AI fusion processing method, system and equipment for state monitoring data of power transmission and transformation equipment and storage medium

Abstract

The invention relates to the technical field of power equipment state monitoring, in particular to an AI fusion processing method, system and equipment for power transmission and transformation equipment state monitoring data and a storage medium. Collecting multi-source monitoring data of the power transmission and transformation equipment, constructing an equipment health portrait, and identifying key feature parameters and a coupling relationship among the parameters through association mining; performing multi-dimensional consistency check on the key feature parameters through interactive verification to obtain an initial health degree evaluation value, extracting a historical feature reference and performing deviation degree calculation, and combining the initial health degree evaluation value to obtain a comprehensive health degree and a corresponding threshold value; and based on the time sequence change of the comprehensive health degree, identifying a degradation evolution trajectory, determining a degradation type and a degradation rate, predicting the future state evolution trend of the equipment, and generating graded early warning information. Fusion and multi-dimensional accurate evaluation of multi-source heterogeneous data are obtained, and static evaluation is converted into dynamic prediction.

Description

Technical Field

[0001] This invention relates to the field of power equipment condition monitoring technology, and in particular to an AI fusion processing method, system, device and storage medium for power transmission and transformation equipment condition monitoring data. Background Technology

[0002] The operational status of power transmission and transformation equipment has always been one of the safety goals for power grid stability. With the continuous growth of electricity demand and the extension of equipment service life, the condition monitoring and fault early warning of power transmission and transformation equipment have also become important issues in power system operation and maintenance. In recent years, the monitoring data of power transmission and transformation equipment has shown the characteristics of multi-source heterogeneity. For example, on the one hand, electrical quantity monitoring data such as voltage, current, and power factor reflect the electrical performance of the equipment; on the other hand, non-electrical quantity monitoring data such as temperature, humidity, vibration, and sound reflect the direct status of the equipment. However, monitoring data from different sensors and different detection conditions are often independent of each other and lack fusion analysis; or they are simply displayed side by side or simply weighted averaged, which cannot show the inherent correlation and coupling relationship between different parameters.

[0003] Existing equipment condition assessment methods have several limitations: first, they rely on a single assessment dimension, making it difficult to cross-verify equipment condition from multiple perspectives; second, they lack deep learning of historical normal operating patterns, resulting in an inability to accurately determine the degree of deviation from the current condition; and third, they lack sufficient understanding of the evolutionary patterns of equipment degradation, making it difficult to identify degradation types and rates. Current monitoring systems can only provide simple threshold alarms and cannot offer accurate condition predictions or tiered early warnings.

[0004] In addition, existing equipment also has shortcomings in data processing. How to achieve the fusion of heterogeneous data from multiple sources over time is a technical problem. The evolution of equipment status is a gradual process, which requires time series analysis to capture the deterioration trend. However, existing methods focus more on instantaneous status and are not good at tracking the status evolution trajectory. Summary of the Invention

[0005] In view of the problems existing in the prior art, the present invention is proposed.

[0006] Therefore, the problem to be solved by this invention is how to address the issues of insufficient multi-source data fusion, single dimension of state assessment, and insufficient ability to identify degradation trends in the prior art.

[0007] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, embodiments of the present invention provide an AI fusion processing method for power transmission and transformation equipment status monitoring data, which includes collecting multi-source monitoring data of power transmission and transformation equipment and constructing a health profile of the equipment, performing correlation mining based on the health profile of the equipment, and identifying key characteristic parameters of equipment operation and coupling relationships between parameters. By conducting multi-dimensional consistency checks on key feature parameters through interactive verification, the initial health assessment value of the equipment is obtained. Feature benchmarks of the equipment's historical normal operation are extracted, and the deviation between the current operating status and the feature benchmarks is calculated. Combined with the initial health assessment value, the overall health of the equipment and its corresponding threshold are obtained. Based on the temporal changes in comprehensive health status, the degradation evolution trajectory is identified, the degradation type and degradation rate are determined, and the degradation evolution trajectory and comprehensive health status threshold are combined to predict the future state evolution trend of the equipment and generate graded early warning information.

[0008] As a preferred embodiment of the AI ​​fusion processing method for power transmission and transformation equipment status monitoring data described in this invention, the step of collecting multi-source monitoring data of power transmission and transformation equipment and constructing equipment health profiles includes collecting online monitoring data and offline detection data of the equipment to form a multi-dimensional monitoring dataset. By integrating monitoring information from different sources through the mapping relationship of multi-source data, an initial health profile of the equipment can be established.

[0009] As a preferred embodiment of the AI ​​fusion processing method for power transmission and transformation equipment status monitoring data according to the present invention, the identification of key characteristic parameters of equipment operation and the coupling relationship between parameters includes extracting characteristic parameters reflecting the equipment status from the equipment health profile. Analyze the temporal correlation between characteristic parameters to determine the leader and lag relationships between parameters; The coupling strength between parameters is quantified, and a coupling relationship matrix is ​​established.

[0010] As a preferred embodiment of the AI ​​fusion processing method for power transmission and transformation equipment status monitoring data according to the present invention, the step of performing multi-dimensional consistency verification through interactive verification includes setting electrical performance dimension, thermal characteristic dimension and mechanical characteristic dimension as evaluation dimensions. Calculate the status indicators for each evaluation dimension; cross-compare the status indicators for each dimension, adjust the weights of each dimension, and obtain the initial health assessment value.

[0011] The beneficial effects of this preferred technical solution are as follows: by setting electrical performance, thermal characteristics, and mechanical characteristics as evaluation dimensions, the equipment status can be comprehensively evaluated from multiple physical domains; the status indicators under each evaluation dimension are calculated separately, realizing the quantitative characterization of different physical characteristics of the equipment; by cross-comparing the status indicators of each dimension and adjusting the weight of each dimension, abnormal data can be identified through mutual verification between different dimensions, improving the reliability of the evaluation results; at the same time, the importance of each dimension is dynamically adjusted according to the actual operating characteristics of the equipment, resulting in a more accurate initial health assessment value.

[0012] As a preferred embodiment of the AI ​​fusion processing method for power transmission and transformation equipment status monitoring data according to the present invention, the step of calculating the deviation between the current operating status and the feature benchmark includes statistically analyzing the distribution range of feature parameters during the historical normal operation of the equipment. Calculate the distance metric between the current feature parameters and the historical distribution center; based on the distance metric and the initial health assessment value, determine the comprehensive health level and the classification threshold.

[0013] As a preferred embodiment of the AI ​​fusion processing method for power transmission and transformation equipment status monitoring data according to the present invention, the step of identifying the deterioration evolution trajectory based on the time-series changes of comprehensive health includes recording the change values ​​of comprehensive health over a continuous time period. Fit a time-series curve of overall health status to identify the trend of the curve; determine the type of degradation based on the slope and curvature of the curve, and determine the degradation rate based on the rate of change.

[0014] The beneficial effects of this preferred technical solution are as follows: by recording the changes in the overall health status over a continuous period of time, the dynamic evolution of the equipment status over time can be captured; by fitting the time-series curve of the overall health status and identifying the trend of the curve, discrete monitoring data can be transformed into a continuous evolution trajectory, making it easier to intuitively judge the development direction of equipment degradation; the degradation type can be judged based on the slope and curvature of the curve, and the degradation rate can be determined based on the rate of change, which can distinguish whether the equipment is in a slow degradation, accelerated degradation or sudden abnormality, and quantify the degree of degradation, providing a basis for early warning decision-making.

[0015] As a preferred embodiment of the AI ​​fusion processing method for power transmission and transformation equipment condition monitoring data according to the present invention, the prediction of the equipment's condition evolution trend over a future time period includes extrapolating the future changes in the equipment's overall health based on the deterioration evolution trajectory. The system compares the predicted overall health level with the threshold; when the predicted overall health level approaches or exceeds the threshold, it generates a warning message of the corresponding level.

[0016] Secondly, embodiments of the present invention provide an AI fusion processing system for power transmission and transformation equipment status monitoring data, which includes a data association analysis module, used to collect multi-source monitoring data of power transmission and transformation equipment and construct a health profile of the equipment, perform association mining based on the health profile of the equipment, and identify key characteristic parameters of equipment operation and coupling relationships between parameters. The evaluation module is used to perform multi-dimensional consistency checks on key feature parameters through interactive verification, obtain the initial health assessment value of the equipment, extract the feature benchmark of the equipment's historical normal operation, calculate the deviation between the current operating status and the feature benchmark, and combine the initial health assessment value to obtain the comprehensive health of the equipment and the corresponding threshold. The early warning module is used to identify the degradation evolution trajectory based on the time-series changes in comprehensive health, determine the degradation type and degradation rate, combine the degradation evolution trajectory with the comprehensive health threshold, predict the status evolution trend of the equipment in the future time period, and generate graded early warning information.

[0017] Thirdly, embodiments of the present invention provide a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program instructions are executed by the processor, the steps of the AI ​​fusion processing method for power transmission and transformation equipment status monitoring data as described in the first aspect of the present invention are implemented.

[0018] Fourthly, embodiments of the present invention provide a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program instructions are executed by a processor, they implement the steps of the AI ​​fusion processing method for power transmission and transformation equipment status monitoring data as described in the first aspect of the present invention.

[0019] The beneficial effects of this invention are as follows: By constructing a health profile of equipment and performing correlation mining, this invention unifies multi-source heterogeneous monitoring data from different sensors and physical domains into a fused feature map, establishing a mapping relationship and coupling relationship matrix between monitoring parameters, thus solving the problem of independent data sources and lack of correlation analysis in traditional monitoring methods; by establishing interactive verification of three evaluation dimensions—electrical performance, thermal characteristics, and mechanical characteristics—the invention cross-compares the status indicators of each dimension and dynamically adjusts the weights, avoiding the one-sidedness and misjudgment risk of single-dimensional evaluation, and calculating the deviation degree by combining the characteristic benchmarks of the equipment's historical normal operation, thus evaluating the current state in the context of the equipment's entire life cycle; by recording the temporal changes of comprehensive health and fitting evolution curves, the invention identifies the type and rate of degradation, transforming static state evaluation into dynamic trend prediction, extrapolating future changes in comprehensive health based on the degradation evolution trajectory, and generating early warning information by comparing with the graded threshold; thus realizing the transformation from post-maintenance to preventive maintenance, improving the operational reliability of power transmission and transformation equipment, and reducing maintenance costs. Attached Figure Description

[0020] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 A flowchart of an AI fusion processing method for condition monitoring data of power transmission and transformation equipment; Figure 2 A computer device diagram for an AI fusion processing method for condition monitoring data of power transmission and transformation equipment. Detailed Implementation

[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0023] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0024] Secondly, the term "an embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places throughout this specification does not necessarily refer to the same embodiment, nor is it a single embodiment or an embodiment selectively excluded from other embodiments.

[0025] Example 1 Reference Figure 1 - Figure 2 As a first embodiment of the present invention, this embodiment provides an AI fusion processing method for power transmission and transformation equipment condition monitoring data, including, S100: Collects multi-source monitoring data of power transmission and transformation equipment and constructs equipment health profiles. Based on the equipment health profiles, it performs correlation mining to identify key characteristic parameters of equipment operation and the coupling relationships between parameters.

[0026] S200: Through interactive verification, the key feature parameters are checked for consistency in multiple dimensions to obtain the initial health assessment value of the equipment. The feature benchmark of the equipment's historical normal operation is extracted, the deviation between the current operating status and the feature benchmark is calculated, and the overall health of the equipment and the corresponding threshold are obtained by combining the initial health assessment value.

[0027] S300: Based on the time-series changes in comprehensive health status, identify the degradation evolution trajectory, determine the degradation type and degradation rate, combine the degradation evolution trajectory with the comprehensive health status threshold, predict the status evolution trend of the equipment in the future time period, and generate graded early warning information.

[0028] It should be noted that the fusion of feature maps is not a simple data splicing, but rather the establishment of mapping relationships between different monitoring parameters, enabling data from different sensors and different physical domains to be correlated and analyzed within a unified framework; by identifying the coupling relationships between parameters; in the process of multi-dimensional consistency verification, interactive verification avoids misjudgment of a single dimension through cross-comparison; and by calculating the deviation by comparing with historical feature benchmarks, the current state is evaluated in the context of the entire life cycle of the equipment.

[0029] This invention addresses the challenge of unified analysis of multi-source heterogeneous data by constructing a fused feature map and identifying parameter coupling relationships in step S100; it enables multi-dimensional assessment of equipment status by establishing interactive verification and deviation calculation in step S200; and it transforms static assessment into dynamic prediction by identifying degradation evolution trajectories and predicting status evolution trends in step S300. This achieves comprehensive monitoring and accurate early warning of the status of power transmission and transformation equipment, solving the technical problems of insufficient fusion of multi-source data, single status assessment dimensions, and insufficient ability to identify degradation trends in existing technologies.

[0030] Example 2 Reference Figure 1 - Figure 2 This is the second embodiment of the present invention.

[0031] In this embodiment, step S100 involves collecting multi-source monitoring data of power transmission and transformation equipment and constructing an equipment health profile. Based on the equipment health profile, correlation mining is performed to identify key characteristic parameters of equipment operation and the coupling relationships between parameters, including the following steps A1-A2: A1: Collecting multi-source monitoring data of power transmission and transformation equipment and constructing equipment health profiles includes collecting online monitoring data and offline detection data of equipment to form a multi-dimensional monitoring dataset; By integrating monitoring information from different sources through the mapping relationship of multi-source data, an initial health profile of the equipment can be established.

[0032] Specifically, the multi-source monitoring data of power transmission and transformation equipment includes two main categories: online monitoring data and offline detection data. Online monitoring data is collected in real time by sensors deployed on the equipment, including electrical quantity monitoring data and non-electrical quantity monitoring data. Electrical quantity monitoring data includes three-phase voltage, three-phase current, active power, reactive power, power factor, harmonic content, etc., with a sampling frequency set to once per second and data accuracy maintained within 0.1%. Non-electrical quantity monitoring data includes equipment body temperature, ambient temperature, ambient humidity, vibration signals, sound signals, partial discharge signals, etc., with temperature data sampled once per minute, vibration and sound signals sampled at 1kHz, and partial discharge signals sampled at 10MHz. Offline detection data is obtained through specialized testing equipment during periodic power outages for maintenance, including insulation resistance test data, dielectric loss factor test data, oil chromatography analysis data, infrared thermal imaging data, etc., with the testing cycle ranging from one to six months depending on the importance of the equipment.

[0033] When establishing the initial health profile of the equipment, monitoring information from different sources is integrated through mapping relationships of multi-source data to create a multi-dimensional feature space. Electrical quantity monitoring data is used as the first dimension, thermal characteristic monitoring data as the second dimension, and mechanical characteristic monitoring data as the third dimension. In this multi-dimensional feature space, principal component analysis is used to reduce the dimensionality of high-dimensional features, retaining principal components with a cumulative variance contribution rate of over 95%. A feature mapping matrix is ​​constructed to establish the correlation between different monitoring parameters. The feature mapping matrix takes the form of a correlation coefficient matrix, where the element in the i-th row and j-th column represents the correlation between the i-th and j-th monitoring parameters, calculated using the following formula: in, Let be the correlation coefficient between the i-th parameter and the j-th parameter. Let be the value of the i-th parameter at time k. Let be the mean of the i-th parameter, and n be the total number of sampling points.

[0034] The initial health profile is expressed as a feature vector, which contains the normalized values ​​of all monitored parameters at the current moment, the time-domain statistical characteristics of each parameter, and the frequency-domain characteristics. The time-domain statistical characteristics include mean, standard deviation, peak value, peak-to-peak value, skewness, and kurtosis, while the frequency-domain characteristics are extracted using Fast Fourier Transform, including dominant frequency and spectral energy distribution. All features are grouped according to feature type: electrical features, thermal features, and mechanical features each form a sub-vector. These three sub-vectors are concatenated to form the complete equipment health profile feature vector.

[0035] A2: Identify key characteristic parameters of equipment operation and the coupling relationship between parameters, including extracting characteristic parameters reflecting the equipment status from the equipment health profile; Analyze the temporal correlation between characteristic parameters to determine the leader and lag relationships between parameters; The coupling strength between parameters is quantified, and a coupling relationship matrix is ​​established.

[0036] Specifically, firstly, a device status labeling system is established, dividing the historical operating status of the device into four levels: normal, alert, abnormal, and faulty. Historical health profile data for each status is collected to form a labeled dataset. A random forest algorithm is used to train a feature importance evaluation model. The model input is the health profile feature vector, and the output is the importance score of each feature for device status classification. The random forest model contains 100 decision trees, with a maximum depth of 15 layers per tree. When splitting a node, the number of randomly selected features is the square root of the total number of features. The model training uses 80% of the historical data as the training set and 20% as the validation set, with cross-validation used to determine the model parameters. After training, features are sorted according to their importance scores, and the top 30% of features are selected as key feature parameters.

[0037] Cross-correlation analysis is used to identify the correlation and time lag relationship between two time series. In the calculation, one time series is shifted relative to the other, and the correlation coefficient is calculated at different time lags. The time lag corresponding to the maximum correlation coefficient is the lag time between the two parameters. Granger causality tests are used to determine whether historical information of one time series helps predict the future value of another time series, thereby determining the leader-lag relationship between parameters. In practice, Granger causality tests are performed on each pair of key characteristic parameters, with lag orders set from 1 to 10 and a significance level set to 0.05. When the p-value of the Granger causality test for parameter A on parameter B is less than 0.05, parameter A is considered a Granger cause of parameter B, meaning parameter A leads parameter B in time. By performing pairwise tests on all key characteristic parameters, a time-series leader-lag relationship graph is established between the parameters.

[0038] When quantifying the coupling strength between parameters, mutual information and transfer entropy methods are used. Mutual information measures the statistical dependency between two random variables, reflecting how much information one variable contains about the other. The formula for calculating mutual information is: in, Mutual information between variables X and Y Let X be the joint probability distribution of X and Y, and p(x) and p(y) be the marginal probability distributions of X and Y, respectively. A larger mutual information value indicates a stronger coupling between the two parameters. Transfer entropy measures the amount of information transferred from one time series to another, and can identify the direction and strength of causal relationships between parameters. The mutual information and transfer entropy values ​​between all key feature parameters are calculated to form a coupling strength matrix.

[0039] When the coupling strength between two parameters exceeds a set threshold of 0.3, the corresponding position in the matrix is ​​marked as 1, indicating the existence of a coupling relationship; otherwise, it is marked as 0, indicating the absence of a coupling relationship. The coupling matrix provides a clear visual representation of the complex network of relationships between equipment operating parameters, offering a basis for subsequent multi-dimensional consistency checks. Furthermore, network analysis based on the coupling matrix identifies key node parameters—those with the most frequent coupling relationships with other parameters. These key node parameters have a significant impact on the overall equipment status and are given higher weight in subsequent evaluations.

[0040] In this embodiment, step S200 involves performing multi-dimensional consistency checks on key feature parameters through interactive verification to obtain the initial health assessment value of the device. Feature benchmarks of the device's historical normal operation are extracted, and the deviation between the current operating state and the feature benchmarks is calculated. Combined with the initial health assessment value, the overall health of the device and its corresponding threshold are obtained, including the following steps B1-B2: B1: Conduct multi-dimensional consistency checks through interactive verification, including setting electrical performance, thermal characteristics, and mechanical characteristics as evaluation dimensions; Calculate the status indicators for each evaluation dimension; cross-compare the status indicators for each dimension, adjust the weights of each dimension, and obtain the initial health assessment value.

[0041] Specifically, when setting electrical performance, thermal characteristics, and mechanical characteristics as evaluation dimensions, each dimension includes several key characteristic parameters. The electrical performance dimension includes parameters such as voltage stability, current balance, power factor, and harmonic distortion rate, which reflect the electrical operating status of the equipment. The thermal characteristics dimension includes parameters such as equipment body temperature, temperature rise rate, hot spot temperature, and ambient temperature correction, which reflect the thermal state of the equipment. The mechanical characteristics dimension includes parameters such as vibration intensity, vibration spectrum characteristics, sound intensity, and partial discharge amplitude, which reflect the mechanical and insulation state of the equipment.

[0042] For the electrical performance dimension, each parameter within this dimension is first scored, with the scoring rules based on the normal operating range of the parameter. When the parameter value is within the normal range, the score is 100 points; when the parameter value deviates from the normal range but does not exceed the attention threshold, the score is 80-100 points, with the specific score inversely proportional to the degree of deviation; when the parameter value exceeds the attention threshold but does not exceed the abnormal threshold, the score is 60-80 points; and when the parameter value exceeds the abnormal threshold, the score is 0-60 points. To calculate the state index for the electrical performance dimension, the scores of all parameters within this dimension are multiplied by their corresponding weights and then summed to obtain the comprehensive state index for the electrical performance dimension. The calculation methods for the state indices of the thermal and mechanical characteristics dimensions are the same as those for the electrical performance dimension.

[0043] Consistency check: If the overall equipment condition is good, all status indicators should be high; if the equipment has a fault or abnormality, multiple status indicators may decline simultaneously. First, calculate the variance of the three status indicators. The variance formula is: in, For variance, Let i be the state index of the i-th dimension. This represents the mean of the three-dimensional status indicators. A smaller variance indicates good consistency in the evaluation results across the dimensions; a larger variance indicates significant differences in the evaluation results across the dimensions, requiring further analysis to determine the reasons for these differences.

[0044] When the status indicator of a certain dimension remains stable and at a high level for a long period, the weight of that dimension should be appropriately reduced, and more attention should be paid to dimensions that may show a deteriorating trend. The weight adjustment uses an exponential smoothing method, and the adjustment formula is as follows: in, Let be the weight of the i-th dimension at time t. The adjusted weights are calculated based on the changing trends of the state indicators, with α being a smoothing coefficient of 0.2. The weight is the weight from the previous time step.

[0045] When obtaining the initial health assessment value, the three dimensions of status indicators are multiplied by their corresponding dynamic weights and then summed. The calculation formula is as follows: in, This is the initial health assessment value. (t) represents the weight of the i-th dimension at the current time. This represents the status indicator of the i-th dimension at the current moment. The initial health assessment value ranges from [0, 100], with a higher value indicating a better overall device status. When the initial health assessment value is greater than 90, the device is in a healthy state; when the initial health assessment value is between 70 and 90, the device is in a good state; when the initial health assessment value is between 50 and 70, the device is in a state of alert and requires enhanced monitoring; when the initial health assessment value is less than 50, the device is in an abnormal state and requires detailed inspection.

[0046] B2: Calculate the deviation between the current operating status and the characteristic benchmark, including the distribution range of characteristic parameters during the historical normal operation of the equipment; Calculate the distance metric between the current feature parameters and the historical distribution center; based on the distance metric and the initial health assessment value, determine the comprehensive health level and the classification threshold.

[0047] Specifically, when analyzing the distribution range of characteristic parameters during the historical normal operation of equipment, the time period of normal operation is first determined. Normal operation is defined as the period within the first three years after equipment commissioning without any failures or anomalies, or the period within the year before major overhaul without any failures. All key characteristic parameter data for the normal operation period are extracted from the historical database, and a probability distribution model is established for each parameter. The kernel density estimation method is used to perform nonparametric estimation of the parameter probability distribution. Kernel density estimation does not assume that the data follows a specific parameter distribution and can more accurately reflect the actual distribution characteristics of the parameters. The central location of the distribution for each parameter is calculated, and the median is used as the distribution center. Compared to the mean, the median is less sensitive to outliers and better reflects the typical characteristics of normal operation.

[0048] The feature baseline vector has the same dimension as the number of key feature parameters. The i-th element of the vector corresponds to the baseline value of the i-th key feature parameter, which is the median of that parameter during normal operation. The feature baseline vector serves as a reference standard for the equipment's health status, and all subsequent status assessments are compared against this benchmark. When the equipment's operating parameters are close to the feature baseline, the equipment is considered to be in good condition; when the equipment's operating parameters deviate from the feature baseline, the greater the deviation, the worse the equipment's condition.

[0049] When calculating the distance between the current feature parameter and the historical distribution center, Mahalanobis distance considers the correlation between feature parameters and the differences in the dispersion of each parameter, thus providing a more accurate measure of distance in multidimensional space. The formula for Mahalanobis distance is: in, Let X be the Mahalanobis distance, X be the feature parameter vector at the current time, μ be the feature reference vector, and Σ be the covariance matrix of the feature parameters. This is the inverse of the covariance matrix. The covariance matrix is ​​calculated based on historical data during normal operation. The element in the i-th row and j-th column of the matrix represents the covariance between the i-th and j-th parameters. The advantage of Mahalanobis distance is that when the normal operating range of a parameter is narrow, a small deviation of that parameter will lead to a significant increase in the Mahalanobis distance; while when the normal operating range of a parameter is wide, a large deviation of that parameter will have a relatively small impact on the Mahalanobis distance.

[0050] The initial health assessment value reflects the overall performance of the equipment across various assessment dimensions, while the Mahalanobis distance reflects the degree of deviation of the equipment's current state from the normal operating baseline. The Mahalanobis distance is converted into a deviation score using the following formula: in, For the deviation score, λ is the attenuation coefficient with a value of 0.1. The Mahalanobis distance is used. When the Mahalanobis distance is 0, the deviation score is 100, indicating complete compliance with normal operating standards; as the Mahalanobis distance increases, the deviation score decreases exponentially. The formula for calculating the overall health score is: in, For overall health, This is the initial health assessment value. The deviation score is used for evaluation. The weighting coefficients of 0.6 and 0.4 indicate the relative importance of the initial health assessment value and the deviation score in the overall evaluation. The initial health assessment value has a larger weight because it comprehensively considers multiple dimensions of status indicators, while the deviation score mainly reflects the comparison with the historical benchmark.

[0051] When determining the grading thresholds, the status levels are divided according to the numerical range of the overall health score. When the overall health score is greater than or equal to 90, the equipment is in excellent condition. In this state, the equipment operates stably, all parameters are normal, and no failure is expected in the short term. When the overall health score is between 75 and 90, the equipment is in good condition. In this state, the equipment operates normally overall, but some parameters may fluctuate slightly, and continued monitoring is recommended. When the overall health score is between 60 and 75, the equipment is in a watchful state. In this state, the equipment shows some signs of deterioration, and some parameters deviate from the normal range. It is recommended to increase the monitoring frequency and arrange a detailed inspection. When the overall health score is between 45 and 60, the equipment is in a warning state. In this state, the equipment exhibits obvious abnormal characteristics, and multiple parameters deviate from the normal range. It is recommended to arrange a power outage for maintenance as soon as possible. When the overall health score is less than 45, the equipment is in a dangerous state. In this state, the equipment may fail at any time, and immediate measures must be taken.

[0052] In this embodiment, step S300 identifies the degradation evolution trajectory based on the time-series changes in overall health, determines the degradation type and rate, and combines the degradation evolution trajectory with the overall health threshold to predict the future state evolution trend of the equipment and generate graded early warning information, including the following steps C1-C2: C1: Identifying the deterioration and evolution trajectory based on the temporal changes in comprehensive health includes recording the changes in comprehensive health over a continuous time period; Fit a time-series curve of overall health status to identify the trend of the curve; determine the type of degradation based on the slope and curvature of the curve, and determine the degradation rate based on the rate of change.

[0053] Specifically, a time-series database of comprehensive health scores is established to record changes in overall health scores over continuous time periods. This database stores the calculated comprehensive health score values ​​for each moment in chronological order, with the storage interval matching the data collection cycle. For online monitoring systems, the comprehensive health score is updated hourly, corresponding to a one-hour storage interval. In addition to storing the overall health score values, the time-series database also stores intermediate results used in calculating the comprehensive health score, such as the initial health assessment value, deviation score, and status indicators for each dimension, facilitating subsequent traceability and analysis.

[0054] When identifying the trend of a curve, the fitted curve is differentiated. The first derivative reflects the rate of change of overall health, while the second derivative reflects the rate of change of that rate, i.e., acceleration. When the first derivative is negative, it indicates that the overall health is declining and the equipment condition is deteriorating; when the first derivative is positive, it indicates that the overall health is rising and the equipment condition is improving; when the first derivative is close to zero, it indicates that the overall health remains stable. The sign of the second derivative reflects the characteristics of the deterioration trend. When the second derivative is negative, it indicates that the rate of decline of overall health is accelerating and the equipment deteriorates at an accelerating rate; when the second derivative is positive, it indicates that the rate of decline of overall health is slowing down and the equipment deteriorates at a decelerating rate.

[0055] When determining the type of degradation based on the slope and curvature of the curve, a degradation type discrimination rule is established. Degradation types are divided into four categories: linear degradation, exponential degradation, abrupt degradation, and fluctuating degradation. Linear degradation is characterized by an approximately linear decrease in overall health over time, with a negative first derivative that remains roughly constant, and a second derivative close to zero. This type of degradation typically corresponds to the normal aging process of equipment. Exponential degradation is characterized by a gradually accelerating rate of decline in overall health, with a negative first derivative whose absolute value increases over time, and a negative second derivative. This type of degradation typically corresponds to the development of a certain fault, such as insulation degradation or accelerated corrosion. Abrupt degradation is characterized by a rapid decline in overall health within a short period, with a sudden increase in the absolute value of the first derivative. This type of degradation typically corresponds to sudden faults or external shocks, such as short-circuit shocks or overvoltages.

[0056] Specifically, when determining the degradation rate based on the rate of change, the average slope of the overall health change is used as a quantitative indicator of the degradation rate. The formula for calculating the degradation rate is: in, Let H(t) be the degradation rate, and H(t) be the overall health status at the current moment. The overall health status is represented by the time interval Δt, typically 30 or 90 days. A negative degradation rate indicates equipment degradation, with a larger absolute value indicating faster degradation. A positive degradation rate indicates improved equipment condition. Degradation is categorized into three levels based on the degradation rate: minor, moderate, and severe. A degradation rate with an absolute value less than 1 (i.e., a monthly decrease in overall health of no more than 1 point) is considered minor degradation; a rate between 1 and 3 is considered moderate degradation; and a rate greater than 3 is considered severe degradation.

[0057] C2: Predict the future state evolution trend of the equipment, including extrapolating the future changes in the overall health of the equipment based on the degradation evolution trajectory; The system compares the predicted overall health level with the threshold; when the predicted overall health level approaches or exceeds the threshold, it generates a warning message of the corresponding level.

[0058] Specifically, when extrapolating future changes in the overall health of equipment based on the degradation trajectory, the appropriate prediction model is selected according to the identified degradation type. For linear degradation types, a linear extrapolation model is used, and the prediction formula is: in, The predicted overall health status at a future time t+τ Assess the overall health status at the current moment. Given the current rate of degradation, To predict the time span, for exponential degradation types, an exponential decay model is used, fitting the parameters of the exponential function based on historical data, and then extrapolating the future overall health. For abrupt degradation types, due to the unpredictability of abrupt changes, a conservative estimation strategy is adopted, assuming that the overall health will continue its current rapid decline.

[0059] When comparing the predicted overall health status with the threshold, the predicted overall health status is compared with the grading thresholds determined in step B2. The overall health status is predicted for the next 7, 30, and 90 days, and the prediction results are compared with the corresponding grading thresholds. If the predicted overall health status consistently remains above the good condition threshold, it indicates that the equipment is stable in the short term and requires no special handling. If the predicted overall health status will drop below the attention threshold within the next 7 days, a Level 1 warning is generated. If the predicted overall health status will drop below the warning threshold within the next 30 days, a Level 2 warning is generated. If the predicted overall health status will drop below the danger threshold within the next 90 days, a Level 3 warning is generated.

[0060] When generating tiered early warning information, the warning information includes four elements: warning level, warning time, warning reason, and recommended measures. The warning level is determined based on the relationship between the predicted overall health and a threshold, and is divided into Level 1, Level 2, and Level 3 warnings. The warning time indicates the expected moment when the equipment condition will cross a certain threshold, providing a time reference for maintenance personnel to schedule maintenance plans. The warning reason is determined by analyzing the main contributing factors to the decline in overall health, tracing back to specific assessment dimensions and key characteristic parameters, indicating whether the main cause of equipment condition deterioration is electrical performance degradation, abnormal thermal characteristics, or mechanical degradation. Recommended measures are formulated based on the warning reason and warning level. For Level 1 warnings, it is recommended to increase monitoring frequency and schedule a detailed inspection during the next power outage maintenance; for Level 2 warnings, it is recommended to schedule a power outage maintenance within one month to detect and address abnormal parts; for Level 3 warnings, it is recommended to immediately develop an emergency plan and, if necessary, schedule a power outage maintenance in advance to prevent failures.

[0061] In summary, a random forest algorithm is used to train a feature importance assessment model to screen key feature parameters. Granger causality tests are used to determine the temporal leading-lag relationships between parameters. Mutual information and transfer entropy methods are used to quantify coupling strength and establish a coupling relationship matrix, thus obtaining the intrinsic correlation network between equipment operating parameters. By setting three assessment dimensions and performing consistency tests on the state indicators of each dimension, the variance is calculated to determine the consistency of the assessment results. The weights of each dimension are dynamically adjusted using an exponential smoothing method. Mahalanobis distance is used to measure the deviation between the current state and historical feature benchmarks. The deviation score is weighted and fused with the initial health assessment value to obtain the comprehensive health level, which is divided into five status levels: excellent, good, attention, warning, and danger, based on the numerical range. By fitting the comprehensive health time series curve and performing derivative analysis, four types of degradation are identified: linear degradation, exponential degradation, abrupt degradation, and fluctuating degradation. Linear extrapolation or exponential decay models are used to predict future changes in comprehensive health level. By comparing the classification thresholds, graded warning information including warning level, warning time, warning cause, and recommended measures is generated, providing a complete technical implementation path for the operation and maintenance of power transmission and transformation equipment.

[0062] Example 3 The above is an illustrative scheme of an AI fusion processing method for power transmission and transformation equipment condition monitoring data. It should be noted that the technical solution of this AI fusion processing system for power transmission and transformation equipment condition monitoring data belongs to the same concept as the technical solution of the aforementioned AI fusion processing method for power transmission and transformation equipment condition monitoring data. Details not described in detail in this embodiment of the AI ​​fusion processing system for power transmission and transformation equipment condition monitoring data can be found in the description of the aforementioned AI fusion processing method for power transmission and transformation equipment condition monitoring data.

[0063] This embodiment also provides an AI fusion processing system for power transmission and transformation equipment status monitoring data, including: The data correlation analysis module is used to collect multi-source monitoring data of power transmission and transformation equipment and build equipment health profiles. Based on the equipment health profiles, correlation mining is performed to identify key characteristic parameters of equipment operation and the coupling relationship between parameters. The evaluation module is used to perform multi-dimensional consistency checks on key feature parameters through interactive verification, obtain the initial health assessment value of the equipment, extract the feature benchmark of the equipment's historical normal operation, calculate the deviation between the current operating status and the feature benchmark, and combine the initial health assessment value to obtain the comprehensive health of the equipment and the corresponding threshold. The early warning module is used to identify the degradation evolution trajectory based on the time-series changes in comprehensive health, determine the degradation type and degradation rate, combine the degradation evolution trajectory with the comprehensive health threshold, predict the status evolution trend of the equipment in the future time period, and generate graded early warning information.

[0064] This embodiment also provides an electronic device suitable for AI fusion processing of power transmission and transformation equipment status monitoring data, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the AI ​​fusion processing method for power transmission and transformation equipment status monitoring data proposed in the above embodiment.

[0065] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the AI ​​fusion processing method for power transmission and transformation equipment status monitoring data as proposed in the above embodiments.

[0066] The storage medium proposed in this embodiment and the AI ​​fusion processing method for realizing the status monitoring data of power transmission and transformation equipment proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0067] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0068] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. An AI fusion processing method for power transmission and transformation equipment state monitoring data, characterized in that: The method comprises the following steps: collecting multi-source monitoring data of power transmission and transformation equipment and constructing an equipment health portrait; performing correlation mining based on the equipment health portrait; identifying key characteristic parameters of equipment operation and coupling relationships between the parameters; Through multi-dimensional consistency verification by interaction verification, an initial health degree evaluation value of the equipment is obtained, a characteristic benchmark of normal historical operation of the equipment is extracted, a deviation degree is calculated between the current operating state and the characteristic benchmark, and a comprehensive health degree of the equipment and a corresponding threshold value are obtained in combination with the initial health degree evaluation value; Based on the time sequence change of the comprehensive health degree, a degradation evolution trajectory is identified, the degradation type and degradation rate are determined, the state evolution trend of the equipment in a future time period is predicted in combination with the degradation evolution trajectory and the comprehensive health degree threshold value, and graded early warning information is generated.

2. The AI fusion processing method for power transmission equipment condition monitoring data according to claim 1, characterized in that: The step of collecting multi-source monitoring data of power transmission and transformation equipment and constructing an equipment health portrait comprises the following steps: collecting online monitoring data and offline detection data of the equipment to form a multi-dimensional monitoring data set; Different sources of monitoring information are integrated through the mapping relationship of multi-source data, and an initial health portrait of the equipment is established.

3. The AI fusion processing method for power transmission equipment condition monitoring data according to claim 2, characterized in that: The step of identifying key characteristic parameters of equipment operation and coupling relationships between the parameters comprises the following steps: extracting characteristic parameters reflecting the state of the equipment from the equipment health portrait; The time sequence correlation between the characteristic parameters is analyzed, and the leading and lagging relationships between the parameters are determined; The coupling strength between the parameters is quantified, and a coupling relationship matrix is established.

4. The AI fusion processing method for power transmission equipment condition monitoring data according to claim 3, characterized in that: The step of performing multi-dimensional consistency verification by interaction verification comprises the following steps: setting electrical performance dimension, thermal characteristic dimension and mechanical characteristic dimension as evaluation dimensions; State indicators under each evaluation dimension are calculated respectively, the state indicators of each dimension are cross-compared, the weights of each dimension are adjusted, and an initial health degree evaluation value is obtained.

5. The AI fusion processing method for power transmission equipment condition monitoring data according to claim 4, characterized in that: The step of calculating the deviation degree between the current operating state and the characteristic benchmark comprises the following steps: counting the distribution range of the characteristic parameters during the normal historical operation of the equipment; The distance measure between the current characteristic parameters and the historical distribution center is calculated; Based on the distance measure and the initial health degree evaluation value, a comprehensive health degree and a grading threshold value are determined.

6. The AI fusion processing method for power transmission equipment condition monitoring data according to claim 5, characterized in that: The step of identifying the degradation evolution trajectory based on the time sequence change of the comprehensive health degree comprises the following steps: recording the change values of the comprehensive health degree in consecutive time periods; The time sequence curve of the comprehensive health degree is fitted, the change trend of the curve is identified, the degradation type is determined according to the slope and curvature of the curve, and the degradation rate is determined according to the change rate.

7. The AI fusion processing method for power transmission equipment condition monitoring data according to claim 6, characterized in that: The step of predicting the state evolution trend of the equipment in a future time period comprises the following steps: extrapolating the future change of the comprehensive health degree of the equipment based on the degradation evolution trajectory; The relationship between the predicted comprehensive health degree and the threshold value is compared; when the predicted comprehensive health degree approaches or exceeds the threshold value, corresponding graded early warning information is generated.

8. An AI fusion processing system for power transmission and transformation equipment state monitoring data, based on the AI fusion processing method for power transmission and transformation equipment state monitoring data according to any one of claims 1-7, characterized in that: The method further comprises the following steps: a data correlation analysis module is used to collect multi-source monitoring data of power transmission and transformation equipment and construct an equipment health portrait, perform correlation mining based on the equipment health portrait, and identify key characteristic parameters of equipment operation and coupling relationships between the parameters. An evaluation module is configured to perform multi-dimensional consistency test on the key characteristic parameters through interactive verification, obtain an initial health degree evaluation value of the equipment, extract a characteristic benchmark of normal historical operation of the equipment, calculate a deviation degree between the current operation state and the characteristic benchmark, and obtain a comprehensive health degree of the equipment and a corresponding threshold value in combination with the initial health degree evaluation value; An early warning module is configured to identify a degradation evolution track based on a time sequence change of the comprehensive health degree, determine a degradation type and a degradation rate, predict a state evolution trend of the equipment in a future time period in combination with the degradation evolution track and the comprehensive health degree threshold value, and generate graded early warning information. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is characterized in that: The processor implements the steps of the AI fusion processing method of the power transmission and transformation equipment state monitoring data according to any one of claims 1-7 when executing the computer program.

10. A computer readable storage medium having stored thereon a computer program, characterized in that: The computer program is executed by the processor to implement the steps of the AI fusion processing method of the power transmission and transformation equipment state monitoring data according to any one of claims 1-7.

Images