Monitoring terminal status control method, device, equipment and storage medium

Abstract

The present invention relates to the technical field of monitoring terminal control, and discloses a state control method, device, equipment and storage medium for a monitoring terminal. The method comprises: collecting multidimensional environmental data sets and equipment state data of multiple monitoring terminals; performing nonlinear time-varying function calculation to obtain a terminal urgency value of each monitoring terminal; performing threshold comparison and adaptive backoff analysis to obtain a trigger signal and priority for updating the state of each monitoring terminal; performing nonlinear state transfer model training to obtain an adaptive state estimator; obtaining an initial state estimation value and performing mixed uncertainty modeling and multi-objective optimization calculation to obtain an optimal state estimation value; performing distributed consistency optimization and abnormal event propagation analysis to obtain a global collaborative control strategy for multiple monitoring terminals. The present invention realizes efficient management and intelligent decision-making of a large-scale monitoring network, and greatly improves the performance and reliability of the monitoring system.

Description

Technical Field

[0001] The present invention relates to the field of monitoring terminal control technology, and in particular to a monitoring terminal state control method, device, equipment and storage medium. Background Art

[0002] With the rapid development of the Internet of Things and smart cities, monitoring terminals are playing an increasingly important role in urban management, public safety, and environmental monitoring. However, traditional monitoring terminal status control methods often use fixed sampling frequencies and single threshold judgment strategies, which are difficult to adapt to complex and changing environments and network conditions. This method is not only prone to waste of resources, but may also lead to the loss or delayed transmission of key information.

[0003] In addition, existing monitoring terminals usually adopt an independent operation mode and lack an effective coordination mechanism. When faced with a large-scale, distributed monitoring network, it is difficult to achieve global optimal resource allocation and information processing. This isolated operation mode limits the overall performance of the monitoring system and cannot give full play to the advantages of networked monitoring. Another prominent problem is that the current monitoring terminal state control method has a relatively slow response to abnormal events and lacks the ability to actively predict and quickly adjust. When faced with emergencies or drastic changes in the environment, the system often shows low sensitivity and adaptability, and it is difficult to take effective countermeasures in time. This not only affects the reliability of the monitoring system, but also increases the difficulty of management and maintenance. Summary of the invention

[0004] The present invention provides a state control method, device, equipment and storage medium for a monitoring terminal, which realizes efficient management and intelligent decision-making of a large-scale monitoring network and greatly improves the performance and reliability of the monitoring system.

[0005] In a first aspect, the present invention provides a state control method for a monitoring terminal, the state control method for a monitoring terminal comprising:

[0006] Collect multi-dimensional environmental data sets and equipment status data from multiple monitoring terminals;

[0007] Performing nonlinear time-varying function calculation on the multidimensional environment data set and the device status data to obtain a terminal urgency value of each monitoring terminal;

[0008] Perform threshold comparison and adaptive backoff analysis on the terminal urgency value of each monitoring terminal to obtain the trigger signal and priority of each monitoring terminal status update;

[0009] Performing nonlinear state transfer model training on the plurality of monitoring terminals according to the trigger signal and the priority to obtain an adaptive state estimator;

[0010] Obtaining the initial state estimation value output by the adaptive state estimator and performing hybrid uncertainty modeling and multi-objective optimization calculation to obtain the optimal state estimation value of each monitoring terminal;

[0011] Distributed consistency optimization and abnormal event propagation analysis are performed on the optimal state estimation value to obtain a global collaborative control strategy for the multiple monitoring terminals.

[0012] In a second aspect, the present invention provides a state control device for a monitoring terminal, the state control device for the monitoring terminal comprising:

[0013] A collection module, used to collect multi-dimensional environmental data sets and equipment status data from multiple monitoring terminals;

[0014] A calculation module, used for performing nonlinear time-varying function calculation on the multi-dimensional environment data set and the device status data to obtain a terminal emergency value of each monitoring terminal;

[0015] An analysis module, used to perform threshold comparison and adaptive backoff analysis on the terminal urgency value of each monitoring terminal, and obtain a trigger signal and priority for updating the status of each monitoring terminal;

[0016] A training module, used for performing nonlinear state transfer model training on the plurality of monitoring terminals according to the trigger signal and the priority to obtain an adaptive state estimator;

[0017] A modeling module, used to obtain the initial state estimation value output by the adaptive state estimator and perform hybrid uncertainty modeling and multi-objective optimization calculation to obtain the optimal state estimation value of each monitoring terminal;

[0018] The optimization module is used to perform distributed consistency optimization and abnormal event propagation analysis on the optimal state estimation value to obtain a global collaborative control strategy for the multiple monitoring terminals.

[0019] The third aspect of the present invention provides a computer device, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the computer device executes the above-mentioned state control method of the monitoring terminal.

[0020] A fourth aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the above-mentioned state control method of a monitoring terminal.

[0021] In the technical solution provided by the present invention, by calculating the nonlinear time-varying function of the multidimensional environmental data set and the device status data, combined with the adaptive sampling strategy, the environmental changes can be captured more accurately, the collection of redundant data can be reduced, and the data quality and resource utilization efficiency can be improved. The dynamic threshold and adaptive backoff analysis are adopted, so that the system can automatically adjust the state update strategy according to the real-time situation, and the response speed and adaptability of the monitoring terminal to environmental changes are improved. Through the training of nonlinear state transfer model and hybrid uncertainty modeling, combined with multi-objective optimization calculation, a more accurate state estimation value can be obtained, providing a reliable basis for subsequent decision-making. Through distributed consistency optimization and abnormal event propagation analysis, the system can realize the optimal allocation of resources and efficient processing of information in a global scope, and improve the collaborative efficiency of the entire monitoring network. Combined with complex network propagation analysis and multi-agent reinforcement learning, the system can actively predict the impact range of abnormal events and quickly generate distributed emergency response strategies, thereby improving the security and reliability of the monitoring network. By adopting a hierarchical communication matrix and a robust control gain design, the system can still maintain stable performance in the face of network topology changes or local failures, thereby enhancing the reliability and anti-interference ability of the entire monitoring system. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0023] Figure 1 A schematic diagram of the steps of a method for controlling the state of a monitoring terminal according to an embodiment of the present invention;

[0024] Figure 2 Schematic diagram of the structure of the state control device of the monitoring terminal in an embodiment of the present invention. DETAILED DESCRIPTION

[0025] Embodiments of the present invention provide a state control method, device, equipment and storage medium for a monitoring terminal. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or precedence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0026] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1 , an embodiment of the state control method of the monitoring terminal in the embodiment of the present invention includes:

[0027] Step S1, collecting multi-dimensional environmental data sets and equipment status data of multiple monitoring terminals;

[0028] It is understandable that the execution subject of the present invention may be a state control device of a monitoring terminal, or may be a terminal or a server, which is not limited here. The embodiment of the present invention is described by taking a server as the execution subject as an example.

[0029] Specifically, the sensor arrays of multiple monitoring terminals are periodically activated to obtain the original environmental data stream, and the data sampling frequency is adaptively adjusted according to the original environmental data stream to obtain a dynamic sampling strategy to ensure that the collected data has high real-time and accuracy. The collected environmental data is subjected to multi-dimensional feature extraction and noise reduction processing to obtain a structured environmental feature vector. In this process, signal processing technology and machine learning algorithms are used to filter out noise and redundant information in the original data and extract representative and critical features. According to the structured environmental feature vector, the abnormal event detection model is updated in real time to obtain an optimized abnormal detection threshold. By continuously adjusting and optimizing the detection model, the sensitivity and accuracy of the system to abnormal events can be improved, and potential risks can be discovered and responded to in a timely manner. The hardware status parameters of the monitoring terminal are periodically scanned and encoded to obtain the device status code. According to the device status code, the terminal resource allocation scheme is dynamically optimized to obtain a resource utilization efficiency indicator. Through resource allocation, the overall efficiency and reliability of the system are maximized to ensure the smooth execution of various tasks. The structured environmental feature vector and the device status code are timestamped and encrypted to obtain a secure data packet. Timestamps can ensure the timing consistency of data, while encryption can effectively prevent data leakage and tampering during transmission, ensuring data security and integrity. According to resource utilization efficiency indicators and network conditions, the transmission of secure data packets is optimized to obtain multi-dimensional environmental data sets and device status data. By optimizing the transmission strategy, efficient data transmission is achieved in different network environments, ensuring that the status control system of the monitoring terminal can obtain and process necessary data in real time, achieving accurate and efficient status control.

[0030] Step S2, performing nonlinear time-varying function calculation on the multi-dimensional environment data set and the device status data to obtain the terminal emergency value of each monitoring terminal;

[0031] Specifically, the multidimensional environmental data set is subjected to principal component analysis and feature normalization. Through principal component analysis, the variables with high correlation in the original data are converted into a set of unrelated principal components to reduce the dimension of the data; then the environmental feature vector after dimension reduction is subjected to feature normalization to ensure that different features are compared and calculated at the same scale. The environmental feature vector after dimension reduction is evaluated for environmental abnormality. By evaluating the abnormality of each feature vector, the environmental abnormality is obtained, which can reflect the degree of deviation of the current environment from the normal state. At the same time, the state urgency of the equipment status data is evaluated. According to the operating status and performance indicators of the equipment, the terminal state urgency index is calculated, which can reflect the current health status and urgency of the equipment. The historical network quality data of the monitoring terminal is analyzed and predicted in time series. By analyzing the changing trend of network quality in the past period of time, the future network quality is predicted to obtain the network quality factor. The task importance analysis is performed on the monitoring task types of the monitoring terminal, and the task importance weight matrix is ​​established according to the priority and importance of different tasks. A multi-input single-output initial urgency calculation model is constructed by comprehensively considering the environmental anomaly, terminal status urgency index, network quality factor and task importance weight matrix. The model calculates the initial urgency of each monitoring terminal by comprehensively considering the weights and influences of various indicators. The initial urgency calculation model is jointly trained, and the model parameters are continuously adjusted so that the model can more accurately reflect the relationship and influence between various indicators, and the optimized urgency calculation model is obtained. According to the optimized urgency calculation model, the multi-dimensional environmental data set and device status data are calculated respectively, and the terminal urgency value of each monitoring terminal is calculated by comprehensively considering multiple factors such as environment, device status, network quality and task importance.

[0032] Step S3: performing threshold comparison and adaptive backoff analysis on the terminal urgency value of each monitoring terminal to obtain a trigger signal and priority for updating the status of each monitoring terminal;

[0033] Specifically, the terminal urgency value of each monitoring terminal is subjected to sequence analysis and clustering processing, the distribution characteristics and trends of the urgency value are identified, and the initial interval of the dynamic threshold is obtained. The terminal urgency value is classified according to the initial interval of the dynamic threshold to obtain a multi-level threshold classification result, which reflects the emergency situation of each monitoring terminal in more detail. The multi-level threshold classification result is subjected to entropy weight analysis, and the importance weight of each level of threshold is obtained by calculating the entropy weight of each level of threshold, and the initial interval of the dynamic threshold is adaptively adjusted according to the importance weight to obtain an optimized dynamic threshold set, so that terminals with different urgency levels can obtain corresponding processing priorities. The terminal urgency value of each monitoring terminal is compared with the optimized dynamic threshold set to obtain an initial trigger signal. The initial trigger signal reflects whether each monitoring terminal needs to update its status in the current state and the urgency of the update. According to the initial trigger signal and the historical network load data, an adaptive backoff strategy is constructed, and the use of network resources is understood by analyzing the historical network load data, so that when the network load is heavy, some status update operations are appropriately delayed to avoid network congestion. The optimal backoff time analysis of the adaptive backoff strategy is performed to obtain the optimal backoff time sequence, which guides the system on how to reasonably arrange the status update operation under different network load conditions, ensuring that the utilization efficiency of network resources is maximized while ensuring timely handling of emergencies. Priority calculation and conflict detection are performed based on the optimal backoff time sequence and the initial trigger signal, and the urgency of each terminal and the network load are comprehensively considered to obtain the trigger signal and priority of the status update of each monitoring terminal. Through the above steps, the status update operations of each monitoring terminal are effectively coordinated to ensure the stable operation and efficient response of the system in various complex environments.

[0034] Step S4: training a nonlinear state transfer model for multiple monitoring terminals according to the trigger signal and the priority to obtain an adaptive state estimator;

[0035] Specifically, nonlinear feature extraction and dimensionality reduction processing are performed on the device state data to extract the key features in the data and reduce the dimension of the data to obtain a low-dimensional state representation vector. Based on the low-dimensional state representation vector, an initial nonlinear state transfer model is constructed, which includes an input layer, a recurrent neural network and an output layer. The input layer is used to receive the low-dimensional state representation vector, and the recurrent neural network processes the time series data through its internal loop structure to capture the dynamic changes and dependencies in the data. The output layer generates the corresponding state prediction results. The trigger signal and priority are encoded and embedded to obtain the state update trigger feature. Through encoding and embedding processing, the trigger signal and priority are converted into numerical features suitable for the input model, and the model's understanding and utilization of these information are enhanced. According to the state update trigger feature, the initial nonlinear state transfer model is trained by conditional generative adversarial network to improve the generalization ability and robustness of the model. The conditional generative adversarial network continuously optimizes the state transfer model through adversarial training of the generator and the discriminator, so that it can more accurately capture and predict the state changes of the monitoring terminal, and obtain an enhanced state transfer model. Hyperparameter optimization is performed on the enhanced state transfer model, and the optimal model structure and parameter set are obtained by adjusting and optimizing the hyperparameters of the model. Through hyperparameter optimization, the most suitable model configuration and parameter settings for the current task are found. Based on the optimal model structure and parameter set, the initial state estimator is generated. Based on the optimized model structure and parameter set, the estimator can efficiently estimate and predict the state of the monitoring terminal. Online transfer learning and incremental learning are performed on the initial state estimator, so that the estimator can continuously adapt to new data and environmental changes and improve its dynamic response ability. Online transfer learning can enable the model to quickly adjust its own parameters and structure when facing new data, maintain efficient estimation performance, while incremental learning can gradually accumulate new knowledge and experience without forgetting existing knowledge. Based on online transfer learning and incremental learning, a dynamic update mechanism is formed, so that the model can be continuously updated and optimized. According to the dynamic update mechanism and the distribution of historical estimation errors, uncertainty quantification analysis is performed. By analyzing the distribution characteristics of historical estimation errors, the uncertainty of the estimation results is quantified, the reliability and accuracy of the estimator are improved, and the adaptive state estimator is obtained.

[0036] Step S5, obtaining the initial state estimation value output by the adaptive state estimator and performing hybrid uncertainty modeling and multi-objective optimization calculation to obtain the optimal state estimation value of each monitoring terminal;

[0037] Specifically, the initial state estimation value output by the adaptive state estimator is subjected to interval analysis and probability statistics to determine the upper and lower limits of the initial state estimation value and its probability distribution, and obtain the set bounded uncertainty description and random uncertainty distribution. Based on the set bounded uncertainty description and random uncertainty distribution, a hybrid uncertainty representation is constructed. The hybrid uncertainty representation can comprehensively reflect the different sources and properties of uncertainty. Information entropy analysis and integral calculation are performed on the hybrid uncertainty representation to quantify the comprehensive degree of uncertainty and obtain a comprehensive uncertainty measurement index. Information entropy analysis can capture the amount of information of uncertainty, while integral calculation can integrate the influence of various uncertainties, so that the comprehensive uncertainty measurement index can fully reflect the uncertainty level of the current state. Based on the comprehensive uncertainty measurement index and the physical constraints of the monitoring terminal, a multi-objective optimization problem is constructed to obtain an optimization objective function set. The construction of the multi-objective optimization problem can simultaneously consider multiple optimization objectives, ensure that a balanced solution is found between different objectives, and achieve the optimal state estimation. The optimization objective function set is initialized by non-dominated sorting genetic initialization to obtain the initial Pareto front solution set. The non-dominated sorting genetic algorithm can effectively find multiple non-dominated solutions to the optimization problem by simulating the natural selection process, so that the initial Pareto front solution set can cover multiple excellent areas of the solution space. Adaptive differential evolution is performed based on the initial Pareto front solution set. Through continuous mutation and selection, the solution set can gradually tend to the global optimum and obtain the target Pareto optimal solution set. The adaptive differential evolution algorithm makes the evolution process more flexible and efficient by adaptively adjusting parameters, so that it can quickly converge to the optimal solution set. The optimal state estimation is solved for the target Pareto optimal solution set, and the optimal state estimation candidate value is determined by specific analysis and evaluation of each solution. The candidate value represents the most likely optimal state estimation under the current conditions. Exponential smoothing prediction is performed on the optimal state estimation candidate value. By weighted averaging of historical data, the prediction result is made smoother and more stable, and the optimal state estimation value of each monitoring terminal is obtained. The exponential smoothing prediction method can better capture the changing trend of the current state and improve the accuracy and timeliness of the prediction by giving higher weights to the latest data.

[0038] Step S6: Perform distributed consistency optimization and abnormal event propagation analysis on the optimal state estimate to obtain a global collaborative control strategy for multiple monitoring terminals.

[0039] Specifically, the network topology of multiple monitoring terminals is analyzed by graph theory to identify the connection relationship and hierarchical structure between nodes in the network, and a hierarchical communication matrix is ​​obtained. The hierarchical communication matrix can accurately describe the communication path and hierarchical relationship between each monitoring terminal. According to the hierarchical communication matrix, the optimal state estimate is processed by distributed average consistency. By exchanging and averaging the state information of each monitoring terminal, the consistency state is obtained to ensure that the state of each node in the entire network is consistent, thereby improving the stability and coordination of the system. Robust control gain design and stability analysis are performed on the consistency state. By designing appropriate control gains and parameters, the system's resistance to external interference and internal fluctuations is enhanced to ensure that the system can remain stable in various complex environments, and the convergence speed optimization parameters are obtained. Based on the convergence speed optimization parameters, a target consistency protocol is generated, which is used to guide each monitoring terminal on how to adjust and update its own state during the distributed control process to achieve the purpose of rapid convergence and stable control. The target consistency protocol is subjected to execution process anomaly detection and fault diagnosis. By real-time monitoring of various abnormal conditions and faults during the execution of the protocol, possible problems in the system can be discovered and identified in a timely manner, and the abnormal event feature vector is obtained. Perform complex network propagation analysis on the feature vector of abnormal events. By analyzing the propagation path and impact range of abnormal events in the network, predict the impact that abnormal events may have on the system and obtain the prediction results of the impact range of abnormal events. Perform multi-agent reinforcement learning on the prediction results of the impact range of abnormal events. By simulating the interaction and learning process of multiple agents in a complex environment, optimize the emergency response strategy of the system and obtain a distributed emergency response strategy. Multi-agent reinforcement learning enables the system to quickly adjust and optimize its own control strategy in the face of various emergencies, thereby improving the emergency response capability of the system. Perform decision analysis based on the distributed emergency response strategy and consistency state. By comprehensively considering the current system state and emergency response strategy, the optimal control decision is formulated to obtain a global collaborative control strategy for multiple monitoring terminals.

[0040] In the embodiment of the present invention, by calculating the nonlinear time-varying function of the multidimensional environmental data set and the device status data, combined with the adaptive sampling strategy, the environmental changes can be captured more accurately, the collection of redundant data can be reduced, and the data quality and resource utilization efficiency can be improved. The dynamic threshold and adaptive backoff analysis are adopted, so that the system can automatically adjust the state update strategy according to the real-time situation, and the response speed and adaptability of the monitoring terminal to environmental changes are improved. Through the training of nonlinear state transfer model and hybrid uncertainty modeling, combined with multi-objective optimization calculation, a more accurate state estimation value can be obtained, providing a reliable basis for subsequent decision-making. Through distributed consistency optimization and abnormal event propagation analysis, the system can achieve optimal resource allocation and efficient information processing in a global scope, and improve the collaborative efficiency of the entire monitoring network. Combined with complex network propagation analysis and multi-agent reinforcement learning, the system can actively predict the impact range of abnormal events and quickly generate distributed emergency response strategies, thereby improving the security and reliability of the monitoring network. By adopting a hierarchical communication matrix and robust control gain design, the system can still maintain stable performance in the face of network topology changes or local failures, thereby enhancing the reliability and anti-interference ability of the entire monitoring system.

[0041] In a specific embodiment, the process of executing step S1 may specifically include the following steps:

[0042] (1) The sensor arrays of multiple monitoring terminals are periodically activated to obtain the original environmental data stream, and the data sampling frequency is adaptively adjusted according to the original environmental data stream to obtain a dynamic sampling strategy;

[0043] (2) Perform multi-dimensional feature extraction and noise reduction on the collected environmental data to obtain a structured environmental feature vector, and update the abnormal event detection model in real time according to the structured environmental feature vector to obtain an optimized abnormality detection threshold;

[0044] (3) Periodically scan and encode the hardware status parameters of the monitoring terminal to obtain the device status code, and dynamically optimize the terminal resource allocation plan based on the device status code to obtain the resource utilization efficiency index;

[0045] (4) The structured environmental feature vectors and device status codes are timestamped and encrypted to obtain secure data packets, and the secure data packets are optimized for transmission based on resource utilization efficiency indicators and network conditions to obtain multidimensional environmental data sets and device status data.

[0046] Specifically, the sensor array of the monitoring terminal is periodically activated at a preset time interval to capture various parameters in the environment, such as temperature, humidity, light intensity, air quality, etc., to generate an original environmental data stream. According to the collected original environmental data stream, the data sampling frequency is adjusted in real time using an adaptive adjustment algorithm. The adaptive adjustment algorithm can dynamically adjust the sampling frequency according to the change rate and fluctuation of the environmental data to ensure that the sampling frequency is increased when the data changes drastically to capture more details, and the sampling frequency is reduced when the data changes steadily to reduce data redundancy and resource consumption. For example, the following formula is used for adaptive adjustment:

[0047]

[0048] Where f(t) represents the sampling frequency at time t, f0 is the initial sampling frequency, k is the adjustment coefficient, ΔE(t) is the change in environmental data at the current moment, and E avg is the average value of the environmental data. In this way, the data sampling frequency is adaptively adjusted to obtain a dynamic sampling strategy. The collected environmental data is subjected to multi-dimensional feature extraction and noise reduction processing to obtain a structured environmental feature vector. Multi-dimensional feature extraction uses principal component analysis, independent component analysis and other methods to extract representative and distinguishing feature dimensions from the original data; noise reduction processing can use filtering, detrending and other methods to remove noise and interference in the data. For example, using principal component analysis for feature extraction, its mathematical expression is:

[0049] Z = XW;

[0050] Among them, Z is the eigenvector after dimensionality reduction, X is the original environmental data matrix, and W is the eigenvector matrix. In this way, a structured environmental feature vector is obtained. According to the structured environmental feature vector, the abnormal event detection model is updated in real time to obtain the optimized abnormal detection threshold. The abnormal event detection model can use machine learning algorithms, such as support vector machines, neural networks, etc., and update model parameters and thresholds in real time to enable it to accurately identify abnormal situations in environmental data. For example, assuming that a support vector machine is used for anomaly detection, its decision function is:

[0051] f(x)=w T x+b;

[0052] Among them, w is the weight vector, x is the input feature vector, and b is the bias term.

[0053] By continuously updating the values ​​of w and b, the model can detect abnormal events more accurately. The hardware status parameters of the monitoring terminal are periodically scanned and encoded to obtain the device status code. Hardware status parameters include CPU usage, memory usage, hard disk space, battery power, etc. These parameters can reflect the operating status of the monitoring terminal. By periodically scanning and encoding these parameters, a device status code is generated to intuitively represent the current status of the device. For example, a hash function can be used to encode the hardware status parameters to generate a unique status code. According to the device status code, the terminal resource allocation plan is dynamically optimized to obtain a resource utilization efficiency indicator. The resource allocation plan may include CPU time slice allocation, memory allocation, network bandwidth allocation, etc. By dynamically adjusting these resource allocation plans according to the device status code, the resource utilization efficiency can be improved. For example, a linear programming method can be used to optimize resource allocation, and its objective function is:

[0054]

[0055] Among them, u i is the utilization efficiency of the i-th resource, x i is the allocation of the i-th resource. By solving the linear programming problem, the resource utilization efficiency index is obtained. The structured environmental feature vector and device status code are timestamped and encrypted to obtain a secure data packet. Timestamping can ensure the temporal consistency of the data, and encryption can protect the privacy and security of the data. Symmetric encryption algorithms (such as AES) or asymmetric encryption algorithms (such as RSA) are used to encrypt the data. For example, the process of encryption using AES is as follows:

[0056] C=AES encrypt (P, K);

[0057] Among them, C is the encrypted data packet, P is the original data packet, and K is the encryption key. According to the resource utilization efficiency index and network status, the transmission of the secured data packet is optimized to obtain a multi-dimensional environmental data set and device status data. Transmission optimization can include selecting the optimal transmission path, adjusting the transmission rate, and using data compression techniques to ensure the efficiency and reliability of data during transmission. For example, a dynamic routing algorithm is used to select the optimal transmission path, and its objective function is:

[0058]

[0059] Among them, d i is the delay of the ith path. By solving the optimization problem, the optimal transmission path is determined and the multi-dimensional environmental data set and device status data are obtained.

[0060] In a specific embodiment, the process of executing step S2 may specifically include the following steps:

[0061] (1) Perform principal component analysis and feature normalization on the multidimensional environmental data set to obtain the reduced-dimensional environmental feature vector, and then evaluate the environmental anomaly degree of the reduced-dimensional environmental feature vector to obtain the environmental anomaly degree;

[0062] (2) Evaluate the status urgency of the device status data to obtain the terminal status urgency index, and perform time series analysis and prediction on the historical network quality data of the monitoring terminal to obtain the network quality factor;

[0063] (3) Perform task importance analysis on the monitoring task types of the monitoring terminal to obtain a task importance weight matrix;

[0064] (4) Based on the environmental abnormality, terminal status urgency index, network quality factor, and task importance weight matrix, a multi-input single-output initial urgency calculation model is constructed;

[0065] (5) The initial urgency calculation model is jointly trained to obtain an optimized urgency calculation model, and the multi-dimensional environmental data set and the device status data are calculated according to the optimized urgency calculation model to obtain the terminal urgency value of each monitoring terminal.

[0066] Specifically, principal component analysis is performed on environmental data to retain the most representative features of the data by projecting high-dimensional data into low-dimensional space. Suppose there is an environmental data matrix X, each row of which is a sample and each column is a feature. The goal of principal component analysis is to find a transformation matrix W that maximizes the variance of the projected data Z = XW. Calculate the covariance matrix of the data matrix X:

[0067]

[0068] Then perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and eigenvectors. Select the eigenvectors corresponding to the first k largest eigenvalues ​​as the transformation matrix W to obtain the data after dimensionality reduction. Perform feature normalization on the environmental feature vector after dimensionality reduction so that each feature has the same scale to avoid unreasonable impact of certain features on model training due to different dimensions. Perform mean normalization and variance normalization on each eigenvector to obtain a standardized environmental feature vector. Perform environmental anomaly assessment on the environmental feature vector after dimensionality reduction. Environmental anomaly assessment can adopt distance-based methods, such as anomaly detection based on Euclidean distance. Calculate the Euclidean distance between each sample point and the center of the data set. The larger the distance, the higher the anomaly of the sample point. Let the data after dimensionality reduction be Z and the center of the data set be:

[0069]

[0070] Then the abnormality of the i-th sample δ i It can be expressed as Evaluate the state urgency of the device state data to obtain the terminal state urgency index. The state urgency evaluation can be based on the device's operating parameters, such as CPU usage, memory usage, battery power, etc. Define a comprehensive index η, for example, by combining various parameters in a weighted average manner. Suppose the device state data is S, each row is a sample, each column is a state parameter, and a weight vector w is defined, then the state urgency η i It can be expressed as η i =w T s i . Perform time series analysis and prediction on the historical network quality data of the monitoring terminal to obtain the network quality factor. Time series analysis can use methods such as the autoregressive moving average model to predict the network quality in the future. Assuming that the historical network quality data is N, the autoregressive moving average model can be expressed as:

[0071]

[0072] Where N t is the network quality at time t, φ i and θ j are the autoregressive and moving average coefficients, ∈ t is white noise. By fitting the autoregressive moving average model, the future network quality is predicted to obtain the network quality factor. The task importance analysis is performed on the monitoring task types of the monitoring terminal to obtain the task importance weight matrix. The task importance analysis is based on factors such as the urgency of the task and the scope of influence, and the weight matrix T is defined. For example, the importance weight of each task is determined by expert scoring or analysis based on historical data. According to the environmental abnormality δ i , terminal status urgency index η i , Network quality factor N t and the task importance weight matrix T, construct a multi-input single-output initial urgency calculation model. This model can use a neural network model such as a multi-layer perceptron. Suppose the input feature is x i =[δ i , η i , N t , T], the output is the terminal urgency value τ i , then the forward propagation process of the model can be expressed as τ i =f(Wx i +b), where W and b are model parameters and f is the activation function. The initial urgency calculation model is jointly trained and the loss function is minimized through optimization algorithms such as gradient descent:

[0073]

[0074] where τ i is the actual urgency value, is the model prediction value. After training, the optimized urgency calculation model is obtained. According to the optimized urgency calculation model, the multi-dimensional environment data set and the device status data are calculated respectively to obtain the terminal urgency value τ of each monitoring terminal i .

[0075] In a specific embodiment, the process of executing step S3 may specifically include the following steps:

[0076] (1) performing sequence analysis and clustering processing on the terminal urgency value of each monitoring terminal to obtain a dynamic threshold initial interval, and classifying the terminal urgency value according to the dynamic threshold initial interval to obtain a multi-level threshold classification result;

[0077] (2) Performing entropy weight analysis on the multi-level threshold classification results to obtain the importance weights of each level of thresholds, and adaptively adjusting the initial interval of the dynamic threshold according to the importance weights to obtain the optimized dynamic threshold set;

[0078] (3) Compare the terminal urgency value of each monitored terminal with the optimized dynamic threshold set to obtain an initial trigger signal;

[0079] (4) construct an adaptive backoff strategy based on the initial trigger signal and historical network load data, and perform optimal backoff time analysis on the adaptive backoff strategy to obtain the optimal backoff time series;

[0080] (5) Priority calculation and conflict detection are performed based on the optimal backoff time sequence and the initial trigger signal to obtain the trigger signal and priority for the status update of each monitoring terminal.

[0081] Specifically, the terminal urgency value of each monitoring terminal is analyzed in time series to identify the trend and periodic changes of the urgency value. The time series data is smoothed by moving average method, exponential smoothing method and other methods to extract the trend component of the data. Assume that the terminal urgency value sequence is {τ t}, the smoothed sequence is in α is the smoothing coefficient. Clustering is performed on the smoothed urgency value sequence to obtain the initial interval of the dynamic threshold. Clustering can use methods such as K-means clustering and Gaussian mixture model to divide the data into several clusters, each cluster corresponding to a threshold interval. Suppose there are k clusters, and the cluster centers are {c1, c2, ..., c k}, then the threshold interval of each cluster is [c i -δ i , c i+δ i ], where δ i is the radius of cluster i. In this way, the initial interval of the dynamic threshold is obtained. According to the initial interval of the dynamic threshold, the terminal urgency values ​​are classified to obtain the multi-level threshold classification result. i According to the threshold interval to which they belong, they are classified into different levels. Assuming there are L levels, the classification result of each level is {C1, C2, ..., C L}. Perform entropy weight analysis on the multi-level threshold classification results to obtain the importance weights of each level of threshold. Entropy weight analysis is a commonly used weight determination method. By calculating the information entropy of each classification result, its importance weight is determined. Suppose the classification result of the i-th level threshold is C i , its information entropy is:

[0082] H i =-∑ j p ij log(p ij );

[0083] where p ij is the probability of the jth sample in the classification result of the i-th level threshold. According to the information entropy, the importance weights of each level of threshold are calculated as:

[0084]

[0085] The importance weights of each level of thresholds are obtained through the formula. According to the importance weights, the initial interval of the dynamic threshold is adaptively adjusted to obtain the optimized dynamic threshold set. According to the importance weights of each level of thresholds, the upper and lower limits of the initial interval are adjusted to better protect the more important threshold intervals. Assume that the initial interval is [a i , b i ], the adjusted interval is [a′ i , b′ i ], then a′ i =a i -w i δ i , b′ i =b i +w i δ i In this way, the optimized dynamic threshold set is obtained. The terminal urgency value of each monitored terminal is compared with the optimized dynamic threshold set to obtain the initial trigger signal. When the terminal urgency value exceeds the optimized threshold interval to which it belongs, a trigger signal is generated. Let the trigger signal be S t ,when Then S t =1, otherwise S t= 0. Based on the initial trigger signal and historical network load data, an adaptive backoff strategy is constructed. The adaptive backoff strategy can be dynamically adjusted based on the historical data of network load to avoid network congestion through a reasonable backoff time. Suppose the network load data is {L t}, the backoff time is Δt, then the backoff strategy is Δt=f(L t ), where f is an adaptive function, which is fitted by regression analysis or machine learning methods. The optimal backoff time analysis is performed on the adaptive backoff strategy to obtain the optimal backoff time series. The optimal backoff time can be solved by an optimization algorithm, such as genetic algorithm, particle swarm optimization, etc. Suppose the objective function is:

[0086] J=∑ t g(Δt t , L t );

[0087] Where g is the objective function, which represents the relationship between backoff time and network load. By optimizing J, the optimal backoff time sequence is obtained. According to the optimal backoff time sequence and the initial trigger signal, priority calculation and conflict detection are performed. Priority calculation can be performed based on the combination of backoff time and trigger signal, and the priority of each trigger signal is determined by setting the priority weight. Let the priority be {P t},but:

[0088]

[0089] Where h is the priority function, which can be designed according to the specific application scenario. Through conflict detection, the trigger signal and priority of each monitoring terminal status update are determined. Conflict detection can avoid conflicts caused by simultaneous triggering of multiple signals by analyzing the relationship between trigger signals. Suppose the conflict detection result is {C t}, then C t =φ(P t , P t-1 , ..., P t-n ), where φ is the conflict detection function, which determines the final trigger signal and priority by analyzing the priorities of multiple consecutive moments.

[0090] In a specific embodiment, the process of executing step S4 may specifically include the following steps:

[0091] (1) Performing nonlinear feature extraction and dimensionality reduction processing on the device state data to obtain a low-dimensional state representation vector, and constructing an initial nonlinear state transfer model based on the low-dimensional state representation vector. The initial nonlinear state transfer model includes an input layer, a recurrent neural network, and an output layer;

[0092] (2) Encoding and embedding the trigger signal and priority to obtain the state update trigger feature, and performing conditional generative adversarial network training on the initial nonlinear state transfer model according to the state update trigger feature to obtain an enhanced state transfer model;

[0093] (3) performing hyperparameter optimization on the enhanced state transition model to obtain an optimal model structure and parameter set, and generating an initial state estimator based on the optimal model structure and parameter set;

[0094] (4) Perform online transfer learning and incremental learning on the initial state estimator to obtain a dynamic update mechanism, and perform uncertainty quantification analysis based on the dynamic update mechanism and historical estimation error distribution to obtain an adaptive state estimator.

[0095] Specifically, through deep learning methods such as autoencoders, nonlinear features are extracted from device state data. Autoencoders are unsupervised learning algorithms that can extract key features of data by compressing input data and reconstructing output data. Assuming that the device state data matrix is ​​X, the encoder part of the autoencoder maps the high-dimensional input X to a low-dimensional space to obtain a low-dimensional state representation vector Z. The function of the encoder part can be expressed as Z = f(X; W e , b e ), where f represents the nonlinear activation function, W e and b e are the weight and bias of the encoder respectively. The initial nonlinear state transfer model is constructed based on the low-dimensional state representation vector. The initial nonlinear state transfer model includes an input layer, a recurrent neural network, and an output layer. The recurrent neural network can capture the dynamic change relationship in time series data. Assuming that the input layer receives the low-dimensional state representation vector Z, the hidden state h of each time step is calculated through the recursive structure of the recurrent neural network. t The recursive calculation formula is h t =g(Z t ,h t-1 ; W r , b r ), where g is the recursive function of the recurrent neural network, W r and b r are the weights and biases of the recurrent network respectively. The output layer transforms the hidden state h t Map to the target output space to get the predicted state transition result Where h is the nonlinear activation function of the output layer, W o and b oare the weight and bias of the output layer respectively. The trigger signal and priority are encoded and embedded to obtain the state update trigger feature. The trigger signal and priority are discrete data, which are mapped to the continuous feature space through the embedding layer. Assuming the trigger signal is S and the priority is P, the output of the embedding layer can be expressed as

[0096] E S =Embed(S;W s ) and E P =Embed(P;W p ), where Embed is the embedding function, W s and W p are the embedding matrices of the trigger signal and priority respectively. These two embedded features are concatenated together to obtain the state update trigger feature E = [E S , E P ]. According to the state update trigger feature, the initial nonlinear state transition model is trained by conditional generative adversarial network. The conditional generative adversarial network introduces conditional variables to incorporate more information into the generation model and improve the quality of the generated results. Let the input of the generator be the noise vector Z and the conditional variable E, and the output be the generated state transition result The function of the generator is expressed as Where G is the generator network, W g and b g are the weight and bias of the generator respectively. The input of the discriminator is the real state data X and the conditional variable E, and the output is the true or false discrimination result D(X, E; W d , b d ), where D is the discriminator network, W d and b d are the weight and bias of the discriminator respectively. Through adversarial training of the generator and the discriminator, the parameters of the generator are optimized so that the state transition results generated by it are closer to the real data, and an enhanced state transition model is obtained. Hyperparameter optimization is performed on the enhanced state transition model. Hyperparameter optimization refers to finding the parameter combination that optimizes the model performance by searching and adjusting the hyperparameters of the model. Hyperparameter optimization methods include grid search, random search, and Bayesian optimization. Suppose the hyperparameter of the model is Θ, then the optimization goal is to minimize the loss function on the validation set By optimizing Get the optimal model structure and parameter set Θ * Based on the optimal model structure and parameter set, an initial state estimator is generated. The initial state estimator can use the aforementioned enhanced state transition model combined with the optimal hyperparameter Θ * , perform state estimation on the input data. Assume that the input of the state estimator is the current state data X t and trigger feature E t , the output is the estimated state at the next moment Then the function of the state estimator is expressed as In order to make the initial state estimator have better adaptability and accuracy in practical applications, online transfer learning and incremental learning are performed. Online transfer learning means that after the model is deployed, it is updated by continuously receiving new data so that the model can adapt to the new data distribution. Incremental learning means gradually accumulating new knowledge and improving the performance of the model without forgetting the existing knowledge. Through fine-tuning methods, the parameters of the state estimator are slightly adjusted to adapt it to the new data. Through online transfer learning and incremental learning, a dynamic update mechanism is obtained. The dynamic update mechanism can enable the state estimator to always maintain high accuracy and adaptability when facing a changing data environment. Suppose the new input data is X′t and E′ t , then the updated state estimator is According to the dynamic update mechanism and the distribution of historical estimation errors, uncertainty quantification analysis is performed. By analyzing the distribution characteristics of historical estimation errors, the uncertainty of state estimation is quantified and the reliability of estimation results is improved. Uncertainty quantification analysis can be achieved through methods such as Bayesian method or Monte Carlo simulation. Assume that the historical estimation error is {∈ t},in The uncertainty quantification index can be expressed as:

[0097]

[0098] Among them, σ ∈ is the standard deviation of the estimation error.

[0099] In a specific embodiment, the process of executing step S5 may specifically include the following steps:

[0100] (1) Perform interval analysis and probability statistics on the initial state estimate output by the adaptive state estimator to obtain the collective bounded uncertainty description and random uncertainty distribution;

[0101] (2) Based on the bounded uncertainty description of the set and the random uncertainty distribution, a hybrid uncertainty representation is constructed, and information entropy analysis and integral calculation are performed on the hybrid uncertainty representation to obtain a comprehensive uncertainty measurement index;

[0102] (3) Construct a multi-objective optimization problem based on the comprehensive uncertainty measurement index and the physical constraints of the monitoring terminal to obtain the optimization objective function set;

[0103] (4) Performing non-dominated sorting genetic initialization on the optimization objective function set to obtain the initial Pareto front solution set, and performing adaptive differential evolution based on the initial Pareto front solution set to obtain the target Pareto optimal solution set;

[0104] (5) Solving the optimal state estimation for the target Pareto optimal solution set to obtain the optimal state estimation candidate value, and performing exponential smoothing prediction on the optimal state estimation candidate value to obtain the optimal state estimation value of each monitoring terminal.

[0105] Specifically, an interval analysis is performed on the initial state estimate output by the adaptive state estimator. Assume that the output of the adaptive state estimator is By analyzing the distribution of the estimated value at multiple times, the upper and lower limits of the estimated value are determined to form a bounded uncertainty description. The confidence interval is calculated by statistical methods and the confidence interval is set to [L t , U t ], where L t and U t are the lower and upper limits of the estimated value, respectively. Perform probability statistics on the estimated value of the initial state to obtain the random uncertainty distribution. By calculating the probability density function of the estimated value, the random distribution of the estimated value is described. Assume that the probability distribution of the estimated value of the initial state is Then, we can fit it through kernel density estimation and other methods to obtain random uncertainty distribution. Based on the bounded uncertainty description of the set and the random uncertainty distribution, we construct a hybrid uncertainty representation. The hybrid uncertainty representation combines bounded uncertainty and random uncertainty to provide a more comprehensive uncertainty description. Let the hybrid uncertainty representation be The information entropy analysis and integral calculation of the mixed uncertainty representation are performed to obtain the comprehensive uncertainty measurement index. Information entropy is an important indicator for measuring uncertainty. By calculating the entropy of the mixed uncertainty representation, the degree of uncertainty can be quantified. Let the information entropy be H, then By integrating the information entropy, we can get the comprehensive uncertainty metric: The formula is:

[0106]

[0107] Based on the comprehensive uncertainty measurement index and the physical constraints of the monitoring terminal, a multi-objective optimization problem is constructed. The optimization problem needs to consider multiple objectives at the same time, such as minimizing uncertainty and satisfying physical constraints. Suppose the optimization objective function set is F = {f1, f2, ..., f n}, where f i represents the i-th optimization objective. The objective function can be expressed as Where C is a physical constraint. The optimization objective function set is initialized by non-dominated sorting genetic algorithm to obtain the initial Pareto front solution set. The non-dominated sorting genetic algorithm is a multi-objective optimization algorithm that finds multiple non-dominated solutions by simulating the natural selection process. The initial Pareto front solution set represents a set of equilibrium solutions found on the optimization objective function set. Let the initial Pareto front solution set be According to the initial Pareto front solution set, adaptive differential evolution is performed to obtain the target Pareto optimal solution set. The adaptive differential evolution algorithm continuously optimizes the solution set through mutation, crossover and selection operations, and finally converges to the global optimal solution. Suppose the target Pareto optimal solution set is The optimal state estimation is performed on the target Pareto optimal solution set to obtain the optimal state estimation candidate value. By performing state estimation on each solution in the optimal solution set, a set of optimal state estimation candidate values ​​is obtained. Let the optimal state estimation candidate value be Exponential smoothing is performed on the optimal state estimation candidate value to obtain the optimal state estimation value of each monitoring terminal. Exponential smoothing prediction is a time series prediction method that smoothes the estimation result by weighted averaging historical data to reduce the impact of noise. Assuming the smoothing coefficient is α, the prediction formula is:

[0108] In a specific embodiment, the process of executing step S6 may specifically include the following steps:

[0109] (1) Graph theory analysis is performed on the network topology of multiple monitoring terminals to obtain a hierarchical communication matrix, and distributed average consistency processing is performed on the optimal state estimation value according to the hierarchical communication matrix to obtain a consistent state;

[0110] (2) Perform robust control gain design and stability analysis on the consistency state to obtain convergence speed optimization parameters, and generate the target consistency protocol based on the convergence speed optimization parameters;

[0111] (3) Perform anomaly detection and fault diagnosis on the target consistency protocol during execution to obtain the abnormal event feature vector, and perform complex network propagation analysis on the abnormal event feature vector to obtain the prediction result of the impact range of the abnormal event;

[0112] (4) Multi-agent reinforcement learning is performed on the prediction results of the impact range of abnormal events to obtain a distributed emergency response strategy. Decision analysis is performed based on the distributed emergency response strategy and the consistency state to obtain a global collaborative control strategy for multiple monitoring terminals.

[0113] Specifically, the network topology is analyzed by graph theory. The network topology can be represented as a graph G = (V, E), where V represents the set of monitoring terminals and E represents the communication links between terminals. Through graph theory analysis, the degree, centrality and other parameters of each node are calculated to reflect the overall structure of the network and the importance of each node. Based on the network topology, a hierarchical communication matrix is ​​constructed. The hierarchical communication matrix can be represented as a matrix A, where A ij represents the communication weight between node i and node j. If there is a direct communication link between the two nodes, then A ij =1; otherwise, A ij=0. By calculating the hierarchical communication matrix, the communication paths and hierarchical structures between nodes in the network are determined. According to the hierarchical communication matrix, the optimal state estimate is processed by distributed average consistency. Distributed average consistency processing can be implemented through an iterative algorithm to make the states of all nodes tend to be consistent. Assume that the initial state of each node is x i (0), then in each iteration step t, the state update rule of node i is:

[0114]

[0115] Among them, N i represents the set of neighbor nodes of node i, and α is the step size parameter. Through iterative updates, the consistency state is obtained Perform robust control gain design and stability analysis on the consistency state. Robust control gain design is to improve the stability of the system in an uncertain environment. By designing appropriate control gain, the system can converge quickly. Assuming the control gain is K, the closed-loop state update equation of the system is:

[0116] x(t+1)=(I-αLK)x(t);

[0117] Among them, L is the Laplace matrix and I is the unit matrix. Through stability analysis, the optimal value of the control gain K is determined, and the convergence speed optimization parameters are obtained. According to the convergence speed optimization parameters, the target consistency protocol is generated. The target consistency protocol is used to guide the system on how to adjust the state of each node in a distributed environment to achieve global consistency. The core of the protocol is to dynamically adjust the step size α and the control gain K according to the convergence speed optimization parameters to ensure that the system can maintain efficient and stable state updates under different conditions. After the target consistency protocol is generated, it is performed with execution process anomaly detection and fault diagnosis. By monitoring the state changes during the execution of the protocol, the possible abnormal conditions and faults in the system can be discovered and identified in time. Anomaly detection and fault diagnosis can adopt machine learning-based methods to identify the feature vector of abnormal events by training models. Assume that the feature vector of the abnormal event is f, then the abnormal event label is obtained through the detection of model M:

[0118] y = M(f);

[0119] The complex network propagation analysis is performed on the feature vector of the abnormal event to obtain the prediction result of the impact range of the abnormal event. Complex network propagation analysis can predict the possible impact by simulating the propagation path and impact range of the abnormal event in the network. Assume that the propagation model is The predicted impact range is:

[0120]

[0121] Where R represents the predicted impact range. Multi-agent reinforcement learning is performed on the prediction results of the impact range of abnormal events to obtain a distributed emergency response strategy. Multi-agent reinforcement learning optimizes the system's emergency response strategy through the interaction and learning of multiple agents in a simulated environment. Let the strategy of each agent be π i , through the reinforcement learning algorithm, we get the global distributed emergency response strategy:

[0122] Π={π1,π2,...,π n};

[0123] According to the distributed emergency response strategy and consistency state, a decision analysis is performed to obtain the global collaborative control strategy of multiple monitoring terminals. The decision analysis combines the current consistency state and emergency response strategy to determine the control operation of each monitoring terminal. Let the decision function be D, then the global collaborative control strategy is C = D (Π, x * ).

[0124] The above describes the state control method of the monitoring terminal in the embodiment of the present invention. The following describes the state control device of the monitoring terminal in the embodiment of the present invention. Figure 2 , an embodiment of the state control device of the monitoring terminal in the embodiment of the present invention includes:

[0125] A collection module, used to collect multi-dimensional environmental data sets and equipment status data from multiple monitoring terminals;

[0126] A calculation module is used to perform nonlinear time-varying function calculation on the multi-dimensional environmental data set and the device status data to obtain the terminal emergency value of each monitoring terminal;

[0127] An analysis module, used to perform threshold comparison and adaptive backoff analysis on the terminal urgency value of each monitoring terminal, and obtain a trigger signal and priority for updating the status of each monitoring terminal;

[0128] A training module, used for training a nonlinear state transfer model for multiple monitoring terminals according to a trigger signal and a priority, to obtain an adaptive state estimator;

[0129] A modeling module is used to obtain the initial state estimation value output by the adaptive state estimator and perform mixed uncertainty modeling and multi-objective optimization calculation to obtain the optimal state estimation value of each monitoring terminal;

[0130] The optimization module is used to perform distributed consistency optimization and abnormal event propagation analysis on the optimal state estimation value to obtain the global collaborative control strategy of multiple monitoring terminals.

[0131] Through the collaboration of the above components, by calculating the nonlinear time-varying function of the multidimensional environmental data set and the device status data, combined with the adaptive sampling strategy, the environmental changes can be captured more accurately, the collection of redundant data can be reduced, and the data quality and resource utilization efficiency can be improved. The dynamic threshold and adaptive backoff analysis enable the system to automatically adjust the state update strategy according to the real-time situation, which improves the response speed and adaptability of the monitoring terminal to environmental changes. Through the training of nonlinear state transfer models and hybrid uncertainty modeling, combined with multi-objective optimization calculations, more accurate state estimates can be obtained, providing a reliable basis for subsequent decision-making. Through distributed consistency optimization and abnormal event propagation analysis, the system can achieve optimal resource allocation and efficient information processing in a global scope, and improve the collaborative efficiency of the entire monitoring network. Combined with complex network propagation analysis and multi-agent reinforcement learning, the system can actively predict the impact range of abnormal events and quickly generate distributed emergency response strategies, which improves the security and reliability of the monitoring network. By adopting a hierarchical communication matrix and robust control gain design, the system can still maintain stable performance in the face of network topology changes or local failures, enhancing the reliability and anti-interference ability of the entire monitoring system.

[0132] The present invention also provides a computer device, which includes a memory and a processor. The memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, the processor executes the steps of the state control method of the monitoring terminal in the above-mentioned embodiments.

[0133] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are executed on a computer, the computer executes the steps of the state control method of the monitoring terminal.

[0134] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0135] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program codes.

[0136] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A state control method for a monitoring terminal, characterized in that: The state control method of the monitoring terminal includes: Collect multi-dimensional environmental data sets and equipment status data from multiple monitoring terminals; Performing nonlinear time-varying function calculation on the multidimensional environment data set and the device status data to obtain a terminal urgency value of each monitoring terminal; Performing threshold comparison and adaptive backoff analysis on the terminal urgency value of each monitoring terminal to obtain a trigger signal and priority for updating the status of each monitoring terminal; specifically including: performing sequence analysis and clustering processing on the terminal urgency value of each monitoring terminal to obtain an initial interval of a dynamic threshold, and classifying the terminal urgency value according to the initial interval of the dynamic threshold to obtain a multi-level threshold classification result; performing entropy weight analysis on the multi-level threshold classification result to obtain the importance weights of each level of thresholds, and adaptively adjusting the initial interval of the dynamic threshold according to the importance weights to obtain an optimized dynamic threshold set; performing a comparison operation on the terminal urgency value of each monitoring terminal and the optimized dynamic threshold set to obtain an initial trigger signal; constructing an adaptive backoff strategy based on the initial trigger signal and historical network load data, and performing an optimal backoff time analysis on the adaptive backoff strategy to obtain an optimal backoff time sequence; performing priority calculation and conflict detection based on the optimal backoff time sequence and the initial trigger signal to obtain a trigger signal and priority for updating the status of each monitoring terminal; Performing nonlinear state transfer model training on the plurality of monitoring terminals according to the trigger signal and the priority to obtain an adaptive state estimator; Obtaining the initial state estimation value output by the adaptive state estimator and performing hybrid uncertainty modeling and multi-objective optimization calculation to obtain the optimal state estimation value of each monitoring terminal; Distributed consistency optimization and abnormal event propagation analysis are performed on the optimal state estimation value to obtain a global collaborative control strategy for the multiple monitoring terminals.

2. The state control method of the monitoring terminal according to claim 1, characterized in that: The collecting of multi-dimensional environment data sets and equipment status data of multiple monitoring terminals includes: Periodically activating the sensor arrays of multiple monitoring terminals to obtain an original environmental data stream, and adaptively adjusting the data sampling frequency according to the original environmental data stream to obtain a dynamic sampling strategy; Perform multi-dimensional feature extraction and noise reduction processing on the collected environmental data to obtain a structured environmental feature vector, and update the abnormal event detection model in real time according to the structured environmental feature vector to obtain an optimized abnormality detection threshold; Periodically scan and encode the hardware status parameters of the monitoring terminal to obtain a device status code, and dynamically optimize the terminal resource allocation plan according to the device status code to obtain a resource utilization efficiency index; The structured environmental feature vector and the device status code are timestamped and encrypted to obtain a secure data packet, and the secure data packet is optimized for transmission according to the resource utilization efficiency index and network conditions to obtain a multidimensional environmental data set and device status data.

3. The state control method of the monitoring terminal according to claim 1, characterized in that: The performing of nonlinear time-varying function calculation on the multi-dimensional environment data set and the device status data to obtain the terminal emergency level value of each monitoring terminal includes: Performing principal component analysis and feature normalization processing on the multidimensional environmental data set to obtain a dimensionally reduced environmental feature vector, and performing environmental anomaly evaluation on the dimensionally reduced environmental feature vector to obtain an environmental anomaly degree; Performing status urgency evaluation on the device status data to obtain a terminal status urgency index, and performing time series analysis and prediction on the historical network quality data of the monitoring terminal to obtain a network quality factor; Perform task importance analysis on the monitoring task types of the monitoring terminal to obtain a task importance weight matrix; Constructing a multi-input single-output initial urgency calculation model according to the environmental abnormality, the terminal status urgency index, the network quality factor and the task importance weight matrix; The initial urgency calculation model is jointly trained to obtain an optimized urgency calculation model, and the multidimensional environmental data set and the device status data are respectively calculated according to the optimized urgency calculation model to obtain the terminal urgency value of each monitoring terminal.

4. The state control method of the monitoring terminal according to claim 1, characterized in that: The performing nonlinear state transfer model training on the plurality of monitoring terminals according to the trigger signal and the priority to obtain an adaptive state estimator comprises: Performing nonlinear feature extraction and dimensionality reduction processing on the device state data to obtain a low-dimensional state representation vector, and constructing an initial nonlinear state transfer model based on the low-dimensional state representation vector, wherein the initial nonlinear state transfer model includes an input layer, a recurrent neural network, and an output layer; Encoding and embedding the trigger signal and the priority to obtain a state update trigger feature, and performing conditional generative adversarial network training on the initial nonlinear state transfer model according to the state update trigger feature to obtain an enhanced state transfer model; Performing hyperparameter optimization on the enhanced state transition model to obtain an optimal model structure and parameter set, and generating an initial state estimator according to the optimal model structure and parameter set; Online transfer learning and incremental learning are performed on the initial state estimator to obtain a dynamic update mechanism, and uncertainty quantification analysis is performed based on the dynamic update mechanism and historical estimation error distribution to obtain an adaptive state estimator.

5. The state control method of the monitoring terminal according to claim 1, characterized in that: The obtaining of the initial state estimation value output by the adaptive state estimator and performing mixed uncertainty modeling and multi-objective optimization calculation to obtain the optimal state estimation value of each monitoring terminal includes: Performing interval analysis and probability statistics on the initial state estimation value output by the adaptive state estimator to obtain a collective bounded uncertainty description and a random uncertainty distribution; Constructing a hybrid uncertainty representation according to the set bounded uncertainty description and the random uncertainty distribution, and performing information entropy analysis and integral calculation on the hybrid uncertainty representation to obtain a comprehensive uncertainty measurement index; Constructing a multi-objective optimization problem according to the comprehensive uncertainty measurement index and the physical constraints of the monitoring terminal to obtain an optimization objective function set; Performing non-dominated sorting genetic initialization on the optimization objective function set to obtain an initial Pareto front solution set, and performing adaptive differential evolution based on the initial Pareto front solution set to obtain a target Pareto optimal solution set; The optimal state estimation is solved for the target Pareto optimal solution set to obtain the optimal state estimation candidate value, and the optimal state estimation candidate value is predicted by exponential smoothing to obtain the optimal state estimation value of each monitoring terminal.

6. The state control method of the monitoring terminal according to claim 1, characterized in that: The performing distributed consistency optimization and abnormal event propagation analysis on the optimal state estimation value to obtain a global collaborative control strategy for the multiple monitoring terminals includes: Performing graph theory analysis on the network topology structure of the plurality of monitoring terminals to obtain a hierarchical communication matrix, and performing distributed average consistency processing on the optimal state estimation value according to the hierarchical communication matrix to obtain a consistency state; Performing robust control gain design and stability analysis on the consistency state to obtain convergence speed optimization parameters, and generating a target consistency protocol according to the convergence speed optimization parameters; Performing execution process anomaly detection and fault diagnosis on the target consistency protocol to obtain an abnormal event feature vector, and performing complex network propagation analysis on the abnormal event feature vector to obtain a prediction result of the impact range of the abnormal event; Multi-agent reinforcement learning is performed on the prediction results of the impact range of the abnormal event to obtain a distributed emergency response strategy, and decision analysis is performed based on the distributed emergency response strategy and the consistency state to obtain a global collaborative control strategy for the multiple monitoring terminals.

7. A state control device for a monitoring terminal, characterized in that: Used to execute the state control method of the monitoring terminal according to any one of claims 1 to 6, the state control device of the monitoring terminal comprises: A collection module, used to collect multi-dimensional environmental data sets and equipment status data from multiple monitoring terminals; A calculation module, used for performing nonlinear time-varying function calculation on the multi-dimensional environment data set and the device status data to obtain a terminal emergency value of each monitoring terminal; An analysis module is used to perform threshold comparison and adaptive backoff analysis on the terminal urgency value of each monitoring terminal to obtain a trigger signal and priority for updating the status of each monitoring terminal; specifically comprising: performing sequence analysis and clustering processing on the terminal urgency value of each monitoring terminal to obtain an initial interval of a dynamic threshold, and classifying the terminal urgency value according to the initial interval of the dynamic threshold to obtain a multi-level threshold classification result; performing entropy weight analysis on the multi-level threshold classification result to obtain the importance weights of each level of thresholds, and adaptively adjusting the initial interval of the dynamic threshold according to the importance weights to obtain an optimized dynamic threshold set; performing a comparison operation on the terminal urgency value of each monitoring terminal and the optimized dynamic threshold set to obtain an initial trigger signal; constructing an adaptive backoff strategy based on the initial trigger signal and historical network load data, and performing an optimal backoff time analysis on the adaptive backoff strategy to obtain an optimal backoff time sequence; performing priority calculation and conflict detection based on the optimal backoff time sequence and the initial trigger signal to obtain a trigger signal and priority for updating the status of each monitoring terminal; A training module, used for performing nonlinear state transfer model training on the plurality of monitoring terminals according to the trigger signal and the priority to obtain an adaptive state estimator; A modeling module, used to obtain the initial state estimation value output by the adaptive state estimator and perform hybrid uncertainty modeling and multi-objective optimization calculation to obtain the optimal state estimation value of each monitoring terminal; The optimization module is used to perform distributed consistency optimization and abnormal event propagation analysis on the optimal state estimation value to obtain a global collaborative control strategy for the multiple monitoring terminals.

8. A computer device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program executable on the processor, and wherein the processor implements the state control method of the monitoring terminal according to any one of claims 1 to 6 when executing the computer program.

9. A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the processor is enabled to execute the state control method of the monitoring terminal according to any one of claims 1 to 6.

Images