A monitoring and early warning method and system for substation power equipment

Abstract

The present invention relates to the technical field of power equipment monitoring and early warning, and particularly to a monitoring and early warning method and system for substation power equipment. It includes a data acquisition module, an intelligent diagnosis module, a risk assessment module, a collaborative early warning module, and an interactive control module. The data acquisition module collects equipment operation data, including current, voltage, temperature and humidity, and vibration signals. The intelligent diagnosis module identifies equipment anomalies through a generative adversarial network and a graph neural network. The risk assessment module generates a risk assessment curve based on a fractional-order differential model and a reinforcement learning algorithm. The collaborative early warning module uses edge computing to generate a hierarchical early warning plan. The interactive control module displays the status through a human-machine interface and optimizes the control strategy. The present invention improves the intelligent monitoring and management level of substation power equipment, extends the service life of the equipment, and reduces the operation and maintenance costs.

Description

Technical Field

[0001] The present invention relates to the technical field of power equipment monitoring and early warning, and particularly to a monitoring and early warning method and system for substation power equipment. Background Art

[0002] With the rapid development of the power industry and the continuous expansion of the power grid scale, as an important node of the power grid, the safe operation of substations directly affects the reliability and stability of the power system. Power equipment in substations (such as transformers, circuit breakers, capacitors, etc.) is a key component of the operation of the power system. However, due to complex operating environments, equipment aging, or improper operations, etc., potential faults often occur in substation power equipment, such as overheating, partial discharge, mechanical damage, etc. If these potential faults are not discovered and processed in time, they may lead to equipment failures and even large-scale power outages of the power grid, causing significant economic losses and social impacts.

[0003] In the prior art, traditional substation monitoring mainly relies on manual inspections and regular maintenance. However, manual inspections have defects such as long detection cycles, poor real-time performance, and insufficient coverage, making it difficult to timely discover potential problems during equipment operation. In addition, although some equipment is equipped with single-parameter monitoring functions (such as temperature, humidity, current, etc.), multi-parameter comprehensive analysis and fault early warning cannot be achieved, resulting in a low discovery rate of potential faults and difficulty in accurately locating the specific location where the fault occurs; the evaluation models of the prior art have low accuracy and lack the ability of dynamic adjustment, making it difficult to reflect the changes in equipment risks in real time. Summary of the Invention

[0004] To solve the above problems, the present invention provides a monitoring and early warning method and system for substation power equipment, which solves the problems of how to realize real-time monitoring of the operating status of substation power equipment and multi-parameter comprehensive analysis, timely identify potential fault hazards and accurately locate the fault location, and at the same time improve the dynamic adaptability of risk assessment to ensure the safe and reliable operation of the power system, thereby improving the intelligent monitoring and management level of substation power equipment.

[0005] To achieve the above object, the technical solution adopted by the present invention is:

[0006] On the one hand, a monitoring and early warning system for substation power equipment includes a data acquisition module, an intelligent diagnosis module, a risk assessment module, a collaborative early warning module, and an interactive control module that are communicatively connected in sequence;

[0007] The data acquisition module is used to collect the operating data of substation power equipment, and the operating data includes the current, voltage, temperature, humidity, and vibration signals of the power equipment;

[0008] The intelligent diagnosis module is used to deeply analyze the operation characteristics of power equipment and identify abnormal patterns based on the operation data by using a generative adversarial network combined with a graph neural network algorithm, and generate a power equipment diagnosis report;

[0009] The risk assessment module is used to, based on the power equipment diagnosis report, adopt a dynamic assessment framework combining a fractional-order differential model and a reinforcement learning algorithm. Through the fractional-order differential model, it highly accurately models the change rate and trend of the operation state of power equipment, and uses reinforcement learning to dynamically optimize the equipment risk level assessment index, generating a real-time risk assessment curve;

[0010] The collaborative warning module is used to, by constructing a collaborative architecture based on edge computing, perform multi-node collaborative processing on the real-time risk assessment curve, and generate a hierarchical warning plan according to the area where the equipment is located and the operation level;

[0011] The interactive control module is used to, based on an adaptive optimization algorithm of a fuzzy neural network, display the equipment operation state, fault information, and warning suggestions through a human-machine interaction interface, and provide feedback to adjust parameters based on the interface.

[0012] Further, the operation process of the intelligent diagnosis module includes the following steps:

[0013] Perform time series segmentation and feature fusion on the operation data, and construct a multi-modal feature vector based on time-frequency domain analysis methods;

[0014] Based on the multi-modal feature vector, use a generative adversarial network. The generator simulates the potential fault state of the equipment by perturbing the collected data, and the discriminator dynamically optimizes the normal and abnormal data to generate fault-sensitive features;

[0015] Based on the fault-sensitive features, according to the actual spatial distribution and operation state of substation equipment, map the power equipment into a graph structure. The graph nodes represent single power equipment, and the graph edges represent electrical or physical associations between equipment, generating a regionalized graph structure;

[0016] Based on the regionalized graph structure, use a graph neural network algorithm to analyze the association characteristics and fault propagation patterns between power equipment, and identify abnormal nodes and abnormal paths;

[0017] Based on the results of abnormal pattern analysis, classify and identify the abnormal types and influence ranges in the equipment operation state, and generate a power equipment diagnosis report.

[0018] Furthermore, the fault-sensitive features include the frequency characteristics, vibration modes, electrical parameters, and environmental coupling characteristics of power equipment.

[0019] Further, the operation process of the risk assessment module includes the following steps:

[0020] Based on the power equipment diagnosis report, establish the mapping relationship between abnormal patterns and potential risk factors, and extract the risk factor characteristics;

[0021] Based on the risk factor characteristics, use the fractional differential model to perform high-precision dynamic modeling on the equipment operating state, capture the rate and trend of the power equipment state change, and generate the power equipment operating state characteristic curve;

[0022] Based on the power equipment operating state characteristic curve, dynamically optimize the risk assessment index through the reinforcement learning algorithm, and construct a multi-level risk assessment framework including operating reliability, failure probability, and risk propagation degree;

[0023] Based on the multi-level risk assessment framework and the power equipment diagnosis report, dynamically analyze the current operating state and potential risks of the equipment, and generate the equipment risk level;

[0024] Based on the dynamic risk analysis results, generate a real-time risk assessment curve.

[0025] Furthermore, the formula of the fractional differential model is as follows:

[0026]

[0027] Among them, represents the fractional derivative of the power equipment operating state x(t), that is, the change rate of the power equipment operating state; x(t) represents the operating state eigenvalue of the power equipment; t represents the current moment; q and represent the fractional order, q is the dominant order, is the fractional order corresponding to other characteristic variables; represents the contribution degree of different operating data to the overall state dynamic change; represents the disturbance intensity of environmental fluctuations on the equipment operating state; represents the time decay rate of environmental impact; represents the response intensity of the equipment operating state to different time delays; represents the power equipment at the time point receives an external input signal; represents uncontrollable random factors.

[0028] Furthermore, the formula of the reinforcement learning algorithm is as follows:

[0029]

[0030] Among them, represents the effect of power equipment risk assessment optimization; is the state parameter at time step t; P represents the update weight, which controls the proportion of the historical Q value and the current optimization benefit; represents the measurement of the action the direct contribution to the current risk assessment optimization; D represents the discount factor; M represents the number of multi-level risk assessment indicators; represents the weight of the i-th layer of risk assessment indicators; represents the next state at the action under the optimization effect of the i-th layer of indicators.

[0031] Furthermore, based on the geographical distribution characteristics and communication delay of the devices, the collaborative warning module uses an edge node hierarchical mechanism to dynamically optimize the allocation of collaborative computing tasks, and by constructing a node priority queue, it performs task load balancing allocation according to the computing power, network stability, and operational importance of regional devices, and at the same time combines the regional risk level to achieve real-time response of high-priority nodes.

[0032] Furthermore, the interactive control module uses a multi-objective decision-making optimization algorithm based on a fuzzy neural network to adaptively analyze the device operation state and risk assessment results, generate an optimized control strategy, and dynamically adjust the strategy execution path through a visual interface.

[0033] On the other hand, a monitoring and warning method for substation electrical equipment includes the following steps:

[0034] Collect the operation data of substation electrical equipment;

[0035] Based on the operation data, use a generative adversarial network combined with a graph neural network algorithm to deeply analyze the operation characteristics of electrical equipment and identify abnormal patterns, and generate a diagnostic report for electrical equipment;

[0036] Based on the diagnostic report of the electrical equipment, use a dynamic evaluation framework combining a fractional-order differential model and a reinforcement learning algorithm to generate a real-time risk assessment curve;

[0037] Through constructing a collaborative architecture based on edge computing, perform multi-node collaborative processing on the real-time risk assessment curve, and generate a hierarchical warning plan according to the device location and operation level;

[0038] Based on an adaptive optimization algorithm of a fuzzy neural network, display the device operation state, fault information, and warning suggestions through a human-machine interaction interface, and adjust the parameters based on the feedback provided by the interface.

[0039] The beneficial effects of the present invention are as follows:

[0040] In the present invention, the data acquisition module can comprehensively obtain the key data of the operation of power equipment, providing a high-quality data basis for subsequent analysis; the intelligent diagnosis module combining the generative adversarial network and the graph neural network algorithm deeply analyzes the operation data and identifies abnormal patterns, which can quickly and accurately discover potential problems and improve the diagnosis accuracy. Through the risk assessment module, the fractional-order differential model is used to accurately model the rate of change and trend of the equipment state, and the risk assessment index is dynamically optimized by reinforcement learning, realizing the dynamic and real-time adjustment of the risk level, generating a high-precision risk assessment curve, and providing effective support for the safe operation of the equipment. The collaborative warning module constructs a multi-node collaborative architecture based on edge computing, which can efficiently distribute and process the risk assessment results, and generate a hierarchical warning plan according to the equipment area and operation level, improving the response speed and pertinence of the warning, and enhancing the reliability and expandability of the system. The interactive control module, through the adaptive optimization algorithm of the fuzzy neural network and combined with an intuitive human-machine interface, real-time displays the equipment operation status, fault information and warning suggestions, and at the same time supports the user to adjust parameters through feedback, so as to realize the real-time optimization of the operation status and enhance the usability and controllability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 is a schematic diagram of the modules of a monitoring and warning system for substation power equipment according to the present invention.

[0042] Figure 2 is a schematic flowchart of the operation process of the risk assessment module provided by an embodiment of the present invention.

[0043] Figure 3 is a schematic flowchart of a monitoring and warning method for substation power equipment according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0044] Please refer to Figures 1 - 3 as shown, the present invention relates to a monitoring and warning method and system for substation power equipment.

[0045] Embodiment 1

[0046] A monitoring and warning system for substation power equipment includes a data acquisition module, an intelligent diagnosis module, a risk assessment module, a collaborative warning module and an interactive control module that are sequentially communicatively connected;

[0047] The data acquisition module is used to collect the operation data of substation power equipment, and the operation data includes the current, voltage, temperature, humidity and vibration signals of the power equipment;

[0048] Specifically, by deploying various types of sensors and acquisition devices, the key operation parameters of substation power equipment are real-time monitored, including current, voltage, temperature, humidity and vibration signals.

[0049] Current and voltage acquisition: By installing high-precision current transformers and voltage transformers, accurate acquisition of the electrical parameters of the equipment is achieved, and the noise in the acquired data is eliminated through filtering and anti-interference algorithms.

[0050] Temperature and humidity acquisition: Multiple temperature and humidity sensors are arranged inside and around the equipment, and a distributed acquisition scheme is adopted to monitor the changing trends of the temperature and humidity in the equipment environment.

[0051] Vibration signal acquisition: By installing acceleration sensors or MEMS sensors, the vibration signals of the equipment are obtained, and the vibration spectrum characteristics are extracted by combining with the FFT (Fast Fourier Transform) algorithm to determine whether there are mechanical abnormalities in the equipment.

[0052] Communication interface: The data acquisition module integrates multiple communication protocols (such as Modbus, IEC 61850, MQTT), and local data storage and preliminary processing are carried out through edge computing devices to ensure the real-time performance and integrity of data transmission.

[0053] The intelligent diagnosis module is used to deeply analyze the operation characteristics of power equipment and identify abnormal patterns based on the operation data by using a generative adversarial network combined with a graph neural network algorithm, and generate a diagnosis report for the power equipment;

[0054] Among them, the operation process of the intelligent diagnosis module includes the following steps:

[0055] Perform time series segmentation and feature fusion on the operation data, and construct a multi-modal feature vector based on time-frequency domain analysis methods;

[0056] Based on the multi-modal feature vector, use a generative adversarial network. The generator simulates the potential fault states of the equipment by perturbing the acquired data, and the discriminator dynamically optimizes the normal and abnormal data for generating fault-sensitive features; the fault-sensitive features include the frequency characteristics, vibration modes, electrical parameters, and environmental coupling characteristics of power equipment.

[0057] Specifically, input the multi-modal feature vector, and the generator of the generative adversarial network (GAN) perturbs the feature data to simulate the potential fault states of the equipment. The generator adopts a structure combining a fully connected neural network and a convolutional neural network (CNN) to generate feature samples containing potential fault characteristics. The discriminator learns to distinguish the characteristics of normal data and abnormal data by inputting the generated feature samples and the actually acquired operation data. The discriminator dynamically adjusts the weights to enhance the extraction of abnormal features and suppress the invalid perturbations of the generator.

[0058] Compare the features of the generated abnormal data and normal data, and extract sensitive features, including:

[0059] Frequency characteristics: harmonic frequency variations, harmonic distortions, etc. during the operation of the equipment.

[0060] Vibration mode: vibration amplitude and spectrum variations of mechanical components.

[0061] Electrical parameters: fluctuation characteristics of current and voltage, and overload phenomena.

[0062] Environmental coupling characteristics: correlation variations between the operation of the equipment and environmental humidity and temperature.

[0063] Based on the fault-sensitive characteristics, according to the actual spatial distribution and operation status of substation equipment, map power equipment into a graph structure, where graph nodes represent single power equipment and graph edges represent electrical or physical associations between equipment, generating a regionalized graph structure;

[0064] Specifically, according to the spatial distribution and operation status of the equipment, map multiple power equipment into a regionalized graph structure:

[0065] Node definition: Each power equipment corresponds to a node, and node attributes include the fault-sensitive characteristics of the equipment.

[0066] Edge definition: Based on the electrical network or physical structure, define the associated edges between equipment (such as current coupling, mechanical connection).

[0067] Describe the association strength between equipment through a weighted adjacency matrix, and the weights are determined by the distance, transmission power, and fault correlation between equipment. For example, the weight of the electrical coupling edge between a transformer and a switch is relatively high, while the weight of the physical connection edge between remote equipment is relatively low. According to the geographical distribution or functional partition of the equipment (such as the main transformer area, high-voltage switch area), divide the global graph into multiple regional subgraphs. Improve the calculation efficiency through subgraph processing and reduce the interference of edge equipment.

[0068] Based on the regionalized graph structure, use the graph neural network algorithm to analyze the association characteristics and fault propagation patterns between power equipment, and identify abnormal nodes and abnormal paths;

[0069] Specifically, use the graph convolutional network (GCN) to model the graph structure: aggregate the features of neighbor nodes through graph convolutional layers and update the global attributes of the nodes. Dynamically adjust the weights of graph edges in combination with fault-sensitive characteristics to enhance the significance of abnormal propagation paths. Simulate the propagation path of faults in the graph structure and analyze the possible influence range. Identify high-risk nodes (abnormal equipment) and critical edges (connections that may cause cascading faults). Through the output layer of the GCN, classify the operation status of each node and mark abnormal nodes. Based on the shortest path algorithm or random walk method of the graph, locate the critical path of abnormal propagation.

[0070] Based on the results of abnormal pattern analysis, classify and identify the types of abnormalities and the scope of influence in the operating state of the device, and generate a power equipment diagnostic report.

[0071] It should be noted that, combining the analysis results of generative adversarial networks and graph neural networks, classify the operating states of abnormal nodes:

[0072] Electrical abnormalities: such as current overload and voltage fluctuation.

[0073] Mechanical abnormalities: such as excessive vibration and running jamming.

[0074] Environmental abnormalities: such as insulation reduction caused by high humidity.

[0075] Based on the analysis results of the fault propagation path, evaluate the scope of equipment that may be affected by the abnormality. Mark key equipment and areas, determine the priority order of handling. Output a diagnostic report, including an overview of equipment status, types of abnormalities, scope of influence, analysis of fault causes and repair suggestions. Provide visual diagnostic results, such as equipment operating state distribution maps, abnormal propagation path maps, etc.

[0076] The risk assessment module is used to generate a real-time risk assessment curve based on the power equipment diagnostic report, a dynamic assessment framework combining a fractional differential model and a reinforcement learning algorithm. The fractional differential model is used to accurately model the change rate and trend of the operating state of power equipment, and the reinforcement learning is used to dynamically optimize the equipment risk level assessment index.

[0077] Among them, the operation process of the risk assessment module includes the following steps:

[0078] Based on the power equipment diagnostic report, establish a mapping relationship between abnormal patterns and potential risk factors, and extract risk factor characteristics.

[0079] Specifically, obtain key data such as historical operation data, abnormal records, and fault reports from the power equipment diagnostic report. Preprocess these data, including removing noise, filling missing values, and normalizing. Use clustering algorithms to classify the patterns of abnormal data and label different types of abnormal types (such as temperature abnormalities, current abnormalities, or vibration abnormalities). Establish a mapping with potential risk factors according to the abnormal types. For example, temperature abnormalities are associated with thermal failure, current abnormalities are associated with overload, and vibration abnormalities are associated with mechanical failure risks. Extract characteristic indicators corresponding to each abnormality, such as average temperature, temperature rise rate, current fluctuation amplitude, etc. Construct a risk factor characteristic matrix, including time series information, characteristic statistical values, and importance weights, to provide input for subsequent dynamic modeling.

[0080] Based on the characteristics of the risk factors, a fractional differential model is used to perform high-precision dynamic modeling of the equipment operation status, capture the rate and trend of the power equipment status change, and generate the characteristic curve of the power equipment operation status;

[0081] It should be noted that key input variables (such as temperature rise rate, vibration amplitude, load change amplitude) are selected from the risk factor matrix. These variables should have high temporal correlation and physical meaning. The sliding window technique (for example, the window length is set to 5 minutes) is used to segment the real-time data, and fractional differential modeling is performed on each segment of data. The parameters fitted within the window include the trend (such as rising / falling), the rate (the change amount per unit time), etc. The slope and curvature of the curve are output through fractional differentiation to capture the rate and trend of the equipment status change. For example: when the curve slope is positive and the curvature increases, it indicates that the equipment status is deteriorating rapidly; when the curve slope is negative and the curvature decreases, it indicates that the status is returning to normal. The fractional order q is optimized using gradient descent or genetic algorithms to ensure the minimization of the curve fitting error. The accuracy of the curve fitting is verified. The result curve of the fractional differential modeling is visualized, and key points (such as risk trigger points, change acceleration points) are marked with different colors. Combining with the diagnostic report, annotate the operation trend and potential risk points of the equipment.

[0082] Based on the characteristic curve of the power equipment operation status, the risk assessment index is dynamically optimized through a reinforcement learning algorithm, and a multi-level risk assessment framework including operation reliability, failure occurrence probability, and risk propagation degree is constructed;

[0083] Specifically, define the initial risk assessment index set, including operation reliability (such as MTTF / MTBF), failure occurrence probability (such as historical data statistics), and risk propagation degree (such as the mutual influence coefficient between devices). Set the initial weight value for each index. Use a reinforcement learning algorithm (such as deep Q-learning or Actor-Critic algorithm), with the equipment historical data as the training set, to construct a risk assessment index optimization model. Through the dynamic relationship between the equipment operation status characteristic curve and the risk factors, set the reward and punishment mechanism (such as giving a positive reward for reducing the risk level, and giving a positive reward for improving the prediction failure accuracy). According to the optimal weights output by the reinforcement learning, optimize the weight allocation of each risk assessment index in the framework. Conduct a comprehensive analysis of operation reliability, failure occurrence probability, and risk propagation degree according to the hierarchical structure (device level, system level, network level) to form a multi-level assessment framework.

[0084] Based on the multi-level risk assessment framework and the power equipment diagnostic report, dynamically analyze the current operation status and potential risks of the equipment to generate the equipment risk level;

[0085] Based on the results of the dynamic risk analysis, generate a real-time risk assessment curve.

[0086] Furthermore, the formula of the fractional-order differential model is as follows:

[0087]

[0088] Wherein, represents the fractional derivative of the operating state x(t) of the power equipment, that is, the change rate of the operating state of the power equipment; x(t) represents the eigenvalue of the operating state of the power equipment; t represents the current time; q and represent the fractional order, q is the dominant order, is the fractional order corresponding to other characteristic variables; represents the contribution degree of different operating data to the dynamic change of the overall state; represents the perturbation intensity of environmental fluctuations on the equipment operating state; represents the time decay rate of environmental impact; represents the response intensity of the equipment operating state to different time delays; represents the external input signal received by the power equipment at time point ; represents uncontrollable random factors.

[0089] Kernel function The specific formula is as follows:

[0090]

[0091] Wherein, represents the Gamma function, which is used for normalization; represents the memory degree parameter, which represents the sensitivity of the equipment to the past operating state; when is larger, the equipment is more sensitive to the relatively recent historical operating state (such as the most recent current or vibration signal); when is smaller, the equipment has a stronger cumulative memory of the more distant historical operating data (such as temperature fluctuations and humidity effects).

[0092] The formula of the reinforcement learning algorithm is as follows:

[0093]

[0094] Wherein, represents the effect of the power equipment risk assessment optimization; is the state parameter at time step t; P represents the update weight, which controls the ratio of the historical Q value to the current optimization benefit, and the value range is ; represents the direct contribution of measuring the action to the current risk assessment optimization; D represents the discount factor; M represents the number of multi-level risk assessment indicators, such as equipment operation stability, failure probability, and risk propagation range; denotes the weight of the risk assessment index at the i-th layer; denotes the next state in the action the optimization effect of the i-th layer index is as follows:

[0095]

[0096] where denotes the optimization increment of the i-th layer index; denotes the maximum benchmark value of the i-th layer index; denotes the risk adjustment coefficient of the i-th layer index, which is used to adjust the sensitivity of the index to the risk assessment result; denotes the risk value corresponding to the i-th layer index, which changes dynamically according to the equipment operation state.

[0097] The collaborative early warning module is used to perform multi-node collaborative processing on the real-time risk assessment curve by constructing a collaborative architecture based on edge computing, and generate a hierarchical early warning plan according to the equipment location area and operation level; the collaborative early warning module dynamically optimizes and allocates collaborative computing tasks based on the geographical distribution characteristics and communication delays of the equipment, and through constructing a node priority queue, performs task load balancing allocation according to the computing power, network stability of the nodes and the operation importance of the regional equipment, and at the same time combines the regional risk level to achieve real-time response of high-priority nodes.

[0098] In one embodiment, within the substation area, multiple edge computing nodes (such as high-performance edge gateways or edge servers) are deployed according to the distribution characteristics of power equipment. These nodes have independent computing capabilities, can receive data, analyze risks, and generate local early warnings. The edge nodes are interconnected through fiber optic communication, 5G network or wireless private network to achieve low-latency and highly reliable data exchange. Each edge node performs preliminary processing on the equipment risk assessment curve it is responsible for, generates preliminary early warning data. Through real-time communication between edge nodes, cross-node data is fused, the overall risk trend of the region is analyzed, and an optimized global early warning plan is generated. Distributed storage technology is used to ensure data consistency. Even if a certain node fails, other nodes can quickly take over its tasks to avoid information loss.

[0099] Task levels (e.g., emergency, high priority, general) are determined based on the operating status of the equipment and the risk level of the area. Emergency tasks include real-time responses to high-risk equipment; general tasks involve routine monitoring of low-risk equipment. Each node maintains a task queue and prioritizes tasks for high-risk equipment or critical areas. Task priorities are dynamically updated based on the equipment's geographic location, operational importance, and communication latency. Edge nodes dynamically adjust task allocation based on computing power and real-time network conditions to ensure efficient utilization of computing resources. A load prediction-based allocation mechanism is introduced to avoid node overload or resource waste. A reinforcement learning algorithm is used to adaptively optimize the task allocation strategy, improving task allocation efficiency by continuously learning about edge node performance and task execution results.

[0100] The warning level is generated based on indicators such as the power equipment health index, real-time risk assessment value, and regional equipment importance. The specific classification is as follows:

[0101] High-priority warnings: Generate real-time warnings for critical equipment (such as main transformers and switchgear) and equipment with higher operating risks, including fault type predictions and recommended emergency response strategies.

[0102] Medium priority warning: Provides trend forecasts and maintenance recommendations for equipment with fluctuating operating conditions.

[0103] Low-priority warning: For equipment that is operating normally, only monitoring data is recorded and a regular operation report is generated.

[0104] Generate diverse response plans for different warning levels:

[0105] High-priority alerts directly notify operations and maintenance personnel, linking them to the control module for rapid adjustments. Medium- and low-priority alerts are stored as reference information for subsequent equipment optimization or regular maintenance. Comprehensive analysis of risks across multiple devices in a region assesses the overall regional risk level. Regional-level alert plans are generated to guide the coordinated management of equipment within the region.

[0106] Edge nodes are classified into different levels based on their performance (e.g., CPU, memory), network latency, and reliability. High-performance nodes (e.g., servers with greater computing power) prioritize complex tasks, while low-performance nodes handle basic tasks. Node priorities are dynamically adjusted based on real-time device status changes and regional risk levels. A real-time priority adjustment algorithm is developed to optimize task allocation based on the real-time status of edge nodes. A backup node is set up for each critical task. When the primary node fails, the backup node immediately takes over the task, ensuring continuity of early warning services. When the risk level in a substation area increases, edge nodes in surrounding areas can collaboratively support its computing tasks, forming a cross-regional collaborative early warning network.

[0107] Push the generated warning information to the interactive control module for operation and maintenance personnel to view and provide feedback. The operation and maintenance personnel can confirm or adjust the warning suggestions through the interface, and the system will use the feedback results for model optimization. Regularly evaluate the warning accuracy, response time, and task assignment efficiency. Continuously optimize the warning algorithm through reinforcement learning to improve the system performance.

[0108] The interactive control module is used for the adaptive optimization algorithm based on the fuzzy neural network, to display the device operation status, fault information, and warning suggestions through the human-machine interaction interface, and to adjust the feedback parameters based on the interface; the interactive control module adopts the multi-objective decision-making optimization algorithm based on the fuzzy neural network, to perform adaptive analysis on the device operation status and risk assessment results, generate an optimized control strategy, and dynamically adjust the strategy execution path through the visualization interface.

[0109] In one embodiment, use dynamic charts (such as line charts, bar charts, heat maps) to display the device operation status, including core parameters such as current, voltage, temperature, humidity, and vibration signals. Dynamically update the risk assessment curve, and combine different colors to mark the risk levels (for example, green indicates low risk, yellow indicates medium risk, and red indicates high risk). Display the health summary of all monitored devices on the main interface, including the device operation status (healthy, abnormal, faulty), risk assessment value (represented by the health index HI), and warning level. The user can click on a specific device to enter the detailed page to view the diagnostic report and suggestions. Intuitively display the operation status and geographical distribution of each device through the 3D substation device model (digital twin). Real-time superimpose warning information on the device model (for example, the area with too high temperature is highlighted in red).

[0110] Classify and evaluate the device status based on the fuzzy rule base (for example, "if the temperature is high and the vibration is large, then the health index is low").

[0111] Fuzzy logic rule definition:

[0112] Input variables: operating parameters (such as current, voltage, temperature, vibration) and risk assessment values.

[0113] Output variable: optimized control strategy (such as adjusting operating parameters or recommending warning solutions).

[0114] Use the neural network to adaptively adjust the fuzzy rule weights and dynamically optimize the decision-making results. Weigh and optimize among multiple objectives (such as operating efficiency, maintenance cost, and fault risk): preferentially recommend stopping the operation of high-risk devices; suggest adjusting the operating parameters (such as reducing the load) for medium-risk devices; maintain normal operation for low-risk devices and reduce interference.

[0115] The optimization control strategy generation process includes the following steps: Obtain the current device operating status and risk assessment data. Based on the output of the fuzzy neural network, generate preliminary optimization suggestions in combination with the actual operating parameters of the device. Input the preliminary optimization suggestions into the multi-objective decision-making model. Optimize the execution path in combination with operating objectives (such as reducing risks, extending device life, etc.). For example: Optimization path 1: Reduce device load → Adjust the operating time period → Enhance heat dissipation. Optimization path 2: Increase the device tolerance threshold → Adjust the maintenance plan. Use simulation technology to predict the execution effect of the optimization strategy (such as reducing temperature, reducing vibration, etc.). The interface displays the prediction effect curves of different strategies to facilitate users to select the optimal solution.

[0116] Embodiment 2

[0117] A monitoring and early warning method for substation power equipment includes the following steps:

[0118] Collect the operating data of substation power equipment;

[0119] Based on the operating data, adopt a generative adversarial network combined with a graph neural network algorithm to deeply analyze the operating characteristics of power equipment and identify abnormal patterns, and generate a power equipment diagnosis report;

[0120] Based on the power equipment diagnosis report, a dynamic evaluation framework combining a fractional-order differential model and a reinforcement learning algorithm is used to generate a real-time risk assessment curve;

[0121] By constructing a collaborative architecture based on edge computing, perform multi-node collaborative processing on the real-time risk assessment curve, and generate a hierarchical early warning plan according to the device location area and operating level;

[0122] Based on the adaptive optimization algorithm of the fuzzy neural network, display the device operating status, fault information, and early warning suggestions through a human-computer interaction interface, and adjust parameters based on the feedback provided by the interface.

[0123] In this embodiment, a monitoring and early warning method for substation power equipment is applied to a monitoring and early warning system for substation power equipment described in Embodiment 1, which will not be elaborated here.

[0124] In summary, the present invention uses multiple types of sensors (such as current transformers, voltage transformers, temperature and humidity sensors, and acceleration sensors) to achieve high-precision acquisition of key operating parameters such as current, voltage, temperature and humidity, and vibration signals. Combined with filtering and anti-interference algorithms, it ensures the accuracy and integrity of data, providing a reliable basis for subsequent analysis. The algorithm combining a generative adversarial network (GAN) and a graph neural network (GNN) can not only simulate potential fault states, but also identify the complex correlation characteristics and fault propagation patterns of power equipment. Through multi-modal feature fusion and dynamic optimization, the accuracy and comprehensiveness of fault diagnosis are improved.

[0125] Based on the fractional-order differential model, the present invention performs high-precision modeling on the change rate and trend of the device operation state, capturing the dynamic characteristics of potential risks. Combining with the reinforcement learning algorithm, it dynamically optimizes the risk assessment indicators to generate a real-time risk assessment curve, ensuring real-time control of the device health state. Through the edge computing architecture, it realizes multi-node collaborative processing, dynamically optimizes the task allocation, and reduces the communication delay. Based on the regional risk level and device importance, it generates a hierarchical early warning scheme to ensure that high-risk devices are given priority responses, improving the overall early warning efficiency and reliability.

[0126] The present invention displays the device operation state, fault information, and risk curve through a visualization interface, providing clear and intuitive device health information. At the same time, combining with the fuzzy neural network adaptive optimization algorithm, users can adjust the system parameters in real time, improving the decision-making efficiency and optimizing the operation strategy. The system optimizes the utilization rate of computing resources through the task load prediction and dynamic allocation mechanism, and introduces a backup node mechanism to ensure the continuous operation ability of the system, maintaining a stable early warning service even in case of node failures.

[0127] According to the device operation state and diagnostic report, the system classifies and analyzes the types of anomalies (such as electrical, mechanical, environmental anomalies), providing detailed fault causes and influence ranges. The early warning scheme covers different levels, from emergency tasks to routine monitoring, ensuring flexible responses in different risk scenarios. By providing early warnings for high-risk devices, optimizing the operation parameters and fault response strategies, the system can effectively reduce the device operation risks, extend the device service life, and reduce the maintenance cost and the probability of operation interruption.

[0128] The above embodiments are only descriptions of the preferred embodiments of the present invention, and do not limit the scope of the present invention. Without departing from the design spirit of the present invention, various deformations and improvements made by those of ordinary engineering and technical personnel in the art to the technical solutions of the present invention shall fall within the protection scope determined by the claims of the present invention.

Claims

1. A monitoring and early warning system for substation power equipment, characterized in that, It includes a data acquisition module, an intelligent diagnosis module, a risk assessment module, a collaborative early warning module, and an interaction control module that are communicatively connected in sequence; The data acquisition module is used to acquire the operation data of substation power equipment, and the operation data includes the current, voltage, temperature, humidity, and vibration signals of the power equipment; The intelligent diagnosis module is used to deeply analyze the operation characteristics of power equipment and identify abnormal patterns based on the operation data by using a generative adversarial network combined with a graph neural network algorithm, and generate a power equipment diagnosis report; The risk assessment module is used to, based on the power equipment diagnosis report, adopt a dynamic assessment framework combining a fractional differential model and a reinforcement learning algorithm, perform high-precision modeling on the change rate and trend of the operation state of power equipment through the fractional differential model, and dynamically optimize the equipment risk level assessment index by using reinforcement learning to generate a real-time risk assessment curve; The collaborative early warning module is used to perform multi-node collaborative processing on the real-time risk assessment curve by constructing a collaborative architecture based on edge computing, and generate a hierarchical early warning plan according to the equipment location area and operation level; The interaction control module is used to, based on an adaptive optimization algorithm of a fuzzy neural network, display the equipment operation state, fault information, and early warning suggestions through a human-computer interaction interface, and provide feedback adjustment parameters based on the interface; The operation process of the risk assessment module includes the following steps: Based on the power equipment diagnosis report, establish a mapping relationship between abnormal patterns and potential risk factors, and extract risk factor characteristics; Based on the risk factor characteristics, use the fractional differential model to perform high-precision dynamic modeling on the equipment operation state, capture the change rate and trend of the power equipment state, and generate a power equipment operation state characteristic curve; Based on the power equipment operation state characteristic curve, dynamically optimize the risk assessment index through a reinforcement learning algorithm, and construct a multi-level risk assessment framework including operation reliability, fault occurrence probability, and risk propagation degree; Based on the multi-level risk assessment framework and the power equipment diagnosis report, dynamically analyze the current operation state and potential risks of the equipment, and generate an equipment risk level; Based on the dynamic risk analysis result, generate a real-time risk assessment curve; The formula of the fractional differential model is as follows: ； Among them, represents the fractional derivative of the operation state x(t) of the power equipment, that is, the change rate of the operation state of the power equipment; x(t) represents the operation state eigenvalue of the power equipment; t represents the current time; q and represent the fractional orders, q is the dominant order, is the fractional order corresponding to other characteristic variables; represents the contribution degree of different operation data to the dynamic change of the overall state; represents the perturbation intensity of environmental fluctuations on the operation state of the equipment; represents the time decay rate of environmental impact; represents the response intensity of the operation state of the equipment to different time delays; represents the power equipment at the time point receives an external input signal; represents uncontrollable random factors; The formula of the reinforcement learning algorithm is as follows: ； Among them, represents the effect of optimizing the risk assessment of power equipment; is the state parameter at time step t; P represents the update weight, controlling the proportion of the historical Q value and the current optimization benefit; represents the measure of the action 's direct contribution to the current risk assessment optimization; D represents the discount factor; M represents the number of multi-level risk assessment indicators; represents the weight of the i-th layer risk assessment indicator; represents the next state under the action and the optimization effect of the i-th layer indicator.

2. The monitoring and early warning system for substation power equipment according to claim 1, characterized in that, The operation process of the intelligent diagnosis module includes the following steps: Perform time series segmentation and feature fusion on the operation data, and construct a multi-modal feature vector based on the time-frequency domain analysis method; Based on the multi-modal feature vector, use a generative adversarial network. The generator simulates the potential fault state of the equipment by perturbing the collected data, and the discriminator dynamically optimizes the normal and abnormal data through confrontation to generate fault-sensitive features; Based on the fault-sensitive features, map the power equipment into a graph structure according to the actual spatial distribution and operation state of substation equipment. The graph nodes represent a single power equipment, and the graph edges represent electrical or physical associations between equipment, generating a regionalized graph structure; Based on the regionalized graph structure, use the graph neural network algorithm to analyze the association characteristics and fault propagation modes between power equipment, and identify abnormal nodes and abnormal paths; Based on the results of abnormal pattern analysis, classify and identify the types of abnormalities and the scope of influence in the operating state of the device, and generate a diagnostic report for power equipment.

3. The monitoring and early warning system for substation power equipment according to claim 2, wherein, The fault-sensitive features include the frequency characteristics, vibration modes, electrical parameters, and environmental coupling characteristics of power equipment.

4. The monitoring and early warning system for substation power equipment according to claim 1, characterized in that, Based on the geographical distribution characteristics and communication delays of the devices, the collaborative early warning module dynamically optimizes the allocation of collaborative computing tasks using an edge node hierarchical mechanism. By constructing a node priority queue, it balances the task load according to the computing power of the nodes, network stability, and the operational importance of regional devices. At the same time, it combines the regional risk level to achieve real-time response of high-priority nodes.

5. The monitoring and early warning system for substation power equipment according to claim 1, wherein, The interactive control module uses a multi-objective decision-making optimization algorithm based on a fuzzy neural network to adaptively analyze the operating state of the device and the results of risk assessment, generate an optimized control strategy, and dynamically adjust the strategy execution path through a visual interface.

6. A monitoring and early warning method for substation power equipment, characterized in that, The method is applied to a monitoring and early warning system for substation power equipment as described in any one of claims 1-5, and includes the following steps: Collect the operating data of substation power equipment; Based on the operating data, use a generative adversarial network combined with a graph neural network algorithm to deeply analyze the operating characteristics of power equipment and identify abnormal patterns, and generate a diagnostic report for power equipment; Based on the diagnostic report of the power equipment, generate a real-time risk assessment curve based on a dynamic evaluation framework combining a fractional-order differential model and a reinforcement learning algorithm; Through the construction of a collaborative architecture based on edge computing, perform multi-node collaborative processing on the real-time risk assessment curve, and generate a hierarchical early warning plan according to the region where the device is located and the operating level; Based on the adaptive optimization algorithm of the fuzzy neural network, display the operating state of the device, fault information, and early warning suggestions through a human-computer interaction interface, and adjust the parameters based on the feedback provided by the interface.

Images