Optimization method for anti-interference intelligent diagnosis of power transmission and transformation equipment under complex environment through infrared thermal imaging

Abstract

The application discloses a complex environment power transmission and transformation equipment infrared thermal imaging anti-interference intelligent diagnosis optimization method, and belongs to the technical field of infrared thermal imaging.The complex environment power transmission and transformation equipment infrared thermal imaging anti-interference intelligent diagnosis optimization method comprises the following steps: complex environment infrared thermal imaging image anti-interference pretreatment; multi-scale self-adaptive feature enhancement defect identification algorithm optimization; multi-source operation data fusion modeling; and realization of defect grading, fault positioning and life prediction.The application solves the problems of weak anti-interference ability, insufficient diagnosis precision and single research and judgment dimension in the prior art, and can improve the anti-interference ability of complex environment, accurately identify early hidden heating defects by relying on light-weight feature extraction and dynamic feature enhancement, reduce the false detection rate and temperature measurement error, realize defect grading, millimeter-level positioning and life prediction, complete the pre-risk prediction upgrade, guide differentiated maintenance, and reduce the power grid operation and maintenance cost.

Description

Technical Field

[0001] This invention relates to the field of infrared thermal imaging technology, specifically to an intelligent diagnostic optimization method for anti-interference of infrared thermal imaging of power transmission and transformation equipment under complex environments. Background Technology

[0002] Infrared thermal imaging detection technology, with its advantages of being non-contact, uninterrupted operation, high efficiency, and full coverage, has become a core technology for condition monitoring and fault diagnosis of power transmission and transformation equipment. It is widely used in the daily inspection and condition assessment of substations, transmission lines, and distribution equipment. With the continuous advancement of smart grid construction, infrared intelligent diagnostic technology is gradually replacing traditional manual inspection methods, significantly improving the efficiency of power equipment detection. Several existing patents disclose fault diagnosis solutions for power equipment based on infrared thermal imaging, providing technical support for intelligent detection in the industry.

[0003] Chinese Patent Publication No. CN202512006112.X discloses a method and system for diagnosing thermal faults in power equipment based on infrared images. The method includes: acquiring an infrared image of a covered area and extracting contour features from the infrared image; acquiring a base image corresponding to the covered area and extracting contour features from the base image; correcting the contour features of the infrared image using the contour features from the base image to obtain a corrected infrared image contour; separating the infrared image using the corrected infrared image contour to obtain a separated infrared image; performing basic processing on the separated infrared image; and using the temperature features of the separated infrared image to determine the type of thermal fault in the power equipment.

[0004] The aforementioned patents have technical shortcomings in practical use, such as weak anti-interference capability in complex environments, insufficient diagnostic accuracy, and single judgment dimensions. They are difficult to adapt to the high-precision, multi-dimensional, and forward-looking intelligent diagnostic needs of power transmission and transformation equipment under complex operating conditions of smart grids. Therefore, we propose an optimized method for anti-interference intelligent diagnosis of power transmission and transformation equipment under complex environments using infrared thermal imaging. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent diagnostic optimization method for infrared thermal imaging of power transmission and transformation equipment under complex environments. This method addresses infrared image distortion and temperature drift through layered noise reduction, grayscale compensation, and clutter removal. Utilizing lightweight feature extraction and dynamic feature enhancement, it accurately identifies early-stage latent heating defects, reducing false detection rates and temperature measurement errors. Simultaneously, it aggregates multi-source heterogeneous data to construct a fusion diagnostic model, breaking down data silos and achieving defect classification, millimeter-level localization, and lifespan prediction, thus upgrading pre-emptive risk assessment. Furthermore, the algorithm has low computational overhead, can guide differentiated maintenance, and reduce power grid operation and maintenance costs, solving the problems mentioned in the background section.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an intelligent diagnostic optimization method for anti-interference of infrared thermal imaging in power transmission and transformation equipment under complex environments, comprising the following steps: S1: Real-time acquisition of infrared thermal imaging images and corresponding environmental parameters from relevant devices for anti-interference preprocessing; S2: Employs a multi-scale feature-enhanced infrared defect recognition algorithm to perform feature analysis on the preprocessed standardized infrared image, thereby completing the identification and feature marking of equipment defects; S3: Real-time acquisition and preprocessing of multi-source operation data of power transmission and transformation equipment; construction of multi-source data fusion diagnostic model based on time-series dynamic weight association mechanism. S4: Based on the multi-source data fusion diagnostic model, the identified equipment defects are classified to achieve accurate location of fault points and output the remaining operating life and aging trend of the equipment.

[0007] Preferably, the implementation process of S1 specifically includes: Real-time acquisition of infrared thermal imaging images and corresponding environmental parameters of power transmission and transformation equipment under complex outdoor, enclosed cabinet, and extreme high and low temperature conditions; The acquired raw infrared images are preprocessed frame by frame to output high-fidelity, high-contrast standardized infrared thermal imaging images.

[0008] Preferably, the specific process of preprocessing the acquired raw infrared image frame by frame includes: To address interference from high and low temperatures and high humidity condensation environments: An adaptive bilateral filtering algorithm is used to perform layered noise reduction on the original infrared image. Through a dual screening mechanism of spatial neighborhood constraints and grayscale value range constraints, Gaussian noise, salt-and-pepper noise, and fog diffusion noise caused by complex environments are accurately filtered out while fully preserving the edge contours of equipment defects and temperature gradient features. Combined with real-time environmental temperature and humidity parameters, a pixel-level grayscale dynamic compensation mechanism is established to address image fogging and detail blurring caused by high humidity condensation and low temperature fogging. The image attenuation features are repaired pixel by pixel to restore the true temperature distribution details of the equipment. To address outdoor lighting and background noise heat source interference: an infrared and visible light image registration and comparison algorithm is used to extract the effective target area of ​​the power transmission and transformation equipment and eliminate invalid background heat source interference such as tree obstruction, sunlight reflection, and ground high temperature noise; based on the equipment standard temperature reference library, dynamic correction is performed on the deviation of the equipment surface detection temperature caused by environmental temperature fluctuations to eliminate the impact of environmental temperature drift on temperature measurement accuracy. Output standardized infrared images: Perform contrast adaptive enhancement processing on the calibrated images to improve the image recognition of subtle temperature difference defects in the equipment, and generate standardized infrared thermal imaging images with uniform size, balanced grayscale and clear features.

[0009] Preferably, the specific construction process of the multi-scale feature-enhanced infrared defect recognition algorithm in S2 includes: A four-layer progressive multi-scale feature analysis link is established, which divides the analysis levels according to the order of overall macro structure of equipment, local structure of components, obvious minor defects and micro temperature difference anomalies, and predefines the extraction boundaries of features at different scales. Based on the hierarchical parsing results, the original multi-layer parallel redundant feature extraction logic is simplified, and two types of redundant operation steps, namely cross-scale invalid convolutional alignment and repeated texture calculation, are eliminated. The original layered features output by the lightweight architecture are processed a second time. Based on the intensity of real-time image illumination, clutter, and humidity interference, feature weights are dynamically allocated, and features such as weak temperature difference and latent heat generation are actively gained.

[0010] Preferably, the construction of the multi-scale feature-enhanced infrared defect recognition algorithm further includes: summarizing feature recognition deviation data under complex working conditions in all scenarios, iteratively optimizing the four-layer scale resolution boundary, feature enhancement gain coefficient, and background weight suppression threshold, and continuously correcting the entire set of recognition link parameters.

[0011] Preferably, the specific process of multi-source operational data fusion modeling in S3 includes: Multi-source data classification and acquisition: Simultaneously acquire image dimension data, equipment operation dimension data, environmental condition dimension data, and maintenance history dimension data; Heterogeneous data preprocessing: For numerical operational data, range normalization is used to unify the data dimensions; for image feature data, feature dimensionality reduction is used; and for text-based operational data, label quantization is performed. Interpolation is used to fill in short-term missing data, and the 3σ criterion is used to remove abnormal and sudden sensor data to ensure the effectiveness and consistency of multi-source data. Constructing a spatiotemporal fusion correlation model: Based on the time-series dynamic weight correlation mechanism, we deeply explore the temporal coupling law between equipment operation fluctuations, environmental condition evolution and infrared defect temperature characteristics, construct the correlation mapping rules between image features and equipment operation status, and finally integrate them to form a multi-source data fusion diagnostic model.

[0012] Preferably, the implementation process of S4 specifically includes: Based on a multi-source data fusion diagnostic model, combined with a preset threshold standard for classifying defects in power transmission and transformation equipment, the identified equipment defects are accurately classified. By mapping the pixel coordinates of infrared images and matching them with the three-dimensional topology model of the equipment, the fault location can be accurately pinpointed. Based on the aging patterns of equipment operation, defect development trends, and historical failure data, a prediction model for the remaining service life of equipment is constructed, which dynamically outputs the remaining service life and aging degradation trend of the equipment.

[0013] Preferably, the specific criteria for accurately classifying the identified equipment defects include: A Level 1 minor defect is defined as a local temperature difference of ≤5℃ in the equipment, normal operation under load, no abnormal parameters, no tendency for the defect to expand, and no impact on the normal operation of the equipment. Level 2 general defects are characterized by a local temperature difference of 5℃-15℃ in the equipment, slight fluctuations in equipment operating parameters, slow defect development rate, and no safety hazards in the short term. Level 3 serious defects are characterized by a local temperature difference of 15℃-30℃ in the equipment, a decrease in the equipment's load capacity, abnormal fluctuations in operating parameters, continuous expansion of the defect, and potential safety hazards. A Level 4 emergency defect is defined as a local temperature difference of ≥30℃ in the equipment, severely abnormal equipment operating parameters, and a risk of overheating breakdown and short circuit.

[0014] Preferably, the specific implementation method for achieving precise location of the fault point is as follows: Establish a three-dimensional topological digital model of the power transmission and transformation equipment, and accurately register the pixel coordinates of the preprocessed infrared image with the spatial coordinates of the equipment's three-dimensional model; By combining the coordinates of the feature areas identified by defect identification, the spatial location of the fault point can be completed with a positioning accuracy of millimeters. Simultaneously, the faulty equipment type, faulty component, fault location, and defect range are labeled to generate a visual fault location map.

[0015] Preferably, the specific process of equipment life prediction in S4 includes: Construct an equipment aging characteristic index system: with equipment running time, load cycle count, defect severity level, environmental stress degree and maintenance frequency as core aging indicators, and quantify the impact weight of each indicator on equipment insulation aging, mechanical loss and electrical performance degradation. Establish a lifespan prediction model: Based on the LSTM time series prediction network, combined with historical aging data of equipment, defect development time series data and real-time operating status data, train the remaining lifespan prediction model of the equipment and fit the aging and decay curve of the equipment. Dynamic lifespan correction output: Based on real-time equipment defect classification results, fluctuations in operating parameters, and changes in environmental conditions, the lifespan prediction parameters are dynamically corrected, the remaining operating life of the equipment is updated in real time, and the equipment aging risk points and lifespan degradation warning levels are output.

[0016] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention improves the anti-interference capability in complex environments by using layered adaptive bilateral noise reduction, pixel-level grayscale dynamic compensation, and target area clutter removal to repair condensation and fogging images, offset ambient temperature measurement deviations, and remove invalid background heat sources. Combined with lightweight layered feature extraction and dynamic weight adaptive enhancement mechanism, it makes up for the shortcomings of existing algorithms in weak temperature difference feature extraction, specifically identifying early hidden heating defects. It makes up for the anti-interference shortcomings from two layers: image preprocessing and feature recognition, reducing the false detection rate and temperature measurement error in complex scenes. 2. This application constructs a multi-source data fusion diagnostic model by collecting multi-source data in four dimensions, standardizing heterogeneous data, and dynamically weighting time series data. This model supplements the three categories of data for analysis: equipment operation, environment, and maintenance, breaking down the silos of single image data. Based on this fusion model, it sequentially achieves four-level defect quantitative classification, infrared pixel and three-dimensional topology millimeter-level fault location, and multi-index time series aging life prediction. This addresses the shortcomings of existing technologies that can only determine the presence or absence of faults and cannot conduct refined maintenance analysis, thus upgrading from post-fault detection to pre-risk prediction. 3. This invention reduces redundant computations by using lightweight links, adapts to the low computing power requirements of field edge devices, guides differentiated maintenance through refined hierarchical positioning results, reduces blind power outages and maintenance, lowers grid maintenance costs, and matches the actual needs of forward-looking smart grid maintenance. Attached Figure Description

[0017] Figure 1 This is a flowchart of the intelligent diagnostic optimization method for infrared thermal imaging anti-interference of power transmission and transformation equipment under complex environments, as presented in this invention. Figure 2 This is a flowchart of the fault location and life prediction process of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] To address the shortcomings of existing technologies in practical applications, such as weak anti-interference capabilities in complex environments, insufficient diagnostic accuracy, and limited analytical dimensions, please refer to [link to relevant documentation]. Figures 1-2 This embodiment provides the following technical solution: An intelligent diagnostic optimization method for anti-interference of infrared thermal imaging in power transmission and transformation equipment under complex environments includes the following steps: S1: Real-time acquisition of infrared thermal imaging images and corresponding environmental parameters from relevant devices for anti-interference preprocessing; S2: Employs a multi-scale feature-enhanced infrared defect recognition algorithm to perform feature analysis on the preprocessed standardized infrared image, thereby completing the identification and feature marking of equipment defects; S3: Real-time acquisition of multi-source operation data of power transmission and transformation equipment and preprocessing. Based on the time-series dynamic weight association mechanism, in-depth mining of the time-series coupling law of equipment operation fluctuation, environmental condition evolution and infrared defect temperature characteristics, construction of image feature and equipment operation status association mapping rules, and finally integration to form a multi-source data fusion diagnostic model. S4: Based on a multi-source data fusion diagnostic model and combined with preset defect classification threshold standards for power transmission and transformation equipment, the identified equipment defects are classified into four precise levels: Level 1 minor defects, Level 2 general defects, Level 3 severe defects, and Level 4 emergency defects. By matching the pixel coordinates of infrared images with the three-dimensional topology model of the equipment, the fault location is accurately located at the millimeter level. Based on the aging patterns of equipment operation, defect development trends, and historical failure data, a prediction model for the remaining service life of the equipment is constructed, and the remaining service life and aging degradation trend of the equipment are dynamically output.

[0020] The implementation process of S1 specifically includes: This system acquires real-time infrared thermal images and corresponding environmental parameters of power transmission and transformation equipment under complex outdoor, enclosed cabinet, and extreme high and low temperature conditions. These environmental parameters include ambient temperature, ambient humidity, light intensity, and background clutter and heat source data. The acquired raw infrared images undergo frame-by-frame preprocessing, sequentially performing noise filtering, condensation blur repair, clutter and heat source removal, and temperature deviation correction. Utilizing a comprehensive anti-interference preprocessing mechanism that incorporates layered adaptive noise reduction, pixel-level dynamic grayscale compensation, and precise multi-scene clutter removal, it eliminates image blurring, detail loss, and temperature detection deviations caused by complex environments at their source. This effectively improves the infrared imaging quality and temperature measurement accuracy under complex conditions, outputting high-fidelity, high-contrast standardized infrared thermal images.

[0021] The specific process of preprocessing the acquired raw infrared images frame by frame includes: Addressing interference from high and low temperatures and high humidity condensation environments: An adaptive bilateral filtering algorithm is used to perform layered noise reduction on the raw infrared images. Through a dual screening mechanism of spatial neighborhood constraints and grayscale value range constraints, Gaussian noise, salt-and-pepper noise, and fog-like diffusion noise caused by complex environments are accurately filtered out while fully preserving the edge contours of equipment defects and temperature gradient characteristics. Combined with real-time environmental temperature and humidity parameters, a pixel-level grayscale dynamic compensation mechanism is established to address image fogging and detail blurring caused by high humidity condensation and low temperature fogging. Pixel-by-pixel image attenuation features are repaired to restore the true temperature distribution details of the equipment and avoid imaging distortion caused by environmental humidity. Addressing interference from outdoor lighting and background clutter heat sources: An infrared and visible light image registration and comparison algorithm is used to extract the effective target area of ​​the power transmission and transformation equipment, eliminating invalid background heat source interference such as tree obstruction, sunlight reflection, and ground high-temperature clutter. Based on the equipment's standard temperature reference library, dynamic correction is performed on the deviation of the equipment surface detection temperature caused by environmental temperature fluctuations to eliminate the impact of environmental temperature drift on temperature measurement accuracy. Output standardized infrared images: The corrected images undergo adaptive contrast enhancement processing to improve the image recognition of subtle temperature difference defects in the equipment, generating standardized infrared thermal imaging images with uniform size, balanced grayscale, and clear features, providing high-quality input data for subsequent defect identification algorithms. This step abandons the traditional preprocessing mode of fixed parameter noise reduction and single correction, and can adapt to various complex interference scenarios such as high humidity condensation and fogging, high and low temperature drift, outdoor light reflection, background clutter, and heat sources.

[0022] The specific construction process of the multi-scale feature-enhanced infrared defect recognition algorithm in S2 includes: A four-layer progressive multi-scale feature parsing chain is established, dividing the parsing levels according to the order of overall macroscopic structure of equipment, local structure of components, obvious minor defects, and microscopic temperature anomalies. Extraction boundaries for features at different scales are pre-defined to adapt to the full-scale defect identification needs of power transmission and transformation equipment, providing a hierarchical framework for subsequent feature extraction. Based on the hierarchical parsing results, a lightweight step-by-step feature extraction architecture is built, simplifying the original multi-layer parallel redundant feature extraction logic. Two redundant computational steps—invalid cross-scale convolutional comparison and repetitive texture calculation—are eliminated. The computation chain is compressed without losing feature information at all four scales, reducing overall computational power consumption and ensuring real-time image feature extraction efficiency, adapting to various routine inspection scenarios for power transmission and transformation equipment. The original layered features output by the lightweight architecture undergo secondary processing. Based on the intensity of real-time image illumination, clutter, and humidity interference, feature weights are dynamically allocated. Features with weak temperature differences and low-response defects such as latent heating are actively amplified, while high-response invalid features such as vegetation and ground background are adaptively suppressed, eliminating false anomalies caused by environmental interference. This method abandons the traditional static threshold recognition mode, dynamically adjusting feature enhancement gain and filtering weights to strengthen the recognition capability of defects such as 1-2℃ small temperature differences, latent contact defects, and local insulation aging, effectively reducing the probability of missed and false detections. The equipment temperature measurement error can be stably controlled within ±0.5℃. Scene adaptability parameter iterative closed-loop optimization: Feature recognition deviation data under complex working conditions across all scenarios are summarized, and the four-layer scale parsing boundary, feature enhancement gain coefficient, and background weight suppression threshold are iteratively optimized in reverse. Based on massive complex working condition samples, parameters are continuously iteratively optimized, enabling the algorithm to adapt to different regions, seasons, and equipment types of operation and maintenance scenarios, significantly improving the recognition stability, generalization, and robustness under extreme interference scenarios.

[0023] The specific process of multi-source operational data fusion modeling in S3 includes: Multi-source data classification and acquisition: Simultaneously acquire image dimension data, equipment operation dimension data, environmental condition dimension data, and maintenance history dimension data. Among them, image dimension data includes pre-processed infrared image defect features and temperature distribution data; equipment operation dimension data includes real-time voltage, current, load rate, operating frequency, and equipment temperature rise rate data; environmental condition dimension data includes real-time temperature, humidity, wind speed, light intensity, and rain / snow weather data; and maintenance history dimension data includes equipment commissioning time, maintenance records, defect history, and failure case data.

[0024] Heterogeneous data preprocessing: For numerical operational data, range normalization is used to unify the data dimensions; for image feature data, feature dimensionality reduction is used; and for text-based operational data, label quantization is performed. Interpolation is used to fill in short-term missing data, and the 3σ criterion is used to remove abnormal and sudden sensor data to ensure the effectiveness and consistency of multi-source data.

[0025] A spatiotemporal fusion correlation model is constructed: Based on a time-series dynamic weight correlation mechanism, the model deeply explores the temporal coupling patterns between equipment operation fluctuations, environmental condition evolution, and infrared defect temperature characteristics. It establishes correlation mapping rules between image features and equipment operating status, ultimately integrating them to form a multi-source data fusion diagnostic model. The time-series attention mechanism dynamically assigns weights to multi-source data according to the time series, prioritizing data on sudden equipment operation changes and sudden environmental interference. This effectively captures early latent defect characteristics caused by short-term operating condition changes and sudden environmental stresses, establishing a precise coupling correlation between equipment defect development and operating status and environmental changes. This breaks through the limitations of traditional methods relying solely on single infrared image temperature information for judgment, achieving comprehensive and refined judgment of equipment status.

[0026] The implementation process of S4 specifically includes: Based on a multi-source data fusion diagnostic model, combined with a preset threshold standard for power transmission and transformation equipment defect classification, the identified equipment defects are accurately classified; by mapping the pixel coordinates of infrared images with the three-dimensional topology model of the equipment, the fault location is accurately located; based on the aging patterns of equipment operation, defect development trends, and historical failure data, a prediction model for the remaining service life of the equipment is constructed, and the remaining service life and aging degradation trend of the equipment are dynamically output.

[0027] The specific criteria for accurately classifying identified equipment defects include: Level 1 Minor Defect: Local temperature difference ≤ 5℃, normal load operation, no abnormal parameters, no tendency for the defect to expand, does not affect normal equipment operation, only requires periodic monitoring; Level 2 General Defect: Local temperature difference 5℃-15℃, slight fluctuation in equipment operating parameters, slow defect development rate, no safety hazards in the short term, requires planned maintenance; Level 3 Severe Defect: Local temperature difference 15℃-30℃, decreased equipment load capacity, abnormal fluctuations in operating parameters, continuous expansion of the defect, potential safety hazards, requires time-limited shutdown for maintenance; Level 4 Emergency Defect: Local temperature difference ≥ 30℃, severely abnormal equipment operating parameters, risk of overheating breakdown and short circuit, requires immediate shutdown to prevent safety accidents.

[0028] The specific implementation method for achieving precise fault location is as follows: A three-dimensional topological digital model of the power transmission and transformation equipment is established, and the pixel coordinates of the preprocessed infrared image are accurately registered with the spatial coordinates of the equipment's three-dimensional model. Combined with the feature area coordinates of defect identification, the spatial location of the fault point is completed with a positioning accuracy of millimeters. At the same time, the fault equipment type, fault component, fault location and defect range are labeled to generate a visualized fault location map.

[0029] The specific process for predicting equipment lifespan in S4 includes: A system of equipment aging characteristic indicators is constructed: Core aging indicators include equipment operating time, load cycle count, defect severity level, environmental stress level, and maintenance frequency. The weight of each indicator's impact on equipment insulation aging, mechanical loss, and electrical performance degradation is quantified. A lifespan prediction model is established: Based on an LSTM time-series prediction network, combined with historical equipment aging data, defect development time-series data, and real-time operating status data, the model is trained to predict the remaining lifespan of the equipment and fit the equipment aging degradation curve. Dynamic lifespan correction output: Based on real-time equipment defect classification results, operating parameter fluctuations, and changes in environmental conditions, the lifespan prediction parameters are dynamically corrected, and the remaining operating life of the equipment is updated in real time. Simultaneously, equipment aging risk points and lifespan degradation warning levels are output. This part can predict equipment aging degradation risks in advance, guide maintenance personnel to conduct differentiated maintenance, avoid safety accidents such as equipment overheating breakdown and short-circuit shutdowns, and reduce power grid maintenance costs and equipment failure losses.

[0030] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for intelligent diagnosis and optimization of power transmission and transformation equipment under complex environments using infrared thermal imaging to resist interference, characterized in that... Includes the following steps: S1: Real-time acquisition of infrared thermal imaging images and corresponding environmental parameters from relevant devices for anti-interference preprocessing; S2: Employs a multi-scale feature-enhanced infrared defect recognition algorithm to perform feature analysis on the preprocessed standardized infrared image, thereby completing the identification and feature marking of equipment defects; S3: Real-time acquisition and preprocessing of multi-source operation data of power transmission and transformation equipment; construction of multi-source data fusion diagnostic model based on time-series dynamic weight association mechanism. S4: Based on the multi-source data fusion diagnostic model, the identified equipment defects are classified to achieve accurate location of fault points and output the remaining operating life and aging trend of the equipment.

2. The infrared thermal imaging anti-interference intelligent diagnosis and optimization method for power transmission and transformation equipment under complex environments according to claim 1, characterized in that, The implementation process of S1 specifically includes: Real-time acquisition of infrared thermal imaging images and corresponding environmental parameters of power transmission and transformation equipment under complex outdoor, enclosed cabinet, and extreme high and low temperature conditions; The acquired raw infrared images are preprocessed frame by frame to output high-fidelity, high-contrast standardized infrared thermal imaging images.

3. The infrared thermal imaging anti-interference intelligent diagnosis and optimization method for power transmission and transformation equipment under complex environments according to claim 2, characterized in that, The specific process of preprocessing the acquired raw infrared images frame by frame includes: To address interference from high and low temperatures and high humidity condensation environments: An adaptive bilateral filtering algorithm is used to perform layered noise reduction on the original infrared image. Through a dual screening mechanism of spatial neighborhood constraints and grayscale value range constraints, Gaussian noise, salt-and-pepper noise, and fog diffusion noise caused by complex environments are accurately filtered out while fully preserving the edge contours of equipment defects and temperature gradient features. Combined with real-time environmental temperature and humidity parameters, a pixel-level grayscale dynamic compensation mechanism is established to address image fogging and detail blurring caused by high humidity condensation and low temperature fogging. The image attenuation features are repaired pixel by pixel to restore the true temperature distribution details of the equipment. To address outdoor lighting and background noise heat source interference: an infrared and visible light image registration and comparison algorithm is used to extract the effective target area of ​​the power transmission and transformation equipment and eliminate invalid background heat source interference such as tree obstruction, sunlight reflection, and ground high temperature noise; based on the equipment standard temperature reference library, dynamic correction is performed on the deviation of the equipment surface detection temperature caused by environmental temperature fluctuations to eliminate the impact of environmental temperature drift on temperature measurement accuracy. Output standardized infrared images: Perform contrast adaptive enhancement processing on the calibrated images to improve the image recognition of subtle temperature difference defects in the equipment, and generate standardized infrared thermal imaging images with uniform size, balanced grayscale and clear features.

4. The infrared thermal imaging anti-interference intelligent diagnosis and optimization method for power transmission and transformation equipment under complex environments according to claim 1, characterized in that, The specific construction process of the multi-scale feature-enhanced infrared defect recognition algorithm in S2 includes: A four-layer progressive multi-scale feature analysis link is established, which divides the analysis levels according to the order of overall macro structure of equipment, local structure of components, obvious minor defects and micro temperature difference anomalies, and predefines the extraction boundaries of features at different scales. Based on the hierarchical parsing results, the original multi-layer parallel redundant feature extraction logic is simplified, and two types of redundant operation steps, namely cross-scale invalid convolutional alignment and repeated texture calculation, are eliminated. The original layered features output by the lightweight architecture are processed a second time. Based on the intensity of real-time image illumination, clutter, and humidity interference, feature weights are dynamically allocated, and features such as weak temperature difference and latent heat generation are actively gained.

5. The infrared thermal imaging anti-interference intelligent diagnosis and optimization method for power transmission and transformation equipment under complex environments according to claim 4, characterized in that, The construction of the multi-scale feature-enhanced infrared defect recognition algorithm also includes: summarizing feature recognition deviation data under complex working conditions in all scenarios, iteratively optimizing the four-layer scale parsing boundary, feature enhancement gain coefficient, and background weight suppression threshold, and continuously correcting the entire set of recognition link parameters.

6. The infrared thermal imaging anti-interference intelligent diagnosis and optimization method for power transmission and transformation equipment under complex environments according to claim 1, characterized in that, The specific process of multi-source operational data fusion modeling in S3 includes: Multi-source data classification and acquisition: Simultaneously acquire image dimension data, equipment operation dimension data, environmental condition dimension data, and operation and maintenance history dimension data; Heterogeneous data preprocessing: For numerical operational data, range normalization is used to unify the data dimensions; for image feature data, feature dimensionality reduction is used; and for text-based operational data, label quantization is performed. Interpolation is used to fill in short-term missing data, and the 3σ criterion is used to remove abnormal and sudden sensor data to ensure the effectiveness and consistency of multi-source data. Constructing a spatiotemporal fusion correlation model: Based on the time-series dynamic weight correlation mechanism, we deeply explore the temporal coupling law between equipment operation fluctuations, environmental condition evolution and infrared defect temperature characteristics, construct the correlation mapping rules between image features and equipment operation status, and finally integrate them to form a multi-source data fusion diagnostic model.

7. The infrared thermal imaging anti-interference intelligent diagnosis and optimization method for power transmission and transformation equipment under complex environments according to claim 1, characterized in that, The implementation process of S4 specifically includes: Based on a multi-source data fusion diagnostic model, combined with a preset threshold standard for classifying defects in power transmission and transformation equipment, the identified equipment defects are accurately classified. By mapping the pixel coordinates of infrared images and matching them with the three-dimensional topology model of the equipment, the fault location can be accurately pinpointed. Based on the aging patterns of equipment operation, defect development trends, and historical failure data, a prediction model for the remaining service life of equipment is constructed, which dynamically outputs the remaining service life and aging degradation trend of the equipment.

8. The infrared thermal imaging anti-interference intelligent diagnosis and optimization method for power transmission and transformation equipment under complex environments according to claim 6, characterized in that, The specific criteria for accurately classifying identified equipment defects include: A Level 1 minor defect is defined as a local temperature difference of ≤5℃ in the equipment, normal operation under load, no abnormal parameters, no tendency for the defect to expand, and no impact on the normal operation of the equipment. Level 2 general defects are characterized by a local temperature difference of 5℃-15℃ in the equipment, slight fluctuations in equipment operating parameters, slow defect development rate, and no safety hazards in the short term. Level 3 serious defects are characterized by a local temperature difference of 15℃-30℃ in the equipment, a decrease in the equipment's load capacity, abnormal fluctuations in operating parameters, continuous expansion of the defect, and potential safety hazards. A Level 4 emergency defect is defined as a local temperature difference of ≥30℃ in the equipment, severely abnormal equipment operating parameters, and a risk of overheating breakdown or short circuit.

9. The infrared thermal imaging anti-interference intelligent diagnosis and optimization method for power transmission and transformation equipment in complex environments according to claim 6, characterized in that, The specific implementation method for achieving precise location of the fault point is as follows: Establish a three-dimensional topological digital model of the power transmission and transformation equipment, and accurately register the pixel coordinates of the preprocessed infrared image with the spatial coordinates of the equipment's three-dimensional model; By combining the coordinates of the feature areas identified by defect identification, the spatial location of the fault point can be completed with a positioning accuracy of millimeters. Simultaneously, the faulty equipment type, faulty component, fault location, and defect range are labeled to generate a visual fault location map.

10. The infrared thermal imaging anti-interference intelligent diagnosis and optimization method for power transmission and transformation equipment under complex environments according to claim 1, characterized in that, The specific process for predicting equipment lifespan in S4 includes: Construct an equipment aging characteristic index system: with equipment running time, load cycle count, defect severity level, environmental stress degree and maintenance frequency as core aging indicators, and quantify the impact weight of each indicator on equipment insulation aging, mechanical loss and electrical performance degradation. Establish a lifespan prediction model: Based on the LSTM time series prediction network, combined with historical aging data of equipment, defect development time series data and real-time operating status data, train the remaining lifespan prediction model of the equipment and fit the aging and decay curve of the equipment. Dynamic lifespan correction output: Based on real-time equipment defect classification results, fluctuations in operating parameters, and changes in environmental conditions, the lifespan prediction parameters are dynamically corrected, the remaining operating life of the equipment is updated in real time, and the equipment aging risk points and lifespan degradation warning levels are output.

Images