A visual control management system and method for switch cabinet

Abstract

The present invention relates to the field of smart grid technology, and in particular to a visual control management system and method for a switch cabinet, comprising multimodal data acquisition, edge computing processing, feature extraction and reasoning, collaborative optimization control, and augmented reality interaction and decision-making modules; the system collects environmental, electrical and mechanical vibration data, and generates high-quality feature data using a multimodal fusion algorithm; performs data preprocessing and feature extraction at edge computing nodes, and implements strategy optimization in combination with reinforcement learning of intelligent agents; visualizes the system operating status through an augmented reality platform, and uses graph convolutional networks and Bayesian networks for fault prediction and decision support; the present invention implements dynamic adjustment through feedback loop optimization, improves the stability and response speed of the system, and enhances the operating efficiency and safety of the power system.

Description

Technical Field

[0001] The present invention relates to the technical field of smart grids, and in particular to a visual control management system and method for a switch cabinet. Background Art

[0002] Switchgear is a vital component of the power system. Its main function is to control and protect power lines and equipment. With the development of smart grids, the demand for intelligent management of switchgear is growing. Realizing visual control and management of switchgear can significantly improve the reliability and efficiency of the power system, identify and handle faults in a timely manner, reduce maintenance costs, and improve the safety and stability of the power system.

[0003] The existing technology (Chinese invention patent, publication number: CN117891160B, name: A switch cabinet intelligent control system and method) mainly uses the traditional PID control algorithm combined with the BP neural network to perform intelligent control of the switch cabinet. This method relies on the sliding window mechanism to optimize the learning rate of the BP neural network, thereby indirectly optimizing the parameters of the PID controller. However, this technology has some shortcomings:

[0004] Due to electromagnetic interference, there is a lot of random noise in the data collected by the sensor, which will affect the training effect of the BP neural network and the accuracy of PID control.

[0005] A fixed learning rate is difficult to adapt to the quality changes of all operating data, resulting in unstable model training and affecting the accuracy of PID parameters.

[0006] The PID algorithm is suitable for linear systems, but its regulation performance is poor when dealing with complex and nonlinear systems, and parameters need to be adjusted repeatedly to achieve the best effect. Summary of the invention

[0007] In view of the many problems existing in the above-mentioned prior art, the present invention provides a visual control management system and method for a switch cabinet. The present invention realizes visual intelligent control of the switch cabinet by integrating multimodal data acquisition, edge computing processing, intelligent agent collaborative optimization and augmented reality interaction technology; the system uses a graph convolutional network to perform in-depth analysis of power data, combines a Bayesian network to predict faults, and provides real-time maintenance prediction and decision support; the adaptive feedback control network dynamically adjusts the operating parameters of the switch cabinet to ensure the efficient operation of the system in a complex environment; the system not only improves the fault response speed and prediction accuracy, but also enhances the user experience through an intuitive augmented reality interface.

[0008] A visual control and management system for a switch cabinet, comprising:

[0009] A multimodal data acquisition module is used to collect environmental status data, electrical parameter data and mechanical vibration data, and use an adaptive multimodal fusion algorithm combined with an automatic encoder model to fuse and denoise the data to generate fused feature data;

[0010] An edge computing processing module, used for receiving the fused feature data, and performing real-time preprocessing on the edge computing node to generate a preprocessed data summary;

[0011] A feature extraction and reasoning module, for extracting multi-level feature data from the preprocessed data summary, using a multi-layer variational autoencoder and a deep belief network, and generating a fuzzy rule set based on the feature data, generating fuzzy reasoning parameters through real-time analysis by a fuzzy logic reasoning engine, and optimizing the feature extraction process and the fuzzy rule set in combination with a feedback loop;

[0012] A collaborative optimization control module includes an intelligent agent deployed on an edge computing node, wherein the intelligent agent uses a reinforcement learning algorithm to optimize the initial strategy, adaptively adjusts the real-time data stream to generate optimized agent strategy data, and dynamically optimizes the global strategy through a multi-agent collaborative mechanism, a genetic algorithm, and a particle swarm optimization algorithm to generate collaborative optimization strategy data;

[0013] The augmented reality interaction and decision-making module is used to integrate the collaborative optimization strategy data in the augmented reality platform, generate augmented reality visualization data, use graph convolutional networks and Bayesian networks to analyze the real-time data stream for fault prediction, generate maintenance prediction data and decision support data, adjust the operating parameters of the switch cabinet in real time through an adaptive feedback control network, and generate feedback control data.

[0014] Preferably, the multimodal data acquisition module includes:

[0015] A multimodal sensor array for simultaneously collecting environmental status data, electrical parameter data, and mechanical vibration data, wherein each data type corresponds to a different sensor, including an optical fiber sensor, an acoustic sensor, and an infrared sensor;

[0016] A data fusion unit is configured to integrate raw data from different sensors using an adaptive multimodal fusion algorithm, wherein the algorithm assigns different weights to each data type. , generate fusion feature data, the formula is:

[0017]

[0018] in, To fuse feature data; For the The weight of each sensor data; For Data from sensors; is the number of sensors.

[0019] Preferably, the edge computing processing module includes:

[0020] A real-time processor receives the fused feature data and performs data denoising using an autoencoder model, wherein the autoencoder model removes redundant information in the data through an autoencoder neural network;

[0021] The data summarizer optimizes the format and compresses the denoised data to generate a pre-processed data summary suitable for fast transmission and analysis.

[0022] Preferably, the feature extraction and reasoning module includes:

[0023] A multi-layer variational autoencoder is configured to extract multi-level feature data from the preprocessed data summary. The multi-layer variational autoencoder captures complex features in the data through a multi-level network structure and is optimized using the following variational loss function:

[0024]

[0025] in, is the loss function; and are the network parameters of the autoencoder; For input data; is a hidden variable; is the variational distribution; To generate distribution; represents the Kullback-Leibler divergence between the variational distribution and the generating distribution;

[0026] A deep belief network for further analyzing the nonlinear relationships in the multi-level feature data and generating a fuzzy rule set in combination with the extracted features;

[0027] The fuzzy rule generator optimizes the fuzzy rule set based on historical data and expert knowledge base, so that the rule set can adapt to different system states.

[0028] Preferably, the fuzzy logic inference engine comprises:

[0029] A fuzzification module converts the feature data into fuzzy variables;

[0030] The reasoning module performs real-time reasoning based on the fuzzy rule set and generates fuzzy reasoning parameters. Its reasoning formula is:

[0031]

[0032] in, is the fuzzy reasoning parameter; Fuzzy rules The membership function of For rules In Features The application results on is the number of fuzzy rules;

[0033] The defuzzification module converts the fuzzy reasoning parameters into specific operation instructions so as to dynamically adjust the operating parameters of the switch cabinet.

[0034] Preferably, a feedback loop optimization unit is further included, and the feedback loop optimization unit includes:

[0035] Response collector, which collects system response data and performance indicators in real time;

[0036] The optimization engine performs adaptive optimization adjustments on the feature extraction unit and the fuzzy rule generator based on the collected data.

[0037] Preferably, the collaborative optimization control module includes:

[0038] The intelligent agent system is configured to be deployed on each edge computing node. The intelligent agent uses a reinforcement learning algorithm to continuously optimize its initial strategy;

[0039] The adaptive adjustment mechanism obtains the latest system status information through real-time data streams, thereby adjusting and optimizing proxy strategy data in real time.

[0040] Preferably, the multi-agent collaboration mechanism realizes strategy sharing and global optimization through the following steps:

[0041] Genetic algorithms, which use selection, crossover, and mutation operations to generate multiple optimization strategy solutions;

[0042] The particle swarm optimization algorithm iteratively updates the position and velocity of particles, and its optimization formula is:

[0043]

[0044]

[0045] in, For particles In the The speed of generation; For particles In the The position of the generation; is the inertia weight; and is the learning factor; and For A random number between For particles The best position in history; is the global best position;

[0046] This generates collaborative optimization strategy data.

[0047] Preferably, the augmented reality interaction and decision-making module includes:

[0048] An augmented reality platform integrates the collaborative optimization strategy data and generates visual augmented reality data, so that users can monitor the operating status of the switch cabinet in real time through augmented reality devices;

[0049] The prediction analysis system uses graph convolutional networks and Bayesian networks to analyze real-time data streams and generate maintenance prediction data. Its prediction formula is:

[0050]

[0051] in, is the predicted output; is the adjacency matrix; is the node feature matrix; is the normalized adjacency matrix; and is the weight matrix; is the activation function;

[0052] The decision support system generates decision support data by combining the prediction analysis results and adjusts the operating parameters of the switchgear through an adaptive feedback control network to generate feedback control data.

[0053] A method for implementing the visual control management system of the switch cabinet comprises the following steps:

[0054] Collect environmental status data, electrical parameter data and mechanical vibration data through a multimodal data acquisition module, and fuse and denoise the data using an adaptive multimodal fusion algorithm combined with an autoencoder model to generate fused feature data;

[0055] The fused feature data is received through an edge computing processing module, and real-time preprocessing is performed on the edge computing node to generate a preprocessed data summary;

[0056] Extracting multi-level feature data from the preprocessed data summary through a feature extraction and reasoning module, using a multi-layer variational autoencoder and a deep belief network, and generating a fuzzy rule set based on the feature data, generating fuzzy reasoning parameters through real-time analysis by a fuzzy logic reasoning engine, and optimizing the feature extraction process and the fuzzy rule set in combination with a feedback loop;

[0057] Through the collaborative optimization control module, an intelligent agent is deployed on the edge computing node. The intelligent agent uses a reinforcement learning algorithm to optimize the initial strategy, adaptively adjusts the real-time data stream to generate optimized agent strategy data, and dynamically optimizes the global strategy through a multi-agent collaborative mechanism, a genetic algorithm, and a particle swarm optimization algorithm to generate collaborative optimization strategy data.

[0058] Through the augmented reality interaction and decision-making module, the collaborative optimization strategy data is integrated into the augmented reality platform to generate augmented reality visualization data. The real-time data stream is analyzed using graph convolutional networks and Bayesian networks to predict faults, generate maintenance prediction data and decision support data, and adjust the operating parameters of the switch cabinet in real time through an adaptive feedback control network to generate feedback control data.

[0059] Compared with the prior art, the advantages and beneficial effects of the present invention are:

[0060] The present invention achieves more reliable noise elimination and feature extraction effects through multimodal data fusion technology;

[0061] The present invention achieves more flexible and real-time control strategy optimization through edge computing and intelligent agent technology;

[0062] The present invention achieves high accuracy and flexibility in fault prediction by combining graph convolutional networks and Bayesian networks;

[0063] The present invention realizes more intuitive operation status monitoring and decision support through an augmented reality interactive platform;

[0064] The present invention realizes adaptive learning and continuous optimization of the system through a feedback loop optimization mechanism, and significantly improves the stability and reliability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 It is a structural block diagram of the system of the present invention;

[0066] Figure 2 It is a schematic diagram of data collection and fusion in the present invention;

[0067] Figure 3 A schematic diagram of augmented reality interaction and decision making in the present invention;

[0068] Figure 4 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0069] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present disclosure. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, it is obvious that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present disclosure.

[0070] The terms used herein are only for describing specific embodiments and are not intended to limit the present disclosure. The terms "comprise", "include", etc. used herein indicate the existence of the features, steps, operations and / or components, but do not exclude the existence or addition of one or more other features, steps, operations or components.

[0071] All terms (including technical and scientific terms) used herein have the meanings commonly understood by those skilled in the art unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0072] like Figure 1 As shown, a visual control and management system for a switch cabinet includes:

[0073] like Figure 2 As shown, the multimodal data acquisition module is used to collect environmental state data, electrical parameter data and mechanical vibration data, and use an adaptive multimodal fusion algorithm combined with an automatic encoder model to fuse and denoise the data to generate fused feature data;

[0074] Preferably, the multimodal data acquisition module includes:

[0075] A multimodal sensor array for simultaneously collecting environmental status data, electrical parameter data, and mechanical vibration data, wherein each data type corresponds to a different sensor, including an optical fiber sensor, an acoustic sensor, and an infrared sensor;

[0076] A data fusion unit is configured to integrate raw data from different sensors using an adaptive multimodal fusion algorithm, wherein the algorithm assigns different weights to each data type. , generate fusion feature data, the formula is:

[0077]

[0078] in, To fuse feature data; For the The weight of each sensor data; For Data from sensors; is the number of sensors. Through this data fusion method, the system can extract valuable information from multiple data sources and improve the reliability and integrity of the data.

[0079] The multimodal data acquisition module uses a multimodal sensor array, including fiber optic sensors, acoustic sensors, and infrared sensors. Each sensor collects different types of data. For example, fiber optic sensors can be used to monitor temperature and strain changes, acoustic sensors are used to capture mechanical vibration and noise signals, and infrared sensors are used to detect heat distribution and abnormal hot spots. The purpose of this sensor combination is to ensure all-round and multi-angle monitoring of the environment around the switch cabinet.

[0080] One of the core technologies in this module is the adaptive multimodal fusion algorithm, which processes the collected data in combination with the autoencoder model. The working principle of the adaptive multimodal fusion algorithm is to assign different weights to each data type according to its importance and characteristics. These weights can be dynamically adjusted through historical data or expert systems. This algorithm achieves data denoising and maximum information extraction by integrating the raw data from different sensors.

[0081] The main effect of the multimodal data acquisition module is to improve the accuracy of the data and the response speed of the system. Due to the collaborative work of multiple sensors and the adaptive data fusion strategy, the module can effectively eliminate the errors that may exist in a single sensor and ensure that the data acquired by the system has higher accuracy and consistency.

[0082] This fusion strategy has shown great advantages in the operation monitoring of switchgear. For example, when detecting mechanical faults, the acoustic sensor can capture abnormal vibration signals, while the optical fiber sensor can simultaneously detect the strain changes caused by the fault. Through the data fusion unit, these signals are integrated into a highly reliable diagnostic data, which enables the system to warn of potential faults earlier and take corresponding measures.

[0083] In practical applications, for example, the above multimodal data acquisition module is deployed in the switch cabinet operation monitoring system of a certain substation. In this system, the acoustic sensor monitors the vibration state of the switch cabinet in real time, while the optical fiber sensor detects the strain and temperature changes of the cabinet. Through the adaptive multimodal fusion algorithm, the system can update the weight configuration of the sensor in real time, so that the fused feature data always reflects the most accurate equipment operation status.

[0084] The application of this module significantly improves the accuracy and reliability of switchgear monitoring and reduces false alarms or missed alarms caused by sensor failures or data errors. Ultimately, this solution effectively extends the maintenance cycle of the switchgear in actual operation, reduces unnecessary downtime, and thus improves the overall operating efficiency of the system.

[0085] In summary, the multimodal data acquisition module provides a solid data foundation for the visual control and intelligent decision-making of the switchgear through comprehensive collection and intelligent fusion of environmental data, demonstrating its great potential and wide application value in industrial applications.

[0086] An edge computing processing module, used for receiving the fused feature data, and performing real-time preprocessing on the edge computing node to generate a preprocessed data summary;

[0087] Preferably, the edge computing processing module includes:

[0088] A real-time processor receives the fused feature data and performs data denoising using an autoencoder model, wherein the autoencoder model removes redundant information in the data through an autoencoder neural network;

[0089] The data summarizer optimizes the format and compresses the denoised data to generate a pre-processed data summary suitable for fast transmission and analysis.

[0090] The edge computing processing module includes a real-time processor and a data summarizer. The main task of the real-time processor is to receive the fused feature data transmitted from the multimodal data acquisition module. In order to improve the clarity and accuracy of the data, the real-time processor uses an autoencoder model for data denoising. An autoencoder is an unsupervised learning model that uses a neural network for data compression and feature extraction, and removes redundant information by learning a low-dimensional representation of the data. The core of this model in the denoising process lies in its autoencoder neural network structure, which separates the noise from the useful signal in the data through training and retains the key feature information.

[0091] The data summarizer is responsible for formatting and compressing the denoised data. This process not only reduces the storage space required for the data, but also improves the efficiency of data transmission. The compressed data forms a pre-processed data summary suitable for fast transmission and subsequent analysis. This process ensures that the system can efficiently process large amounts of data under limited bandwidth conditions and maintain a high response speed.

[0092] Through the edge computing processing module, the system significantly improves data processing efficiency and response speed. The denoising function of the real-time processor effectively improves the quality of the data, making subsequent feature extraction and analysis more accurate. This advantage is particularly significant in scenarios with high-frequency data transmission and processing. For example, in the power system, the status information of the switch cabinet needs to be transmitted and analyzed in real time. The edge computing processing module can ensure that this data has been denoised and compressed before being transmitted to the central server, thereby greatly reducing the delay in data transmission.

[0093] In actual applications, the above-mentioned edge computing processing module is deployed in a switch cabinet monitoring system of a smart grid. In this system, after receiving the fused feature data, the real-time processor immediately denoises the data through the autoencoder model. For example, when detecting the overload of the power line, the denoised current data is clearer and can accurately reflect the actual load of the line. The subsequent data summary generator compresses the denoised data to make the data packet smaller and easier to transmit.

[0094] The system has achieved remarkable results after deployment, especially during peak load periods, where the delay in data transmission has been greatly reduced, while the accuracy of data has been improved by about 15%. Due to the efficient operation of the edge computing processing module, the system can monitor the status changes of the switch cabinet in real time, identify potential problems in a timely manner and issue warnings, reducing the frequency of accidents and improving the reliability and stability of the power grid.

[0095] In summary, the edge computing processing module provides efficient data support and decision-making basis for the visual control management of switch cabinets through real-time data processing and intelligent denoising, combined with data summary generation technology, demonstrating its important value in the fields of industry and energy management.

[0096] A feature extraction and reasoning module, for extracting multi-level feature data from the preprocessed data summary, using a multi-layer variational autoencoder and a deep belief network, and generating a fuzzy rule set based on the feature data, generating fuzzy reasoning parameters through real-time analysis by a fuzzy logic reasoning engine, and optimizing the feature extraction process and the fuzzy rule set in combination with a feedback loop;

[0097] The feature extraction and inference module first extracts features from the preprocessed data summary through a multi-layer variational autoencoder (VAE). VAE is a probabilistic generative model that can capture the implicit features of data and map high-dimensional data into a low-dimensional latent space. In this process, VAE uses its generative model and encoder structure to accurately extract representative and compressed feature information, providing the core expression of the data for subsequent analysis.

[0098] After extracting multi-level feature data, the module uses a deep belief network (DBN) to further analyze the nonlinear relationships in these feature data. DBN is a deep learning model composed of multiple layers of restricted Boltzmann machines (RBMs) that excels at capturing complex multi-level associations and nonlinear features. In switchgear management, DBN can identify deep patterns in data, such as small changes in equipment status and their potential impact.

[0099] Based on these extracted feature data, the module generates a fuzzy rule set. The fuzzy rule set allows the system to make smarter decisions in an uncertain environment by fuzzifying the relationship between feature data and operating parameters. Through the fuzzy logic reasoning engine, the system can parse feature data in real time and generate fuzzy reasoning parameters, which are used to guide the specific operation adjustments of the switch cabinet.

[0100] In order to continuously improve the accuracy of feature extraction and reasoning, the module combines a feedback loop mechanism. This mechanism continuously monitors the system output and actual operation results, and uses the feedback information to optimize the feature extraction process and fuzzy rule set, so that the system can dynamically adapt to changing environments and data characteristics.

[0101] Through the feature extraction and reasoning module, the visual control and management system of the switch cabinet has significantly enhanced its intelligent decision-making ability and adaptability. The combination of VAE and DBN enables the system to extract the most representative features from massive and complex data, reducing the dependence on raw data and improving the efficiency and effectiveness of data analysis. At the same time, fuzzy logic reasoning provides a way to deal with uncertainty and complex decision-making environments, allowing the system to flexibly respond to changing operating conditions.

[0102] In actual applications, the switchgear control system of a smart grid adopts this feature extraction and reasoning module. During the monitoring process of the high-voltage power grid, the system extracts the key features of current, voltage and temperature from the preprocessed data summary through VAE, and then uses DBN to further analyze the relationship between these features to detect potential overload and temperature anomalies. The fuzzy logic reasoning engine generates reasoning parameters based on these features and adjusts the operating mode of the switchgear in real time to avoid possible failures and damage.

[0103] Through the feedback loop optimization mechanism, the system continuously adjusts and improves itself during operation. Every time the system makes a decision, it immediately monitors its effect and feeds the operation results back to the model for further optimization. This closed-loop feedback mechanism ensures the stability and reliability of the system and improves the safety of power grid operation.

[0104] In summary, the feature extraction and reasoning module provides in-depth analysis and efficient decision-making support for the visual control management of switchgear through advanced machine learning technology and intelligent reasoning algorithms, greatly improving the automation level and responsiveness of the system.

[0105] Preferably, the feature extraction and reasoning module includes:

[0106] A multi-layer variational autoencoder is configured to extract multi-level feature data from the preprocessed data summary. The multi-layer variational autoencoder captures complex features in the data through a multi-level network structure and is optimized using the following variational loss function:

[0107]

[0108] in, is the loss function; and are the network parameters of the autoencoder; For input data; is a hidden variable; is the variational distribution; To generate distribution; represents the Kullback-Leibler divergence between the variational distribution and the generative distribution; this optimization strategy ensures that the model can not only accurately reconstruct the input data, but also perform effective feature representation in the latent space;

[0109] A deep belief network for further analyzing the nonlinear relationships in the multi-level feature data and generating a fuzzy rule set in combination with the extracted features;

[0110] The fuzzy rule generator optimizes the fuzzy rule set based on historical data and expert knowledge base, so that the rule set can adapt to different system states.

[0111] Multi-layer variational autoencoder is a deep generative model that uses a multi-layer network structure to capture complex features in data. Hidden variables compressed into latent space , and then through the decoder network from Reconstruct the original data. The core of this process is to model the potential characteristics of the data through variational inference techniques.

[0112] Through multi-layer variational autoencoders, the visual control management system of the switchgear can efficiently extract multi-level features from preprocessed data. These feature data not only retain the key information of the original data, but also simplify the data structure, making subsequent analysis and processing more efficient. This feature extraction technology significantly improves the system's ability to recognize complex data patterns, thereby optimizing the operation monitoring and anomaly detection of the switchgear.

[0113] In a smart grid management application, a multi-layer variational autoencoder is used to analyze the real-time operating data of the switchgear. In the system, the data collected by the sensor is transmitted to the VAE for processing, and the VAE converts this data into a feature representation in the latent space through its encoder. For example, when monitoring parameters such as current and temperature, the VAE can extract the potential relationship between these data and identify abnormal patterns or signs of failure.

[0114] This processing method enables the system to respond to abnormal situations more quickly. For example, when VAE identifies a potential correlation between current changes and temperature rise, the system can issue an alarm in advance to prompt maintenance personnel to check for possible equipment failures. By using VAE, the system not only improves the accuracy of fault prediction, but also reduces the occurrence of false positives and false negatives.

[0115] In summary, the application of multi-layer variational autoencoders in feature extraction and reasoning modules provides powerful data processing capabilities for visual control and management of switchgear through advanced deep learning technology, improving the intelligence level and response speed of the system. Its capabilities in data compression and feature extraction provide important support for the processing and analysis of complex data.

[0116] Preferably, the fuzzy logic inference engine comprises:

[0117] A fuzzification module converts the feature data into fuzzy variables;

[0118] The reasoning module performs real-time reasoning based on the fuzzy rule set and generates fuzzy reasoning parameters. Its reasoning formula is:

[0119]

[0120] in, is the fuzzy reasoning parameter; Fuzzy rules The membership function of For rules In Features The application results on is the number of fuzzy rules; this formula combines the influence of all relevant rules; and generates corresponding output parameters for each input.

[0121] The defuzzification module converts the fuzzy reasoning parameters into specific operation instructions so as to dynamically adjust the operating parameters of the switch cabinet.

[0122] The fuzzy logic inference engine includes fuzzification module, inference module and defuzzification module. These modules work together to transform data features into specific operation instructions through fuzzy logic methods to achieve dynamic control.

[0123] The fuzzification module converts the input feature data into fuzzy variables. Fuzzification is achieved by defining fuzzy sets and membership functions. The membership function maps the exact input value to a fuzzy set to indicate the degree to which the input value is in the fuzzy set. For example, temperature data can be fuzzified into fuzzy sets such as "low", "medium", and "high".

[0124] The inference module performs real-time inference on the fuzzified variables according to a pre-set set of fuzzy rules. Fuzzy rules usually take the form of "if...then..." to encode the relationship between the input fuzzy variables and the output variables. During the inference process, each fuzzy rule is evaluated by its membership function and applied to the input features. , calculate the corresponding fuzzy inference parameters.

[0125] The defuzzification module converts fuzzy inference parameters into specific operating instructions. Defuzzification is the process of converting fuzzy variables into executable precise instructions to dynamically adjust the operating parameters of the switch cabinet. This process is usually achieved through methods such as weighted average or central average, converting fuzzy control signals into specific values ​​to guide actual operations.

[0126] Through the fuzzy logic reasoning engine, the switchgear system can process complex and uncertain information and make flexible decisions. This ability greatly improves the adaptability and response speed of the system in a changing environment. Fuzzy logic reasoning not only simplifies the design of rules, but also enables the system to maintain efficient operation when faced with inaccurate or incomplete data.

[0127] In a switchgear control application for a certain power system, a fuzzy logic inference engine is used to adjust the operating parameters of the switchgear in real time. The system obtains real-time temperature, current, and vibration data through sensors, which are fuzzified into corresponding fuzzy variables. The inference module analyzes these variables based on a set of fuzzy rules and generates fuzzy inference parameters. For example, a rule can be set as: "If the temperature is high and the vibration is strong, reduce the load power." This type of rule guides the system to adjust the load of the switchgear in real time through the calculated fuzzy inference parameters.

[0128] The defuzzification module converts the inference results into specific operating instructions to ensure that the switch cabinet operates within a safe operating range. In this process, the system's dynamic adjustment capability is significantly improved, which can effectively reduce the failure rate and maintenance costs.

[0129] In summary, the fuzzy logic inference engine provides strong support for the intelligent control of switchgear by converting complex data into clear operation instructions. It not only improves the decision-making accuracy and real-time performance of the system, but also enhances the system's adaptive ability in complex environments, becoming an important part of the visual management of switchgear.

[0130] Preferably, a feedback loop optimization unit is further included, and the feedback loop optimization unit includes:

[0131] Response collector, which collects system response data and performance indicators in real time;

[0132] The optimization engine performs adaptive optimization adjustments on the feature extraction unit and the fuzzy rule generator based on the collected data.

[0133] The feedback loop optimization unit consists of two parts: the response collector and the optimization engine. Its main task is to continuously optimize the system's feature extraction and reasoning process through the feedback mechanism.

[0134] The response collector is responsible for collecting the system's response data and performance indicators in real time. These data include information such as the operating status of the switch cabinet, equipment performance, system output, etc. These feedback data are a direct reflection of the system's operating effect and can provide timely and accurate basis for the optimization process.

[0135] The optimization engine uses the collected response data to adaptively optimize the feature extraction unit and fuzzy rule generator. The optimization process uses an adaptive algorithm to adjust system parameters and models based on feedback information to improve the accuracy and response speed of the system. For example, the optimization engine can optimize the feature extraction process by adjusting the network parameters of the multi-layer variational autoencoder to make it more adaptable to changes in real-time data. At the same time, the optimization engine can also update the fuzzy rule set based on the actual operation situation to ensure that the rule set can accurately reflect the current operation requirements.

[0136] The introduction of the feedback loop optimization unit enables the switch cabinet control system to have adaptive learning capabilities and continuously optimize its performance in a dynamic environment. By collecting and analyzing feedback information in real time, the system can quickly identify and correct potential problems and maintain efficient and stable operation. This closed-loop optimization mechanism not only improves the overall performance of the system, but also reduces the occurrence of failures and increases the service life of the equipment.

[0137] In a smart grid application scenario, the feedback loop optimization unit is applied to the intelligent management system of the switch cabinet. During the operation of the system, the response collector monitors the key performance indicators of the switch cabinet such as current, voltage, temperature, etc. in real time. Once an abnormal fluctuation is detected, such as excessive current or abnormal temperature rise, the system will immediately feed back this data to the optimization engine.

[0138] Based on this feedback information, the optimization engine dynamically adjusts the parameters of the feature extraction unit to improve the sensitivity of anomaly detection. At the same time, the optimization engine optimizes the fuzzy rule generator and updates the rule set to better adapt it to the current operating status and operational requirements. Through this adaptive optimization, the system can warn before a fault occurs and automatically take protective measures to reduce losses.

[0139] This optimization mechanism significantly improves the response speed and accuracy of the system, ensuring that the switch cabinet always maintains efficient and safe operation in a complex and changing environment. In long-term operation, the failure rate of the system has been significantly reduced, and the maintenance cost has been effectively controlled, reflecting the important value of the feedback loop optimization unit in intelligent control.

[0140] In summary, the feedback loop optimization unit provides strong support for the visual control management of the switch cabinet through real-time collection and dynamic optimization. It plays a key role in enhancing the system's adaptability and intelligence, and has become an indispensable part of modern smart grid management.

[0141] A collaborative optimization control module includes an intelligent agent deployed on an edge computing node, wherein the intelligent agent uses a reinforcement learning algorithm to optimize the initial strategy, adaptively adjusts the real-time data stream to generate optimized agent strategy data, and dynamically optimizes the global strategy through a multi-agent collaborative mechanism, a genetic algorithm, and a particle swarm optimization algorithm to generate collaborative optimization strategy data;

[0142] The core of the collaborative optimization control module lies in the intelligent agent deployed on the edge computing node. The intelligent agent is a computing unit that can make autonomous decisions and learn in a specific environment. It optimizes the initial strategy through the reinforcement learning algorithm, which is a machine learning method that continuously updates the strategy through interaction with the environment to maximize long-term benefits. In the switchgear control, the intelligent agent adjusts its operating strategy in real time through the reinforcement learning algorithm to adapt to the dynamically changing environment.

[0143] The intelligent agent obtains information about the state of the environment from real-time data streams and adaptively adjusts its strategy based on this information. This adjustment is made through a reinforcement learning model that updates the strategy in real time, enabling the agent to maintain efficient decision-making capabilities in a constantly changing environment.

[0144] In addition, the collaborative optimization control module adopts a multi-agent collaborative mechanism, combined with genetic algorithms and particle swarm optimization algorithms, to achieve dynamic optimization of global strategies. The multi-agent collaborative mechanism allows multiple intelligent agents to share information and strategies, and improve the decision-making level of the overall system through collaborative work. The genetic algorithm generates new strategy solutions by simulating the selection, crossover and mutation operations in the natural evolution process; the particle swarm optimization algorithm adjusts the agent's strategy by simulating social behavior, so that it tends to the global optimal when solving problems.

[0145] The application of collaborative optimization control modules has significantly improved the intelligence level and global optimization capabilities of the switchgear control system. Through the adaptive learning of intelligent agents and the collaborative optimization of multiple agents, the system can respond to environmental changes more quickly, optimize control strategies, and improve overall operating efficiency. This modular design also enhances the robustness and flexibility of the system and adapts to different operating conditions and environmental changes.

[0146] In the present invention, the collaborative optimization control module is used to optimize the power distribution strategy. The intelligent agent is deployed on each edge computing node to monitor key parameters such as current load and temperature in real time. During system operation, the intelligent agent optimizes the current load distribution strategy through reinforcement learning algorithm. For example, when an increase in the load on a certain line is detected, the intelligent agent can quickly adjust the distribution strategy to avoid overload.

[0147] At the same time, through the multi-agent coordination mechanism, each intelligent agent shares information and improves the overall decision-making level. Genetic algorithm and particle swarm optimization algorithm further optimize the overall strategy, making power distribution more balanced and efficient. This coordination mechanism has achieved significant energy-saving effects in practical applications and effectively reduced the risk of equipment overload.

[0148] Through this optimized control, the switch cabinet system not only improves energy utilization, but also reduces the potential risk of equipment damage and extends the service life of the equipment. In long-term operation, the system's maintenance cost and failure rate are significantly reduced.

[0149] In summary, the collaborative optimization control module provides advanced technical support for the intelligent control and global optimization of switchgear through the adaptive learning and collaborative mechanism of intelligent agents, greatly improving the operating efficiency and stability of the system. Its application value in smart grid management is huge, providing strong support for the intelligent development of modern power systems.

[0150] Preferably, the collaborative optimization control module includes:

[0151] The intelligent agent system is configured to be deployed on each edge computing node. The intelligent agent uses a reinforcement learning algorithm to continuously optimize its initial strategy;

[0152] The adaptive adjustment mechanism obtains the latest system status information through real-time data streams, thereby adjusting and optimizing proxy strategy data in real time.

[0153] The intelligent agent system is the basic component of the collaborative optimization control module. Each intelligent agent is configured on the edge computing node and is responsible for monitoring the environment and operating status of its area of ​​responsibility. The intelligent agent continuously optimizes its initial strategy through reinforcement learning algorithm, which is a trial-and-error-based learning method that updates the strategy through interaction with the environment to achieve the best decision. In switchgear management, the intelligent agent adjusts the operating parameters of the switchgear based on real-time monitoring data such as current load, temperature and vibration to achieve efficient and safe operation.

[0154] The core of the reinforcement learning algorithm is that the intelligent agent can continuously improve its strategy through reward and punishment mechanisms. Each intelligent agent obtains feedback information based on environmental changes and learns how to make better decisions in similar situations in the future through algorithmic models. This self-learning ability enables the intelligent agent to maintain efficient responsiveness when facing complex and changing power environments.

[0155] The adaptive adjustment mechanism obtains the latest system status information through real-time data streams, ensuring that the system can quickly adapt to dynamic changes. This mechanism adjusts and optimizes agent policy data in real time based on the data collected by the intelligent agent. The real-time data stream provides the latest information about the current state of the system, allowing the intelligent agent to adjust its strategy in time to adapt to new situations. Adaptive adjustment not only ensures the stability of the system, but also optimizes the overall performance of the system.

[0156] The collaborative optimization control module achieves efficient operation and dynamic optimization of the system through an intelligent agent system and adaptive adjustment mechanism. This module improves the system's adaptability to environmental changes and enhances the accuracy and real-time nature of decision-making. By deploying intelligent agents on edge nodes, the system can process information and make decisions in a distributed manner, reducing the burden on central processing.

[0157] In a smart grid management application, the collaborative optimization control module is used for load management of switch cabinets. Each intelligent agent is deployed on different edge computing nodes to monitor the current, voltage, temperature and other parameters of the switch cabinet it is responsible for in real time. When a node detects that the load is too high, the intelligent agent immediately adjusts its strategy through the reinforcement learning algorithm to reduce the load pressure. At the same time, the adaptive adjustment mechanism updates the strategy of the intelligent agent according to the real-time data stream, so that it can still operate efficiently in the new environment.

[0158] For example, if a line has an increased load due to abnormal weather, the intelligent agent can adjust the load distribution strategy in a timely manner through real-time monitoring and data analysis, and transfer the overloaded part to other lines, thus avoiding the failure and damage that may be caused by the overload. This mechanism not only improves the efficiency of power distribution, but also effectively reduces the risk of accidents.

[0159] This distributed optimization control strategy significantly improves the reliability and operating efficiency of the system and reduces the failure rate caused by load imbalance. In long-term operation, the intelligent agent continuously optimizes the control strategy through adaptive adjustment to ensure that the system can maintain efficient and stable operation in any environment.

[0160] In summary, the collaborative optimization control module provides strong support for the intelligent management of switchgear through the intelligent agent system and adaptive adjustment mechanism. This technology not only improves the decision-making ability and response speed of the system, but also enhances the adaptability and robustness of the system in complex environments. It is one of the indispensable key technologies in modern smart grid management.

[0161] Preferably, the multi-agent collaboration mechanism realizes strategy sharing and global optimization through the following steps:

[0162] Genetic algorithms, which use selection, crossover, and mutation operations to generate multiple optimization strategy solutions;

[0163] The particle swarm optimization algorithm iteratively updates the position and velocity of particles, and its optimization formula is:

[0164]

[0165]

[0166] in, For particles In the The speed of generation; For particles In the The position of the generation; is the inertia weight; and is the learning factor; and For A random number between For particles The best position in history; is the global best position;

[0167] This generates collaborative optimization strategy data.

[0168] The original intention of the multi-agent coordination mechanism is to achieve information sharing and strategy optimization among multiple intelligent agents. The mechanism is implemented through two main algorithms: genetic algorithm and particle swarm optimization algorithm.

[0169] Genetic algorithm is an optimization algorithm based on the principle of biological evolution. It simulates the natural selection process and generates multiple optimization strategy solutions through selection, crossover and mutation operations. In switchgear management, genetic algorithms are used to generate a set of initial strategies that represent different control parameter configurations. The algorithm selects the strategies with better performance through selection operations, and then generates new strategy combinations through crossover and mutation operations. This method can find the global optimal solution in a larger search space and avoid falling into the local optimum.

[0170] Particle swarm optimization (PSO) is an optimization algorithm based on swarm intelligence, which simulates the foraging behavior of bird flocks and achieves optimization through information sharing between individuals (particles). Each particle represents a possible solution, and the particle iteratively updates its speed and position to get closer to the optimal solution of the problem.

[0171] The PSO algorithm adjusts the speed and position of particles so that the entire particle swarm gradually converges to the optimal solution. This process helps to find a better strategy combination globally.

[0172] Through the multi-agent collaborative mechanism, the system can achieve fast and effective global optimization, ensuring that the switch cabinet is always in the optimal operating state in a complex dynamic environment. The combination of the global search capability of the genetic algorithm and the fast convergence characteristics of the particle swarm optimization enables the system to take into account both exploration and utilization, improving the efficiency of strategy optimization.

[0173] In actual applications, a power company uses a multi-agent collaborative mechanism to optimize the load distribution strategy of each switch cabinet in the power grid. The system first generates an initial strategy plan through a genetic algorithm. After a round of selection, crossover and mutation operations, these strategies form an initial particle swarm. The particle swarm optimization algorithm further adjusts each strategy on this basis to ensure that the switch cabinet can maintain efficient and safe operation when facing changes in the power grid load.

[0174] For example, during peak power periods, the system can identify imbalances in load distribution in the power grid through real-time monitoring and data analysis. Through a multi-agent collaborative mechanism, the system can dynamically adjust the load distribution of each switch cabinet to avoid the risk of overloading certain lines. The particle swarm optimization algorithm enables rapid strategy adjustments to ensure the stability of power distribution.

[0175] The implementation of this mechanism has significantly improved the stability and efficiency of the power grid and reduced failures and losses caused by load imbalance. In long-term operation, the system has maintained the efficiency and reliability of the power system by continuously optimizing strategies.

[0176] In summary, the multi-agent collaborative mechanism provides strong support for intelligent control and global optimization of switchgear through advanced algorithm combination. It plays a key role in improving system decision-making efficiency and adaptability, and is an important technical means in smart grid management.

[0177] like Figure 3 As shown, the augmented reality interaction and decision-making module is used to integrate the collaborative optimization strategy data in the augmented reality platform, generate augmented reality visualization data, use graph convolutional networks and Bayesian networks to analyze the real-time data stream for fault prediction, generate maintenance prediction data and decision support data, adjust the operating parameters of the switch cabinet in real time through an adaptive feedback control network, and generate feedback control data.

[0178] Preferably, the augmented reality interaction and decision-making module includes:

[0179] An augmented reality platform integrates the collaborative optimization strategy data and generates visual augmented reality data, so that users can monitor the operating status of the switch cabinet in real time through augmented reality devices;

[0180] The prediction analysis system uses graph convolutional networks and Bayesian networks to analyze real-time data streams and generate maintenance prediction data. Its prediction formula is:

[0181]

[0182] in, is the predicted output; is the adjacency matrix; is the node feature matrix; is the normalized adjacency matrix; and is the weight matrix; is the activation function;

[0183] The decision support system generates decision support data by combining the prediction analysis results and adjusts the operating parameters of the switchgear through an adaptive feedback control network to generate feedback control data.

[0184] The augmented reality interaction and decision-making module includes three main components: augmented reality platform, predictive analysis system and decision support system.

[0185] The augmented reality platform is responsible for converting collaborative optimization strategy data into visual augmented reality data. Through augmented reality devices, users can monitor the operating status of the switchgear in real time. This process achieves intuitive display and interaction of equipment status by superimposing virtual information on the real-world switchgear view. Users can observe the operating parameters, equipment health status, and visualization of potential faults of the switchgear through the augmented reality interface, so as to monitor and manage more efficiently.

[0186] The predictive analysis system uses graph convolutional networks (GCNs) and Bayesian networks to analyze real-time data streams to generate maintenance prediction data. Graph convolutional networks are deep learning models that can perform convolution operations on graph-structured data and are suitable for processing complex network data such as power grids. In this system, GCN makes predictions using the following formula:

[0187]

[0188] in, is the predicted output; is the adjacency matrix; is the node feature matrix; is the normalized adjacency matrix; and is the weight matrix; is the activation function; GCN combined with the causal reasoning ability of Bayesian network can accurately predict and analyze the faults of the power system and identify potential risks and fault points.

[0189] The decision support system combines the results of predictive analysis to generate decision support data. The system uses an adaptive feedback control network to adjust the operating parameters of the switchgear in real time to generate feedback control data. This process ensures that when a fault occurs or an abnormality is detected, the system can automatically adjust the operating strategy to maintain the optimal operating state of the equipment.

[0190] The augmented reality interaction and decision-making module greatly improves the efficiency and safety of switchgear management by providing an intuitive visual interface and intelligent decision support. Users can quickly obtain equipment status information and optimization suggestions through the augmented reality platform, reducing the possibility of manual judgment errors. The predictive analysis system ensures that the system can identify and respond to faults in advance, reducing downtime and maintenance costs.

[0191] In a large power system, the augmented reality interaction and decision-making module is applied to the daily monitoring and maintenance of switchgear. On-site operation and maintenance personnel can view the operating status and operation suggestions of each switchgear in real time by wearing augmented reality equipment. When an abnormal vibration signal appears in a certain equipment, the system identifies the possible fault caused by the signal through GCN and generates maintenance prediction data.

[0192] The decision support system immediately provides corresponding operational adjustment suggestions, and the operation and maintenance personnel can intuitively understand the effect and specific operation steps after the adjustment through the augmented reality interface. This mechanism significantly improves the operation and maintenance efficiency, reduces the frequency of equipment failures, and improves the stability of the system.

[0193] Through this module that integrates multiple advanced technologies, the system not only achieves efficient management of switch cabinets, but also provides a model for the intelligent upgrade of future power grids. The augmented reality interaction and decision-making module has shown great application potential in improving equipment safety and reducing maintenance costs, and has become an indispensable part of smart grid management.

[0194] like Figure 4 As shown, a method for implementing the visual control management system of the switch cabinet comprises the following steps:

[0195] The multimodal data acquisition module collects environmental status data, electrical parameter data and mechanical vibration data, and uses an adaptive multimodal fusion algorithm combined with an autoencoder model to fuse and denoise the data to generate fused feature data; the intelligent management of the switch cabinet begins with the accurate monitoring of multiple physical quantities. The multimodal data acquisition module uses a variety of sensors to collect environmental status data (such as temperature, humidity), electrical parameters (such as current, voltage) and mechanical vibration data. These data are fused and denoised by an adaptive multimodal fusion algorithm combined with an autoencoder model. The adaptive multimodal fusion algorithm assigns different weights according to the characteristics of the data, thereby improving the reliability and accuracy of the data and removing noise interference. The fused feature data generated by this data fusion provides high-quality basic information for subsequent analysis.

[0196] The fused feature data is received by the edge computing processing module, and real-time preprocessing is performed on the edge computing node to generate a preprocessed data summary; the fused feature data is transmitted to the edge computing processing module, where real-time preprocessing is performed. The advantage of edge computing is that it can process data nearby, reduce latency, and improve response speed. Real-time preprocessing generates data summaries suitable for fast transmission through compression and summarization. These data summaries retain key information, reduce the amount of data, and improve transmission and computing efficiency.

[0197] The multi-level feature data is extracted from the preprocessed data summary through the feature extraction and reasoning module, and a multi-layer variational autoencoder and a deep belief network are used to generate a fuzzy rule set based on the feature data. The fuzzy reasoning parameters are generated by real-time analysis through the fuzzy logic reasoning engine, and the feature extraction process and the fuzzy rule set are optimized in combination with the feedback loop; in the feature extraction and reasoning module, the system uses a multi-layer variational autoencoder (VAE) and a deep belief network (DBN) to extract multi-level feature data from the preprocessed data summary. VAE is used to identify implicit structures in the data, and DBN further analyzes the deep features of the data. Based on the extracted feature data, a fuzzy rule set is generated, and the fuzzy reasoning parameters are generated by real-time analysis through the fuzzy logic reasoning engine. These parameters are used to describe and predict the system state. Combined with the feedback loop mechanism, the system can continuously optimize the feature extraction process and the fuzzy rule set to make the reasoning results more accurate.

[0198] Through the collaborative optimization control module, an intelligent agent is deployed on the edge computing node. The intelligent agent uses a reinforcement learning algorithm to optimize the initial strategy, adaptively adjusts the real-time data stream to generate optimized agent strategy data, and dynamically optimizes the global strategy through a multi-agent collaborative mechanism, a genetic algorithm, and a particle swarm optimization algorithm to generate collaborative optimization strategy data; the collaborative optimization control module deploys an intelligent agent on the edge computing node and optimizes the initial strategy through a reinforcement learning algorithm. The intelligent agent adaptively adjusts according to the real-time data stream to generate optimized agent strategy data. The multi-agent collaborative mechanism allows multiple intelligent agents to share and exchange information, and dynamically optimizes the global strategy through a genetic algorithm and a particle swarm optimization algorithm to generate collaborative optimization strategy data. This module ensures that the switch cabinet system can flexibly respond to various dynamic environmental changes and optimize its operating efficiency.

[0199] Through the augmented reality interaction and decision-making module, the collaborative optimization strategy data is integrated into the augmented reality platform to generate augmented reality visualization data, and the real-time data stream is analyzed by using graph convolutional networks and Bayesian networks for fault prediction, maintenance prediction data and decision support data are generated, and the operating parameters of the switch cabinet are adjusted in real time through the adaptive feedback control network to generate feedback control data. Through the augmented reality interaction and decision-making module, the system integrates the collaborative optimization strategy data into the augmented reality platform to generate augmented reality visualization data. Users can monitor the operating status of the switch cabinet in real time through augmented reality devices to perform fault prediction and decision support. Graph convolutional networks and Bayesian networks are used to analyze real-time data streams to generate maintenance prediction data and decision support data. The adaptive feedback control network combines these analysis results to adjust the operating parameters of the switch cabinet in real time to ensure that the system operates in the best state.

[0200] In actual application at a large power substation, after the system was successfully deployed, the real-time data collected by the sensors was quickly processed and analyzed. Through augmented reality equipment, operation and maintenance personnel can monitor the status of each switch cabinet in real time, graph convolutional networks help them predict potential failures, and Bayesian networks provide detailed decision support. When a switch cabinet approaches the load limit, the system immediately optimizes the control strategy through intelligent agents and performs load balancing adjustments, thereby avoiding equipment overload and potential failures.

[0201] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware.

[0202] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A visual control and management system for a switch cabinet, characterized in that: include: A multimodal data acquisition module is used to collect environmental status data, electrical parameter data and mechanical vibration data, and use an adaptive multimodal fusion algorithm combined with an automatic encoder model to fuse and denoise the data to generate fused feature data; An edge computing processing module, used for receiving the fused feature data, and performing real-time preprocessing on the edge computing node to generate a preprocessed data summary; A feature extraction and reasoning module is used to extract multi-level feature data from the preprocessed data summary, use a multi-layer variational autoencoder and a deep belief network, and generate a fuzzy rule set based on the feature data, generate fuzzy reasoning parameters for guiding the specific operation adjustment of the switch cabinet in real time through a fuzzy logic reasoning engine, and optimize the feature extraction process and the fuzzy rule set in combination with a feedback loop; the feature extraction and reasoning module includes: A multi-layer variational autoencoder is configured to extract multi-level feature data from the preprocessed data summary. The multi-layer variational autoencoder captures complex features in the data through a multi-level network structure and is optimized using the following variational loss function: in, is the loss function; and are the network parameters of the autoencoder; For input data; is a hidden variable; is the variational distribution; To generate distribution; represents the Kullback-Leibler divergence between the variational distribution and the generating distribution; Represents hidden variables The prior distribution of A deep belief network is used to further analyze the nonlinear relationship in the multi-level feature data and generate a fuzzy rule set in combination with the extracted features; a fuzzy rule generator is used to optimize the fuzzy rule set based on historical data and expert knowledge base so that the rule set can adapt to different system states; A collaborative optimization control module includes an intelligent agent deployed on an edge computing node, wherein the intelligent agent uses a reinforcement learning algorithm to optimize the initial strategy, adaptively adjusts the real-time data stream to generate optimized agent strategy data, and dynamically optimizes the global strategy through a multi-agent collaborative mechanism, a genetic algorithm, and a particle swarm optimization algorithm to generate collaborative optimization strategy data; The augmented reality interaction and decision-making module is used to integrate the collaborative optimization strategy data in the augmented reality platform, generate augmented reality visualization data, use graph convolutional networks and Bayesian networks to analyze the real-time data stream for fault prediction, generate maintenance prediction data and decision support data, adjust the operating parameters of the switch cabinet in real time through an adaptive feedback control network, and generate feedback control data.

2. The visual control and management system of the switch cabinet according to claim 1 is characterized in that: The multimodal data acquisition module comprises: A multimodal sensor array for simultaneously collecting environmental status data, electrical parameter data, and mechanical vibration data, wherein each data type corresponds to a different sensor, including an optical fiber sensor, an acoustic sensor, and an infrared sensor; A data fusion unit is configured to integrate raw data from different sensors using an adaptive multimodal fusion algorithm, wherein the algorithm assigns different weights to each data type. , generate fusion feature data, the formula is: in, To fuse feature data; For the The weight of each sensor data; For the Data from sensors; is the number of sensors.

3. The visual control and management system of the switch cabinet according to claim 1 is characterized in that: The edge computing processing module includes: A real-time processor receives the fused feature data and performs data denoising using an autoencoder model, wherein the autoencoder model removes redundant information in the data through an autoencoder neural network; The data summarizer optimizes the format and compresses the denoised data to generate a pre-processed data summary suitable for fast transmission and analysis.

4. The visual control and management system of the switch cabinet according to claim 1, characterized in that: The fuzzy logic inference engine comprises: A fuzzification module converts the feature data into fuzzy variables; The reasoning module performs real-time reasoning based on the fuzzy rule set and generates fuzzy reasoning parameters. Its reasoning formula is: in, is the fuzzy reasoning parameter; Fuzzy rules The membership function of For rules In Features The application results on is the number of fuzzy rules; The defuzzification module converts the fuzzy reasoning parameters into specific operation instructions so as to dynamically adjust the operating parameters of the switch cabinet.

5. The visual control and management system of the switch cabinet according to claim 1, characterized in that: It also includes a feedback loop optimization unit, which includes: Response collector, which collects system response data and performance indicators in real time; The optimization engine performs adaptive optimization adjustments on the feature extraction unit and the fuzzy rule generator based on the collected data.

6. The visual control and management system of the switch cabinet according to claim 1, characterized in that: The collaborative optimization control module includes: The intelligent agent system is configured to be deployed on each edge computing node. The intelligent agent uses a reinforcement learning algorithm to continuously optimize its initial strategy; The adaptive adjustment mechanism obtains the latest system status information through real-time data streams, thereby adjusting and optimizing proxy strategy data in real time.

7. The visual control and management system of the switch cabinet according to claim 6, characterized in that: The multi-agent collaboration mechanism achieves strategy sharing and global optimization through the following steps: Genetic algorithms, which use selection, crossover, and mutation operations to generate multiple optimization strategy solutions; The particle swarm optimization algorithm iteratively updates the position and velocity of particles, and its optimization formula is: in, For particles In the The speed of generation; For particles In the The position of the generation; is the inertia weight; and is the learning factor; and For A random number between For particles The best position in history; is the global best position; This generates collaborative optimization strategy data.

8. The visual control and management system of the switch cabinet according to claim 1, characterized in that: The augmented reality interaction and decision-making module includes: An augmented reality platform integrates the collaborative optimization strategy data and generates visual augmented reality data, so that users can monitor the operating status of the switch cabinet in real time through augmented reality devices; The prediction analysis system uses graph convolutional networks and Bayesian networks to analyze real-time data streams and generate maintenance prediction data. Its prediction formula is: in, is the predicted output; is the adjacency matrix; is the node feature matrix; is the normalized adjacency matrix; and is the weight matrix; is the activation function; The decision support system generates decision support data by combining the prediction analysis results and adjusts the operating parameters of the switchgear through an adaptive feedback control network to generate feedback control data.

9. A method for implementing the visual control management system of a switch cabinet according to any one of claims 1 to 8, characterized in that: The following steps are involved: Collect environmental status data, electrical parameter data and mechanical vibration data through a multimodal data acquisition module, and fuse and denoise the data using an adaptive multimodal fusion algorithm combined with an autoencoder model to generate fused feature data; The fused feature data is received through an edge computing processing module, and real-time preprocessing is performed on the edge computing node to generate a preprocessed data summary; Extracting multi-level feature data from the preprocessed data summary through a feature extraction and reasoning module, using a multi-layer variational autoencoder and a deep belief network, and generating a fuzzy rule set based on the feature data, generating fuzzy reasoning parameters through real-time analysis by a fuzzy logic reasoning engine, and optimizing the feature extraction process and the fuzzy rule set in combination with a feedback loop; Through the collaborative optimization control module, an intelligent agent is deployed on the edge computing node. The intelligent agent uses a reinforcement learning algorithm to optimize the initial strategy, adaptively adjusts the real-time data stream to generate optimized agent strategy data, and dynamically optimizes the global strategy through a multi-agent collaborative mechanism, a genetic algorithm, and a particle swarm optimization algorithm to generate collaborative optimization strategy data. Through the augmented reality interaction and decision-making module, the collaborative optimization strategy data is integrated into the augmented reality platform to generate augmented reality visualization data. The real-time data stream is analyzed using graph convolutional networks and Bayesian networks to predict faults, generate maintenance prediction data and decision support data, and adjust the operating parameters of the switch cabinet in real time through an adaptive feedback control network to generate feedback control data.

Images