the system lacks the ability to effectively respond to the dynamic changes of electrical equipment and ecological Internet of Things nodes, affecting the system's accurate monitoring and analysis of the data of each node of the ecological Internet of Things, thereby limiting the real-time and reliability of the system, and the system lacks the ability to intelligently process tasks, and cannot fully understand and process the relationship between different tasks, thereby affecting the overall intelligence of the system.

[0006] To achieve the above-mentioned purpose, the present invention provides the following technical solutions: an electrical automation ecological Internet of Things system based on meta-learning polymorphic graph network, comprising an interface integration module, an ecological Internet of Things module, a polymorphic graph network module, an adaptive meta-learning module, an electrical automation module, and a real-time monitoring module, characterized in that: the interface integration module is used to coordinate and integrate system interfaces to ensure efficient communication and data transmission between the system and different Internet of Things nodes; the ecological Internet of Things module comprises a sensor data acquisition unit and a data transmission and storage unit; the sensor data acquisition unit is used to collect various data from all nodes in the ecological Internet of Things in real time, and calibrate the sensors to ensure the accuracy and reliability of the data acquired by the system; the data transmission and storage unit is used to transmit the sensor data to the system, and perform reliable storage and efficient retrieval; the polymorphic graph network module proposes a dynamic adaptation algorithm based on the polymorphic graph network to handle the dynamic changes of different nodes in the ecological Internet of Things; Ensure the real-time adaptability of the system's graph structure to support the system's subsequent meta-learning and orderly switching of tasks. The adaptive meta-learning module proposes an adaptive task adjustment algorithm based on meta-learning, which is used for transfer learning between different electrical tasks and adaptive adjustment of the system when responding to different tasks, so that the system can quickly adapt to new tasks and effectively adapt to the requirements of various tasks. The electrical automation module includes a task identification and switching unit and an electrical automation control unit. The task identification and switching unit is used to identify the current task in real time. Based on the results of meta-learning and multi-variable graph networks, the task execution and switching strategies are designed to ensure that the system achieves a smooth transition between different tasks. The electrical automation control unit is used to execute automation control strategies and adjust the operating status of electrical equipment to meet the requirements of the current task. The real-time monitoring module is used to provide users with a real-time monitoring interface, display electrical automation and ecological Internet of Things data, so that users can monitor the system status in real time and make necessary adjustments and feedback.

[0007] Preferably, the interface integration module ensures smooth data transmission and efficient communication by coordinating and integrating the interfaces between the system and each node of the ecological Internet of Things, so as to achieve collaborative work and unified management of the system.

[0008] Preferably, the ecological Internet of Things module includes a sensor data acquisition unit, which is responsible for real-time collection of various data in the ecological Internet of Things by deploying various sensor devices, and ensures the accuracy and timeliness of the data, providing real-time multi-feature input for subsequent modules.

[0009] Preferably, the ecological Internet of Things module includes a data transmission and storage unit, which ensures the rapid transmission of data within the system by designing an efficient data transmission mechanism, and uses reliable storage solutions, including cloud storage and local database storage, to ensure the stable and long-term preservation of the collected data, so that the system can efficiently retrieve historical data for analysis, and support users' long-term monitoring and management of system data.

[0010] Preferably, the variable graph network module proposes a dynamic adaptation algorithm based on the variable graph network. By real-time management of the nodes and edges in the graph network, it ensures that the graph structure in the system can flexibly adapt to the dynamic changes of data of various electrical equipment and ecological Internet of Things nodes, and ensures that the system can maintain a high degree of flexibility and adaptability in different tasks and environments, providing reliable data support for the subsequent meta-learning of the system and the orderly switching of tasks.

[0011] Specifically, the dynamic adaptive algorithm based on the multi-variable graph network is as follows: First, the sensor data in the ecological Internet of Things is modeled and graphed, and the internal logical relationship between the sensor data of each node in the ecological Internet of Things is dynamically represented by the graph neural network. The modeling formula of the sensor data of each node is specifically expressed as follows:

[0012]

[0013] Among them, A ij Represents the adjacency matrix of the graph, which represents the connection strength between node i and node j in the graph. The Gaussian kernel function is used to model the relationship between sensor data. i 、x j With x k represents the sensor data corresponding to nodes i, j, and k. A graph is constructed through the data to reflect the relationship between nodes. exp represents the exponential function, ||·|| 2 represents the Euclidean distance between sensor data, ∑ k≠i exp(-||x i -x k || 2) represents the sum of Gaussian kernel functions for all nodes k that are different from node i, which is used for normalization processing to ensure a reasonable range of connection strength. i, j, and k all represent the indexes of nodes in the graph, corresponding to the actual sensor devices and data sources. i and j represent the indexes of two nodes used to construct the edges in the graph, respectively. k represents the index used to sum all nodes in the graph that are different from node i when calculating the Gaussian kernel function. By constructing the adjacency matrix A of the graph ij , to reflect the dynamic relationship between sensor data. The Gaussian kernel function is used in the construction process to emphasize the similarity between sensor data, which is conducive to learning the dynamic characteristics of data. The dynamic graph structure is used to model the relationship between sensor data, so that the system can adapt to different tasks and environments more flexibly, providing a basis for subsequent dynamic adaptive learning. Secondly, a dynamic adaptive mechanism is introduced to learn the potential relationship between sensor data through the graph neural network, and dynamically adjust the network structure at runtime to adapt to changes in different tasks and environments. The specific formula of the graph neural network training process is expressed as:

[0014]

[0015]

[0016] Among them, node represents each node of sensor data in the ecological Internet of Things, edge represents the edge of the connection relationship between sensor data, node→edge represents the information transmission from node to edge, and edge→node represents the information transmission from edge to node. Represents the embedding vector of edge edge(i,j) in layer l, which is used to enable the graph neural network to better understand and express the dynamic relationship between sensor data. l represents the layer index of the graph neural network, and edge(i,j) represents the edge related to nodes i and j. Represents the node edge j The embedding vector in the l+1th layer is used to learn the relationship between node j and adjacent sensor data. σ represents the sigmoid activation function. represents the weight matrix for the linear transformation of the embedded feature vector of edge edge(i,j), represents the weight matrix for the linear transformation of the embedded feature vector of node j, and Respectively represent the embedding vectors of node i and node j in layer l, x (i,j) represents the sensor data that exists between nodes i and j, N j represents the set of nodes adjacent to node j, represents the bias term associated with the edge edge of the lth layer, Represents the bias term associated with the edge node of the lth layer, and the adjacency matrix A of the graph ij Introduced into the graph neural network training, the model can more flexibly consider the similarities between sensor data when learning the intrinsic relationship of sensor data, so as to improve the adaptability of the model to dynamic environments. Then, by constructing a variational autoencoder to limit the distribution of potential relationships, the system's learning performance of complex nonlinear relationships between sensor data of each node in the ecological Internet of Things is improved. At the same time, the multi-layer perceptron decoder is used to accurately predict the system state. The specific formula of the constructed variational autoencoder is expressed as:

[0017]

[0018] Among them, p(x) represents the distribution of the potential relationship, the probability distribution function constructed by the variational autoencoder, ψ represents the hyperparameter that controls the distribution, which is used to balance the smoothness of the potential relationship and the complexity of the model, m represents the dimension of the potential relationship, and s represents the index of the potential relationship dimension. represents the standard deviation of the potential relationship on each dimension, represents the mean of the potential relationship in each dimension, log represents the logarithmic function, and the potential relationship is restricted by introducing the regularization mechanism of the variational autoencoder to improve the system's learning performance of the complex nonlinear relationship of sensor data. The prediction formula of the multilayer perceptron decoder is specifically expressed as:

[0019]

[0020] in, represents the predicted state of node i and node j in the t+1th iteration process, It represents the state of the actual sensor data of node i and node j in the tth iteration process, represents the mapping function in the multilayer perceptron decoder, which is used to map the edge embedding vector to the node state prediction, Represents the embedded vector of edge edge(i,j) during the tth iteration, represents the information stored in edge edge(i,j) during the tth iteration. Through the prediction of the multi-layer perceptron decoder, the system uses the information based on the learned edges to predict the state of the node, further improving the system's ability to predict complex dynamic changes. Finally, the loss function is used to balance the model's learning of potential relationships and accurate prediction of system states, ensuring that the graph neural network can not only judge the inherent logical relationship between sensor data during training, but also accurately predict the system state. The specific formula of the loss function is expressed as:

[0021]

[0022] Among them, L represents the loss function, c represents the weight coefficient of the Kulbach-Leibler divergence in the total loss, which is used to balance the contribution of the Kulbach-Leibler divergence and the mean square error in the loss function. The expected function of the Kulbaklei-Buler divergence representing the potential relationship represents the difference between the distribution of the potential relationship learned by the graph neural network and the standard normal distribution. P(x) represents the distribution of the potential variables learned by the graph neural network through training. Q(x) represents the standard normal distribution. N represents the total sample size of the sensor data of each node in the ecological Internet of Things. n represents the sample quantity index of the sensor data of each node in the ecological Internet of Things. x(n) represents the actual data of the sensors after integration of each node in the ecological Internet of Things. y(n) represents the predicted data of the sensors after integration of each node in the ecological Internet of Things. The dynamic adaptation algorithm based on the multi-variable graph network has the advantage of high adaptability in the electrical automation ecological Internet of Things system based on the meta-learning multi-variable graph network. By comprehensively considering the dynamic characteristics and multi-dimensional logical relationships of the sensor data of each node, it provides strong support for the intelligence and flexibility of the system.

[0023] Preferably, the adaptive meta-learning module proposes an adaptive task adjustment algorithm based on meta-learning, which automatically adjusts the system's learning strategy and model parameters by analyzing the characteristics and requirements of different tasks to achieve transfer learning and adaptive adjustment between different electrical tasks, ensuring that the system can learn and adjust quickly and efficiently when faced with new tasks.

[0024] Specifically, the meta-learning-based adaptive task adjustment algorithm is as follows: First, the tasks and environments in the electrical automation ecological Internet of Things system are modeled so that the system can understand and adapt to the changing electrical tasks and environmental conditions. Task modeling enables the system to understand the operation of electrical equipment, and environmental modeling enables the system to effectively respond to task switching in unknown scenarios and the conditions of the ecological Internet of Things in different locations. The specific formulas of the task measurement model and the environmental model are expressed as follows:

[0025] y i =σ(Φ(x i )ω+b)+n i

[0026] f(x i )=σ(Φ(x i )ω+b)

[0027] Among them, y i represents the measured value of the sensor at position i in the ecological Internet of Things, σ(·) represents the activation function that introduces nonlinear relationships, Φ(x i ) represents the actual sensor data x of the unknown scene function at position i i The basis function set of the features at x irepresents the actual sensor data at position i, i represents the location index of the sensor in the ecological Internet of Things, ω represents the weight vector, b represents the bias term, and n i represents the Gaussian noise at position i, f(x i ) represents the unknown scenario function, and a flexible nonlinear model is used to better judge the complexity of the behavior of electrical equipment. Secondly, a meta-learning model is constructed so that the system can learn knowledge from previous tasks and quickly adapt to new tasks. The specific formula of the meta-learning model update mechanism is expressed as:

[0028]

[0029] Among them, ω t+1 represents the weight vector in the t+1th iteration process, represents the parameters of the meta-learning model, arg represents the variable operation, represents the variable u that minimizes the expression in (·), u represents the temporary variable for weight update, N represents the number of tasks involved in the system, n represents the index of the task involved in the system, t represents the iteration number index of the meta-learning model, and L n (u) represents the loss function of the nth task, representing the prediction error on the task, γ represents the regularization parameter, which is used to balance the task loss and the regularization term, Γ(u) represents the regularization term, which is used to control the sparsity of the weight, η t represents the learning rate of the meta-learning model during the tth iteration, ω t represents the weight vector in the tth iteration process, Represents the estimated gradient during the tth iteration, which is used to guide the weight ω t Update direction, Represents the square calculation of the Euclidean norm. The weight vector is updated through the meta-learning model to adapt to the new task, and the sparsity of the weight vector is maintained in this process. The system can learn how to adjust the weight to adapt to different tasks to improve the adaptability of the entire system. Then, the relationship between the task model performance and the communication cost in the ecological Internet of Things is weighed by introducing a reward and penalty mechanism in the meta-learning process to effectively deal with the problem of increased communication costs in the overall ecological Internet of Things when switching tasks. The specific formula of the constructed reward and penalty is expressed as:

[0030] P(ω)=λL(ω)+(1-λ)Γ(ω)

[0031] Among them, P(ω) represents the loss function jointly affected by the task model loss and communication cost, L(ω) represents the performance function of the system when executing the task, Γ(ω) represents the communication overhead function generated by the system during perception and learning, λ represents the reward parameter that weighs the task model performance and communication cost. The reward part focuses on the loss of the task model, (1-λ) is the penalty parameter, which focuses on the system communication cost. Finally, by establishing an adaptive perception scheme to intelligently adjust the data collection strategy of the sensor, reduce the spatial sampling of the ecological Internet of Things, and improve data utilization, so as to adapt to the switching between different tasks and improve the overall work efficiency of the system. The specific formula of the adaptive perception scheme is expressed as:

[0032]

[0033] Among them, π t represents the sampling distribution, represents the perception scheme adopted by the system in the tth iteration, ρ represents the adjustment parameter used to balance uniform sampling and sampling based on relative importance, π represents a fixed uniform sampling distribution, and g t (n) represents the relative information value of task n during the tth iteration. Through the adaptive perception scheme, the system can actively determine the perception position with the most information value, thereby significantly reducing the spatial sampling of the ecological Internet of Things and improving the efficiency and adaptability of the system. The adaptive task adjustment algorithm based on meta-learning enables the electrical automation ecological Internet of Things system to adapt to various complex tasks and dynamic environments more intelligently and efficiently through intelligent learning and adaptive adjustment, which helps to enhance the generalization and adaptability of the overall system.

[0034] Preferably, the electrical automation module includes a task identification and switching unit, which ensures that the system can quickly and accurately identify and understand the current electrical automation ecological Internet of Things tasks by utilizing the knowledge information learned by the adaptive meta-learning module, and make accurate decisions for subsequent automation control.

[0035] Preferably, the electrical automation module includes an electrical automation control unit, which performs intelligent control on each electrical device by executing an adaptive adjustment strategy to ensure that the system can effectively control the state and behavior of each electrical device according to the requirements of the current task.

[0036] Preferably, the real-time monitoring module ensures that users can obtain real-time feedback on the system operation status and task execution status by providing users with a real-time monitoring interface, which helps users to promptly identify potential problems, make decisions and adjust system parameters to ensure stable operation of the system and flexible adaptation to different scenarios.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] 1. The multi-variable graph network module proposes a dynamic adaptation algorithm based on a multi-variable graph network. First, the dynamic adaptation algorithm of the multi-variable graph network realizes the modeling of the inherent logical relationship between the sensor data of each node of the ecological Internet of Things through the graph neural network. In the constructed graph neural network, the nodes of the graph represent the sensor data, and the edges represent the relationship between different sensor data. In this way, the dynamic adaptation algorithm based on the multi-variable graph network can not only pay attention to the time characteristics of the sensor data of each node, but also infer the logical relationship between data of different dimensions, thereby improving the system's ability to model complex relationships. Secondly, the algorithm introduces a dynamic adaptability mechanism, which allows the system to dynamically adjust the graph neural network structure during runtime to adapt to different tasks and environments. This dynamic adaptability enables the system to adapt to electrical automation tasks and ecological environments more flexibly. Changes in the IoT environment enhance the adaptability and robustness of the system. Furthermore, the algorithm uses variational autoencoders and Gumbel sampling techniques to limit the distribution of potential relationships. Through the coordination of the encoder and decoder, the encoder learns the potential relationship between sensor data, and the decoder predicts the system state to improve the learning performance of the algorithm, so that the system can more accurately learn the complex nonlinear relationship between the sensor data of each node, so as to better adapt to meta-learning tasks and sensor data in different environments. In general, the dynamic adaptation algorithm based on the multi-variable graph network has the advantage of high adaptability in the electrical automation ecological IoT system based on the meta-learning multi-variable graph network. By comprehensively considering the dynamic characteristics and multi-dimensional logical relationships of the sensor data of each node, it provides strong support for the intelligence and flexibility of the system.

[0039] 2. Adaptive meta-learning module proposes an adaptive task adjustment algorithm based on meta-learning. Through the meta-learning mechanism, the algorithm enables the system to learn and continuously optimize the task processing strategy to adapt to the ever-changing tasks and environmental conditions in the electrical automation ecological Internet of Things system. The meta-learning mechanism of the algorithm enables the system to dynamically adjust the control algorithm and parameters for different electrical tasks, thereby improving the automation control efficiency of electrical equipment. Through real-time learning and adaptive methods, the system can quickly adapt to new tasks, reduce response time, and improve control accuracy, thereby improving the adaptability of the overall system. Using the idea of ​​meta-learning, the data collection strategy of the sensor is intelligently adjusted through an adaptive perception scheme. The meta-learning mechanism enables the system to actively determine the perception position with the most information value, thereby significantly reducing the spatial sampling of the ecological Internet of Things. Through the learning process, The system can identify important node data, optimize data collection methods, improve data utilization, and reduce communication costs, laying the foundation for the efficient operation of the overall system. At the same time, a machine learning model based on stochastic gradient descent is used to predict the task model to reduce prediction errors, while the reinforcement meta-learning model optimizes perception decisions by rewarding the error reduction of the obtained samples and punishing the corresponding communication costs. This enables the system to more intelligently adjust task processing and perception strategies, realize transfer learning between tasks, and improve the adaptability and robustness of the system. Overall, the meta-learning-based adaptive task adjustment algorithm enables the electrical automation ecological Internet of Things system to adapt to various complex tasks and dynamic environments more intelligently and efficiently through intelligent learning and adaptive adjustment, which helps to enhance the generalization and adaptability of the overall system. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The invention is further illustrated by the accompanying drawings, but the embodiments in the accompanying drawings do not constitute any limitation to the invention. A person skilled in the art can obtain other drawings based on the following drawings without creative work.

[0041] Figure 1 It is a schematic diagram of the structure of the present invention. DETAILED DESCRIPTION

[0042] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0043] See also Figure 1The present invention provides an electrical automation ecological Internet of Things system based on meta-learning multi-variable graph network, including an interface integration module, an ecological Internet of Things module, a multi-variable graph network module, an adaptive meta-learning module, an electrical automation module, and a real-time monitoring module. The interface integration module is used to coordinate and integrate system interfaces to ensure efficient communication and data transmission between the system and different Internet of Things nodes. The ecological Internet of Things module includes a sensor data acquisition unit and a data transmission and storage unit. The sensor data acquisition unit is used to collect various data from all nodes in the ecological Internet of Things in real time, and calibrate the sensors to ensure the accuracy and reliability of the data acquired by the system. The data transmission and storage unit is used to transmit the sensor data to the system, and perform reliable storage and efficient retrieval. The multi-variable graph network module proposes a dynamic adaptation algorithm based on the multi-variable graph network to process the dynamic changes of different nodes in the ecological Internet of Things and ensure the real-time monitoring of the graph structure in the system. The adaptive meta-learning module proposes an adaptive task adjustment algorithm based on meta-learning, which is used for transfer learning between different electrical tasks and adaptive adjustment of the system when dealing with different tasks, so that the system can quickly adapt to new tasks and effectively adapt to the requirements of various tasks. The electrical automation module includes a task identification and switching unit and an electrical automation control unit. The task identification and switching unit is used to identify the current task in real time. Based on the results of meta-learning and multi-variable graph networks, the task execution and switching strategies are designed to ensure smooth transition between different tasks. The electrical automation control unit is used to execute automation control strategies and adjust the operating status of electrical equipment to meet the requirements of the current task. The real-time monitoring module is used to provide users with a real-time monitoring interface, display electrical automation and ecological Internet of Things data, so that users can monitor the system status in real time and make necessary adjustments and feedback.

[0044] See also Figure 1 ,Furthermore, the interface integration module ensures the smooth ,transmission and efficient communication of data by coordinating and integrating the ,interfaces between the system and the nodes of the ecological Internet of ,Things to achieve the collaborative work and unified management of the ,system.

[0045] See also Figure 1 ,Furthermore, the ecological Internet of Things module includes a sensor data acquisition unit, ,which is responsible for real-time collection of various data in the ecological Internet of Things by ,deploying various types of sensor devices, and ensuring the accuracy and ,timeliness of the data, providing real-time multi-feature input for ,subsequent modules.

[0046] See also Figure 1Furthermore, the ecological Internet of Things module includes a data transmission and storage unit, which ensures the rapid transmission of data within the system by designing an efficient data transmission mechanism, and uses reliable storage solutions, including cloud storage and local database storage, to ensure the stable and long-term preservation of the collected data, so that the system can efficiently retrieve historical data for analysis and support users' long-term monitoring and management of system data.

[0047] See also Figure 1 ,Furthermore, the multi-variable graph network module proposes a dynamic ,adaptation algorithm based on the multi-variable graph network. ,By real-time management of the nodes and edges in the graph network, ,it ensures that the graph structure in the system can flexibly ,adapt to the dynamic changes of the data of different ,electrical equipment and the ecological Internet of Things nodes, and ,ensures that the system can maintain a high flexibility and adaptability in different ,tasks and environments, and provides reliable data support for the subsequent ,meta-learning and orderly switching of tasks of the system.

[0048] See also Figure 1 ,Furthermore, the dynamic adaptation algorithm based on the multi-variable graph network is as follows: ,First, the sensor data in the ecological Internet of Things is modeled and ,graphed. The internal logical relationship between the sensor data of each ,node in the ecological Internet of Things is dynamically represented by the ,graph neural network. The modeling formula of the sensor data of each node is ,specifically expressed as:

[0049]

[0050] Among them, A ij Represents the adjacency matrix of the graph, which represents the connection strength between node i and node j in the graph. The Gaussian kernel function is used to model the relationship between sensor data. i 、x j With x k represents the sensor data corresponding to nodes i, j, and k. A graph is constructed through the data to reflect the relationship between nodes. exp represents the exponential function, ||·|| 2 represents the Euclidean distance between sensor data, ∑ k≠i exp(-||x i -x k || 2 ) represents the sum of Gaussian kernel functions for all nodes k that are different from node i, which is used for normalization processing to ensure a reasonable range of connection strength. i, j, and k all represent the indexes of nodes in the graph, corresponding to the actual sensor devices and data sources. i and j represent the indexes of two nodes used to construct the edges in the graph, respectively. k represents the index used to sum all nodes in the graph that are different from node i when calculating the Gaussian kernel function. By constructing the adjacency matrix A of the graph ij, to reflect the dynamic relationship between sensor data. The Gaussian kernel function is used in the construction process to emphasize the similarity between sensor data, which is conducive to learning the dynamic characteristics of data. The dynamic graph structure is used to model the relationship between sensor data, so that the system can adapt to different tasks and environments more flexibly, providing a basis for subsequent dynamic adaptive learning. Secondly, a dynamic adaptive mechanism is introduced to learn the potential relationship between sensor data through the graph neural network, and dynamically adjust the network structure at runtime to adapt to changes in different tasks and environments. The specific formula of the graph neural network training process is expressed as:

[0051]

[0052]

[0053] Among them, node represents each node of sensor data in the ecological Internet of Things, edge represents the edge of the connection relationship between sensor data, node→edge represents the information transmission from node to edge, and edge→node represents the information transmission from edge to node. Represents the embedding vector of edge edge(i,j) in layer l, which is used to enable the graph neural network to better understand and express the dynamic relationship between sensor data. l represents the layer index of the graph neural network, and edge(i,j) represents the edge related to nodes i and j. Represents the node edge j The embedding vector in the l+1th layer is used to learn the relationship between node j and adjacent sensor data. σ represents the sigmoid activation function. represents the weight matrix for the linear transformation of the embedded feature vector of edge edge(i,j), represents the weight matrix for the linear transformation of the embedded feature vector of node j, and Respectively represent the embedding vectors of node i and node j in layer l, x (i,j) represents the sensor data that exists between nodes i and j, N j represents the set of nodes adjacent to node j, represents the bias term associated with the edge edge of the lth layer, Represents the bias term associated with the edge node of the lth layer, and the adjacency matrix A of the graph ijIntroduced into the graph neural network training, the model can more flexibly consider the similarities between sensor data when learning the intrinsic relationship of sensor data, so as to improve the adaptability of the model to dynamic environments. Then, by constructing a variational autoencoder to limit the distribution of potential relationships, the system's learning performance of complex nonlinear relationships between sensor data of each node in the ecological Internet of Things is improved. At the same time, the multi-layer perceptron decoder is used to accurately predict the system state. The specific formula of the constructed variational autoencoder is expressed as:

[0054]

[0055] Among them, p(x) represents the distribution of the potential relationship, the probability distribution function constructed by the variational autoencoder, ψ represents the hyperparameter that controls the distribution, which is used to balance the smoothness of the potential relationship and the complexity of the model, m represents the dimension of the potential relationship, and s represents the index of the potential relationship dimension. represents the standard deviation of the potential relationship on each dimension, represents the mean of the potential relationship in each dimension, log represents the logarithmic function, and the potential relationship is restricted by introducing the regularization mechanism of the variational autoencoder to improve the system's learning performance of the complex nonlinear relationship of sensor data. The prediction formula of the multilayer perceptron decoder is specifically expressed as:

[0056]

[0057] in, represents the predicted state of node i and node j in the t+1th iteration process, It represents the state of the actual sensor data of node i and node j in the tth iteration process, represents the mapping function in the multilayer perceptron decoder, which is used to map the edge embedding vector to the node state prediction, Represents the embedded vector of edge edge(i,j) during the tth iteration, represents the information stored in edge edge(i,j) during the tth iteration. Through the prediction of the multi-layer perceptron decoder, the system uses the information based on the learned edges to predict the state of the node, further improving the system's ability to predict complex dynamic changes. Finally, the loss function is used to balance the model's learning of potential relationships and accurate prediction of system states, ensuring that the graph neural network can not only judge the inherent logical relationship between sensor data during training, but also accurately predict the system state. The specific formula of the loss function is expressed as:

[0058]

[0059] Among them, L represents the loss function, c represents the weight coefficient of the Kulbach-Leibler divergence in the total loss, which is used to balance the contribution of the Kulbach-Leibler divergence and the mean square error in the loss function. The expected function of the Kulbaklei-Buler divergence representing the potential relationship represents the difference between the distribution of the potential relationship learned by the graph neural network and the standard normal distribution. P(x) represents the distribution of the potential variables learned by the graph neural network through training. Q(x) represents the standard normal distribution. N represents the total sample size of the sensor data of each node in the ecological Internet of Things. n represents the sample quantity index of the sensor data of each node in the ecological Internet of Things. x(n) represents the actual data of the sensors after integration of each node in the ecological Internet of Things. y(n) represents the predicted data of the sensors after integration of each node in the ecological Internet of Things. The dynamic adaptation algorithm based on the multi-variable graph network has the advantage of high adaptability in the electrical automation ecological Internet of Things system based on the meta-learning multi-variable graph network. By comprehensively considering the dynamic characteristics and multi-dimensional logical relationships of the sensor data of each node, it provides strong support for the intelligence and flexibility of the system.

[0060] See also Figure 1 ,Furthermore, the adaptive meta-learning module proposes an adaptive task ,adjustment algorithm based on meta-learning, which automatically adjusts the system’s ,learning strategy and model parameters by analyzing the characteristics and ,requirements of different tasks to achieve transfer learning and adaptive ,adjustment between different electrical tasks, ensuring that the system can ,learn and adjust quickly and efficiently when facing new tasks.

[0061] See also Figure 1 ,Furthermore, the adaptive task adjustment algorithm based on meta-learning is as follows: First, the tasks and environment in the electrical automation ecological Internet of Things system are modeled so that the system can understand and adapt to the changing electrical tasks and environmental conditions. Task modeling enables the system to understand the operation of electrical equipment, and environmental modeling enables the system to effectively respond to task switching in unknown scenarios and the situation of its ecological Internet of Things in different locations. The specific formulas of the task measurement model and the environmental model are expressed as:

[0062] y i =σ(Φ(x i )ω+b)+n i

[0063] f(x i )=σ(Φ(x i )ω+b)

[0064] Among them, y i represents the measured value of the sensor at position i in the ecological Internet of Things, σ(·) represents the activation function that introduces nonlinear relationships, Φ(x i ) represents the actual sensor data x of the unknown scene function at position i iThe basis function set of the features at x i represents the actual sensor data at position i, i represents the location index of the sensor in the ecological Internet of Things, ω represents the weight vector, b represents the bias term, and n i represents the Gaussian noise at position i, f(x i ) represents the unknown scenario function, and a flexible nonlinear model is used to better judge the complexity of the behavior of electrical equipment. Secondly, a meta-learning model is constructed so that the system can learn knowledge from previous tasks and quickly adapt to new tasks. The specific formula of the meta-learning model update mechanism is expressed as:

[0065]

[0066] Among them, ω t+1 represents the weight vector in the t+1th iteration process, represents the parameters of the meta-learning model, arg represents the variable operation, represents the variable u that minimizes the expression in (·), u represents the temporary variable for weight update, N represents the number of tasks involved in the system, n represents the index of the task involved in the system, t represents the iteration number index of the meta-learning model, and L n (u) represents the loss function of the nth task, representing the prediction error on the task, γ represents the regularization parameter, which is used to balance the task loss and the regularization term, Γ(u) represents the regularization term, which is used to control the sparsity of the weight, η t represents the learning rate of the meta-learning model during the tth iteration, ω t represents the weight vector in the tth iteration process, Represents the estimated gradient during the tth iteration, which is used to guide the weight ω t Update direction, Represents the square calculation of the Euclidean norm. The weight vector is updated through the meta-learning model to adapt to the new task, and the sparsity of the weight vector is maintained in this process. The system can learn how to adjust the weight to adapt to different tasks to improve the adaptability of the entire system. Then, the relationship between the task model performance and the communication cost in the ecological Internet of Things is weighed by introducing a reward and penalty mechanism in the meta-learning process to effectively deal with the problem of increased communication costs in the overall ecological Internet of Things when switching tasks. The specific formula of the constructed reward and penalty is expressed as:

[0067] P(ω)=λL(ω)+(1-λ)Γ(ω)

[0068] Among them, P(ω) represents the loss function jointly affected by the task model loss and communication cost, L(ω) represents the performance function of the system when executing the task, Γ(ω) represents the communication overhead function generated by the system during perception and learning, λ represents the reward parameter that weighs the task model performance and communication cost. The reward part focuses on the loss of the task model, (1-λ) is the penalty parameter, which focuses on the system communication cost. Finally, by establishing an adaptive perception scheme to intelligently adjust the data collection strategy of the sensor, reduce the spatial sampling of the ecological Internet of Things, and improve data utilization, so as to adapt to the switching between different tasks and improve the overall work efficiency of the system. The specific formula of the adaptive perception scheme is expressed as:

[0069]

[0070] Among them, π t represents the sampling distribution, represents the perception scheme adopted by the system in the tth iteration, ρ represents the adjustment parameter used to balance uniform sampling and sampling based on relative importance, π represents a fixed uniform sampling distribution, and g t (n) represents the relative information value of task n during the tth iteration. Through the adaptive perception scheme, the system can actively determine the perception position with the most information value, thereby significantly reducing the spatial sampling of the ecological Internet of Things and improving the efficiency and adaptability of the system. The adaptive task adjustment algorithm based on meta-learning enables the electrical automation ecological Internet of Things system to adapt to various complex tasks and dynamic environments more intelligently and efficiently through intelligent learning and adaptive adjustment, which helps to enhance the generalization and adaptability of the overall system.

[0071] See also Figure 1 ,Furthermore, the electrical automation module includes a task identification and switching unit, ,which ensures that the system can quickly and accurately identify and understand the ,current electrical automation ecological Internet of Things tasks by utilizing the ,knowledge information learned by the adaptive meta-learning module to make accurate ,decisions for subsequent automation control.

[0072] See also Figure 1 ,Furthermore, the electrical automation module includes an electrical automation control unit, which ,intelligently controls each electrical device by executing an adaptive ,adjustment strategy to ensure that the system can effectively control the ,state and behavior of each electrical device according to the requirements of the ,current task.

[0073] See also Figure 1 ,Furthermore, the real-time monitoring module ensures that users can obtain real-time feedback on the ,system operation status and task execution by providing users with a real-time ,monitoring interface, which helps users to promptly identify potential problems, make ,decisions and adjust system parameters to ensure the ,stable operation of the system and flexible adaptation to different ,scenarios.

[0074] When used specifically, first, the interface integration module coordinates and integrates the system interface to ensure efficient communication and data transmission between the system and different IoT nodes. Then, the ecological IoT module includes a sensor data acquisition unit and a data transmission and storage unit. The sensor data acquisition unit collects various data from all nodes in the ecological IoT in real time, and calibrates the sensor to ensure the accuracy and reliability of the data acquired by the system. The data transmission and storage unit transmits the sensor data to the system, and performs reliable storage and efficient retrieval. Secondly, the multi-variable graph network module proposes a dynamic adaptation algorithm based on a multi-variable graph network to process the dynamic changes of different nodes in the ecological IoT, ensuring the real-time adaptability of the graph structure in the system to support the subsequent meta-learning of the system and the orderly switching of tasks. The adaptive meta-learning module provides An adaptive task adjustment algorithm based on meta-learning is proposed to realize transfer learning between different electrical tasks and adaptive adjustment of the system when responding to different tasks, so that the system can quickly adapt to new tasks and effectively adapt to the requirements of various tasks. Finally, the electrical automation module includes a task identification and switching unit and an electrical automation control unit. The task identification and switching unit identifies the current task in real time and designs task execution and switching strategies based on the results of meta-learning and multi-variable graph networks to ensure smooth transition between different tasks. The electrical automation control unit executes automation control strategies and adjusts the operating status of electrical equipment to meet the requirements of the current task. The real-time monitoring module provides users with a real-time monitoring interface, displays electrical automation and ecological Internet of Things data, so that users can monitor the system status in real time and make necessary adjustments and feedback.

[0075] Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. An electrical automation ecological Internet of Things system based on meta-learning variable graph network, comprising an interface integration module, an ecological Internet of Things module, a variable graph network module, an adaptive meta-learning module, an electrical automation module, and a real-time monitoring module, characterized in that: The interface integration module is used to coordinate and integrate system interfaces to ensure efficient communication and data transfer between the system and different IoT nodes. The ecological IoT module includes a sensor data acquisition unit and a data transmission and storage unit. The sensor data acquisition unit is used to collect various data from all nodes in the ecological IoT in real time, and calibrate the sensors to ensure the accuracy and reliability of the data acquired by the system. The data transmission and storage unit is used to transmit the sensor data to the system, and perform reliable storage and efficient retrieval. The variable graph network module proposes a dynamic adaptation algorithm based on a variable graph network to handle the dynamic changes of different nodes in the ecological IoT, ensure the real-time adaptability of the graph structure in the system, and support the subsequent meta-learning of the system and the orderly switching of tasks. The adaptive meta-learning module proposes a dynamic adaptation algorithm based on a variable graph network. The meta-learning adaptive task adjustment algorithm is used for transfer learning between different electrical tasks and adaptive adjustment of the system in response to different tasks, so that the system can quickly adapt to new tasks and effectively adapt to the requirements of various tasks. The electrical automation module includes a task identification and switching unit and an electrical automation control unit. The task identification and switching unit is used to identify the current task in real time. Based on the results of meta-learning and multi-variable graph networks, the task execution and switching strategy is designed to ensure that the system can smoothly transition between different tasks. The electrical automation control unit is used to execute the automation control strategy and adjust the operating status of the electrical equipment to meet the requirements of the current task. The real-time monitoring module is used to provide a user with a real-time monitoring interface, display electrical automation and ecological Internet of Things data, so that the user can monitor the system status in real time and make necessary adjustments and feedback; The variable graph network module proposes a dynamic adaptation algorithm based on the variable graph network. By real-time management of the nodes and edges in the graph network, it ensures that the graph structure in the system can flexibly adapt to the dynamic changes of data of different electrical equipment and ecological Internet of Things nodes, ensuring that the system can maintain a high degree of flexibility and adaptability in different tasks and environments, and providing reliable data support for the subsequent meta-learning of the system and the orderly switching of tasks; The dynamic adaptive algorithm based on the multivariable graph network is specifically as follows: First, the sensor data in the ecological Internet of Things is modeled and graphed, and the internal logical relationship between the sensor data of each node in the ecological Internet of Things is dynamically represented by the graph neural network; the modeling formula of the sensor data of each node is specifically expressed as: Among them, A ij Represents the adjacency matrix of the graph, which represents the connection strength between node i and node j in the graph. The Gaussian kernel function is used to model the relationship between sensor data. i 、x j With x k represents the sensor data corresponding to nodes i, j, and k. A graph is constructed through the data to reflect the relationship between nodes. exp represents the exponential function, ||·|| 2 represents the Euclidean distance between sensor data, ∑ k≠i exp(-||x i -x k || 2 ) represents the sum of Gaussian kernel functions for all nodes k that are different from node i, which is used for normalization processing to ensure a reasonable range of connection strength. i, j, and k all represent the indexes of nodes in the graph, corresponding to the actual sensor devices and data sources. i and j represent the indexes of two nodes used to construct the edges in the graph, respectively. k represents the index used to sum all nodes in the graph that are different from node i when calculating the Gaussian kernel function. By constructing the adjacency matrix A of the graph ij , to reflect the dynamic relationship between sensor data; Gaussian kernel function is used in the construction process to emphasize the similarity between sensor data, which is conducive to learning the dynamic characteristics of data. Dynamic graph structure is used to model the relationship between sensor data, so that the system can adapt to different tasks and environments more flexibly, providing a basis for subsequent dynamic adaptive learning; Secondly, a dynamic adaptive mechanism is introduced to learn the potential relationship between sensor data through the graph neural network, and dynamically adjust the network structure at runtime to adapt to changes in different tasks and environments; the specific formula of the graph neural network training process is expressed as: Among them, node represents each node of sensor data in the ecological Internet of Things, edge represents the edge of the connection relationship between sensor data, node→edge represents the information transmission from node to edge, and edge→node represents the information transmission from edge to node. Represents the embedding vector of edge edge(i, j) in layer l, which is used to enable the graph neural network to better understand and express the dynamic relationship between sensor data. l represents the layer index of the graph neural network, and edge(i, j) represents the edge related to nodes i and j. Represents the node edge j The embedding vector in the l+1th layer is used to learn the relationship between node j and adjacent sensor data. σ represents the sigmoid activation function. represents the weight matrix for the linear transformation of the embedded feature vector of edge edge(i, j), represents the weight matrix for the linear transformation of the embedded feature vector of node j, and Respectively represent the embedding vectors of node i and node j in layer l, x (i，j) represents the sensor data that exists between nodes i and j, N j represents the set of nodes adjacent to node j, represents the bias term associated with the edge edge of the lth layer, Represents the bias term associated with the edge node of the lth layer; the adjacency matrix A of the graph ij Introducing it into graph neural network training enables the model to more flexibly consider the similarities between sensor data when learning the intrinsic relationships of sensor data, thereby improving the model's adaptability to dynamic environments; Then, a variational autoencoder is constructed to restrict the distribution of potential relationships, improve the system's learning performance of complex nonlinear relationships between sensor data of each node in the ecological Internet of Things, and realize accurate prediction of the system state through a multi-layer perceptron decoder; the specific formula of the constructed variational autoencoder is expressed as: Among them, p(x) represents the distribution of the potential relationship, the probability distribution function constructed by the variational autoencoder, ψ represents the hyperparameter that controls the distribution, which is used to balance the smoothness of the potential relationship and the complexity of the model, m represents the dimension of the potential relationship, and s represents the index of the potential relationship dimension. represents the standard deviation of the potential relationship on each dimension, represents the mean of the potential relationship in each dimension, and log represents the logarithmic function. By introducing the regularization mechanism of the variational autoencoder, the potential relationship is restricted to improve the system's learning performance of the complex nonlinear relationship of sensor data. The prediction formula of the multilayer perceptron decoder is specifically expressed as: in, represents the predicted status of node i and node j in the t+1th iteration process, It represents the state of the actual sensor data of node i and node j in the tth iteration process, represents the mapping function in the multilayer perceptron decoder, which is used to map the edge embedding vector to the node state prediction, The embedded vector of edge(i, j) in the tth iteration process represents the information stored in edge(i, j) in the tth iteration process. Through the prediction of the multi-layer perceptron decoder, the system uses the information based on the learned edge to predict the state of the node, further improving the system's ability to predict complex dynamic changes. Finally, the loss function is used to balance the model's learning of potential relationships and accurate prediction of system states, ensuring that the graph neural network can not only judge the inherent logical relationship between sensor data during training, but also accurately predict the system state; the specific formula of the loss function is expressed as: Among them, L represents the loss function, c represents the weight coefficient of the Kulbach-Leibler divergence in the total loss, which is used to balance the contribution of the Kulbach-Leibler divergence and the mean square error in the loss function. The expected function of the Kulbaklei-Buller divergence representing the potential relationship represents the difference between the distribution of the potential relationship learned by the graph neural network and the standard normal distribution. P(x) represents the distribution of the potential variables learned by the graph neural network through training. Q(x) represents the standard normal distribution. N represents the total sample size of the sensor data of each node in the ecological Internet of Things. n represents the sample number index of the sensor data of each node in the ecological Internet of Things. x(n) represents the actual data of the sensor after integration of each node in the ecological Internet of Things. y(n) represents the predicted data of the sensor after integration of each node in the ecological Internet of Things. The dynamic adaptation algorithm based on the multi-variable graph network has the advantage of high adaptability in the electrical automation ecological Internet of Things system based on the meta-learning multi-variable graph network. By comprehensively considering the dynamic characteristics and multi-dimensional logical relationships of the sensor data of each node, it provides strong support for the intelligence and flexibility of the system. The adaptive meta-learning module proposes an adaptive task adjustment algorithm based on meta-learning. By analyzing the characteristics and requirements of different tasks, it automatically adjusts the system's learning strategy and model parameters to achieve transfer learning and adaptive adjustment between different electrical tasks, ensuring that the system can learn and adjust quickly and efficiently when facing new tasks. The specific adaptive task adjustment algorithm based on meta-learning is as follows: First, the tasks and environment in the electrical automation ecological Internet of Things system are modeled so that the system can understand and adapt to the changing electrical tasks and environmental conditions. Task modeling enables the system to understand the operation of electrical equipment, and environmental modeling enables the system to effectively respond to task switching in unknown scenarios and the situation of the ecological Internet of Things in different locations. The specific formulas of the task measurement model and the environmental model are expressed as follows: y i =σ(Φ(x i )ω+b)+n i f(x i )=σ(Φ(x i )ω+b) Among them, y i represents the measured value of the sensor at position i in the ecological Internet of Things, σ(·) represents the activation function that introduces nonlinear relationships, Φ(x i ) represents the actual sensor data x of the unknown scene function at position i i The basis function set of the features at x i represents the actual sensor data at position i, i represents the location index of the sensor in the ecological Internet of Things, ω represents the weight vector, b represents the bias term, and n i represents the Gaussian noise at position i, f(x i ) represents the unknown scenario function and adopts flexible nonlinear models to better judge the complexity of electrical equipment behavior; Secondly, a meta-learning model is constructed so that the system can learn knowledge from previous tasks and quickly adapt to new tasks; the specific formula of the meta-learning model update mechanism is expressed as: Among them, ω t+1 represents the weight vector in the t+1th iteration process, represents the parameters of the meta-learning model, arg represents the variable operation, represents the variable u that minimizes the expression in (·), u represents the temporary variable for weight update, N represents the number of tasks involved in the system, n represents the index of the task involved in the system, t represents the iteration number index of the meta-learning model, and L n (u) represents the loss function of the nth task, representing the prediction error on the task, γ represents the regularization parameter, which is used to balance the task loss and the regularization term, Γ(u) represents the regularization term, which is used to control the sparsity of the weight, η t represents the learning rate of the meta-learning model during the tth iteration, ω t represents the weight vector in the tth iteration process, Represents the estimated gradient during the tth iteration, which is used to guide the weight ω t Update direction, Represents the square calculation of the Euclidean norm; by updating the weight vector to adapt to new tasks through the meta-learning model and maintaining the sparsity of the weight vector in the process, the system can learn how to adjust the weights to adapt to different tasks to improve the adaptability of the entire system; Then, the reward and penalty mechanism is introduced to weigh the relationship between the task model performance and the communication cost in the ecological Internet of Things during the meta-learning process, so as to effectively deal with the problem of increased communication cost in the overall ecological Internet of Things when switching tasks; the specific formula of the constructed reward and penalty is expressed as: P(ω)=λL(ω)+(1-λ)Γ(ω) Among them, P(ω) represents the loss function jointly affected by the task model loss and communication cost, L(ω) represents the performance function of the system when executing the task, Γ(ω) represents the communication overhead function generated by the system during the perception and learning process, λ represents the reward parameter that weighs the task model performance and communication cost. The reward part focuses on the loss of the task model, and (1-λ) is the penalty parameter, which focuses on the system communication cost. Finally, by establishing an adaptive perception solution to intelligently adjust the data collection strategy of the sensor, reduce the spatial sampling of the ecological Internet of Things, and improve data utilization, so as to adapt to the switching between different tasks and improve the overall work efficiency of the system. The specific formula of the adaptive perception solution is expressed as: Among them, π t represents the sampling distribution, represents the perception scheme adopted by the system in the tth iteration, ρ represents the adjustment parameter used to balance uniform sampling and sampling based on relative importance, π represents a fixed uniform sampling distribution, and g t (n) represents the relative information value of task n during the tth iteration; through the adaptive perception scheme, the system can actively determine the perception position with the most information value, thereby significantly reducing the spatial sampling of the ecological Internet of Things and improving the efficiency and adaptability of the system. The adaptive task adjustment algorithm based on meta-learning enables the electrical automation ecological Internet of Things system to adapt to various complex tasks and dynamic environments more intelligently and efficiently through intelligent learning and adaptive adjustment, which helps to enhance the generalization and adaptability of the overall system.

2. The electrical automation ecological Internet of Things system based on meta-learning multi-variable graph network according to claim 1 is characterized by: The interface integration module coordinates and integrates the interfaces between the system and each node of the ecological Internet of Things to ensure smooth data transmission and efficient communication, so as to achieve collaborative work and unified management of the system.

3. The electrical automation ecological Internet of Things system based on meta-learning multi-variable graph network according to claim 1 is characterized by: The ecological Internet of Things module includes a sensor data acquisition unit. The sensor data acquisition unit is responsible for real-time collection of various data in the ecological Internet of Things by deploying various sensor devices, ensuring the accuracy and timeliness of the data, and providing real-time multi-feature input for subsequent modules.

4. The electrical automation ecological Internet of Things system based on meta-learning multi-variable graph network according to claim 1, characterized in that: The ecological Internet of Things module includes a data transmission and storage unit, which ensures the rapid transmission of data within the system by designing an efficient data transmission mechanism, and uses reliable storage solutions, including cloud storage and local database storage, to ensure the stable and long-term preservation of the collected data, so that the system can efficiently retrieve historical data for analysis and support users' long-term monitoring and management of system data.

5. The electrical automation ecological Internet of Things system based on meta-learning multi-variable graph network according to claim 1, characterized in that: The electrical automation module includes a task identification and switching unit, which ensures that the system quickly and accurately identifies and understands the current electrical automation ecological Internet of Things tasks by utilizing the knowledge information learned by the adaptive meta-learning module, and makes accurate decisions for subsequent automation control.

6. The electrical automation ecological Internet of Things system based on meta-learning multi-variable graph network according to claim 1 is characterized by: The electrical automation module includes an electrical automation control unit, which intelligently controls each electrical device by executing an adaptive adjustment strategy to ensure that the system effectively controls the state and behavior of each electrical device according to the requirements of the current task; The real-time monitoring module ensures that users can obtain real-time feedback on the system operation status and task execution by providing users with a real-time monitoring interface, which helps users to promptly identify potential problems, make decisions and adjust system parameters to ensure the stable operation of the system and flexible adaptation to different scenarios.

Images