Electrical equipment operation risk perception and identification system

Abstract

The invention relates to the technical field of electrical equipment monitoring, and discloses an electrical equipment operation risk perception and identification system. The system comprises a real-time operation state monitoring module which can collect a time sequence data set of voltage fluctuation, current waveform and temperature distribution. The multi-dimensional risk feature extraction module receives the data, performs spatio-temporal feature fusion through a multi-scale convolution kernel group, and generates a time sequence operation feature vector set. And a feature optimization screening module executes feature decoupling based on risk sensitivity on the vector set, and extracts an optimized risk feature subset with significant risk characterization capability. The risk feature migration calculation module inputs the risk feature migration index set into a risk evolution atlas network and calculates a risk feature migration index set of adjacent monitoring periods; and the risk evolution rate quantification module performs time differentiation on the migration index set according to a preset period interval to generate a risk characteristic evolution rate parameter. And the risk decision generation module carries out multi-level matching on the parameter and a predefined risk threshold pedigree, and outputs a risk regulation decision instruction.

Description

Technical Field

[0001] This invention relates to the field of electrical equipment monitoring technology, specifically to an electrical equipment operation risk perception and identification system. Background Technology

[0002] As the core carrier of energy transmission and conversion, the stable operation of electrical equipment is directly related to the continuity of electricity supply for various industrial production and residential use. During long-term operation, potential risks gradually accumulate inside the equipment due to problems such as insulation aging, component wear, and poor contact. At the same time, external environmental factors such as temperature changes, humidity fluctuations, and electromagnetic interference can also exacerbate the instability of the equipment's operating status.

[0003] Currently, a common approach to risk identification for electrical equipment is to deploy a single type of sensor to acquire a specific operating parameter, such as monitoring only current changes or temperature fluctuations. This method only reflects partial information about the equipment's operating status and cannot capture the correlation between different parameters. For example, abnormal voltage fluctuations may be linked to a sudden temperature rise; monitoring a single parameter is unlikely to detect such potential correlations, easily leading to a one-sided risk identification.

[0004] Traditional risk identification methods often employ static analysis, relying on monitoring data at a single moment to determine equipment malfunctions, neglecting the dynamic evolution of equipment status. Equipment risks frequently develop from minor anomalies to serious failures, a gradual process that static analysis struggles to track, often resulting in delayed risk warnings. Furthermore, manual analysis of monitoring data, limited by personnel expertise and processing efficiency, is ill-suited to the real-time monitoring needs of large-scale electrical equipment clusters, significantly compromising the timeliness and accuracy of risk identification.

[0005] With the advancement of smart grids and industrial automation, the operating environment of electrical equipment is becoming more complex, placing higher demands on the real-time, comprehensive, and accurate nature of risk identification. Existing technologies are no longer sufficient to meet the needs of practical applications. Summary of the Invention

[0006] The purpose of this invention is to provide an electrical equipment operation risk perception and identification system to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides an electrical equipment operation risk perception and identification system, the system comprising:

[0008] The real-time operating status monitoring module for electrical equipment is used to continuously collect time-series data sets of voltage fluctuations, current waveforms, and temperature distribution of electrical equipment.

[0009] The multi-dimensional risk feature extraction module is used to receive the voltage fluctuation time series data set, current waveform time series data set, and temperature distribution time series data set output by the real-time operation status monitoring module of the electrical equipment, and perform spatiotemporal feature fusion processing through multi-scale convolution kernel group to generate a set of time series operation feature vectors of electrical equipment.

[0010] The feature optimization and screening module is used to perform a risk-sensitivity-based feature decoupling operation on the set of time-series operation feature vectors of the electrical equipment, and extract an optimized risk feature subset with significant risk characterization capabilities;

[0011] The risk feature migration calculation module is used to input the optimized risk feature subset into the risk evolution graph network and calculate the risk feature migration index set of adjacent monitoring periods.

[0012] The risk evolution rate quantification module is used to perform time differentiation operation on the risk feature migration index set according to a preset monitoring period time interval to generate risk feature evolution rate parameters.

[0013] The risk decision generation module is used to perform multi-level matching analysis between the risk feature evolution rate parameter and the predefined risk threshold spectrum, and output electrical equipment risk control decision instructions.

[0014] Preferably, the system further includes:

[0015] The risk decision verification module is used to perform reverse simulation verification of the risk control decision instructions of the electrical equipment through a virtual operating environment, and generate a risk decision confidence matrix.

[0016] The dynamic decision optimization module is used to perform parameter compensation operations on the electrical equipment risk control decision instructions in conjunction with the risk decision confidence matrix, and generate optimized risk control decision instructions.

[0017] The risk response execution module is used to parse the control logic in the optimized risk control decision instruction and drive the electrical equipment protection device to perform corresponding actions.

[0018] The risk data closed-loop management module is used to store the raw data collected by the real-time operation status monitoring module of the electrical equipment, the risk feature migration index set, and the execution feedback results of the optimized risk control decision instructions.

[0019] Preferably, the feature optimization and filtering module includes:

[0020] The risk feature energy distribution modeling unit is used to construct the energy density distribution model of each feature vector in the set of time-series operation feature vectors of the electrical equipment, and generate a risk feature energy distribution spectrum sequence.

[0021] The feature energy clustering center identification unit is used to calculate the weighted centroid vector of the risk feature energy distribution spectrum sequence and mark it as the risk feature energy clustering center reference.

[0022] The feature energy discretization unit is used to measure the spectral discrete distance between each of the risk feature energy distribution spectra and the risk feature energy clustering center reference, and to generate a set of risk feature energy discretization coefficients.

[0023] The feature selection execution unit is used to mark feature vectors in the set of energy discrete coefficients of the risk features that exceed a preset discrete threshold as high-sensitivity risk features, and form the optimized risk feature subset.

[0024] Preferably, the risk feature energy distribution modeling unit includes:

[0025] The feature covariance analysis subunit is used to calculate the covariance strength between any two feature vectors in the set of time-series operation feature vectors of the electrical equipment, and generate a feature covariance strength matrix.

[0026] The feature energy weight allocation subunit is used to calculate the energy distribution weight value of each feature vector according to the feature covariance correlation strength matrix, and generate the feature energy weight allocation sequence.

[0027] The energy spectrum generation subunit is used to convolve and fuse the feature energy weight allocation sequence with the energy density distribution model of the corresponding feature vector to generate the risk feature energy distribution spectrum sequence.

[0028] Preferably, the feature collaborative correlation analysis subunit includes:

[0029] The feature vector matching unit selects a target feature vector from the set of timing operation feature vectors of the electrical equipment and matches it with all other associated feature vectors.

[0030] A feature interaction response unit is used to calculate the interaction response intensity value between the target feature vector and each associated feature vector, and generate a feature interaction response intensity sequence;

[0031] The association strength synthesis unit is used to perform nonlinear aggregation operations on the feature interaction response strength sequence to generate the association strength vector corresponding to the target feature vector in the feature covariant association strength matrix.

[0032] Preferably, the feature energy discretization unit includes:

[0033] The spectral difference calculation subunit is used to calculate the spectral difference vector between the risk feature energy clustering center benchmark and a single risk feature energy distribution spectrum;

[0034] The covariance matrix generation sub-unit is used to construct the covariance matrix of the risk feature energy distribution spectrum sequence and obtain its inverse matrix;

[0035] The discrete kernel function subunit is used to perform a quadratic operation on the spectrum difference vector and the inverse of the covariance matrix to generate intermediate discrete parameters.

[0036] The discrete coefficient conversion subunit is used to perform a square root operation on the intermediate discreteness parameter and output the corresponding risk characteristic energy discrete coefficient.

[0037] Preferably, the risk decision generation module includes:

[0038] The risk decision space loading unit is used to load a pre-configured multi-dimensional risk decision strategy space;

[0039] The decision association calibration unit is used to perform dynamic association mapping on the multi-dimensional risk decision strategy space according to the risk feature evolution rate parameter, and generate a calibrated risk decision strategy subspace.

[0040] A decision feature activation unit is used to identify decision trigger regions that satisfy the current risk state in the calibrated risk decision strategy subspace.

[0041] The decision fitness evaluation unit is used to measure the matching fitness parameters between each decision strategy and the current operating state of the electrical equipment within the decision triggering area;

[0042] The decision selection execution unit is used to select a decision strategy that satisfies the preset fitness constraints on the matching fitness parameters, and to form the electrical equipment risk control decision instruction.

[0043] Preferably, the decision correlation calibration unit includes:

[0044] The strategy record traversal unit is used to scan the historical risk decision records stored in the multi-dimensional risk decision strategy space;

[0045] The risk status similarity analysis unit is used to calculate the similarity parameter between the current risk feature evolution rate parameter and the risk feature evolution rate in historical risk decision records;

[0046] The strategy correlation synthesis unit is used to perform weighted integration processing on the similarity parameters to generate the strategy correlation coefficient;

[0047] The strategy filtering unit is used to add the corresponding historical risk decision records to the calibrated risk decision strategy subspace when the strategy correlation coefficient exceeds a preset correlation threshold.

[0048] Preferably, the risk decision verification module includes:

[0049] The virtual operating environment construction unit is used to create a three-dimensional virtual operating scene based on the topology of electrical equipment.

[0050] The decision-making and simulation execution unit is used to inject the electrical equipment risk control decision instructions into the three-dimensional virtual operation scenario for dynamic simulation.

[0051] The risk evolution monitoring unit is used to collect the dynamic change trajectory of risk characteristic parameters during the virtual simulation process;

[0052] The confidence assessment unit is used to analyze the degree of deviation between the dynamic change trajectory of the risk characteristic parameters and the expected risk control target, and to generate the risk decision confidence matrix.

[0053] Preferably, the dynamic decision optimization module includes:

[0054] The compensation decision space generation unit is used to create a compensation decision strategy space based on risk evolution prediction data.

[0055] The compensation strategy activation unit is used to identify compensation decision triggering regions that satisfy the current risk state in the compensation decision strategy space.

[0056] The iterative decision generation unit is used to perform multiple rounds of decision selection operations within the compensation decision triggering area based on a preset fitness criterion, and generate a set of compensation decision strategies.

[0057] The decision fusion unit is used to perform feature-level fusion of the compensation decision strategy set and the original electrical equipment risk control decision instructions, and output the optimized risk control decision instructions.

[0058] Compared with the prior art, the beneficial effects of the present invention are:

[0059] The real-time operation status monitoring module for electrical equipment synchronously collects time-series data sets of voltage fluctuations, current waveforms, and temperature distribution, enabling multi-dimensional and continuous perception of equipment operation status. This breaks through the limitations of traditional single-parameter monitoring and can comprehensively capture subtle changes during equipment operation.

[0060] The multi-dimensional risk feature extraction module uses multi-scale convolutional kernel groups for spatiotemporal feature fusion processing, which can mine hidden correlation information in the data from different time scales and spatial dimensions. The generated set of electrical equipment time-series operation feature vectors contains richer risk-related features, providing a more comprehensive foundation for subsequent risk identification.

[0061] The feature optimization and screening module performs feature decoupling based on risk sensitivity, which can eliminate redundant information, extract an optimized subset of risk features with significant risk representation capabilities, reduce the interference of irrelevant features on the identification results, make risk features more focused, and improve the effectiveness of features.

[0062] The risk feature migration calculation module optimizes the risk feature subset input into the risk evolution graph network and calculates the risk feature migration index set of adjacent monitoring periods. This can intuitively reflect the changing trend of risk features over time and show the correlation between risks in different periods.

[0063] The risk evolution rate quantification module performs time differentiation on the risk feature migration index set according to the preset monitoring cycle time interval, and generates risk feature evolution rate parameters, which can quantify the speed of risk change and clearly present the dynamic evolution trend of risk.

[0064] The risk decision generation module performs multi-level matching analysis between the risk characteristic evolution rate parameter and the predefined risk threshold spectrum. The output electrical equipment risk control decision instructions can take corresponding control measures for different risk levels, so that risk identification and control form a complete closed loop, which is more in line with the needs of risk handling in actual application scenarios. Attached Figure Description

[0065] Figure 1 This is a schematic diagram of the working principle of the electrical equipment operation risk perception and identification system described in this invention;

[0066] Figure 2 A flowchart for risk decision verification and closed-loop management;

[0067] Figure 3 A flowchart illustrating the operation of the feature optimization and filtering module;

[0068] Figure 4 A flowchart illustrating the operation of the characteristic energy discretization unit;

[0069] Figure 5 A flowchart illustrating the workflow of the risk decision generation module. Detailed Implementation

[0070] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0071] Please see Figure 1This invention provides an electrical equipment operation risk perception and identification system, comprising: a real-time electrical equipment operation status monitoring module, a multi-dimensional risk feature extraction module, a feature optimization and screening module, a risk feature migration calculation module, a risk evolution rate quantification module, and a risk decision generation module. Specific implementation methods are as follows:

[0072] The real-time operation status monitoring module for electrical equipment continuously collects time-series data sets of voltage fluctuations, current waveforms, and temperature distribution.

[0073] The multi-dimensional risk feature extraction module receives the above data set and performs spatiotemporal feature fusion processing through multi-scale convolutional kernel groups to generate a set of time-series operation feature vectors for electrical equipment.

[0074] The feature optimization and screening module performs a risk-sensitivity-based feature decoupling operation on the set of time-series operation feature vectors of electrical equipment to extract an optimized risk feature subset with significant risk characterization capabilities.

[0075] The risk feature migration calculation module will optimize the risk feature subset input risk evolution graph network and calculate the risk feature migration index set of adjacent monitoring periods.

[0076] The risk evolution rate quantification module performs time differentiation on the risk feature migration index set according to the preset monitoring cycle time interval to generate risk feature evolution rate parameters.

[0077] The risk decision generation module performs multi-level matching analysis between the risk characteristic evolution rate parameter and the predefined risk threshold spectrum, and outputs electrical equipment risk control decision instructions.

[0078] Example 1: See Figure 2The risk decision verification module verifies the effectiveness of electrical equipment risk control decision commands output by the risk decision generation module through a virtual operating environment. This module first uses the virtual operating environment construction unit to establish a high-fidelity 3D virtual operating scenario model based on the actual physical topology, connection relationships, and operating parameter configuration of the target electrical equipment. This model accurately simulates the electromagnetic field distribution, heat conduction path, current loop, and mechanical stress state of the electrical equipment under real operating conditions, forming the basic environment for a digital twin. Subsequently, the decision simulation execution unit converts the electrical equipment risk control decision commands to be verified into virtual control signals and injects them into the 3D virtual operating scenario model. This process requires strict time synchronization to simulate the execution sequence of commands in the actual equipment. After the simulation starts, the risk evolution monitoring unit continuously tracks and captures the preset key risk characteristic parameters in the virtual environment, recording their dynamic changes within the simulation period. The monitored parameters cover core indicators such as the equivalent virtual voltage anomaly mutation, the current harmonic distortion growth trend, and the temperature rise rate spectrum of hotspot areas. All captured change trajectory data is transmitted to the confidence assessment unit in real time. This unit employs a multi-dimensional deviation analysis algorithm to compare the actual evolution trajectory of risk characteristic parameters with the preset risk control targets across time slices. Specifically, it calculates the absolute deviation, relative deviation percentage, and deviation duration between the measured values ​​of key parameters and the expected safety thresholds at each monitoring time point. Combined with the weighting coefficient of this parameter in the risk model, it calculates the individual deviation impact factor for each item. Finally, by integrating and normalizing the deviation impact factors across all monitoring time segments, it outputs a risk decision confidence matrix that quantitatively represents the overall reliability of the decision. This matrix is ​​a multi-dimensional data structure, where rows represent different risk control dimensions (such as overpressure suppression and overheat control), columns represent the time dimension, and matrix element values ​​represent the decision execution confidence level for the corresponding risk dimension and time point.

[0079] The dynamic decision optimization module receives the risk decision confidence matrix generated by the risk decision verification module and the original electrical equipment risk control decision instructions. This module utilizes the risk evolution prediction data stream, including the current risk feature migration index set and predicted evolution trends, to create a compensation decision strategy space. This strategy space has a multi-level structure. The basic layer is a pre-set standard compensation strategy library, including a voltage adjustment gain coefficient set, a current limiting gradient scheme set, and a radiator power adjustment parameter set. The advanced layer uses an online learning engine to dynamically generate an adaptive strategy supplement set based on real-time environmental parameters (such as grid fluctuation status and load mutation prediction values). The compensation strategy activation unit operates within this strategy space. First, it analyzes the risk decision confidence matrix to identify the risk dimensions requiring key compensation (rows with confidence levels below a set warning threshold), and simultaneously associates this with the real-time state vector of the current electrical equipment risk feature evolution rate parameter. Combining these two pieces of information, it performs a region search and matching within the compensation decision strategy space, ultimately locking in the compensation decision trigger region that meets the requirements of the current specific risk state. The iterative decision generation unit performs multiple rounds of strategy screening within this activated region. Each round of selection is based on a preset fitness criterion, which involves multiple objective functions such as strategy execution cost weight, expected risk suppression efficiency weight, and operator response time weight. Each round calculates the composite fitness score for each strategy and selects a subset of strategies to form a temporary set based on the score ranking and predefined perturbation rules. This process gradually converges to generate a set of compensation decision strategies with optimal overall performance through a finite number of iterative loops (e.g., particle swarm optimization or genetic optimization). Finally, the decision fusion unit performs the following operations: on the one hand, it decodes the logical control features of the original electrical equipment risk control decision instructions; on the other hand, it deconstructs the core feature vectors of the compensation decision strategy set. Through feature space alignment and weighted fusion operations, it outputs the fused optimized risk control decision instructions. These instructions correct the parameter bias of the original decision in the low-confidence risk dimension while retaining the control logic of the original decision in the high-confidence dimension.

[0080] The risk response execution module receives optimized risk control decision instructions as input. This module is equipped with a dedicated instruction parsing engine that deconstructs the control logic tree within the optimized instructions layer by layer. This process includes identifying instruction types (such as graded alarms, active power suppression, dynamic load switching, and protective tripping), extracting key control parameters (such as set thresholds, adjustment ranges, action sequences, and backup circuit numbers), confirming the executing entity (such as intelligent circuit breakers, flexible power controllers, cooling fan arrays, and parallel capacitor switching units), and generating a device-level operation instruction sequence. While generating the instruction sequence, the parsing engine rigorously verifies the device readiness status feedback signals to prevent execution conflicts. The sequence is sent to the execution drive unit of the electrical equipment protection device. Based on the absolute or relative timestamp information in the instruction sequence, this drive unit strictly triggers the corresponding protection device's actuator according to a preset timing sequence, such as driving the intelligent circuit breaker's opening and closing coils, adjusting the soft starter's power output curve, and activating the multi-stage cooling fan control system. All actual action status signals fed back by the actuators (such as contact on / off status, real-time output power values, and speed feedback) are recorded and transmitted back.

[0081] The risk data closed-loop management module realizes the full lifecycle recording and flow of system data. This module is configured with a distributed time-series database cluster, possessing high-concurrency data throughput capabilities. The system clearly stores three types of core data: Raw layer data includes voltage fluctuation time-series data sets, current waveform time-series data sets, and temperature distribution time-series data sets collected by the electrical equipment real-time operation status monitoring module, stored in partitions by device ID and timestamp. Processing layer data includes optimized risk feature subsets generated by the feature optimization and screening module and risk feature migration index sets output by the risk feature migration calculation module. Decision execution layer data includes optimized risk control decision instructions generated by the dynamic decision optimization module, corresponding execution instruction sequences recorded by the risk response execution module, and a set of execution result feedback status codes returned by electrical equipment protection devices (including success, failure, timeout, etc.). All stored data items are accompanied by complete metadata descriptions, including the data source module, generation timestamp, associated device ID, and data processing version number. Simultaneously, this module provides standard data interfaces to all other modules in the system, supporting time-range retrieval, device-dimensional aggregation analysis, and key decision execution record traceability requests. Retaining raw data facilitates fault backtracking and reproduction; accumulating risk feature migration indices provides input for the self-learning of evolutionary graph networks; and the associated storage of decision instructions and execution feedback can be used for offline analysis of the matching degree between decision effectiveness and execution efficiency. Data storage is automatically rolled over and cleaned according to a preset strategy, while the system provides data backup and recovery mechanisms.

[0082] Example 2: See Figure 3The core function of the feature optimization and screening module is to process the set of time-series operational feature vectors of electrical equipment generated by the multi-dimensional risk feature extraction module. This set consists of feature vectors of multiple dimensions, each representing the multi-scale operational state of the electrical equipment within a specific time segment. The feature optimization and screening module first activates the risk feature energy distribution modeling unit. This unit constructs an energy density distribution model for each feature vector in the feature vector set. This construction process involves analyzing the numerical distribution pattern of the feature vector within a time window, identifying its frequency density in different amplitude intervals, and fitting a statistically significant energy density probability distribution curve. The energy density distribution models of all feature vectors in the set are arranged by index number, thus generating a risk feature energy distribution spectrum sequence, which constitutes the basic analysis object.

[0083] The risk feature energy distribution modeling unit comprises three sub-processing units. The feature synergistic correlation analysis sub-unit is invoked first, responsible for quantifying the interaction strength between feature vectors. This sub-unit's processing logic is as follows: select one feature vector from the set as the target feature vector (e.g., a feature vector representing a specific voltage harmonic component that appears frequently within the current time window), and iterate through all other feature vectors in the set (e.g., current waveform distortion feature vectors within the corresponding time window, temperature rise feature vectors of key components, etc.) as associated feature vectors. For each target-associated feature vector pair, the feature interaction response unit performs calculations. This calculation captures the sensitivity of one feature vector's change to the other by analyzing the synergy of numerical changes in the two feature vectors along the time axis. Specific operations include calculating the cross-correlation coefficient across time lags and the mutual information entropy values ​​at multiple time scales. This unit outputs a feature interaction response strength sequence for the current target feature vector. The correlation strength synthesis unit then works, receiving this response strength sequence and performing nonlinear aggregation processing. The aggregation method, based on a rule base set according to historical statistical patterns, integrates the interaction response strengths at different time scales and in different correlation directions into a comprehensive correlation strength value. The above process is repeated iteratively across all eigenvectors. The final output is an eigencovariant correlation strength matrix covering all vector pairs in the entire eigenvector set. Each element of this matrix records the combined covariant correlation strength value between the two eigenvectors at the corresponding row and column indices.

[0084] After the matrix is ​​generated, the feature energy weight allocation subunit is activated. This unit parses the aforementioned feature covariance correlation strength matrix and identifies the global correlation pattern of each feature vector in the entire set. The weight calculation is based on a preset logic: the average correlation strength of a feature vector with all other vectors reflects its influence in the entire system. Vectors with high correlation strength values ​​are considered to have higher dominance in energy density distribution. This unit traverses each feature vector index row in the matrix, calculates the weighted average of the correlation strength values ​​in that row, and finally generates a unique feature energy distribution weight value for each feature vector. This sequence of weight values ​​is the feature energy weight allocation sequence. The energy spectrum generation subunit then operates, taking the feature energy weight allocation sequence and the set of energy density distribution models of the original feature vectors as input. This unit uses a convolution fusion algorithm: it fuses the energy density distribution model of the corresponding feature vector with its assigned weight value through a convolution kernel operation to generate a new weighted energy distribution spectrum. The weighted energy distribution spectra of all feature vectors are arranged sequentially to form the final risk feature energy distribution spectrum sequence. This spectrum sequence has internalized the co-correlation information between features.

[0085] After modeling is completed, the feature energy cluster center identification unit is triggered. This unit receives the previously generated risk feature energy distribution spectrum sequence as input. The core task of this unit's computation is to find the overall central trend of the energy distribution of all feature vectors. Its internal mechanism is as follows: the risk feature energy distribution spectrum corresponding to each feature vector is treated as a distribution point in a multi-dimensional space, and a weighted centroid algorithm is used for calculation. The algorithm performs a spatial weighted average operation based on the coordinate position of each feature distribution point in a preset feature space (composed of coordinate components such as the main peak position and distribution width of the spectrum curve), combined with the weight coefficient of that distribution point (transferred from the risk feature energy distribution modeling unit). Finally, a multi-dimensional vector is output, which is marked as the risk feature energy cluster center benchmark. This benchmark represents the global core centroid position of the energy distribution in the current feature vector set.

[0086] The feature energy discretization unit is responsible for measuring the degree of difference between each feature vector distribution and the aforementioned benchmark. The unit operation includes four key steps. The spectral difference calculation subunit first processes the difference relationship between each risk feature energy distribution spectrum and the risk feature energy clustering center benchmark. For a single distribution spectrum, this subunit extracts its key parameter set (e.g., distribution mean, main lobe width, peak shift, etc.) and calculates the vector difference with the corresponding parameters of the benchmark one by one. These differences constitute a multidimensional spectral difference vector. The covariance matrix generation subunit performs statistical analysis on the entire risk feature energy distribution spectrum sequence. Through the statistical fluctuation relationship of all key parameters in the sequence, the covariance characteristics of the parameters are calculated, and a covariance matrix is ​​constructed. This unit then obtains the inverse of this covariance matrix to characterize the overall discrete characteristics of the distribution space. The discreteness kernel function subunit receives the spectral difference vector and the covariance inverse matrix and performs a quadratic operation. This operation takes the difference vector as input and the covariance inverse matrix as the transformation kernel, generating an intermediate scalar value, called the intermediate discreteness parameter, through matrix multiplication. The magnitude of this parameter reflects the relative distance between the current feature vector distribution and the global center in the statistical distribution space. The discrete coefficient transformation subunit performs numerical processing on the intermediate discrete parameter. After scalar operations (including but not limited to nonlinear transformations such as square root extraction), a normalizable and comparable value is output, which is the risk feature energy discrete coefficient of a single feature vector. This quantization process traverses all feature vectors, ultimately generating a set of risk feature energy discrete coefficients covering the entire set.

[0087] The feature selection execution unit ultimately performs the operation, taking as input the previously calculated set of risk feature energy discrete coefficients and a preset discrete threshold parameter. The unit logic is clear: the discrete coefficients quantitatively characterize the degree of deviation of the energy distribution of the feature vector relative to the global energy center. Vectors with large deviations are considered more sensitive to abnormal operating states or changes in risk factors. The unit scans the discrete coefficient values ​​one by one and compares them numerically with the preset discrete threshold. Feature vectors exceeding the discrete threshold are marked (e.g., by setting a high-sensitivity flag). The set of all marked feature vectors is extracted and combined to form an optimized risk feature subset. This subset contains the feature elements that, under the current equipment operating state, have the highest risk representation sensitivity and whose statistical distribution shows significant anomalies relative to the energy center. This optimized risk feature subset, as the output of the feature optimization and screening module, is passed to the downstream risk feature migration calculation module for subsequent cross-cycle risk dynamic migration calculations. The entire process achieves the goal of automatically selecting key risk feature dimensions from the original high-dimensional operating features.

[0088] Example 3: See Figure 4The feature energy discretization unit receives the risk feature energy cluster center benchmark from the feature energy cluster center identification unit and the risk feature energy distribution spectrum sequence generated by the risk feature energy distribution modeling unit. The unit contains four interconnected processing steps, forming a complete discretization calculation chain. The spectral difference calculation subunit starts first, processing a single distribution spectrum in the risk feature energy distribution spectrum sequence. This subunit extracts a set of key feature parameters from the currently processed target distribution spectrum, including the main peak position coordinates μ. p , Distribution half-width ω h Spectral peak shift δ s These parameters together constitute the feature vector V that describes the morphological characteristics of this distribution spectrum. f =(μ p ,ω h ,δ s Simultaneously, the corresponding benchmark parameter vector V is read from the benchmark of the risk characteristic energy concentration center. b The sub-unit calculates the difference vector ΔV = V between these two vectors. f -V b Each component represents the absolute deviation of the corresponding feature parameter. The difference vector ΔV, as a feature description of the deviation of the current distribution spectrum relative to the global energy center, is passed to subsequent processing stages.

[0089] The covariance matrix generating sub-unit is then activated, its input being the set of all feature parameters of the entire risk feature energy distribution spectrum sequence. This unit first constructs a feature parameter matrix X, where each row of the matrix corresponds to a feature parameter vector V of a distribution spectrum. f Based on this matrix, the covariance relationship between the characteristic parameters is calculated, generating the covariance matrix Σ. Matrix elements Σ ij Let represent the covariance intensity between the i-th and j-th characteristic parameters throughout the entire distribution spectrum sequence. Inverting this matrix yields the inverse covariance matrix Σ. -1 The inverse matrix is ​​used to characterize the statistical distribution properties of the feature parameter space. The calculation of the inverse matrix employs a numerically stable decomposition algorithm, ensuring computational accuracy even when there is a high correlation between parameters.

[0090] Discreteness kernel function sub-unit receive spectrum difference vector ΔV and inverse covariance matrix Σ -1 This involves performing the core discrete quantization calculation. The calculation process can be represented as follows:

[0091] D m =ΔV T ·Σ -1 ·ΔV

[0092] Where D m ΔV is the intermediate dispersion parameter. TThis represents the transpose of the difference vector. This quadratic form projectes the multidimensional difference vector onto a statistical space defined by the inverse covariance matrix, yielding a scalarized distance metric. The operation is implemented using optimized matrix multiplication, ensuring computational efficiency even with high feature parameter dimensions. D m The larger the value, the farther the current distribution spectrum is from the energy center in the statistical space, that is, the higher the degree of dispersion.

[0093] Discrete coefficient transformation subunit for intermediate discreteness parameter D m The final transformation process is then performed. This process includes two stages: nonlinear transformation and normalization adjustment. First, D... m The basic dispersion value is obtained by performing the square root operation. Eliminate the dimensional effects of quadratic form calculations. Then, normalize the result based on historical statistical data, and adjust D... b Mapped to the standard range of discrete coefficients. The normalization process considers the overall distribution characteristics of the characteristic parameters of the current operating stage of the electrical equipment, and dynamically adjusts the conversion coefficients using a sliding window statistical method. The final output is the risk characteristic energy discrete coefficient D. c It is a dimensionless scalar value that can be directly used to compare the degree of dispersion across eigenvectors.

[0094] In this embodiment, the feature-co-correlation analysis subunit performs the calculation of the interaction response intensity between feature vectors. This subunit selects the target feature vector V from the set of electrical equipment timing operation feature vectors. t And match all other associated feature vectors V a1 V a2 ,...,V an For each target-association vector pair (V) t V ai The feature interaction response unit calculates the dynamic response relationship between the two vectors across multiple time scales. The calculation process includes two parts: time-domain correlation analysis and frequency-domain coupling analysis. The time-domain analysis calculates the dynamic time-warped distance between the two vectors using a sliding window, capturing the degree of synchronicity in their numerical changes. The frequency-domain analysis extracts the main frequency components using short-time Fourier transform and calculates the phase coupling strength between the components. The results of both analyses are integrated into a single comprehensive interaction response strength value R. i Reflecting V ai For V t Sensitivity to change.

[0095] The association strength synthesis unit receives the sequence of interaction response strength values ​​R1, R2, ..., R of all target-association pairs. nThe process performs nonlinear aggregation operations. The aggregation process considers the differences in physical meaning among different associated feature vectors and employs a weighted fusion strategy. Weight allocation is based on the known associations of feature vectors in the electrical equipment operation model, prioritizing feature pairs with clear physical coupling relationships. The aggregation result generates the target feature vector V corresponding to the feature covariant association strength matrix. t The correlation strength vector A t This vector fully records V t Co-association patterns with all other feature vectors in the set.

[0096] The entire feature energy discretization process forms a closed-loop computational flow within the electrical equipment operation risk perception and identification system. After complete discretization, each feature vector's D... c The values ​​are recorded in a set of discrete coefficients for the risk feature energy. This set serves as input to the feature selection execution unit, ultimately determining which feature vectors should be included in the optimized risk feature subset. The accuracy of the discrete coefficient calculation directly affects the system's ability to identify risk-sensitive features; therefore, the calculation process of each sub-unit employs numerically stable algorithms and is equipped with outlier detection and handling mechanisms. The introduction of feature co-correlation information allows the discrete quantification to consider not only the statistical properties of individual feature vectors but also the interaction network between features, thus providing a more comprehensive assessment of the importance of each feature in risk characterization. During system operation, the calculation parameters of this unit are adaptively adjusted based on the actual operating data of the electrical equipment to maintain the consistency between the discrete quantification results and the actual risk state.

[0097] Example 4: See Figure 5 This paper details the implementation process of the risk decision generation module using a substation main transformer risk control scenario. The monitoring system of a 110kV substation detected the evolution rate parameters of the operational risk characteristics of the #1 main transformer, including an overheating risk evolution rate of 0.15 (temperature rise rate per unit time), an insulation degradation rate of 0.08, a partial discharge growth rate of 0.12, and an abnormal accumulation rate of dissolved gases in the oil of 0.09. The risk decision space loading unit loads a predefined multi-dimensional risk decision strategy space from the knowledge base. This space contains 57 strategy plans, constructed with a three-dimensional index based on risk type (overload, overheating, insulation degradation, discharge fault) and execution level (alarm, power regulation, switching to standby, emergency tripping).

[0098] The decision correlation calibration unit initiates the strategy record traversal unit, scanning 253 historical decision records in the strategy space. The risk state similarity analysis unit compares the current risk characteristic evolution rate with historical records one by one. For example, it finds that the historical event similarity parameters for record ID: CT2023-078 are: overheating risk evolution rate similarity 92%, insulation degradation rate similarity 86%, and overall similarity 89.2%. The strategy correlation synthesis unit calculates the strategy correlation coefficient of this historical record as 0.873 based on the weighted coefficients (overheating risk weight 0.4, insulation degradation weight 0.3, and other parameters weight 0.3), exceeding the preset correlation threshold of 0.85. The strategy screening unit adds 19 highly correlated historical decisions, including this record, to the calibrated risk decision strategy subspace.

[0099] The table below shows the matching results of some key historical records during the decision-related calibration process:

[0100]

[0101] The decision feature activation unit performs region localization within the generated calibrated strategy subspace. The unit detects that the overheating risk evolution rate parameter exceeds the secondary warning line (>0.12) and the insulation degradation rate is within the primary warning zone (0.05-0.1) in the current risk state, automatically locking onto the "overheating + insulation degradation" dual-risk coupled decision region in the corresponding risk dimension of the strategy space. This region contains eight candidate strategies, involving three control types: cooling system enhancement, dynamic load adjustment, and online oil chromatography tracking.

[0102] The decision fitness assessment unit performs matching quantification on eight strategies. For strategy S07 ("Start standby cooling fan + reduce load by 10%), the unit detects real-time parameters such as current ambient temperature of 32℃ (exceeding summer standard temperature), load rate of 82%, and normal cooling system operation. The assessment engine calculates the matching fitness parameters for this strategy: fit to temperature control target 0.92, load adjustment feasibility 0.87, execution resource availability 0.95, and overall fitness 0.913. Strategy S12 ("Switch to standby transformer") fails because the standby transformer is currently under maintenance, resulting in execution resource availability dropping to 0.22 and an overall fitness of 0.496, which does not meet the requirements.

[0103] The decision-making and execution unit selects three compliance strategies based on preset constraints (overall fitness ≥ 0.85 and execution delay < 3 seconds): S07: Start the backup cooling fan + reduce the load by 10% (fitness 0.913), S08: Enhance oil circulation + limit the load by 8% (fitness 0.892), S15: Upgrade online insulation monitoring + reduce the temperature by 5% (fitness 0.861).

[0104] The units are sorted in descending order of fitness and the highest-scoring strategy S07 is selected as the core decision. At the same time, S08 is integrated and output as an alternative plan, and finally the electrical equipment risk control decision instruction is formed: "Execution code CT-TS07: 1. Activate the standby cooling fan units (numbered FAN2-FAN4); 2. Reduce the load by 10% through the AGC system (target value 73.8MW); 3. Upload oil chromatogram data every 15 minutes; Alternative plan CT-B08: If the hot spot temperature does not drop below 85℃ within 30 minutes, execute.

[0105] The entire decision generation process is completed within 1.8 seconds. The system automatically records the decision path: the associated historical records CT2023-078 and CT2024-021 are marked as the decision basis source; the excluded strategy S12 is recorded as the exclusion reason code ER-17 due to resource unavailability. The calibrated strategy subspace data is automatically cached to the edge computing node after the decision is completed, for quick access in subsequent similar risk scenarios. The generation of risk control decision instructions simultaneously triggers the control instruction compilation module, which converts the decision parameters into specific executable equipment control code, and sends it to the station control layer execution terminal after security verification.

[0106] Example 5: Collaborative Implementation Process of Risk Decision Verification Module and Dynamic Decision Optimization Module in Electrical Equipment Operation Risk Perception and Identification System. After the risk decision verification module is activated, the virtual operation environment construction unit performs 3D scene reconstruction based on the physical topology of the target electrical equipment. This unit reads the CAD model database of the equipment body and related components, and analyzes key information such as the spatial distribution of transformer windings, the geometry of the circuit breaker arc-extinguishing chamber, and the busbar connection topology. The physical characteristic configuration engine maps material property parameters to the 3D model, including the permeability curve of silicon steel sheets, the specific heat capacity parameter of insulating oil, and the conductivity temperature coefficient of the conductive busbar. The electromagnetic-thermal coupling field solver sets boundary conditions based on the equipment's rated parameters to generate a basic virtual operation scene. The scene incorporates a monitoring probe matrix, deploying virtual sensor nodes at key locations such as winding hotspot areas, bushing connection points, and radiator flow channels.

[0107] The decision-making and simulation execution unit receives electrical equipment risk control decision instructions to be verified, and performs instruction semantic parsing and virtual signal conversion. For compound instructions such as "reduce load by 15% + start auxiliary cooling," the unit decomposes them into load adjustment instruction codes (target value, slope parameters), cooling control signals (fan number, speed setting), and monitoring instruction sets (sampling period, reporting threshold). The instruction parameters are injected into the virtual operating scenario after being timestamped. The scenario timeline engine synchronously advances the virtual time stream with microsecond-level precision, the load adjustment module dynamically modifies the virtual power grid equivalent impedance parameters based on the instruction curve, and the cooling control drive system adjusts the virtual fan operating status parameters.

[0108] After the risk evolution monitoring unit is activated, a virtual sensor network in the 3D scene is initiated. The unit is configured with a multi-parameter parallel acquisition strategy: capturing the temperature field distribution matrix at key locations every millisecond, sampling instantaneous voltage and current waveform values ​​every cycle, and recording the trajectory of the insulation dielectric loss tangent every five minutes. The partial discharge monitoring channel is configured with a high-speed data buffer to capture the pulse amplitude and phase distribution spectrum. The unit establishes a spatiotemporal correlation database, marking the spatial coordinate attributes and timestamp index of each data item. The data preprocessing workflow automatically filters out system computational noise and retains effective feature change trajectories.

[0109] The confidence assessment unit performs deviation analysis on the collected dynamic dataset of risk characteristics. The unit incorporates a risk control target description framework, such as quantitative constraints like "winding hotspot temperature rise not exceeding 65K" and "total harmonic distortion rate of current controlled within 5%". The analysis engine divides the data into time slices, calculating the absolute deviation and duration percentage between the actual simulated value and the target threshold within each slice. A regional weighted integral algorithm is used for the temperature field data to generate a spatial comprehensive temperature rise deviation index. Waveform distortion data is extracted using fundamental wave separation technology to remove the deviation of each harmonic component. All deviation indices are fused according to predefined risk dimension weighting coefficients, outputting a multidimensional confidence matrix. The vertical axis of this matrix distinguishes categories such as over-temperature risk confidence and insulation degradation confidence, while the horizontal axis records the confidence evolution curves by time panes.

[0110] The dynamic decision optimization module initiates a parameter compensation mechanism based on the confidence matrix. The compensation decision space generation unit calls upon the risk evolution prediction data stream, which includes information such as the predicted temperature gradient map, load fluctuation trend, and harmonic spectrum evolution for the next 30 minutes. The unit generates a two-layer decision space: the basic layer loads 420 records from a pre-set compensation strategy library, covering types such as voltage adjustment, power distribution, and heat dissipation enhancement; the enhancement layer connects to an online strategy generation engine, dynamically generating twelve scenario-adaptive compensation strategies based on real-time virtual simulation data (such as abnormal flow in specific heat dissipation channels).

[0111] The compensation strategy activation unit analyzes the confidence matrix to identify low-confidence risk dimensions (e.g., the over-temperature risk confidence drops to 0.72 within a certain period). This unit is associated with a snapshot of the current equipment operating status, including parameters such as actual load rate, ambient temperature and humidity, and cooling medium flow rate. A three-dimensional index matching is performed within the compensation decision space: the first dimension selects the risk type as temperature control strategies; the second dimension limits the execution resource to available on-site equipment (e.g., the online status of backup fans); and the third dimension matches the control strength with the magnitude of the risk deviation. Finally, nine effective strategies suitable for the current scenario are selected to constitute the compensation decision trigger area.

[0112] The iterative decision generation unit performs multiple rounds of selection operations within this region. Preset fitness criteria include constraints such as execution delay limits, resource cost coefficients, and control effect weights. The first round of screening calculates the coordinates of each strategy in the two-dimensional space of control effect and resource consumption. The second round introduces a time-series constraint factor to eliminate strategies with excessive preparation time. The third round integrates the prediction results of the risk propagation model to evaluate the impact factors on the associated system after strategy execution. After three rounds of iteration, four optimal strategies are generated to form a set of compensation decision strategies, each with a comprehensive fitness score (range 0.82 to 0.91).

[0113] The decision fusion unit integrates the execution features of the original risk control decision instructions and compensation strategy set. The instruction decoding module extracts the control feature vector of the original instructions, including the target setpoint (e.g., a 15% load reduction target), the execution entity identifier (e.g., cooling fan unit No. 3), and control time-domain parameters (e.g., five-minute ramp adjustment). The compensation strategy parsing module generates a compensation feature matrix, including power fine-tuning increments (+2.3% load margin), auxiliary equipment identifiers (addition of temporary air-cooling unit), and monitoring enhancement instructions (increasing the key point temperature sampling frequency to twice per second). The feature alignment engine establishes a mapping relationship between the original control parameters and compensation parameters, and generates the fused control target value through a weighted superposition algorithm. The execution instruction reorganization module reconstructs the control logic sequence to ensure seamless connection between the compensation operation and the original instructions on the timeline, ultimately outputting an optimized risk control decision instruction containing twenty-three operation sub-instructions. This instruction retains the core control intent of the original decision, supplements and strengthens measures in key risk dimensions, and corrects the specific execution value range of some control parameters. The system synchronously records all decision node parameters and selection paths during the compensation strategy fusion process for storage and analysis by the closed-loop management module.

[0114] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0115] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A system for sensing and identifying operational risks of electrical equipment, characterized in that, include: The real-time operating status monitoring module for electrical equipment is used to continuously collect time-series data sets of voltage fluctuations, current waveforms, and temperature distribution of electrical equipment. The multi-dimensional risk feature extraction module is used to receive the voltage fluctuation time series data set, current waveform time series data set, and temperature distribution time series data set output by the real-time operation status monitoring module of the electrical equipment, and perform spatiotemporal feature fusion processing through multi-scale convolution kernel group to generate a set of time series operation feature vectors of electrical equipment. The feature optimization and screening module is used to perform a risk-sensitivity-based feature decoupling operation on the set of time-series operation feature vectors of the electrical equipment, and extract an optimized risk feature subset with significant risk characterization capabilities; The risk feature migration calculation module is used to input the optimized risk feature subset into the risk evolution graph network and calculate the risk feature migration index set of adjacent monitoring periods. The risk evolution rate quantification module is used to perform time differentiation operation on the risk feature migration index set according to a preset monitoring period time interval to generate risk feature evolution rate parameters. The risk decision generation module is used to perform multi-level matching analysis between the risk feature evolution rate parameter and the predefined risk threshold spectrum, and output electrical equipment risk control decision instructions.

2. The electrical equipment operation risk perception and identification system according to claim 1, characterized in that, Also includes: The risk decision verification module is used to perform reverse simulation verification of the risk control decision instructions of the electrical equipment through a virtual operating environment, and generate a risk decision confidence matrix. The dynamic decision optimization module is used to perform parameter compensation operations on the electrical equipment risk control decision instructions in conjunction with the risk decision confidence matrix, and generate optimized risk control decision instructions. The risk response execution module is used to parse the control logic in the optimized risk control decision instruction and drive the electrical equipment protection device to perform corresponding actions. The risk data closed-loop management module is used to store the raw data collected by the real-time operation status monitoring module of the electrical equipment, the risk feature migration index set, and the execution feedback results of the optimized risk control decision instructions.

3. The electrical equipment operation risk perception and identification system according to claim 1, characterized in that, The feature optimization and filtering module includes: The risk feature energy distribution modeling unit is used to construct the energy density distribution model of each feature vector in the set of time-series operation feature vectors of the electrical equipment, and generate a risk feature energy distribution spectrum sequence. The feature energy clustering center identification unit is used to calculate the weighted centroid vector of the risk feature energy distribution spectrum sequence and mark it as the risk feature energy clustering center reference. The feature energy discretization unit is used to measure the spectral discrete distance between each of the risk feature energy distribution spectra and the risk feature energy clustering center reference, and to generate a set of risk feature energy discretization coefficients. The feature selection execution unit is used to mark feature vectors in the set of energy discrete coefficients of the risk features that exceed a preset discrete threshold as high-sensitivity risk features, and form the optimized risk feature subset.

4. The electrical equipment operation risk perception and identification system according to claim 3, characterized in that, The risk characteristic energy distribution modeling unit includes: The feature covariance analysis subunit is used to calculate the covariance strength between any two feature vectors in the set of time-series operation feature vectors of the electrical equipment, and generate a feature covariance strength matrix. The feature energy weight allocation subunit is used to calculate the energy distribution weight value of each feature vector according to the feature covariance correlation strength matrix, and generate the feature energy weight allocation sequence. The energy spectrum generation subunit is used to convolve and fuse the feature energy weight allocation sequence with the energy density distribution model of the corresponding feature vector to generate the risk feature energy distribution spectrum sequence.

5. The electrical equipment operation risk perception and identification system according to claim 4, characterized in that, The feature-based collaborative correlation analysis subunit includes: The feature vector matching unit selects a target feature vector from the set of timing operation feature vectors of the electrical equipment and matches it with all other associated feature vectors. A feature interaction response unit is used to calculate the interaction response intensity value between the target feature vector and each associated feature vector, and generate a feature interaction response intensity sequence; The association strength synthesis unit is used to perform nonlinear aggregation operations on the feature interaction response strength sequence to generate the association strength vector corresponding to the target feature vector in the feature covariant association strength matrix.

6. The electrical equipment operation risk perception and identification system according to claim 3, characterized in that, The feature energy discretization unit includes: The spectral difference calculation subunit is used to calculate the spectral difference vector between the risk feature energy clustering center benchmark and a single risk feature energy distribution spectrum; The covariance matrix generation sub-unit is used to construct the covariance matrix of the risk feature energy distribution spectrum sequence and obtain its inverse matrix; The discrete kernel function subunit is used to perform a quadratic operation on the spectrum difference vector and the inverse of the covariance matrix to generate intermediate discrete parameters. The discrete coefficient conversion subunit is used to perform a square root operation on the intermediate discreteness parameter and output the corresponding risk characteristic energy discrete coefficient.

7. The electrical equipment operation risk perception and identification system according to claim 1, characterized in that, The risk decision generation module includes: The risk decision space loading unit is used to load a pre-configured multi-dimensional risk decision strategy space; The decision association calibration unit is used to perform dynamic association mapping on the multi-dimensional risk decision strategy space according to the risk feature evolution rate parameter, and generate a calibrated risk decision strategy subspace. A decision feature activation unit is used to identify decision trigger regions that satisfy the current risk state in the calibrated risk decision strategy subspace. The decision fitness evaluation unit is used to measure the matching fitness parameters between each decision strategy and the current operating state of the electrical equipment within the decision triggering area; The decision selection execution unit is used to select a decision strategy that satisfies the preset fitness constraints on the matching fitness parameters, and to form the electrical equipment risk control decision instruction.

8. The electrical equipment operation risk perception and identification system according to claim 7, characterized in that, The decision-related calibration unit includes: The strategy record traversal unit is used to scan the historical risk decision records stored in the multi-dimensional risk decision strategy space; The risk status similarity analysis unit is used to calculate the similarity parameter between the current risk feature evolution rate parameter and the risk feature evolution rate in historical risk decision records; The strategy correlation synthesis unit is used to perform weighted integration processing on the similarity parameters to generate the strategy correlation coefficient; The strategy filtering unit is used to add the corresponding historical risk decision records to the calibrated risk decision strategy subspace when the strategy correlation coefficient exceeds a preset correlation threshold.

9. The electrical equipment operation risk perception and identification system according to claim 2, characterized in that, The risk decision verification module includes: The virtual operating environment construction unit is used to create a three-dimensional virtual operating scene based on the topology of electrical equipment. The decision-making and simulation execution unit is used to inject the electrical equipment risk control decision instructions into the three-dimensional virtual operation scenario for dynamic simulation. The risk evolution monitoring unit is used to collect the dynamic change trajectory of risk characteristic parameters during the virtual simulation process; The confidence assessment unit is used to analyze the degree of deviation between the dynamic change trajectory of the risk characteristic parameters and the expected risk control target, and to generate the risk decision confidence matrix.

10. The electrical equipment operation risk perception and identification system according to claim 2, characterized in that, The dynamic decision optimization module includes: The compensation decision space generation unit is used to create a compensation decision strategy space based on risk evolution prediction data. The compensation strategy activation unit is used to identify compensation decision triggering regions that satisfy the current risk state in the compensation decision strategy space. The iterative decision generation unit is used to perform multiple rounds of decision selection operations within the compensation decision triggering area based on a preset fitness criterion, and generate a set of compensation decision strategies. The decision fusion unit is used to perform feature-level fusion of the compensation decision strategy set and the original electrical equipment risk control decision instructions, and output the optimized risk control decision instructions.

Images