Electric power equipment overheating positioning method and device based on infrared hyperspectral imaging

Abstract

The invention belongs to the field of power equipment monitoring, and relates to a power equipment overheating positioning method and device based on infrared hyperspectral imaging, and the method comprises the steps: synchronously collecting infrared and hyperspectral multi-modal data, and carrying out the preprocessing of the multi-modal data; based on the preprocessed multi-modal data, complementary features are extracted from the infrared and hyperspectral data, and feature-level fusion is realized through deep learning; based on surface data inversion equipment internal heat source distribution, three-dimensional space positioning is realized; performing overheating type intelligent diagnosis, and distinguishing normal heating and fault heating of the equipment; generating a visual report and triggering graded early warning; and through online learning and model updating, the calculation efficiency and robustness are optimized. The sensitivity and the positioning precision of overheating defect detection are improved, and particularly, the recognition capability of early weak faults is enhanced; the space coordinates and intensity distribution of the internal abnormal heat source can be quantitatively calculated, visual fault position guidance can be provided, and the troubleshooting time is greatly shortened.

Description

Technical Field

[0001] This invention relates to the field of power equipment monitoring technology, and in particular to a method and apparatus for locating overheated power equipment based on infrared hyperspectral imaging. Background Technology

[0002] During operation, power equipment often experiences localized overheating due to excessive load, poor contact, or insulation aging, which can lead to equipment damage or even power grid failure in severe cases. Traditional monitoring methods mainly rely on infrared thermal imaging technology, but it is susceptible to environmental interference and struggles to penetrate the complex structure of equipment to identify internal heat sources. While hyperspectral imaging provides rich spectral features, it lacks real-time temperature monitoring capabilities. Summary of the Invention

[0003] To address the aforementioned technical problems, this invention provides a method for locating overheated power equipment based on infrared hyperspectral imaging, employing the following technical solution, including the following steps:

[0004] Simultaneously acquire infrared and hyperspectral multimodal data, and preprocess the multimodal data;

[0005] Based on the preprocessed multimodal data, complementary features are extracted from infrared and hyperspectral data, and feature-level fusion is achieved through deep learning;

[0006] Based on surface data, the internal heat source distribution of the equipment is inverted to achieve three-dimensional spatial positioning;

[0007] Perform intelligent diagnosis of overheating types to distinguish between normal equipment overheating and fault overheating;

[0008] Generate visual reports and trigger tiered alerts to guide on-site operations and maintenance;

[0009] Optimize computational efficiency and robustness through online learning and model updates.

[0010] Preferably, the step of simultaneously acquiring infrared and hyperspectral multimodal data and preprocessing the multimodal data specifically includes:

[0011] Perform spatiotemporal calibration of both infrared and hyperspectral sensors to ensure the consistency of the position of the same physical point in the two types of imaging data;

[0012] Perform radiation calibration and atmospheric correction;

[0013] Perform abnormal pixel filtering and data augmentation.

[0014] Preferably, the step of extracting complementary features from infrared and hyperspectral data based on the preprocessed multimodal data and achieving feature-level fusion through deep learning specifically includes:

[0015] Based on the preprocessed multimodal data, infrared temperature distribution features are extracted.

[0016] Perform hyperspectral substance identification feature extraction;

[0017] Cross-modal feature fusion is performed based on infrared temperature distribution characteristics and hyperspectral material identification characteristics.

[0018] Preferably, the step of realizing three-dimensional spatial positioning based on the distribution of internal heat sources of the surface data inversion device specifically includes:

[0019] Construct a heat conduction model;

[0020] Based on the heat conduction model, a deep learning-based inversion solution is performed.

[0021] Monte Carlo Dropout is used to estimate the uncertainty of the inversion results and to calculate the posterior probability distribution of the heat source location.

[0022] Preferably, the step of performing intelligent diagnosis of overheating type and distinguishing between normal overheating and fault overheating of the equipment specifically includes:

[0023] Perform multi-feature joint analysis;

[0024] Train a multi-branch Transformer classifier to build a deep learning classification model;

[0025] A causal discovery algorithm is used to analyze the causal relationship between fever type and characteristics, construct a causal graph model, and verify causal reasoning.

[0026] Preferably, the steps of generating a visual report and triggering tiered early warnings to guide on-site operation and maintenance specifically include:

[0027] Conduct a thermal risk level assessment;

[0028] Augmented reality visualization through AR;

[0029] Adjust the adaptive warning threshold.

[0030] Preferably, the step of optimizing computational efficiency and robustness through online learning and model updates specifically includes:

[0031] Deploy edge-cloud collaborative computing;

[0032] Establish a feedback loop for online learning and model updates;

[0033] Conduct anti-interference robustness tests.

[0034] To address the aforementioned technical problems, this invention also provides an overheating location device for power equipment based on infrared hyperspectral imaging, employing the following technical solution, including:

[0035] A preprocessing module is used to simultaneously acquire infrared and hyperspectral multimodal data and preprocess the multimodal data;

[0036] The fusion module is used to extract complementary features from infrared and hyperspectral data based on the preprocessed multimodal data, and to achieve feature-level fusion through deep learning;

[0037] The positioning module is used to invert the distribution of internal heat sources of the equipment based on surface data to achieve three-dimensional spatial positioning;

[0038] The diagnostic module is used for intelligent diagnosis of overheating types, distinguishing between normal heat generation and fault-related heat generation in equipment;

[0039] The generation module is used to generate visual reports and trigger tiered alerts to guide on-site operations and maintenance.

[0040] The optimization module is used to improve computational efficiency and robustness through online learning and model updates.

[0041] To address the aforementioned technical problems, the present invention also provides a computer device that employs the technical solution described below, comprising a memory and a processor. The memory stores computer-readable instructions, and the processor executes the computer-readable instructions to implement the steps of the above-described method for locating overheated power equipment based on infrared hyperspectral imaging.

[0042] To address the aforementioned technical problems, the present invention also provides a computer-readable storage medium, which employs the technical solution described below. The computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the steps of the above-described method for locating overheated power equipment based on infrared hyperspectral imaging.

[0043] Compared with the prior art, the present invention has the following main advantages:

[0044] (1) Multimodal data fusion breaks through the limitations of a single sensor. By simultaneously collecting infrared thermal radiation and hyperspectral reflectance features, complementary information on the surface temperature distribution and material composition of the equipment can be obtained. Combined with deep learning feature-level fusion technology, the sensitivity and positioning accuracy of overheating defect detection are significantly improved, especially the ability to identify early weak faults is enhanced.

[0045] (2) The three-dimensional heat source inversion technology realizes the accurate mapping from surface temperature to internal thermal field. Based on the inversion model constructed by the heat conduction theory, combined with the equipment structural parameters, the spatial coordinates and intensity distribution of internal abnormal heat sources can be quantitatively calculated, providing operation and maintenance personnel with intuitive fault location guidance and greatly shortening the troubleshooting time.

[0046] (3) The intelligent diagnostic system has the ability to identify fault types. By analyzing the temperature field evolution pattern and material spectral characteristics, it can distinguish between normal load heating and fault types such as insulation aging and poor contact, avoid false alarms and missed alarms, and provide a scientific basis for equipment condition assessment.

[0047] (4) The closed-loop optimization mechanism ensures the long-term reliability of the system. The online learning module continuously absorbs new data and dynamically adjusts the model parameters. While improving the computing efficiency, it enhances the adaptability to complex working conditions. Combined with the hierarchical early warning mechanism, it can realize full-process control from early warning to emergency response, effectively reducing the risk of unplanned equipment downtime. Attached Figure Description

[0048] To more clearly illustrate the solutions in this invention, the accompanying drawings used in the description of the embodiments of this invention will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0049] Figure 1 This is a flowchart of an embodiment of the power equipment overheating location method based on infrared hyperspectral imaging of the present invention;

[0050] Figure 2 This is a schematic diagram of a structure of an embodiment of the power equipment overheat positioning device based on infrared hyperspectral imaging of the present invention;

[0051] Figure 3 This is a schematic diagram of the structure of an embodiment of the computer device of the present invention. Detailed Implementation

[0052] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains; the terminology used herein in the specification is for the purpose of describing particular embodiments only and is not intended to limit the invention; the terms "comprising" and "having," and any variations thereof, in the specification, claims, and foregoing drawings are intended to cover non-exclusive inclusion. The terms "first," "second," etc., in the specification, claims, or foregoing drawings are used to distinguish different objects and not to describe a particular order.

[0053] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0054] To enable those skilled in the art to better understand the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0055] It should be noted that the power equipment overheating location method based on infrared hyperspectral imaging provided in the embodiments of the present invention is generally executed by a server / terminal device, and correspondingly, the power equipment overheating location device based on infrared hyperspectral imaging is generally installed in the server / terminal device.

[0056] It should be understood that the number of terminal devices, networks, and servers is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be used.

[0057] Example 1

[0058] Please refer to Figure 1 The diagram illustrates a flowchart of an embodiment of the power equipment overheating location method based on infrared hyperspectral imaging of the present invention. The power equipment overheating location method based on infrared hyperspectral imaging includes the following steps:

[0059] Step S1: Simultaneously acquire infrared and hyperspectral multimodal data, and preprocess the multimodal data.

[0060] In this embodiment, the electronic device (e.g., a server / terminal device) running on the infrared hyperspectral imaging-based power equipment overheating location method can receive power equipment overheating location requests via wired or wireless connections. It should be noted that the aforementioned wireless connection methods may include, but are not limited to, 3G / 4G / 5G connections, WiFi connections, Bluetooth connections, WiMAXX connections, Zigbee connections, UWB (ultra-wideband) connections, and other currently known or future-developed wireless connection methods.

[0061] In this embodiment, step S1, which involves simultaneously acquiring infrared and hyperspectral multimodal data and preprocessing the multimodal data, specifically includes the following steps:

[0062] S11 performs spatiotemporal calibration of both infrared and hyperspectral sensors to ensure the consistency of the position of the same physical point in the two imaging data.

[0063] A high-precision mechanical support and electronic triggering system are used to rigidly fix an infrared camera (response band 8–14 μm) and a hyperspectral camera (visible-near-infrared band 400–1000 nm) to the same platform, and the acquisition timing is controlled by a GPS synchronization module. The mechanical rigid fixation can greatly reduce the changes in the relative motion between the two cameras (i.e., external parameters R and T) caused by vibration and other factors, ensuring the stability of the spatial relationship.

[0064] Electronic triggering and GPS synchronization ensure that both cameras are exposed at the same time (or within a known minimal time difference). This is crucial for spatiotemporal calibration. If the acquisition times are not synchronized, moving equipment or changing scenes can cause calibration failure.

[0065] Using a checkerboard calibration board, the extrinsic parameters (rotation matrix R and translation vector T) and intrinsic parameters (focal length and distortion coefficients) between the two sensors are calculated to establish a pixel-level mapping relationship.

[0066] Where, x hyper and y hyper x represents the pixel coordinates in the hyperspectral image. ir and y ir K represents the pixel coordinates in the infrared image. ir K is the intrinsic parameter matrix of the infrared camera (containing parameters such as focal length and principal point). hyper R is the intrinsic parameter matrix of the hyperspectral camera, R is the rotation matrix between the two cameras, and T is the translation vector between the two cameras.

[0067] The checkerboard pattern provides a large number of high-contrast, well-defined, and geometrically regular (square size known) feature points. The coordinates of these corner points in the 3D world and their 2D pixel coordinates in the two camera images are known, providing sufficient data constraints for accurate calculation of camera parameters.

[0068] By employing step S11, a geometric correspondence between infrared and hyperspectral images can be established, ensuring the consistency of the position of the same physical point in the two modal data, thus providing a foundation for subsequent multimodal data fusion.

[0069] S12 is used for radiation calibration and atmospheric correction.

[0070] A high-precision surface blackbody (temperature control range 300-500K, covering common fault temperatures of power equipment) was used as a reference radiation source to establish a quantitative relationship between camera pixel grayscale values ​​and known radiance, thus completing radiometric calibration.

[0071] Based on Planck's blackbody radiation law, the calibrated radiance value L λ Convert to the absolute temperature value of the object's surface: Where T is the absolute temperature (unit: K), λ is the wavelength (unit: m), and L is the wavelength. λ Spectral radiance (unit: W·m-) 2 ·sr- 1 ·m- 1 c1 is the first radiation constant, with a value of 3.7418 × 10⁻⁶. -16 W·m 2 c2 is the second radiation constant, with a value of 1.4388 × 10⁻⁶. -2 m·K.

[0072] The purpose of radiometric calibration is to convert the radiance value measured by the infrared camera into an accurate temperature value by using Planck's law to invert the temperature T, thereby achieving non-contact temperature measurement.

[0073] Hyperspectral data were converted for reflectance using the Empirical Linear Model (ELM), and the MODTRAN model was used to eliminate the effects of atmospheric absorption (such as water vapor and carbon dioxide). By inputting information such as imaging time, location, and meteorological parameters, the atmospheric transport process was simulated to obtain the true spectral reflectance data of ground objects.

[0074] The purpose of atmospheric correction is to eliminate differences in sensor response and environmental interference, convert raw data into physical quantities (temperature, reflectivity), and improve data reliability.

[0075] S13 performs abnormal pixel filtering and data augmentation.

[0076] An adaptive median filter is used to process impulse noise in infrared images, and the Savitzky-Golay smoothing algorithm for hyperspectral data is combined to suppress random noise. This algorithm can dynamically adjust the size of the filtering window, effectively suppressing noise while preserving the sharpness and detail of hot spot edges to the maximum extent, preventing blurring of overheated target features.

[0077] For random noise in the spectral dimension, the Savitzky-Golay smoothing algorithm is adopted. This algorithm smooths the noise through local polynomial least squares fitting, which can effectively improve the signal-to-noise ratio while preserving the original shape and width of spectral absorption and reflection characteristics (i.e., maintaining spectral shape preservation) and avoiding the introduction of distortion.

[0078] To address the issue of insufficient samples, a data augmentation algorithm based on Generative Adversarial Networks (GANs) is applied to generate diverse overheating samples. To solve the model overfitting problem caused by the scarcity of on-site overheating samples, GANs are introduced for data augmentation. By training an adversarial game between the generator and discriminator, the distribution characteristics of multimodal data (infrared thermograms and corresponding hyperspectral curves) of power equipment in both normal and overheated states are learned, thereby generating synthetic overheating samples with reasonable physical meaning and diverse forms.

[0079] The purpose of step S13 is to improve the signal-to-noise ratio (SNR≥40dB), expand the training dataset, and enhance the model's generalization ability.

[0080] Step S2: Based on the preprocessed multimodal data, complementary features are extracted from the infrared and hyperspectral data, and feature-level fusion is achieved through deep learning.

[0081] In this embodiment, step S2, which involves extracting complementary features from infrared and hyperspectral data based on the preprocessed multimodal data and achieving feature-level fusion through deep learning, specifically includes the following steps:

[0082] S21, Based on the preprocessed multimodal data, infrared temperature distribution features are extracted.

[0083] The U-Net network is used to segment the power equipment area and extract the statistical features of local temperature gradients and hot spots.

[0084] The temperature gradient magnitude is calculated for each pixel within the device's area, used to quantify the spatial drasticness of temperature changes. Local temperature gradient: Where G(x,y) is the magnitude of the temperature gradient at point (x,y) (unit: K / m), and a larger value indicates a more rapid temperature change at that point. Let x be the rate of temperature change in the x-direction. Let be the rate of temperature change in the y-direction.

[0085] By calculating the local temperature gradient, the severity of temperature changes can be quantified, which helps to identify areas of temperature anomalies.

[0086] Hotspot statistical characteristics include temperature standard deviation, skewness, and kurtosis. Higher-order statistics are calculated within the segmented equipment area or suspected overheated region to describe the distribution pattern of the temperature field.

[0087] Temperature standard deviation characterizes the degree of temperature dispersion within a region; a larger value indicates a more uneven heat distribution.

[0088] Skewness measures the asymmetry of temperature distribution. Positive skewness indicates the presence of a few high-temperature points (right-skewed distribution) and is a strong indicator of overheating.

[0089] Kurtosis reflects how steep the temperature distribution curve is compared to a normal distribution. High kurtosis means the temperature distribution is more concentrated, with a thicker tail, and may contain outliers.

[0090] From a statistical perspective, we can deeply characterize the overall morphology of hot spots and distinguish between normal thermal gradients caused by uneven solar radiation and genuine local fault overheating.

[0091] For dynamically monitored data, calculate the rate of temperature change for a specific area over time. The rate of temperature change over time (if dynamically monitored) is: Where R(t) represents the rate of temperature change at time t (unit: K / s or K / min), ΔT is the amount of temperature change, and Δt is the time interval.

[0092] The purpose of calculating the temperature change rate over time is to characterize the rate of temperature change over time, which helps to identify fault points that are rapidly heating up.

[0093] The purpose of step S21 is to characterize the surface temperature distribution pattern and initially locate abnormal hot areas.

[0094] S22, perform hyperspectral substance identification feature extraction.

[0095] The ResNet-50 network is used to extract deep spectral-spatial features from hyperspectral images. Its deep network structure can automatically and efficiently learn complex spectral-spatial joint features in hyperspectral data cubes, capturing subtle patterns that are difficult for the human eye to recognize. These deep features serve as high-level abstract representations for subsequent classification tasks.

[0096] The deep spectral-spatial features of hyperspectral images include spectral absorption features, spatial texture features, and material classification features.

[0097] Key wavelengths (such as characteristic absorption peaks at 550 nm for CuO and 650 nm for Fe2O3) are extracted using the Continuous Projection Algorithm (SPA) to extract spectral absorption features. To further enhance the interpretability and physical meaning of the model, the SPA algorithm is used for key wavelength selection. SPA accurately selects the most information-rich characteristic wavelengths (such as characteristic absorption valleys around 550 nm for CuO and around 650 nm for Fe2O3) for identifying specific substances by minimizing collinearity between wavelengths. This significantly reduces data dimensionality, highlights the spectral features most relevant to the chemical composition of substances, and provides direct chemical evidence for fault diagnosis.

[0098] Spatial texture features are extracted by using Gabor filter banks to extract texture responses in different directions. Multi-scale, multi-directional Gabor filter banks are used to filter images in specific bands (such as key bands or principal components across the entire band selected by SPA). Gabor filters, similar to the human visual system in frequency and direction, can describe and extract texture features characterizing the surface state of equipment (such as insulator glaze erosion, metal surface roughness variations, and contaminant distribution).

[0099] Support Vector Machines (SVMs) are used to distinguish equipment materials (such as ceramic insulators, copper wires, and silicon steel sheets), extracting material classification features. These extracted deep features, spectral features, and texture features are then fused to construct a high-dimensional feature vector. High-performance classifiers such as SVMs are then used to accurately classify and identify equipment materials (such as ceramics, copper, and silicon steel sheets) and surface conditions (such as oxidation, corrosion, and contamination). This precise identification of material composition and surface conditions provides crucial contextual information for temperature anomaly analysis. For example:

[0100] Distinguish between normal heating (such as the normal temperature rise of copper wires due to their resistance and current) and fault heating (such as abnormal overheating of wire clamps due to contact oxidation and increased resistance).

[0101] Identify the cause of overheating (such as overheating caused by local electric field distortion due to contaminants adhering to the insulator, whose spectral characteristics differ from those of a clean insulator). The output of this step is the core decision-making basis for determining the nature of the overheating (physical normal overheating vs. chemical / fault-related abnormal overheating), greatly improving the accuracy and reliability of fault diagnosis.

[0102] The purpose of step S22 is to identify the material and oxidation state of the equipment and help determine the nature of the heat generation (normal heat generation or fault heat generation).

[0103] S23 performs cross-modal feature fusion based on infrared temperature distribution characteristics and hyperspectral material identification characteristics.

[0104] A Dual-Stream Attention Fusion Network (DSAF) is designed, comprising an infrared stream input of a temperature distribution map and a hyperspectral stream input of cubic data (spatial-spectral dimensions). Using the temperature distribution map (or preprocessed infrared image) extracted in step S21 as input, a CNN module further extracts deep, thermally relevant spatial features F. ir .

[0105] Hyperspectral Stream: Taking atmospherically corrected hyperspectral data cubes (spatial-spectral dimensions) as input, it extracts deep material composition and spatial texture features F through a 3D-CNN or spectral-spatial separated CNN module. hyper .

[0106] Instead of simple concatenation or weighted averaging, this mechanism dynamically calculates the relevance weights between features from two modalities using a cross-attention mechanism. Specifically, this mechanism allows each modality's features to query the most relevant information from the other modality. The formula for calculating feature weights using the cross-attention mechanism is:

[0107] Where, α ij The attention weight between the i-th infrared feature and the j-th hyperspectral feature is... Let i be the i-th infrared feature vector. Let be the j-th hyperspectral feature vector, and sim(·) be the cosine similarity function.

[0108] Calculate the feature weights α ij The formula can adaptively calculate the correlation weights between features of different modalities, enabling targeted feature fusion.

[0109] The weighted and fused features undergo further nonlinear transformation and dimensionality reduction in a fully connected layer, ultimately generating a highly discriminative joint feature vector rich in multimodal information for final fault diagnosis and location. Integrating temperature and material information enhances the ability to infer internal heat sources.

[0110] Step S3: Based on the surface data, the distribution of internal heat sources of the equipment is inverted to achieve three-dimensional spatial positioning.

[0111] In this embodiment, step S3, which involves retrieving the internal heat source distribution of the device based on surface data to achieve three-dimensional spatial positioning, specifically includes the following steps:

[0112] S31, Construct a heat conduction model.

[0113] Based on the equipment structure, establish a three-dimensional partial differential equation (PDE) for heat conduction: Where ρ is the material density (unit: kg / m³) 3 ), c p Specific heat capacity (unit: J / (kg·K)), Let k be the rate of change of temperature over time, and k be the thermal conductivity (unit: W / (m·K)). Let Q be the temperature gradient, and Q be the intensity of the internal heat source (unit: W / m). 3 ).

[0114] It can describe the heat conduction process inside the device and establish a physical model from the internal heat source to the surface temperature distribution.

[0115] The solution is obtained by discretization using the finite element method (FEM), and the mesh generation is combined with the material parameters identified by hyperspectral analysis (e.g., copper: k = 400 W / (m·K), ceramics: k = 1.5 W / (m·K)).

[0116] The purpose of step S31 is to establish a physical mapping model from the internal heat source to the surface temperature distribution.

[0117] S32, based on the heat conduction model, performs inversion solution based on deep learning.

[0118] The inverse problem is transformed into an optimization problem, with the loss function being: in, T is the value of the loss function. surface For the actual measured surface temperature distribution, T model Let ||Q|| be the surface temperature distribution calculated using the heat conduction model based on the currently estimated heat source Q. ||·||2 is the L2 norm, which measures the difference between the prediction and the measurement. ||Q||1 is the L1 norm of the heat source distribution, which promotes sparse solutions. λ is the regularization parameter, which balances the contributions of the two terms.

[0119] The problem of internal heat source inversion is formulated as an optimization problem, and the most likely heat source distribution is estimated by minimizing the loss function.

[0120] The solution is obtained using a Physical Information Neural Network (PINN): the input is surface temperature and hyperspectral material data, and the output is the heat source distribution Q(x,y,z).

[0121] The purpose of step S32 is to overcome the limitations of traditional methods that rely on internal structures and to locate the internal heat source of complex equipment.

[0122] S33 uses Monte Carlo Dropout to estimate the uncertainty of the inversion results and calculates the posterior probability distribution of the heat source location.

[0123] Monte Carlo Dropout (MC-Dropout) is used to estimate the uncertainty of the inversion results, and the posterior probability distribution of the heat source location is calculated: Wherein P(Q|T) surface Let θ be the posterior probability of the heat source distribution given a surface temperature, N be the number of Monte Carlo samplings, and θ be the posterior probability of the heat source distribution. n Let P(Q|T) be the network parameters at the nth sampling. surface ,θ n ) for using parameter θ n The probability of heat source distribution predicted in real time.

[0124] Assess the uncertainty of the inversion results, provide a confidence estimate of the heat source location, and avoid over-reliance on single prediction results.

[0125] The purpose of step S33 is to assess the reliability of the positioning results and avoid overconfidence.

[0126] Step S4: Perform intelligent diagnosis of overheating type to distinguish between normal heat generation and fault-related heat generation.

[0127] In this embodiment, step S4, performing intelligent diagnosis of overheating type and distinguishing between normal heat generation and fault-related heat generation, specifically includes the following steps:

[0128] S41, perform multi-feature joint analysis.

[0129] Extract the following discriminant features, such as thermal pattern features, spectral oxidation index, and temperature rise rate.

[0130] Thermal pattern characteristics refer to the shape of the hot zone (circular shapes are mostly normal, while irregular shapes are mostly faulty).

[0131] The spectral oxidation index refers to the ratio of the absorption depth of Fe2O3 at 650 nm to that of CuO at 550 nm. Among them, R ox A is the oxidation index. 650 The absorption depth at 650 nm (characteristic absorption peak of Fe2O3), A 550 This indicates the absorption depth at 550 nm (characteristic absorption peak of CuO).

[0132] The degree of oxidation of materials is quantified; a high degree of oxidation usually indicates poor contact or aging problems.

[0133] Temperature rise rate refers to the fact that fault-induced heat typically causes a faster temperature rise.

[0134] The purpose of step S41 is to construct interpretable fault discrimination indicators.

[0135] S42, train a multi-branch Transformer classifier to build a deep learning classification model.

[0136] Train a multi-branch Transformer classifier, with the input being the fused feature vector (output of step two) and the heat source inversion result (output of step three). The outputs are normal heating (Class 0), electrical fault (Class 1), and mechanical fault (Class 2). The loss function is weighted cross-entropy (to address class imbalance).

[0137] The purpose of step S42 is to achieve end-to-end overheating type diagnosis with an accuracy of ≥95%.

[0138] S43 employs a causal discovery algorithm to analyze the causal relationship between fever type and characteristics, constructs a causal graph model, and verifies causal reasoning.

[0139] Causal discovery algorithms (such as PC algorithm) are used to analyze the causal relationship between heating type and characteristics, and a causal graph model is constructed. For example, the causal chain of increased oxidation index → ​​increased contact resistance → fault heating is verified.

[0140] The purpose of step S43 is to improve the reliability of the diagnostic results and meet the needs of operation and maintenance decision-making.

[0141] Step S5 generates a visual report and triggers tiered alerts to guide on-site operations and maintenance.

[0142] In this embodiment, step S5, generating a visual report and triggering tiered early warnings to guide on-site operations and maintenance, specifically includes the following steps:

[0143] S51, conduct a thermal risk level assessment.

[0144] Define the risk index R risk : Among them, R risk T represents the heat risk index. max T represents the highest temperature detected. amb T represents ambient temperature. limit Indicates the material's temperature resistance threshold. I represents the rate of temperature rise. ox The oxidation index is represented by w1, w2, and w3, which represent the weights of each indicator and are determined by the Analytic Hierarchy Process (AHP) decision-making method. By comprehensively quantifying overheating risk using multiple indicators, a scientific, tiered early warning system can be achieved.

[0145] The purpose of step S51 is to quantify the risk of overheating and implement graded early warning (normal, attention, warning, danger).

[0146] S52 uses AR (Augmented Reality) visualization.

[0147] An AR system was developed using the Unity3D engine, overlaying overheated areas (rendered in red), internal heat sources (semi-transparent heat clouds), and diagnostic results onto a real-time video stream. SLAM technology was employed for device localization and registration.

[0148] The purpose of step S52 is to visually display the location and extent of overheating, assisting maintenance personnel in quickly locating the problem.

[0149] S53 performs adaptive early warning threshold adjustment.

[0150] Combining historical data with environmental factors (ambient temperature, humidity, load current), Bayesian optimization is used to dynamically adjust the early warning threshold: T threshold =f(T)amb ,I load ,humidity), where T threshold T represents the dynamically adjusted warning threshold. amb Indicates ambient temperature, I load Let represent the load current, humidity represent the ambient humidity, and f(·) represent the function relationship obtained through Bayesian optimization. The warning threshold is adaptively adjusted based on the actual operating environment to reduce the false alarm rate.

[0151] The purpose of step S53 is to reduce the false alarm rate and adapt to different operating conditions.

[0152] Step S6: Optimize computational efficiency and robustness through online learning and model updates.

[0153] In this embodiment, step S6, optimizing computational efficiency and robustness through online learning and model updates, specifically includes the following steps:

[0154] S61 enables edge-cloud collaborative computing deployment.

[0155] Edge computing (embedded GPU): Responsible for data acquisition, preprocessing, and real-time alerts (response time <1s). Cloud computing: Responsible for model training, data storage, and deep analysis. TensorRT is used to accelerate inference, and knowledge distillation technology is employed for model compression.

[0156] The purpose of step S61 is to balance the computational load and meet real-time requirements.

[0157] S62 establishes a feedback loop for online learning and model updates.

[0158] Establish a feedback loop: Operations personnel label false alarms / missed alarms, and the system updates model parameters through online learning. Where, θ t+1 θ represents the updated model parameters. t This represents the current model parameters, and η represents the learning rate. Indicated in the new sample (x) new ,y new The gradient of the loss function on the model is used to enable continuous learning and optimization, adapting to dynamic factors such as equipment aging and environmental changes.

[0159] The purpose of step S62 is to continuously improve the model's adaptability and accuracy.

[0160] S63 was subjected to anti-interference robustness testing.

[0161] The system performance was tested under extreme environments such as heavy rain, strong light, and smog. Adversarial training was used to enhance the robustness of the model. Where θ represents the model parameters, δ represents the added perturbation, and Δ represents the allowable range of the perturbation. This represents the loss function applied to the samples after perturbation. By optimizing the model under worst-case perturbation, the stability and reliability of the system in harsh environments are enhanced.

[0162] The purpose of step S63 is to ensure that the system operates stably in complex environments.

[0163] The beneficial effects of implementing this embodiment are:

[0164] (1) Multimodal data fusion breaks through the limitations of a single sensor. By simultaneously collecting infrared thermal radiation and hyperspectral reflectance features, complementary information on the surface temperature distribution and material composition of the equipment can be obtained. Combined with deep learning feature-level fusion technology, the sensitivity and positioning accuracy of overheating defect detection are significantly improved, especially the ability to identify early weak faults is enhanced.

[0165] (2) The three-dimensional heat source inversion technology realizes the accurate mapping from surface temperature to internal thermal field. Based on the inversion model constructed by the heat conduction theory, combined with the equipment structural parameters, the spatial coordinates and intensity distribution of internal abnormal heat sources can be quantitatively calculated, providing operation and maintenance personnel with intuitive fault location guidance and greatly shortening the troubleshooting time.

[0166] (3) The intelligent diagnostic system has the ability to identify fault types. By analyzing the temperature field evolution pattern and material spectral characteristics, it can distinguish between normal load heating and fault types such as insulation aging and poor contact, avoid false alarms and missed alarms, and provide a scientific basis for equipment condition assessment.

[0167] (4) The closed-loop optimization mechanism ensures the long-term reliability of the system. The online learning module continuously absorbs new data and dynamically adjusts the model parameters. While improving the computing efficiency, it enhances the adaptability to complex working conditions. Combined with the hierarchical early warning mechanism, it can realize full-process control from early warning to emergency response, effectively reducing the risk of unplanned equipment downtime.

[0168] This invention can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This invention can be described in the general context of computer-executable instructions, such as program modules, that are executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This invention can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0169] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by instructing related hardware with computer-readable instructions. These computer-readable instructions can be stored in a computer-readable storage medium. When executed, the program can include the processes of the embodiments of the above methods. The aforementioned storage medium can be a non-volatile storage medium such as a magnetic disk, optical disk, or read-only memory (ROM), or random access memory (RAM).

[0170] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0171] Example 2

[0172] Further reference Figure 2 As a response to the above Figure 1 The present invention provides an embodiment of an overheating location device for power equipment based on infrared hyperspectral imaging, which is similar to the method shown. Figure 1 Corresponding to the method embodiments shown, this device can be specifically applied to various electronic devices.

[0173] like Figure 2 As shown, the power equipment overheating location device 70 based on infrared hyperspectral imaging described in this embodiment includes: a preprocessing module 71, a fusion module 72, a location module 73, a diagnostic module 74, a generation module 75, and an optimization module 76. Wherein:

[0174] The preprocessing module 71 is used to simultaneously acquire infrared and hyperspectral multimodal data and preprocess the multimodal data;

[0175] The fusion module 72 is used to extract complementary features from infrared and hyperspectral data based on the preprocessed multimodal data, and to achieve feature-level fusion through deep learning;

[0176] Positioning module 73 is used to invert the distribution of internal heat sources of the equipment based on surface data to achieve three-dimensional spatial positioning;

[0177] Diagnostic module 74 is used for intelligent diagnosis of overheating type, distinguishing between normal heat generation and fault-related heat generation of equipment;

[0178] Module 75 is used to generate visual reports and trigger tiered alerts to guide on-site operations and maintenance.

[0179] Optimization module 76 is used to optimize computational efficiency and robustness through online learning and model updates.

[0180] The beneficial effects of implementing this embodiment are:

[0181] (1) Multimodal data fusion breaks through the limitations of a single sensor. By simultaneously collecting infrared thermal radiation and hyperspectral reflectance features, complementary information on the surface temperature distribution and material composition of the equipment can be obtained. Combined with deep learning feature-level fusion technology, the sensitivity and positioning accuracy of overheating defect detection are significantly improved, especially the ability to identify early weak faults is enhanced.

[0182] (2) The three-dimensional heat source inversion technology realizes the accurate mapping from surface temperature to internal thermal field. Based on the inversion model constructed by the heat conduction theory, combined with the equipment structural parameters, the spatial coordinates and intensity distribution of internal abnormal heat sources can be quantitatively calculated, providing operation and maintenance personnel with intuitive fault location guidance and greatly shortening the troubleshooting time.

[0183] (3) The intelligent diagnostic system has the ability to identify fault types. By analyzing the temperature field evolution pattern and material spectral characteristics, it can distinguish between normal load heating and fault types such as insulation aging and poor contact, avoid false alarms and missed alarms, and provide a scientific basis for equipment condition assessment.

[0184] (4) The closed-loop optimization mechanism ensures the long-term reliability of the system. The online learning module continuously absorbs new data and dynamically adjusts the model parameters. While improving the computing efficiency, it enhances the adaptability to complex working conditions. Combined with the hierarchical early warning mechanism, it can realize full-process control from early warning to emergency response, effectively reducing the risk of unplanned equipment downtime.

[0185] Example 3

[0186] To address the aforementioned technical problems, embodiments of the present invention also provide a computer device. Please refer to [link / reference needed]. Figure 3 , Figure 3 This is a basic structural block diagram of the computer device in this embodiment.

[0187] The aforementioned computer device 8 includes a memory 81, a processor 82, and a network interface 83 that are interconnected via a system bus. It should be noted that only the computer device 8 with components 81, 82, and 83 is shown in the figure; however, it should be understood that it is not required to implement all the shown components, and more or fewer components can be implemented alternatively. Those skilled in the art will understand that the computer device described herein is a device capable of automatically performing numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes, but is not limited to, microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0188] The aforementioned computer devices can be desktop computers, laptops, handheld computers, and cloud servers, among other computing devices. These devices can facilitate human-computer interaction with users through keyboards, mice, remote controls, touchpads, or voice-activated devices.

[0189] The aforementioned memory 81 includes at least one type of readable storage medium, including flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the aforementioned memory 81 may be an internal storage unit of the aforementioned computer device 8, such as the hard disk or memory of the computer device 8. In other embodiments, the aforementioned memory 81 may also be an external storage device of the aforementioned computer device 8, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the computer device 8. Of course, the aforementioned memory 81 may also include both the internal storage unit and its external storage device of the aforementioned computer device 8. In this embodiment, the aforementioned memory 81 is typically used to store the operating system and various application software installed on the aforementioned computer device 8, such as computer-readable instructions for a power equipment overheating location method based on infrared hyperspectral imaging. In addition, the aforementioned memory 81 can also be used to temporarily store various types of data that have been output or will be output.

[0190] In some embodiments, the processor 82 may be a central processing unit (CPU), controller, microcontroller, microprocessor, or other data processing chip. The processor 82 is typically used to control the overall operation of the computer device 8. In this embodiment, the processor 82 is used to execute computer-readable instructions stored in the memory 81 or to process data, for example, to execute the computer-readable instructions of the above-mentioned method for locating overheated power equipment based on infrared hyperspectral imaging.

[0191] The network interface 83 may include a wireless network interface or a wired network interface, which is typically used to establish a communication connection between the computer device 8 and other electronic devices.

[0192] The beneficial effects of implementing this embodiment are:

[0193] (1) Multimodal data fusion breaks through the limitations of a single sensor. By simultaneously collecting infrared thermal radiation and hyperspectral reflectance features, complementary information on the surface temperature distribution and material composition of the equipment can be obtained. Combined with deep learning feature-level fusion technology, the sensitivity and positioning accuracy of overheating defect detection are significantly improved, especially the ability to identify early weak faults is enhanced.

[0194] (2) The three-dimensional heat source inversion technology realizes the accurate mapping from surface temperature to internal thermal field. Based on the inversion model constructed by the heat conduction theory, combined with the equipment structural parameters, the spatial coordinates and intensity distribution of internal abnormal heat sources can be quantitatively calculated, providing operation and maintenance personnel with intuitive fault location guidance and greatly shortening the troubleshooting time.

[0195] (3) The intelligent diagnostic system has the ability to identify fault types. By analyzing the temperature field evolution pattern and material spectral characteristics, it can distinguish between normal load heating and fault types such as insulation aging and poor contact, avoid false alarms and missed alarms, and provide a scientific basis for equipment condition assessment.

[0196] (4) The closed-loop optimization mechanism ensures the long-term reliability of the system. The online learning module continuously absorbs new data and dynamically adjusts the model parameters. While improving the computing efficiency, it enhances the adaptability to complex working conditions. Combined with the hierarchical early warning mechanism, it can realize full-process control from early warning to emergency response, effectively reducing the risk of unplanned equipment downtime.

[0197] Example 4

[0198] The present invention also provides another embodiment, namely, providing a computer-readable storage medium storing computer-readable instructions that can be executed by at least one processor to cause the at least one processor to perform the steps of the above-described method for locating overheated power equipment based on infrared hyperspectral imaging.

[0199] The beneficial effects of implementing this embodiment are:

[0200] (1) Multimodal data fusion breaks through the limitations of a single sensor. By simultaneously collecting infrared thermal radiation and hyperspectral reflectance features, complementary information on the surface temperature distribution and material composition of the equipment can be obtained. Combined with deep learning feature-level fusion technology, the sensitivity and positioning accuracy of overheating defect detection are significantly improved, especially the ability to identify early weak faults is enhanced.

[0201] (2) The three-dimensional heat source inversion technology realizes the accurate mapping from surface temperature to internal thermal field. Based on the inversion model constructed by the heat conduction theory, combined with the equipment structural parameters, the spatial coordinates and intensity distribution of internal abnormal heat sources can be quantitatively calculated, providing operation and maintenance personnel with intuitive fault location guidance and greatly shortening the troubleshooting time.

[0202] (3) The intelligent diagnostic system has the ability to identify fault types. By analyzing the temperature field evolution pattern and material spectral characteristics, it can distinguish between normal load heating and fault types such as insulation aging and poor contact, avoid false alarms and missed alarms, and provide a scientific basis for equipment condition assessment.

[0203] (4) The closed-loop optimization mechanism ensures the long-term reliability of the system. The online learning module continuously absorbs new data and dynamically adjusts the model parameters. While improving the computing efficiency, it enhances the adaptability to complex working conditions. Combined with the hierarchical early warning mechanism, it can realize full-process control from early warning to emergency response, effectively reducing the risk of unplanned equipment downtime.

[0204] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0205] Obviously, the embodiments described above are merely some embodiments of the present invention, not all embodiments. The accompanying drawings show preferred embodiments of the present invention, but do not limit the patent scope of the present invention. The present invention can be implemented in many different forms; rather, these embodiments are provided to provide a more thorough and complete understanding of the disclosure of the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing specific embodiments, or make equivalent substitutions for some of the technical features. Any equivalent structures made using the content of this specification and drawings, directly or indirectly applied to other related technical fields, are similarly within the patent protection scope of this invention.

Claims

1. A method for locating overheated power equipment based on infrared hyperspectral imaging, characterized in that, Includes the following steps: Simultaneously acquire infrared and hyperspectral multimodal data, and preprocess the multimodal data; Based on the preprocessed multimodal data, complementary features are extracted from infrared and hyperspectral data, and feature-level fusion is achieved through deep learning; Based on surface data, the internal heat source distribution of the equipment is inverted to achieve three-dimensional spatial positioning; Perform intelligent diagnosis of overheating types to distinguish between normal equipment overheating and fault overheating; Generate visual reports and trigger tiered alerts to guide on-site operations and maintenance; Optimize computational efficiency and robustness through online learning and model updates.

2. The method for locating overheated power equipment based on infrared hyperspectral imaging according to claim 1, characterized in that, The steps of simultaneously acquiring infrared and hyperspectral multimodal data and preprocessing the multimodal data specifically include: Perform spatiotemporal calibration of both infrared and hyperspectral sensors to ensure the consistency of the position of the same physical point in the two types of imaging data; Perform radiation calibration and atmospheric correction; Perform abnormal pixel filtering and data augmentation.

3. The method for locating overheated power equipment based on infrared hyperspectral imaging according to claim 1, characterized in that, The step of extracting complementary features from infrared and hyperspectral data based on the preprocessed multimodal data and achieving feature-level fusion through deep learning specifically includes: Based on the preprocessed multimodal data, infrared temperature distribution features are extracted. Perform hyperspectral substance identification feature extraction; Cross-modal feature fusion is performed based on infrared temperature distribution characteristics and hyperspectral material identification characteristics.

4. The method for locating overheated power equipment based on infrared hyperspectral imaging according to claim 1, characterized in that, The steps for achieving three-dimensional spatial positioning based on surface data inversion to determine the internal heat source distribution of the device specifically include: Construct a heat conduction model; Based on the heat conduction model, a deep learning-based inversion solution is performed. Monte Carlo Dropout is used to estimate the uncertainty of the inversion results and to calculate the posterior probability distribution of the heat source location.

5. The method for locating overheated power equipment based on infrared hyperspectral imaging according to claim 1, characterized in that, The steps for performing intelligent diagnosis of overheating types and distinguishing between normal overheating and fault-related overheating specifically include: Perform multi-feature joint analysis; Train a multi-branch Transformer classifier to build a deep learning classification model; A causal discovery algorithm is used to analyze the causal relationship between fever type and characteristics, construct a causal graph model, and verify causal reasoning.

6. The method for locating overheated power equipment based on infrared hyperspectral imaging according to claim 1, characterized in that, The steps for generating a visual report and triggering tiered alerts to guide on-site operations and maintenance specifically include: Conduct a thermal risk level assessment; Augmented reality visualization through AR; Adjust the adaptive warning threshold.

7. The method for locating overheated power equipment based on infrared hyperspectral imaging according to any one of claims 1 to 6, characterized in that, The steps to optimize computational efficiency and robustness through online learning and model updates specifically include: Deploy edge-cloud collaborative computing; Establish a feedback loop for online learning and model updates; Conduct anti-interference robustness tests.

8. A power equipment overheating positioning device based on infrared hyperspectral imaging, characterized in that, include: A preprocessing module is used to simultaneously acquire infrared and hyperspectral multimodal data and preprocess the multimodal data; The fusion module is used to extract complementary features from infrared and hyperspectral data based on the preprocessed multimodal data, and to achieve feature-level fusion through deep learning; The positioning module is used to invert the distribution of internal heat sources of the equipment based on surface data to achieve three-dimensional spatial positioning; The diagnostic module is used for intelligent diagnosis of overheating types, distinguishing between normal heat generation and fault-related heat generation in equipment; The generation module is used to generate visual reports and trigger tiered alerts to guide on-site operations and maintenance. The optimization module is used to improve computational efficiency and robustness through online learning and model updates.

9. A computer device, characterized in that, The device includes a memory and a processor, wherein the memory stores computer-readable instructions, and the processor executes the computer-readable instructions to implement the steps of the power equipment overheating location method based on infrared hyperspectral imaging as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the steps of the power equipment overheating location method based on infrared hyperspectral imaging as described in any one of claims 1 to 7.

Images