Power equipment state information fusion system and method

Abstract

The invention relates to the technical field of information fusion, in particular to a power equipment state information fusion system and method. The method comprises the following steps: obtaining time sequence signal data and infrared thermal image data, and constructing a power equipment heterogeneous fusion data set; acquiring power equipment position and power equipment environment data, performing multi-layer cross-mode data coupling in combination with the power equipment heterogeneous fusion data set, and constructing a power equipment information evaluation model based on the multi-layer cross-mode data coupling; therefore, through multi-source data fusion and cross-level evaluation model construction, the problems of a traditional power equipment monitoring system in the aspects of information island and prediction accuracy are solved, and the health management capability and fault early warning efficiency of power equipment are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of information fusion, and in particular, to a power equipment status information fusion system and method. Background Art

[0002] The technical characteristics of power equipment status include real-time operation data collection, environment and equipment coupling analysis, multi-source data fusion, and fault mode and health prediction. Through time-series signal analysis, thermal anomaly monitoring, and attenuation analysis, the health status and potential risks of the equipment are comprehensively evaluated. Traditional technologies often rely on single types of sensor data, such as current and temperature data, and these data are usually independently collected and analyzed, lacking a comprehensive assessment of the overall state of the equipment. Insufficient multi-source data fusion makes it difficult to accurately reflect the health status of the equipment under different operating conditions. In addition, existing fault diagnosis methods are mostly based on static threshold settings and cannot be dynamically adjusted according to changes in the equipment operating environment or load, thus reducing the accuracy of prediction. Secondly, most existing systems rely on traditional time-series analysis and statistical methods, ignoring the complex interaction between equipment status and environmental factors. Equipment status assessment is mostly limited to single data analysis, lacking effective fusion of heterogeneous data of the equipment, resulting in the potential of information not being fully explored and utilized. Moreover, power equipment health prediction models usually rely on simplified assumptions and lack the support of advanced technologies such as deep learning, making it difficult to comprehensively reflect the dynamic changes and environmental adaptability of the equipment. This makes traditional systems insufficient in accuracy when dealing with complex fault modes and has great limitations in the real-time assessment and continuous optimization of equipment status. Summary of the Invention

[0003] Based on this, it is necessary to provide a power equipment status information fusion system and method to solve at least one of the above technical problems.

[0004] To achieve the above object, a power equipment status information fusion method includes the following steps:

[0005] Step S1: Obtain time-series signal data and infrared thermal image data, and construct a heterogeneous fusion data set of power equipment;

[0006] Step S2: Obtain the location of the power equipment and the power equipment environment data, perform multi-layer cross-modal data coupling in combination with the heterogeneous fusion data set of the power equipment, and construct a power equipment information evaluation model based on the multi-layer cross-modal data coupling;

[0007] Step S3: Obtain historical power equipment status information; calculate the power equipment status evaluation threshold for the historical power equipment status information, and build a multi-level data processing architecture for the power equipment information evaluation model based on the power equipment status evaluation threshold, where the multi-level data processing architecture includes a horizontal hierarchical architecture of power equipment and a vertical hierarchical architecture of power equipment;

[0008] Step S4: Perform device health management prediction according to the multi-level data processing architecture to obtain the predicted value of power device health management; conduct information fusion and comparative analysis on the predicted value of power device health management to generate a power device status information fusion report.

[0009] The beneficial effects of the present invention are as follows. Through the comprehensive evaluation of the power device status and health management prediction, a system architecture based on multi-source data fusion is proposed, which solves the deficiencies in data acquisition and fusion of traditional power device monitoring systems. First, the time-series signal data and infrared thermal image data are real-time acquired through the deployed sensors, and are fused with the heterogeneous data set of power devices, providing a multi-dimensional data basis for the status evaluation of power devices. These data not only include the dynamic operation information of the devices, but also fuse the environmental and thermal image information, providing rich data support for subsequent analysis. Second, by acquiring the location information and environmental data of power devices and performing multi-layer transmembrane data coupling with the previous heterogeneous fusion data set, the power device information evaluation model can simultaneously consider the mutual influence between the device operation status and external environmental changes, thereby improving the accuracy and comprehensiveness of the evaluation model. Further, through the acquisition and analysis of historical power device status information and combined with the calculation of evaluation thresholds, a multi-level data processing architecture is constructed, enabling the model to not only perform real-time status evaluation, but also predict and risk-assess the long-term health status of the devices. Finally, through the fusion and comparative analysis of the predicted values of power device health management, a power device status information fusion report is generated, providing more accurate device health conditions and early warning information for decision-makers. Generally speaking, the present invention can comprehensively improve the real-time performance, accuracy and early warning ability of the power device monitoring system by effectively fusing and deeply analyzing multi-source data, providing strong support for the operation safety and maintenance decision-making of power devices. Therefore, through the construction of multi-source data fusion and cross-level evaluation models, the present invention solves the problems of information islands and prediction accuracy in traditional power device monitoring systems, and improves the health management ability and fault early warning efficiency of power devices.

[0010] Preferably, the construction of the heterogeneous fusion data set of power devices in step S1 includes:

[0011] Collect the current data of the power device through a current sensor and collect the temperature data of the power device through a temperature sensor;

[0012] Perform 1kHz time-series fusion on the current data and temperature data of the power device to obtain time-series signal data; conduct 16-bit entropy evaluation according to the time-series signal data, and screen out the effective information for information entropy calculation to obtain the information entropy evaluation criterion;

[0013] Perform noise removal on the time-series signal data in different frequency bands, where the noise removal range is 0.1 - 500 Hz, and generate preliminary time-series frequency-band noise-reduced data; perform entropy value wavelet basis denoising on the preliminary time-series frequency-band noise-reduced data based on the information entropy evaluation criterion to obtain the processed data of the power time-series signal;

[0014] Perform preliminary spatial feature convolution on the infrared thermal image data, and distinguish hot spots and thermal gradients to obtain the infrared thermal image topological features, where the infrared thermal image topological features include local thermal anomalies and global thermal anomalies;

[0015] Perform spatial feature reconstruction with 95% accuracy on the local thermal anomalies to obtain local spatial coding data;

[0016] Perform analysis from the perspective of heat distribution on the global thermal anomalies to obtain global spatial coding data;

[0017] Perform 1:1 spatial heat topological information fusion on the local thermal anomalies and global thermal anomalies to generate power equipment thermal anomaly data;

[0018] Perform cross-membrane state association with a correlation window of 5 s on the processed data of the power time-series signal and the power equipment thermal anomaly data to obtain a heterogeneous fusion dataset of power equipment.

[0019] The present invention systematically integrates multi-dimensional data of power equipment, proposes an accurate and efficient data processing flow, and effectively improves the accuracy of power equipment status assessment and fault early warning. In the data acquisition stage, current sensors and temperature sensors are used to obtain the current data and temperature data of power equipment respectively, providing a basis for subsequent data fusion and analysis. In the time series signal processing stage, by performing time series fusion on the current data and temperature data at a sampling frequency of 1 kHz, the dynamic changes of power equipment under different working states can be captured in detail. Subsequently, through information entropy evaluation, the time series signal data is screened, the effective information is extracted and calculated to obtain the information entropy evaluation criterion. This process effectively reduces the interference of redundant information and provides more accurate input data for subsequent analysis. Further, the time series signal data is subjected to noise removal in different frequency bands (in the range of 0.1 - 500 Hz), significantly improving the data quality, making the signal smoother and reducing the impact of noise on equipment status assessment. Based on the information entropy evaluation criterion, the entropy value wavelet basis denoising method is used to further process the time series data, making the data more reliable and stable. At the same time, the infrared thermal image data undergoes preliminary spatial feature convolution and distinguishes hot spots and thermal gradients, effectively extracting the local thermal anomaly and global thermal anomaly features of power equipment. The local thermal anomaly undergoes spatial feature reconstruction with 95% accuracy to obtain accurate local spatial coding data, while the global thermal anomaly is analyzed from the perspective of thermal distribution to obtain global spatial coding data. The local and global thermal anomaly data are fused through 1:1 spatial heat topology information, providing a more comprehensive and detailed analysis of the thermal anomaly of power equipment. Finally, by combining the power time series signal processing data and the power equipment thermal anomaly data, heterogeneous data fusion is performed through a 5-second cross-modal state correlation window to generate a heterogeneous fusion data set of power equipment, providing strong data support for the accurate assessment of equipment status and fault diagnosis. This data processing flow can effectively improve the accuracy and reliability of equipment operation status assessment in power equipment monitoring and provide a scientific basis for the optimization of intelligent maintenance and early warning systems.

[0020] Preferably, the multi-layer cross-modal data coupling in step S2 includes:

[0021] Create an equipment node adjacency matrix based on the location of the power equipment, and define the connection weight using a physical distance of 0 m - 100 km to obtain the topological feature data of the power equipment node;

[0022] Couple the environmental sensitivity coefficient of the power equipment node topology map using the environmental data of the power equipment to obtain the environment-equipment coupling data; construct a coupling matrix for the environment-equipment coupling data to obtain the environment-equipment coupling feature matrix;

[0023] Locate the harmonic components in the processed data of the power time-series signal, and perform amplitude-phase distribution correlation insulation degradation analysis. The amplitude range is usually 0 to 1000V, and the phase angle is between 0° and 360°, to obtain the power time-series feature tensor.

[0024] Divide the high-temperature 150°C bushing temperature rise area according to the thermal anomaly data of power equipment, and infer the heat propagation path from 0m to 50m to obtain the thermal anomaly map data.

[0025] The present invention effectively reveals the spatial relationship and interconnectivity between power equipment through the creation of an adjacency matrix based on the location of power equipment and the definition of connection weights of physical distances. This process can provide an accurate network structure basis for subsequent topological feature analysis. Using this topological structure, the mutual influence and cooperation relationship between power equipment can be understood more clearly. Especially in large-scale power systems, it can provide in-depth insights into equipment layout, interaction, and potential fault propagation. Secondly, through the coupled analysis of the environmental data of power equipment, the environmental sensitivity coefficient is introduced into the node topology map to generate an environment-equipment coupling feature matrix. This matrix helps to capture the impact of environmental factors on the operating state of equipment, such as the adjustment effect of environmental variables such as temperature and humidity on equipment load and fault mode. Such a coupling feature matrix not only improves the accuracy of the equipment health assessment model but also provides the ability to predict the equipment state under the comprehensive influence of multiple factors. Further, the harmonic component location and amplitude-phase distribution correlation analysis of the processed data of the power time-series signal help to reveal the insulation degradation inside the power equipment, thus realizing the early identification of faults. Through this process, high-value information can be extracted from the time-series signal to form a power time-series feature tensor, providing data support for subsequent equipment maintenance and decision-making. In addition, through the analysis of dividing the high-temperature 150°C bushing temperature rise area and inferring the heat propagation path, the thermal anomaly data of power equipment are successfully extracted, which provides an effective tool for predicting equipment overheating faults. The inference of the heat propagation path can identify the propagation mode and range of thermal anomalies in the equipment, providing a real-time and accurate monitoring basis for the equipment warning and health management system. In summary, the present invention comprehensively improves the ability of power equipment state assessment and fault diagnosis from different dimensions through the integration and in-depth analysis of multi-source data, and can effectively improve the safety, reliability, and economy of equipment operation.

[0026] Preferably, step S32 includes the following steps:

[0027] Step S321: Screen the key indicators of the historical power equipment state information based on the heterogeneous fusion dataset of power equipment to obtain the strongly correlated indicators of power equipment;

[0028] Step S322: Conduct quartile interval analysis on the strongly related indicators of power equipment, and mark the amplitude and duration of abnormal states to obtain power equipment abnormal marking data; conduct analysis on the strongly related rules of power data for the power equipment abnormal marking data to obtain power strongly related distribution rule data;

[0029] Step S323: Determine the warning line for the power strongly related distribution rule data through normal distribution, and calculate the static threshold for the power strongly related warning data to obtain the static threshold for power status assessment; conduct analysis on the decay tolerance setting of the equipment service life for the power strongly related distribution rule data, and conduct dynamic threshold assessment on the power equipment decay data to obtain the dynamic threshold for power status assessment;

[0030] Step S324: Perform threshold weight merging processing on the static threshold for power status assessment and the dynamic threshold for power status assessment to obtain the threshold for power equipment status assessment.

[0031] Through the screening of key indicators of the heterogeneous fusion dataset of power equipment, the present invention can identify the indicators that are most strongly related to the status of power equipment, thereby providing high-quality data input for subsequent analysis. These key indicators include physical quantities such as current, temperature, and vibration that are closely related to the health status of the equipment, which helps to reveal the potential fault risks and operating abnormalities of the equipment. By conducting quartile interval analysis on the strongly related indicators, outliers in the data can be effectively identified, and by marking the amplitude and duration of abnormal states, the abnormal patterns of the equipment can be further analyzed in depth. This process not only helps to clarify the normal and abnormal boundaries of the equipment status, but also provides an important basis for equipment fault prediction and risk warning. By analyzing the distribution rules of the power equipment abnormal marking data, power strongly related distribution rule data are obtained, which reflect the typical change patterns of the equipment status and help to establish a health assessment model based on historical data in the future. By setting the warning line through normal distribution for the power strongly related distribution rule data, a stable warning benchmark for equipment status assessment is provided, and the normal and abnormal intervals of power status are determined through static threshold calculation. At the same time, by conducting analysis on the decay tolerance setting of the equipment service life, the performance decay characteristics of the equipment with the increase of service years are considered, and the dynamic threshold setting can flexibly adjust the assessment criteria according to the actual use situation of the equipment, making the assessment results more in line with the actual operation situation. Finally, by merging and jointly analyzing the static threshold and the dynamic threshold, the threshold for power equipment status assessment that comprehensively considers time decay and changes in operating status is obtained, and this assessment threshold provides a scientific basis for power equipment health management, fault diagnosis, and maintenance decision-making.

[0032] Preferably, step S33 includes the following steps:

[0033] Step S331: Analyze the correlation between the frequency of over-limit and abnormality of a single device in the power equipment feature interaction layer to obtain the abnormality analysis result of the single device; perform hierarchical structure division processing according to the abnormality analysis result of the single device to obtain the horizontal hierarchical structure of the power equipment;

[0034] Step S332: Performing a logical judgment on the continuous deterioration of multiple indicators of the equipment group on the power equipment status assessment layer to obtain equipment group deterioration trend data; performing equipment group status cluster analysis on the equipment group deterioration trend data to generate equipment group cluster analysis result data; performing a vertical health assessment architecture division process on the equipment group cluster analysis result data to obtain a vertical hierarchical architecture of the power equipment;

[0035] Step S333: constructing a multi-level data processing architecture based on the power equipment state assessment threshold by combining the power equipment horizontal classification architecture and the power equipment vertical classification architecture to obtain a multi-level data processing architecture.

[0036] The present invention can identify the abnormal mode of the equipment under different operating conditions and mine the potential failure risk of the equipment by analyzing the frequency of over-limit and abnormal correlation of a single device in the characteristic interaction layer of the power equipment. This analysis method not only helps to accurately identify abnormal behavior, but also provides a basis for subsequent equipment classification. By dividing the abnormal analysis results into a hierarchical architecture, the state recognition ability of the equipment is further improved, so that the equipment can obtain different evaluation levels under different health states. This process constructs a horizontal hierarchical architecture of power equipment, which provides strong support for equipment comparison and fault risk prediction in multi-equipment systems. By applying the multi-indicator continuous deterioration judgment logic of the equipment group, the overall health trend of the equipment group can be captured, and which equipment groups show a deterioration trend during long-term operation can be identified, avoiding the limitations of single equipment evaluation. The deterioration trend data is classified by cluster analysis method, which not only reveals the heterogeneity of the health status of the equipment group, but also provides more targeted maintenance and early warning strategies for each equipment group. Next, through the vertical health assessment architecture division, combined with the multi-dimensional data of the equipment group, its health status is evaluated from the perspective of time series and equipment life cycle, providing a more refined framework for equipment life prediction and health management. Finally, based on the power equipment status assessment threshold, a multi-level data processing architecture was built by combining the horizontal and vertical hierarchical architectures to form an assessment system with high flexibility and adaptability. Through comprehensive analysis of static and dynamic data, the system can provide accurate status assessment at all stages of equipment operation and adjust the assessment criteria according to different equipment characteristics, thereby improving the reliability and accuracy of the assessment results.

[0037] Preferably, step S4 comprises the following steps:

[0038] Step S41: Obtain real-time device operation data; perform device health management prediction according to the multi-level power equipment status evaluation and processing strategy to obtain the power equipment health management prediction value;

[0039] Step S42: Analyze the status information of the real-time device operation data and the power equipment health management prediction value to obtain the power equipment health status comparison data;

[0040] Step S43: Based on the power equipment health status comparison data, perform report summarization to obtain the power equipment status information fusion report.

[0041] The present invention provides first-hand data for the evaluation of the device health status by obtaining real-time device operation data. These data include key operation parameters such as current, voltage, temperature, load, etc., which can reflect the working status of the device in real time. On this basis, combined with the multi-level power equipment status evaluation and processing strategy, device health management prediction is carried out to obtain the power equipment health management prediction value. This prediction value not only considers the current operation status of the device, but also integrates historical data, failure modes, and device life cycle information to provide a judgment on the future status of the device. This multi-level evaluation method can capture the changing trend of device health, so as to identify potential failure risks in advance. Analyzing the status information of the real-time device operation data and the power equipment health management prediction value can further reveal the anomalies and potential problems in the device operation by comparing the differences between the current device status and the prediction value. Through this comparative analysis, it can accurately judge whether the device is within the normal operation range, and timely discover the devices deviating from the normal state, providing data support for subsequent maintenance decisions. Finally, based on the power equipment health status comparison data, a summary report is constructed to generate the power equipment status information fusion report. This report not only summarizes the current health status of the device, but also combines the prediction data to provide the future operation trend and potential risks of the device, providing an intuitive and comprehensive decision-making basis for the operation and maintenance personnel. By summarizing and integrating multi-dimensional data, the report can effectively reflect the status of the device at different time periods, helping decision-makers formulate reasonable maintenance plans and fault prevention measures according to the device health status.

[0042] Preferably, step S42 includes the following steps:

[0043] Step S421: Analyze the trend similarity of the real-time device operation data and the power equipment health management prediction value to obtain the trend similarity analysis data;

[0044] Step S422: Cluster the abnormal patterns according to the trend similarity analysis data, and perform root cause correlation analysis to obtain the power equipment status correlation data;

[0045] Step S423: Based on the power equipment status correlation data, perform confidence comparison and feedback to obtain the power equipment health status comparison data.

[0046] Through the trend similarity analysis of the real-time device operation data and the power equipment health management prediction values, the present invention can identify the similarity of the device operation trends in different time periods, and then capture the potential abnormal patterns of the device. This step effectively compares the current operation status of the device with the historical prediction trends, enabling early detection and risk identification when the device status deviates slightly. This analysis can reveal the weak change trends existing in the long-term operation process of the device, thus providing strong data support for subsequent fault warning. According to the trend similarity analysis data, perform clustering of 2 to 10 types of abnormal patterns to further classify the abnormal patterns of the device status. This process effectively decomposes the complex abnormal behaviors of the device into several clearly distinguishable categories through the clustering method, enabling each abnormal pattern to be independently identified and deeply analyzed. Clustering analysis not only helps to understand the common fault patterns shown by different devices under similar working conditions, but also helps the operation and maintenance personnel to examine the device health from multiple perspectives and find out the key factors related to the abnormal device status. Combining with the root cause correlation analysis, the root cause leading to the device abnormality can be further clarified, thereby optimizing the subsequent fault prediction and maintenance strategies. Based on the power equipment status correlation data, perform confidence comparison and feedback, which can further improve the reliability of the analysis results by comparing the confidence levels of different source data (such as device operation status and health prediction values). Through the feedback of the confidence level, the comparison data of the device health status can be dynamically adjusted to ensure the minimization of errors in the evaluation process and enhance the accuracy and credibility of the device status evaluation.

[0047] In this specification, a power equipment status information fusion system is provided for performing the above-mentioned power equipment status information fusion method. The power equipment status information fusion system includes:

[0048] A sensor data acquisition and fusion module for acquiring time-series signal data and infrared thermal image data and constructing a heterogeneous fusion data set of power equipment;

[0049] A sensor data acquisition and fusion module for acquiring the location of the power equipment and the power equipment environment data, performing multi-layer cross-modal data coupling in combination with the heterogeneous fusion data set of power equipment, and constructing a power equipment information evaluation model based on the multi-layer cross-modal data coupling;

[0050] Historical data analysis and evaluation model optimization module, which is used to obtain historical power equipment status information; calculate the power equipment status evaluation threshold for the historical power equipment status information, and build a multi-level data processing architecture for the power equipment information evaluation model based on the power equipment status evaluation threshold, where the multi-level data processing architecture includes a horizontal hierarchical architecture of power equipment and a vertical hierarchical architecture of power equipment;

[0051] Equipment health management prediction and reporting module, which is used to perform equipment health management prediction according to the multi-level data processing architecture to obtain the power equipment health management prediction value; perform information fusion and comparative analysis on the power equipment health management prediction value to generate a power equipment status information fusion report.

[0052] The beneficial effects of the present invention are as follows. Through the comprehensive evaluation of the power equipment status and the prediction of equipment health management, a system architecture based on multi-source data fusion is proposed, which solves the deficiencies in data acquisition and fusion of traditional power equipment monitoring systems. First, the time-series signal data and infrared thermal image data are obtained in real time through the deployed sensors, and are fused with the heterogeneous data set of power equipment, providing a multi-dimensional data basis for the status evaluation of power equipment. These data not only include the dynamic operation information of the equipment, but also fuse the environmental and thermal image information, providing rich data support for subsequent analysis. Second, by obtaining the location information and environmental data of the power equipment and performing multi-layer transmembrane data coupling with the previous heterogeneous fusion data set, the power equipment information evaluation model can simultaneously consider the mutual influence between the equipment operation status and the external environmental changes, thereby improving the accuracy and comprehensiveness of the evaluation model. Further, through the acquisition and analysis of historical power equipment status information and combined with the calculation of evaluation thresholds, a multi-level data processing architecture is constructed, enabling the model to not only perform real-time status evaluation, but also predict and risk-assess the long-term health status of the equipment. Finally, through the fusion and comparative analysis of the power equipment health management prediction values, a power equipment status information fusion report is generated, providing more accurate equipment health status and early warning information for decision-makers. Generally speaking, the present invention can comprehensively improve the real-time performance, accuracy and early warning ability of the power equipment monitoring system by effectively fusing and deeply analyzing multi-source data, providing strong support for the operation safety and maintenance decision-making of power equipment. Therefore, the present invention solves the problems of information islands and prediction accuracy in traditional power equipment monitoring systems by constructing multi-source data fusion and cross-level evaluation models, and improves the health management ability and fault early warning efficiency of power equipment. Brief Description of the Drawings

[0053] Figure 1 It is a schematic diagram of the step flow of a method for fusing power equipment status information;

[0054] Figure 2 isFigure 1 Schematic diagram of the detailed implementation steps of step S3 in

[0055] The realization, functional characteristics, and advantages of the object of the present invention will be further described in conjunction with the embodiments with reference to the attached drawings. Specific embodiments

[0056] The technical method of the present invention for patents will be clearly and completely described below with reference to the attached drawings. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of them. All other embodiments obtained by those skilled in the art within the scope of the present invention without creative work belong to the scope of protection of the present invention.

[0057] In addition, the attached drawings are only schematic diagrams of the present invention and are not necessarily drawn to scale. The same reference numerals in the drawings represent the same or similar parts, and thus their repeated description will be omitted. Some of the block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0058] It should be understood that although terms such as "first" and "second" may be used here to describe each unit, these units should not be limited by these terms. These terms are only used to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, the first unit can be called the second unit, and similarly, the second unit can be called the first unit. The term "and / or" used here includes any and all combinations of one or more of the listed associated items.

[0059] To achieve the above object, please refer to Figures 1 to 2 , a method for fusing power equipment status information, the method comprising the following steps:

[0060] Step S1: Obtain time series signal data and infrared thermal image data, and construct a heterogeneous fusion data set of power equipment;

[0061] Step S2: Obtain the location of the power equipment and the power equipment environment data, perform multi-layer cross-modal data coupling in combination with the heterogeneous fusion data set of the power equipment, and construct a power equipment information evaluation model based on the multi-layer cross-modal data coupling;

[0062] Step S3: Obtain historical power equipment status information; calculate the power equipment status evaluation threshold for the historical power equipment status information, and build a multi-level data processing architecture based on the power equipment status evaluation threshold for the power equipment information evaluation model, where the multi-level data processing architecture includes a horizontal classification architecture of power equipment and a vertical classification architecture of power equipment;

[0063] Step S4: Perform equipment health management prediction according to the multi-level data processing architecture to obtain the power equipment health management prediction value; conduct information fusion and comparative analysis on the power equipment health management prediction value to generate a power equipment status information fusion report.

[0064] The beneficial effects of the present invention are as follows. Through the comprehensive evaluation of the power equipment status and health management prediction, a system architecture based on multi-source data fusion is proposed, which solves the deficiencies in data acquisition and fusion of traditional power equipment monitoring systems. First, the time-series signal data and infrared thermal image data are obtained in real time through the deployed sensors, and are fused with the heterogeneous data set of power equipment, providing a multi-dimensional data basis for the status evaluation of power equipment. These data not only include the dynamic operation information of the equipment, but also fuse the environmental and thermal image information, providing rich data support for subsequent analysis. Second, by obtaining the location information and environmental data of the power equipment and performing multi-layer transmembrane data coupling with the previous heterogeneous fusion data set, the power equipment information evaluation model can simultaneously consider the mutual influence between the equipment operation status and external environmental changes, thereby improving the accuracy and comprehensiveness of the evaluation model. Further, through the acquisition and analysis of historical power equipment status information, combined with the calculation of evaluation thresholds, a multi-level data processing architecture is constructed, enabling the model to not only perform real-time status evaluation, but also predict and risk-assess the long-term health status of the equipment. Finally, through the fusion and comparative analysis of the power equipment health management prediction value, a power equipment status information fusion report is generated, providing more accurate equipment health status and early warning information for decision-makers. Generally speaking, the present invention can comprehensively improve the real-time performance, accuracy and early warning ability of the power equipment monitoring system by effectively fusing and deeply analyzing multi-source data, providing strong support for the operation safety and maintenance decision-making of power equipment. Therefore, through the construction of multi-source data fusion and cross-level evaluation models, the present invention solves the problems of information islands and prediction accuracy in traditional power equipment monitoring systems, and improves the health management ability and fault early warning efficiency of power equipment.

[0065] In the embodiment of the present invention, with reference to Figure 1 as shown, in this example, the power equipment status information fusion method includes the following steps:

[0066] Step S1: Obtain time-series signal data and infrared thermal image data, and construct a heterogeneous fusion data set of power equipment;

[0067] In the embodiments of the present invention, by installing multi-type sensors (such as current, voltage, temperature, vibration sensors, etc.) at key parts of power equipment, sequential signal data is collected in real time. These sensors can collect the dynamic changes of the equipment during operation at a high frequency, generating continuous sequential data streams, usually with a sampling frequency reaching the kHz level, so as to be able to reflect the minute changes and potential abnormal behaviors of the equipment in a short period of time. By processing, analyzing and fusing these sequential signal data, the working state, load condition and health status of the equipment can be monitored in real time, providing basic data for subsequent fault diagnosis and early warning. At the same time, an infrared thermal imager is used to monitor the temperature field of the equipment, obtaining the thermal distribution of different parts of the power equipment, especially scanning the surface temperature of the equipment and potential hot spot areas. The infrared thermal image data can not only reflect the operating temperature of the equipment, but also reveal potential thermal anomalies and local overheating problems existing in the equipment. Infrared imaging technology can obtain the temperature distribution on the surface of the equipment in real time without contacting the equipment, thus avoiding the problem that cannot be covered by traditional temperature sensors. In addition, in order to improve the availability and comprehensiveness of the data, the collected sequential signal data and infrared thermal image data are fused to form a heterogeneous fusion data set. Through data fusion technology, data from different sensors can be standardized, time-synchronized and spatially aligned to construct a comprehensive data set with multi-dimensional characteristics. This data set not only covers the dynamic electrical behavior of the equipment, but also includes the temperature state of the equipment and environmental impact factors, making the subsequent state assessment and fault diagnosis more comprehensive and accurate. During the data fusion process, advanced data processing methods, such as interpolation, smoothing, denoising and other technologies, are applied to ensure the consistency and reliability of data from different sources. Finally, this heterogeneous fusion data set of power equipment provides strong data support for the subsequent health assessment, fault diagnosis and construction of prediction models.

[0068] Step S2: Obtain the location and environmental data of the power equipment, perform multi-layer cross-modal data coupling in combination with the heterogeneous fusion data set of the power equipment, and construct a power equipment information evaluation model based on the multi-layer cross-modal data coupling;

[0069] In the embodiments of the present invention, by obtaining the location of power equipment and power equipment environmental data, multi-layer cross-membrane data coupling is performed on the basis of the existing heterogeneous fusion data set of power equipment to construct a power equipment information evaluation model. The specific technical means involve multiple data processing and fusion technologies, aiming to comprehensively integrate real-time information in multiple aspects such as the device body, environment, and location. First, precise location data of power equipment are collected through Geographic Information System (GIS) technology and positioning sensors (such as GPS or RFID tags). These location data are used to calibrate the spatial layout of the equipment. Combining with the working environment information of the power equipment, the performance of the equipment under different geographical locations and environmental conditions can be analyzed, providing a spatial reference for subsequent evaluation. These environmental data include external factors such as temperature, humidity, air pressure, wind speed, vibration, and noise, which are usually collected through environmental sensors. These environmental factors have an important impact on the operating state of power equipment. Especially in extreme weather or high-load conditions, changes in environmental data will significantly affect the performance and lifespan of the equipment. Perform multi-layer cross-membrane data coupling to ensure that data from different sources can be effectively integrated. The core of cross-membrane data coupling technology lies in the time and space alignment of heterogeneous data sources and data standardization processing. Specifically, the device operation data, environmental data, and geographical location information are encoded through a unified standard, so that the state of the power equipment, the location, and the external environmental characteristics at each moment can be comprehensively integrated into a unified data structure. This process is usually achieved through multi-dimensional data fusion technologies, such as using data synchronization algorithms, spatio-temporal interpolation technologies, data cleaning and denoising and other processing methods to ensure the reliability and consistency of various types of data. Through such a coupling process, the power equipment information evaluation model can realize multi-dimensional and multi-level data modeling, reflecting the performance of the equipment under different environments and geographical locations. This process can combine the working state of the power equipment with the changes in its surrounding environment, thus providing more accurate and comprehensive data support for subsequent state evaluation. In addition, data coupling not only helps to improve the richness and diversity of data, but also helps to eliminate the deviations between various types of information, thus providing a more accurate and stable data basis for the evaluation model.

[0070] Step S3: Obtain historical power equipment state information; calculate the power equipment state evaluation threshold for the historical power equipment state information, and build a multi-level data processing architecture for the power equipment information evaluation model based on the power equipment state evaluation threshold, where the multi-level data processing architecture includes a horizontal hierarchical architecture of power equipment and a vertical hierarchical architecture of power equipment;

[0071] In the embodiments of the present invention, by obtaining historical power equipment status information, this process relies on large-scale historical data collection technologies, including equipment status monitoring systems, data recording devices, and storage systems. Historical status information usually covers various operation data of the equipment over a past period of time, such as the operating current, voltage, temperature, vibration, load, etc. of the equipment. These data can be obtained through long-term deployed sensors, monitoring devices, or historical log data. By analyzing these historical data, it can provide an important baseline reference for the health management of power equipment, thus providing effective support for subsequent status assessment and prediction models. After obtaining the historical equipment status information, a power equipment status assessment threshold calculation method is used to analyze these historical data to obtain key assessment thresholds for subsequent evaluation. Specifically, the status assessment thresholds usually include statistical indicators such as the maximum, minimum, mean, and fluctuation range of equipment operation parameters, and these thresholds can effectively reflect the normal operation range of the equipment. Common technical means include data-driven machine learning algorithms (such as clustering analysis, anomaly detection, regression analysis, etc.). By deeply mining the historical status information, the threshold standards in different operation states are calculated. For example, statistical methods such as box plots, standard deviations, and quantile analysis can be used to calculate the boundaries between normal and abnormal states, or dynamic thresholds can be calculated based on the change trends of historical data, so as to adapt to the load characteristics and environmental factors of the equipment changing over time. Then, using the calculated power equipment status assessment thresholds above, a multi-level data processing architecture is built. The design purpose of the multi-level data processing architecture is to systematically and hierarchically process the multi-dimensional information of power equipment, so that the data processing at each layer can perform specialized analysis for different levels of features. At this stage, a hierarchical data processing strategy is adopted, including preprocessing steps such as data cleaning, feature selection, noise reduction, and data aggregation, to ensure that the data input into the evaluation model is accurate, noise-free, and representative. By performing operations such as normalization and standardization on the data and combining the evaluation thresholds for model input, the accuracy and reliability of the evaluation model can be effectively improved.

[0072] Step S4: Perform equipment health management prediction according to the multi-level data processing architecture to obtain the power equipment health management prediction value; conduct information fusion and comparative analysis on the power equipment health management prediction value to generate a power equipment status information fusion report.

[0073] In the embodiments of the present invention, device health management prediction is performed through a multi-level data processing architecture. This process involves comprehensively applying the multi-level data processing framework constructed from the aforementioned information such as historical data, real-time data, and status evaluation thresholds. The role of the multi-level data processing architecture is to deeply mine and analyze different operating states of power equipment, covering data levels from basic equipment operation data to complex dimensions such as time-space, environment, thermal, and vibration. For each level of data, different analysis strategies are adopted for processing. For example, trend prediction is performed on data such as the load, temperature, and vibration of power equipment, and potential health risks are identified through time series modeling or machine learning methods. Common health management prediction methods include algorithms such as regression analysis, neural networks (such as LSTM), and support vector machines (SVM), which can extract valuable information from multi-dimensional data and predict the future operating state and health condition of the device. Based on these predicted values, a quantitative prediction of the device health state can be obtained, providing decision support for device management. Subsequently, information fusion and comparative analysis are performed on the power device health management prediction values. Information fusion technology can comprehensively evaluate the health state of the device more comprehensively and accurately by integrating multiple prediction results from different data sources. At this stage, first, the prediction data from multiple sources need to be compared and analyzed, including data such as current, voltage, and temperature collected by different sensors, as well as health prediction data obtained based on the device status evaluation model. To ensure the accuracy and reliability of the prediction results, information fusion algorithms such as weighted average, Bayesian inference, and Kalman filtering are usually adopted, and the final comprehensive health prediction value is obtained through weighted calculation of various prediction results. At this time, information fusion can not only eliminate the deviation of a single prediction result but also improve the robustness of the prediction, ensuring the effectiveness of the finally output health management prediction value in various scenarios. Finally, based on the fused power device health management prediction data, a power device status information fusion report is generated. The process of generating the report usually relies on data visualization technology, report automatic generation tools, and rule-based report generation algorithms. At this stage, by combining the fused prediction results with information such as historical device data, device status thresholds, and health management strategies, through data visualization means, the current state of the device, predicted health trend, warning results, etc. can be intuitively displayed, providing specific operation suggestions and decision-making basis for device maintenance personnel. The report content not only includes a comprehensive evaluation of the device health condition but also involves key information such as fault warning, health index change trend, and remaining service life of the device, facilitating timely intervention and optimized maintenance of the device by relevant personnel.

[0074] Preferably, the construction of the heterogeneous fusion data set of power equipment in step S1 includes:

[0075] Collecting power equipment current data through a current sensor and collecting power equipment temperature data through a temperature sensor;

[0076] Fuse the current data and temperature data of the power equipment in a 1kHz time series to obtain time series signal data; perform 16-bit entropy evaluation based on the time series signal data, screen out the effective information for information entropy calculation, and obtain the information entropy evaluation criterion.

[0077] Remove the noise in different frequency bands from the time series signal data, where the noise removal range is 0.1 - 500Hz, and generate the preliminary time series frequency band noise reduction data; perform entropy value wavelet basis denoising on the preliminary time series frequency band noise reduction data based on the information entropy evaluation criterion to obtain the processed power time series signal data.

[0078] Perform preliminary spatial feature convolution on the infrared thermal image data, and distinguish hot spots and thermal gradients to obtain the infrared thermal image topological features, where the infrared thermal image topological features include local thermal anomalies and global thermal anomalies.

[0079] Perform spatial feature reconstruction with 95% accuracy on the local thermal anomalies to obtain local spatial coding data.

[0080] Perform analysis from the perspective of heat distribution on the global thermal anomalies to obtain global spatial coding data.

[0081] Perform 1:1 spatial heat topological information fusion on the local thermal anomalies and global thermal anomalies to generate power equipment thermal anomaly data.

[0082] Perform cross-membrane state association with a correlation window of 5s on the processed power time series signal data and the power equipment thermal anomaly data to obtain the heterogeneous fusion data set of the power equipment.

[0083] In the embodiments of the present invention, current data of power equipment is collected through a current sensor, and temperature data of the equipment is collected through a temperature sensor, providing basic input for subsequent data fusion and analysis. To achieve accurate monitoring of the state of power equipment, first, the current data and temperature data are subjected to time-series fusion at 1 kHz. This high sampling rate ensures precise capture of the dynamic behavior of the equipment in real time. The processing of time-series data is the core of subsequent analysis. Subsequently, for the time-series signal data of power equipment, 16-bit entropy evaluation is performed. By calculating the information entropy, the data part containing important information is screened out to form an information entropy evaluation criterion. As a tool for measuring data complexity and uncertainty, information entropy can effectively evaluate the effective information in the signal, remove redundant or useless data, and provide a more concise and efficient data set for subsequent refined processing. In terms of noise removal, to eliminate the noise in the time-series signal caused by environmental interference or other external factors, a band noise removal technique is adopted. The range of noise removal is between 0.1 Hz and 500 Hz, ensuring that the removal of low-frequency and high-frequency noise does not affect the key features of the signal. The signal data after preliminary noise reduction is further subjected to entropy value wavelet basis denoising processing based on the information entropy standard to further remove noise components and optimize data quality, obtaining clear processed data of the power time-series signal. In addition, to obtain the thermal anomaly information of the equipment, infrared thermal image data is used. Through preliminary spatial feature convolution, features of the thermal image are extracted to distinguish hot spots from thermal gradients, thereby obtaining the spatial distribution characteristics of thermal anomalies. In this process, local thermal anomalies and global thermal anomalies become the key analysis objects. By reconstructing the spatial features of local thermal anomalies with 95% accuracy, more accurate local spatial coding data can be obtained, while analyzing the global thermal anomalies from the perspective of thermal distribution can provide global spatial coding data to help reveal the thermal behavior of the entire equipment. Finally, the spatial topology information of local and global thermal anomalies is fused 1:1 to generate thermal anomaly data of power equipment. This thermal anomaly data is fused with the processed data of the power time-series signal through 5-second cross-modal state association, thereby obtaining a more comprehensive heterogeneous fusion data set of power equipment.

[0084] Preferably, the multi-layer cross-modal data coupling in step S2 includes:

[0085] Create an equipment node adjacency matrix based on the location of the power equipment, and define the connection weight using a physical distance of 0 m - 100 km to obtain the topological feature data of the power equipment node;

[0086] Couple the environmental sensitivity coefficient of the topological graph of the power equipment node with the environmental data of the power equipment to obtain environment-equipment coupling data; construct a coupling matrix for the environment-equipment coupling data to obtain an environment-equipment coupling feature matrix;

[0087] Locate the harmonic components in the processed data of power time-series signals, and conduct amplitude-phase distribution correlation insulation degradation analysis. The amplitude range is usually 0 to 1000V, and the phase angle is between 0° and 360°, to obtain the power time-series feature tensor;

[0088] Divide the high-temperature casing temperature rise area according to the thermal anomaly data of power equipment, and conduct thermal propagation path inference from 0m to 50m to obtain thermal anomaly map data.

[0089] In the embodiments of the present invention, the adjacency matrix of device nodes is constructed based on the location information of power devices. The role of this matrix is to provide a mathematical representation for the mutual relationships between power device nodes. The adjacency matrix is a basic tool in graph theory, through which the connectivity and communication paths between devices can be reflected. Further, connection weights are defined based on the physical distances between devices. These weights are quantitative indicators of the intensity or importance of the mutual influence between device nodes. The calculation of the weights is based on the physical space distribution relationship between devices, and the Euclidean distance or other spatial distance measurement methods are adopted, so that the adjacency matrix better conforms to the actual connection relationship between devices. Through this process, the topological feature data of power device nodes can be obtained, revealing the structural relationships between devices and the influence of devices on the power network. The environmental data of power devices is used to couple the environmental sensitivity coefficient to the topological graph of power device nodes, generating an environment-device coupling feature matrix. The environmental data includes factors such as temperature, humidity, air pressure, and wind speed, which directly affect the operating state and failure probability of devices. Through the coupling of the environmental sensitivity coefficient, the potential impact of environmental factors on device operation can be quantified, thereby further optimizing the state evaluation model of devices. This coupling process can accurately capture the impact of environmental changes on the working state of devices and provide more comprehensive reference data for the subsequent health assessment of power devices. For the power time-series signal processing data, harmonic component localization is performed. Harmonic components refer to the frequency components in power signals that are inconsistent with the fundamental frequency, usually representing non-linear loads or device anomalies in the power system. When the time-series signal data undergoes harmonic component localization, the spectral information related to device failures can be extracted, and the distribution correlation analysis of amplitude and phase can be carried out, especially the correlation analysis with insulation deterioration. Insulation deterioration is usually related to the amplitude change and phase shift of harmonic components. Through this analysis, the insulation condition of the device can be revealed, providing key diagnostic indicators for the health state assessment of the device. Finally, all the harmonic analysis results are integrated into a power time-series feature tensor, forming a high-dimensional data structure for multi-dimensional and multi-angle analysis of the operating state of the device. Finally, for the thermal anomaly data of power devices, the high-temperature bushing temperature rise area is divided, and the thermal anomaly is deeply analyzed through thermal propagation path inference. The division of the high-temperature bushing temperature rise area can accurately identify the hot areas exceeding 150°C, which are usually high-risk areas for device failures. The thermal propagation path inference is based on the law of heat conduction, analyzing the heat propagation from the device heat source to the environment or adjacent devices and inferring the heat distribution process.

[0090] Preferably, the construction of the power device information evaluation model in step S2 includes:

[0091] Construct an environment-load joint index from the environment-device coupling feature matrix and the topological feature data of power device nodes, perform vector embedding in combination with the preset model loss parameters, and finally construct an interaction layer to obtain the power device feature interaction layer;

[0092] Capture the temporal evolution law of the power time series feature tensor, introduce the clustering heat anomaly map data for collaborative perception, and finally construct an evaluation layer to obtain the power device status evaluation layer;

[0093] Construct an information evaluation model from the power device feature interaction layer and the power device status evaluation layer to obtain the power device information evaluation model.

[0094] In the embodiments of the present invention, by constructing an environment-load combined index and combining the operating load of power equipment and environmental impact factors, the impact of environmental factors on the load of power equipment can be quantified, and a comprehensive performance evaluation standard can be formed. On this basis, the preset model loss parameters are used to perform vector embedding on these features. This process can effectively reduce the complexity of data and retain key information features by mapping high-dimensional features to a low-dimensional space. The technical means of vector embedding compresses features by utilizing the properties of the embedding space, and then extracts the key operating modes and characteristics of power equipment, thereby obtaining the feature interaction layer of power equipment. This layer optimizes the collaborative working state between devices by mapping the complex relationships between devices and between the environment and devices, and effectively supports subsequent state evaluation and health prediction. For the processing of power time series signals, a method of capturing time series evolution laws is used. This method identifies the operating modes, state changes, and potential trends of power equipment at different times by analyzing the time series feature tensor. Capturing the time series evolution law can reveal the change characteristics of the device in the time dimension and capture the periodic, sudden, or long-term change trends during the operation of the device. This process provides dynamic information in the time series dimension for the health status evaluation of the device through in-depth analysis of time series data. In addition, a clustering method introducing a graph attention mechanism can weight key time points in time series features and improve the perception ability of key thermal anomaly map data. The graph attention mechanism effectively captures abnormal signals such as temperature or current during the operation of the device by adaptively learning the degree of association between different devices or data points, especially focusing on the thermal anomaly area. This collaborative perception strategy models the mutual relationships and influences between devices through a graph model, providing a more accurate dynamic view for the state evaluation of power equipment, thereby obtaining the power equipment state evaluation layer. The power equipment feature interaction layer and the power equipment state evaluation layer are effectively fused to construct an information evaluation model. At this stage, the model combines the key device behavior patterns from the device feature interaction layer and the health evaluation indicators from the state evaluation layer through multi-level data integration to form an all-round power equipment information evaluation framework. The construction of the information evaluation model optimizes the health management ability of the device in multiple dimensions and can comprehensively evaluate the device state, predict faults, and issue risk warnings based on real-time data. This process can accurately evaluate the long-term health trend of the device and provide support for decision-making by comprehensively considering device characteristics, operating environment, historical data, and real-time monitoring data. Finally, through this multi-level evaluation system, the system can provide a high-precision and high-reliability power equipment health management solution.

[0095] As an example of the present invention, with reference to Figure 2 as shown, in this example, step S3 includes:

[0096] Step S31: Obtain historical power equipment state information;

[0097] Step S32: Calculate the power equipment status evaluation threshold for the historical power equipment status information to obtain the power equipment status evaluation threshold;

[0098] Step S33: Based on the power equipment status evaluation threshold, construct a horizontal and vertical hierarchical architecture for the power equipment information evaluation model to obtain a multi-level data processing architecture.

[0099] In the embodiment of the present invention, through the deployed sensors and data acquisition devices, the operation data and status information of the equipment are obtained in real time. These data include various monitoring parameters such as the operation time, current, voltage, temperature, vibration, etc. of the power equipment, which can reflect the characteristic behaviors of the power equipment in different working states. By cleaning, denoising, and normalizing these historical data, the quality and consistency of the data are ensured, laying a solid foundation for subsequent analysis. The system calculates the power equipment status evaluation threshold for the historical power equipment status information, mainly using statistical and machine learning methods. According to the normal operation and fault state characteristics of the power equipment in the historical data, the thresholds of various key indicators are calculated. These thresholds are obtained through statistical analysis of the data distribution (such as mean, standard deviation, quantile, etc.), as well as learning based on regression models or classification models, to obtain a reasonable power equipment status evaluation standard. These evaluation thresholds not only consider the physical properties of the equipment itself, but also combine the environmental change factors in the actual operation of the equipment (such as load fluctuations, temperature changes, etc.) to ensure that the evaluation results can accurately reflect the health status of the equipment. Based on the power equipment status evaluation threshold, the system further constructs the horizontal and vertical hierarchical architecture of the power equipment information evaluation model. The construction of the horizontal hierarchical architecture takes into account the differences between different equipment. By comparing the operation states of different power equipment under the same or similar working conditions, the equipment can be grouped and classified for management. This process forms different levels of equipment status evaluation standards by constructing an adjacency matrix between equipment and using machine learning algorithms (such as K-means clustering, decision trees, etc.) to divide the working states of the equipment. The vertical hierarchical architecture mainly focuses on the health assessment of a single equipment at different life cycle stages, and combines factors such as the service time and operation times of the equipment to dynamically evaluate the health status of the equipment at different usage stages.

[0100] Preferably, step S32 includes the following steps:

[0101] Step S321: Screen the key indicators for the historical power equipment status information based on the power equipment heterogeneous fusion data set to obtain the power equipment strongly related indicators;

[0102] Step S322: Conduct quartile interval analysis on the strongly correlated indicators of power equipment, and mark the amplitude and duration of abnormal states to obtain the abnormal mark data of power equipment; conduct analysis on the strongly correlated laws of power data for the abnormal mark data of power equipment to obtain the strongly correlated distribution law data of power.

[0103] Step S323: Determine the warning line based on the normal distribution for the strongly correlated distribution law data of power, and calculate the static threshold for the strongly correlated warning data of power to obtain the static threshold for power state assessment; conduct analysis on the decay tolerance setting for the equipment service life of the strongly correlated distribution law data of power, and conduct dynamic threshold evaluation on the decay data of power equipment to obtain the dynamic threshold for power state assessment.

[0104] Step S324: Perform threshold weight merging processing on the static threshold for power state assessment and the dynamic threshold for power state assessment to obtain the threshold for power equipment state assessment.

[0105] In the embodiments of the present invention, key indicators are screened for historical power equipment status information based on a heterogeneous fusion dataset of power equipment. In this process, the system integrates data from different sensors and monitoring devices, such as various information like current, voltage, temperature, vibration, load, etc., to construct a heterogeneous dataset of power equipment. The heterogeneous fusion dataset integrates measurement data from multiple data sources, and through feature engineering techniques, key indicators that are most closely related to the equipment operation status are selected, such as equipment load, temperature change, power fluctuation, etc. Then, using statistical methods and correlation analysis (such as Pearson correlation coefficient or Spearman rank correlation analysis), indicators that are strongly correlated with the equipment health status are screened, thereby effectively improving the accuracy of subsequent analysis. Based on the strongly correlated indicators of power equipment screened out, the system conducts box plot interquartile range analysis on them and marks the amplitude and duration of abnormal states. The box plot visually displays the data distribution to identify outliers or extreme points, and can intuitively reflect the central tendency, dispersion degree, and abnormal fluctuation range of each indicator. Marking the amplitude and duration of abnormal states helps capture the abnormal behaviors of the equipment at different time periods and further conduct distribution pattern analysis. Through this analysis, the distribution pattern of abnormal states of power equipment under various operating conditions can be obtained, providing basic data for subsequent threshold setting and early warning. The system determines the early warning line based on the normal distribution of the strongly correlated distribution pattern data of power equipment and calculates the static threshold. This process first conducts a distribution analysis on the historical data of the key indicators of power equipment to determine whether it conforms to the normal distribution or other statistical distributions. Through the fitting of the normal distribution, the normal operation range of the equipment is calculated. Based on this range, the system determines a static early warning line, that is, when the equipment operation parameters exceed this early warning line, the system will trigger an alarm. In addition, for power equipment during long-term service, considering the natural aging and performance decline of the equipment, the system also conducts attenuation tolerance analysis based on the service life of the equipment. This analysis considers the performance degradation of the equipment in different life cycles, and by establishing a dynamic threshold model, evaluates the maximum deviation value that the equipment can tolerate during the decline process, thereby providing a basis for dynamic state assessment. The system combines and jointly analyzes the static threshold and the dynamic threshold through weight merging. By the way of weight merging, combining the advantages of the static and dynamic thresholds, a comprehensive power equipment status assessment threshold is obtained. This threshold can not only reflect the normal operation range of the equipment but also consider the change characteristics of the equipment in different life cycles, making the assessment result more practical and accurate.

[0106] Preferably, step S33 includes the following steps:

[0107] Step S331: Conduct single - device over - limit frequency and anomaly correlation analysis on the power equipment feature interaction layer to obtain the single - device anomaly analysis result; perform hierarchical architecture division processing based on the single - device anomaly analysis result to obtain the horizontal hierarchical architecture of power equipment;

[0108] Step S332: Conduct logic determination on the continuous deterioration of multiple indicators of the device group in the power equipment status evaluation layer to obtain the device - group deterioration trend data; perform device - group status clustering analysis on the device - group deterioration trend data to generate device - group clustering analysis result data; perform vertical health assessment architecture division processing on the device - group clustering analysis result data to obtain the vertical hierarchical architecture of power equipment;

[0109] Step S333: Based on the power equipment status evaluation threshold, build a multi - level data processing architecture for the horizontal and vertical hierarchical architectures of power equipment to obtain a multi - level data processing architecture.

[0110] In the embodiments of the present invention, by analyzing the single-device overlimit frequency and anomaly correlation in the power equipment feature interaction layer, the system can identify and count the abnormal behaviors of a single device in different time periods. By analyzing the overlimit frequency of the device, the fluctuation range of the device under normal working conditions can be found, so as to further judge the abnormal behaviors beyond this range. The anomaly correlation analysis relies on multi-dimensional data, including sensor data such as current, voltage, temperature, vibration, etc., and combines time series analysis and statistical models (such as correlation analysis, regression analysis, etc.) to build the correlation model of the device state. Through this analysis, the abnormal degree and occurrence frequency of the device can be accurately evaluated, and thus provide a basis for the establishment of the hierarchical architecture. On this basis, according to the abnormal analysis results of a single device, the hierarchical architecture division process is carried out to obtain the horizontal hierarchical architecture of the power equipment. By dividing the performance state of the device, this hierarchical architecture can realize the rapid identification and classification of the device state, and effectively improve the efficiency of fault diagnosis and maintenance. The system designs the multi-index continuous deterioration judgment logic for the power equipment group. In this process, the system first analyzes the time series data of multiple key indicators (such as load, temperature, power, etc.) in the equipment group to identify which devices have a continuous deterioration trend. These deterioration trends reflect the performance decline of some devices in the equipment group over a long period of time, thus triggering a warning for the overall health status of the equipment group. The deterioration trend data of the equipment group is determined by applying machine learning algorithms (such as time series prediction, clustering analysis, regression model, etc.), and further conducts the clustering analysis of the equipment group state. Using clustering algorithms (such as K-means, hierarchical clustering, etc.), the equipment group is divided into different categories according to its health status and deterioration trend. According to the clustering analysis results, the system can divide the longitudinal health assessment architecture, and then conduct longitudinal monitoring and early warning of the overall health level of the equipment group. The construction of the longitudinal hierarchical architecture takes into account the health level differences of the equipment group in different dimensions, providing a multi-level perspective for further health management and predictive maintenance. Based on the power equipment state assessment threshold, a multi-level data processing architecture is built for the horizontal and vertical hierarchical architectures of the power equipment. First, by comprehensively applying static and dynamic assessment thresholds, the system can accurately fuse and process data for different hierarchical architectures, and finally obtain a multi-level data processing architecture. This architecture can combine the results of the horizontal and vertical hierarchical architectures for multi-dimensional evaluation and analysis, so as to realize the all-round and three-dimensional health management of the power equipment.

[0111] Preferably, step S4 includes the following steps:

[0112] Step S41: Obtain the real-time device operation data; perform device health management prediction according to the multi-level power equipment state assessment and processing strategy to obtain the power equipment health management prediction value;

[0113] Step S42: Analyze the real-time equipment operation data and the power equipment health management prediction value to obtain power equipment health status comparison data;

[0114] Step S43: Perform report aggregation based on the power equipment health status comparison data to obtain a power equipment status information fusion report.

[0115] In the embodiment of the present invention, real-time data from power equipment is obtained through real-time equipment operation data acquisition technology, and these data usually include various sensor information such as current, voltage, temperature, vibration, etc. The acquisition of real-time data needs to rely on the Internet of Things (IoT) technology, and various sensors and communication modules are used to transmit the equipment operation status to the data center for processing in a timely manner. On this basis, based on the multi-level power equipment status evaluation and processing strategy, the system analyzes and evaluates the equipment status in multiple dimensions, and applies technologies such as time series analysis, regression model, and classification algorithm to predict the health management of the equipment operation status. The predicted value can reflect the health status of the equipment in the future, thereby providing a scientific basis for equipment maintenance and fault warning. This process involves a comprehensive analysis of historical data, real-time data, and environmental data, combined with machine learning algorithms (such as neural networks, support vector machines, etc.) to train the equipment health prediction model, and finally obtain an accurate power equipment health management prediction value. Real-time equipment operation data and power equipment health management prediction value are used for status information analysis. This step uses data fusion technology to compare and analyze different data sources (such as real-time sensor data and health prediction data) to identify the health status of the equipment. To achieve this goal, the system usually adopts data comparison and analysis methods, such as time series comparison based on similarity analysis, distribution comparison based on statistical methods, etc. Through these methods, the system can calculate the health status comparison data of the equipment, thereby finding the difference between the actual operation data and the predicted results, and further evaluating whether the equipment is in normal operation or about to fail. The core technology of status information analysis is to mine the potential correlation between data through algorithms, reveal the performance differences of equipment under different working conditions, and provide a basis for subsequent maintenance decisions. Based on the health status comparison data of power equipment, the system further constructs a summary report to generate a fusion report of power equipment status information. This process involves integrating the multi-dimensional data of equipment status, using data aggregation and report generation technology to integrate the health status comparison data with other related data (such as equipment historical fault records, maintenance records, etc.), and presenting the health status of the equipment through information visualization technology (such as charts, dashboards, etc.). The report not only shows the current health status of the equipment, but also provides future health trends and potential risk assessments of the equipment.

[0116] Preferably, step S42 includes the following steps:

[0117] Step S421: Perform a trend similarity analysis on the real-time device operation data and the power equipment health management prediction values to obtain trend similarity analysis data;

[0118] Step S422: Perform abnormal pattern clustering based on the trend similarity analysis data, and conduct root cause correlation analysis to obtain power equipment status correlation data;

[0119] Step S423: Perform confidence comparison feedback based on the power equipment status correlation data to obtain power equipment health status comparison data.

[0120] In the embodiment of the present invention, by performing a trend similarity analysis on the real-time device operation data and the power equipment health management prediction values, the system first needs to align and preprocess the data from different sources to ensure the temporal consistency of the data. The trend similarity analysis mainly uses time series similarity measurement methods, such as dynamic time warping (DTW) or correlation coefficient and other techniques, to analyze the similarity between the real-time data and the health management prediction values. This process judges the conformity degree between the real-time operation data and the predicted health status by calculating the trend change of the two groups of data in the time dimension, and further reveals the coincidence degree between the change trend of the device operation status and the prediction value. Using this method, the deviation between the device operation status and the predicted health management model can be identified, providing basic data for subsequent anomaly detection and prediction. Perform clustering analysis of abnormal patterns based on the trend similarity analysis data. The clustering analysis method can divide the time series data of the device status into 2 to 10 categories, and each category represents a potential abnormal pattern. Common clustering techniques include K-means clustering, hierarchical clustering, and DBSCAN, etc. Through these methods, the system can classify the device status data at different time points according to their similarity, so as to identify different types of abnormal behaviors. On the basis of clustering, the root cause correlation analysis further explores the relationship between the device abnormal pattern and the root cause of the device failure. Usually, causal relationship modeling, Bayesian network or data mining techniques, such as decision tree, association rule mining and other methods, are used to determine the key factors leading to the device status anomaly. Through this analysis, it can help the device management system accurately identify the specific reasons for the device failure and provide decision support for subsequent maintenance and optimization. Perform confidence comparison feedback based on the power equipment status correlation data to further verify the health status of the device. The confidence comparison feedback is to compare the device status correlation data with the historical health data, and calculate its credibility and accuracy to obtain the device health status comparison data. This process usually uses statistical inference techniques, such as Bayesian inference method, to calculate the confidence interval and probability of the device current state, so as to evaluate the reliability and accuracy of the state prediction.

[0121] In this specification, a power equipment status information fusion system is provided for performing the above-mentioned power equipment status information fusion method. The power equipment status information fusion system includes:

[0122] A sensor data acquisition and fusion module, configured to obtain time-series signal data and infrared thermal image data, and construct a heterogeneous fusion data set of power equipment;

[0123] A sensor data acquisition and fusion module, configured to obtain the location of the power equipment and the power equipment environment data, perform multi-layer transmembrane data coupling in combination with the heterogeneous fusion data set of the power equipment, and construct a power equipment information evaluation model;

[0124] A historical data analysis and evaluation model optimization module, configured to obtain historical power equipment status information; calculate the power equipment status evaluation threshold for the historical power equipment status information, and build a multi-level data processing architecture for the power equipment information evaluation model based on the power equipment status evaluation threshold;

[0125] A device health management prediction and reporting module, configured to perform device health management prediction according to the multi-level data processing architecture to obtain a power equipment health management prediction value; perform information fusion and comparative analysis on the power equipment health management prediction value to generate a power equipment status information fusion report.

[0126] The beneficial effects of the present invention are as follows. Through the comprehensive evaluation and health management prediction of the status of power equipment, a system architecture based on multi-source data fusion is proposed, which solves the deficiencies in data acquisition and fusion of traditional power equipment monitoring systems. First, by deploying sensors, time-series signal data and infrared thermal image data are obtained in real time, and combined with the heterogeneous data sets of power equipment for fusion, providing a multi-dimensional data basis for the status evaluation of power equipment. These data not only include the dynamic operation information of the equipment, but also integrate environmental and thermal image information, providing rich data support for subsequent analysis. Second, by obtaining the location information and environmental data of power equipment and performing multi-layer transmembrane data coupling with the previous heterogeneous fusion data set, the power equipment information evaluation model can simultaneously consider the mutual influence between the equipment operation status and external environmental changes, thereby improving the accuracy and comprehensiveness of the evaluation model. Further, through the acquisition and analysis of historical power equipment status information and combined with the calculation of evaluation thresholds, a multi-level data processing architecture is constructed, enabling the model to not only perform real-time status evaluation, but also predict the long-term health status of the equipment and conduct risk assessment. Finally, through the fusion and comparative analysis of the health management prediction values of power equipment, a power equipment status information fusion report is generated, providing more accurate equipment health status and warning information for decision-makers. Generally speaking, the present invention can comprehensively improve the real-time performance, accuracy and warning ability of the power equipment monitoring system by effectively fusing and deeply analyzing multi-source data, providing strong support for the operation safety and maintenance decision-making of power equipment. Therefore, through the construction of multi-source data fusion and cross-level evaluation models, the present invention solves the problems of information islands and prediction accuracy in traditional power equipment monitoring systems, and improves the health management ability and fault warning efficiency of power equipment.

[0127] Therefore, in any aspect, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, it is intended to cover all changes falling within the meaning and scope of the equivalent elements of the application documents within the present invention.

[0128] The above are only the specific embodiments of the present invention, enabling those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A method for fusing power equipment status information, characterized in that It includes the following steps: Step S1: Obtain time-series signal data and infrared thermal image data, and construct a heterogeneous fusion dataset of power equipment; Step S2: Obtain the location of power equipment and power equipment environment data, perform multi-layer cross-modal data coupling in combination with the heterogeneous fusion dataset of power equipment, and construct a power equipment information evaluation model based on the multi-layer cross-modal data coupling; Step S3: Obtain historical power equipment status information; calculate the power equipment status evaluation threshold for the historical power equipment status information, and build a multi-level data processing architecture for the power equipment information evaluation model based on the power equipment status evaluation threshold, where the multi-level data processing architecture includes a horizontal hierarchical architecture and a vertical hierarchical architecture of power equipment; Step S4: Perform equipment health management prediction according to the multi-level data processing architecture to obtain a power equipment health management prediction value; Conduct information fusion comparison and analysis on the power equipment health management prediction value to generate a power equipment status information fusion report.

2. The method for fusing power equipment status information according to claim 1, characterized in that The construction of the heterogeneous fusion dataset of power equipment described in Step S1 includes: Collect power equipment current data through a current sensor and collect power equipment temperature data through a temperature sensor; Perform 1kHz time-series fusion on the power equipment current data and the power equipment temperature data to obtain time-series signal data; perform 16-bit entropy evaluation on the time-series signal data, and screen out valid information for information entropy calculation to obtain an information entropy evaluation criterion; Remove noise in different frequency bands from the time-series signal data, where the noise removal range is 0.1 - 500Hz, to generate preliminarily frequency-band denoised time-series data; perform entropy value wavelet basis denoising on the preliminarily frequency-band denoised time-series data based on the information entropy evaluation criterion to obtain power time-series signal processing data; Perform preliminary spatial feature convolution on the infrared thermal image data, and distinguish hot spots and thermal gradients to obtain infrared thermal image topological features, where the infrared thermal image topological features include local thermal anomalies and global thermal anomalies; Perform 95% accuracy spatial feature reconstruction on the local thermal anomaly to obtain local spatial coding data; Perform an analysis of the global thermal anomaly from the perspective of heat distribution to obtain global spatial coding data; Perform 1:1 spatial heat topological information fusion on the local thermal anomaly and the global thermal anomaly to generate power equipment thermal anomaly data; Perform 5s cross-modal state association in an associated window on the power time-series signal processing data and the power equipment thermal anomaly data to obtain a heterogeneous fusion dataset of power equipment.

3. The method for fusing power equipment status information according to claim 1, wherein The multi-layer cross-modal data coupling described in Step S2 includes: Create an equipment node adjacency matrix based on the location of power equipment, and define the connection weight using a physical distance of 0m - 100km to obtain power equipment node topological feature data; Couple the environmental sensitivity coefficient of the power equipment node topological graph using the power equipment environment data to obtain environment-equipment coupling data; construct a coupling matrix for the environment-equipment coupling data to obtain an environment-equipment coupling feature matrix; Locate the harmonic components of the power time-series signal processing data, and perform amplitude-phase distribution correlation insulation degradation analysis, where the amplitude range is usually 0 - 1000V and the phase angle is 0° - 360°, to obtain a power time-series feature tensor; Based on the thermal anomaly data of power equipment, the temperature rise area of the bushing at 150°C is divided, and the heat transfer path from 0m to 50m is inferred to obtain the thermal anomaly map data.

4. The method for fusing power equipment status information according to claim 1, wherein The construction of the power equipment information evaluation model described in step S2 includes: Construct an environment-load joint index from the environment-equipment coupling feature matrix and the topological feature data of power equipment nodes, perform vector embedding in combination with the preset model loss parameters, and finally construct an interaction layer to obtain the power equipment feature interaction layer; Capture the time-evolution law of the power time-series feature tensor, introduce the clustered thermal anomaly map data for collaborative perception, and finally construct an evaluation layer to obtain the power equipment state evaluation layer; Construct an information evaluation model from the power equipment feature interaction layer and the power equipment state evaluation layer to obtain the power equipment information evaluation model.

5. The method for fusing power equipment status information according to claim 1, wherein, Step S3 includes the following steps: Step S31: Obtain the historical power equipment state information; Step S32: Calculate the power equipment state evaluation threshold for the historical power equipment state information to obtain the power equipment state evaluation threshold; Step S33: Based on the power equipment state evaluation threshold, construct a horizontal and vertical hierarchical architecture for the power equipment information evaluation model to obtain a multi-level data processing architecture.

6. The method for fusing power equipment status information according to claim 5, characterized in that, Step S32 includes the following steps: Step S321: Screen the key indicators for the historical power equipment state information based on the power equipment heterogeneous fusion data set to obtain the power equipment strongly related indicators; Step S322: Analyze the interquartile range of the box plot for the power equipment strongly related indicators, and mark the abnormal state amplitude and duration to obtain the power equipment abnormal marking data; analyze the strongly related power data law for the power equipment abnormal marking data to obtain the strongly related power distribution law data; Step S323: Determine the warning line for the normal distribution of the strongly related power distribution law data, and calculate the static threshold for the strongly related power warning data to obtain the static threshold for power state evaluation; analyze the attenuation tolerance setting for the equipment service life of the strongly related power distribution law data, and evaluate the dynamic threshold for the power equipment attenuation data to obtain the dynamic threshold for power state evaluation; Step S324: Perform threshold weight merging processing on the static threshold for power state evaluation and the dynamic threshold for power state evaluation to obtain the power equipment state evaluation threshold.

7. The method for fusing power equipment status information according to claim 5, characterized in that, Step S33 includes the following steps: Step S331: Analyze the single equipment overlimit frequency and anomaly relevance for the power equipment feature interaction layer to obtain the single equipment anomaly analysis result; perform hierarchical architecture division processing based on the single equipment anomaly analysis result to obtain the horizontal hierarchical architecture of power equipment; Step S332: Perform logical determination on the continuous deterioration of multiple indicators of the equipment group for the power equipment state evaluation layer to obtain the equipment group deterioration trend data; perform equipment group state clustering analysis on the equipment group deterioration trend data to generate the equipment group clustering analysis result data; perform vertical health evaluation architecture division processing on the equipment group clustering analysis result data to obtain the vertical hierarchical architecture of power equipment; Step S333: Based on the power equipment status evaluation threshold, build a multi-level data processing architecture for the horizontal classification architecture and vertical classification architecture of power equipment to obtain a multi-level data processing architecture.

8. The method for fusing power equipment status information according to claim 1, wherein Step S4 includes the following steps: Step S41: Obtain real-time device operation data; perform device health management prediction according to the multi-level power equipment status evaluation and processing strategy to obtain the power equipment health management prediction value; Step S42: Analyze the status information of the real-time device operation data and the power equipment health management prediction value to obtain the power equipment health status comparison data; Step S43: Based on the power equipment health status comparison data, conduct report summarization to obtain the power equipment status information fusion report.

9. The method for fusing power equipment status information according to claim 8, wherein Step S42 includes the following steps: Step S421: Analyze the trend similarity between the real-time device operation data and the power equipment health management prediction value to obtain the trend similarity analysis data; Step S422: Perform abnormal pattern clustering based on the trend similarity analysis data and conduct root cause correlation analysis to obtain the power equipment status correlation data; Step S423: Based on the power equipment status correlation data, conduct confidence level comparison and feedback to obtain the power equipment health status comparison data.

10. A power equipment status information fusion system, characterized in that, For executing the power equipment status information fusion method as described in claim 1, the power equipment status information fusion system includes: A sensor data acquisition and fusion module, which is used to obtain time series signal data and infrared thermal image data, and construct a heterogeneous fusion data set of power equipment; A sensor data acquisition and fusion module, which is used to obtain the power equipment location and power equipment environment data, perform multi-level cross-modal data coupling in combination with the heterogeneous fusion data set of power equipment, and build a power equipment information evaluation model based on the multi-level cross-modal data coupling; A historical data analysis and evaluation model optimization module, which is used to obtain the historical power equipment status information; calculate the power equipment status evaluation threshold for the historical power equipment status information, and build a multi-level data processing architecture for the power equipment information evaluation model based on the power equipment status evaluation threshold, where the multi-level data processing architecture includes a horizontal classification architecture and a vertical classification architecture of power equipment; A device health management prediction and report module, which is used to perform device health management prediction according to the multi-level data processing architecture to obtain the power equipment health management prediction value; conduct information fusion and comparison analysis on the power equipment health management prediction value to generate a power equipment status information fusion report.

Images