An electrical connection overheating fault early warning method and device based on infrared thermal images

Abstract

The application discloses an electrical connection overheating fault early warning method and device based on infrared thermal images, and relates to the technical field of electrical equipment state monitoring, wherein the method comprises the following steps: collecting an infrared thermal image sequence of an electrical connection area through an infrared thermal imaging device; based on the infrared thermal image sequence, identifying and positioning each connection point, extracting the temperature, binding a time stamp to generate time-series temperature data; based on a heat transfer model and numerical integration, superimposing environmental compensation calculation to calculate a normalized heat accumulation index, and quantifying the heat accumulation degree of the equipment; inputting the index into a deep learning dynamic threshold model trained by historical data, combining the index change trend to make multi-condition fusion determination; if the early warning standard is reached, a structured early warning message with a point position, an index value and a disposal suggestion is generated and output externally, and the intelligent early warning of the electrical joint overheating fault is completed. Through the multi-dimensional feature deep learning threshold model, the application realizes self-adaptive discrimination of working conditions, and solves the defect of poor fixed threshold adaptation.

Description

Technical Field

[0001] This invention relates to the field of electrical equipment condition monitoring technology, and in particular to a method and device for early warning of electrical connection overheating faults based on infrared thermal imaging. Background Technology

[0002] Overheating at electrical connection points is one of the core hidden dangers threatening the safe, stable, and reliable operation of power systems. Prolonged overheating at these connection points can easily lead to component aging, equipment damage, and even power outages. Therefore, accurate and timely early warning of overheating at electrical connection points is a crucial technological step in preventing power equipment failures and ensuring a continuous and stable power supply from the grid. Infrared thermal imaging detection, with its advantages of non-contact, uninterrupted power supply, and real-time monitoring, is now widely used in scenarios involving the perception of electrical equipment operating conditions and temperature monitoring, becoming a core technological means for electrical equipment condition detection.

[0003] Currently, infrared thermal imaging-based electrical connection overheating monitoring technology still suffers from numerous technical shortcomings and application limitations, making it difficult to adapt to the monitoring needs of complex power operation conditions. Existing monitoring and judgment methods mostly rely on static temperature thresholds or single temperature change rates to determine faults, failing to incorporate modeling and analysis of the physical characteristics of heat accumulation during the operation of electrical connection points. This neglects latent overheating faults caused by continuous heat accumulation, significantly reducing the accuracy and effectiveness of overheating fault early warnings. Furthermore, traditional warning thresholds are fixed parameters, unable to adapt to complex operating conditions such as fluctuating ambient temperatures and dynamic changes in equipment load. Environmental interference and load fluctuations can easily cause warning judgments to fail.

[0004] Furthermore, existing monitoring technologies primarily focus on acquiring and determining instantaneous temperature values, lacking in-depth analysis of temperature trends and thermal state evolution. This makes it difficult to predict overheating faults in advance, leading to frequent false alarms and missed alarms. At the data processing level, traditional data noise reduction and correction methods are relatively simplistic, failing to effectively eliminate noise interference and outliers in infrared image acquisition, resulting in insufficient consistency and reliability of monitoring data. Simultaneously, traditional early warning models lack dynamic iteration and adaptive update mechanisms, failing to optimize judgment logic based on long-term equipment operating status and condition changes, resulting in poor long-term monitoring stability and versatility.

[0005] In summary, existing electrical connection overheating early warning technologies suffer from numerous technical drawbacks, including a lack of consideration for thermal accumulation characteristics, weak adaptability to operating conditions, low data processing accuracy, simplistic early warning judgment logic, and insufficient model adaptive capability. These shortcomings make it difficult to meet the requirements for high-precision, high-reliability, and all-condition adaptability overheating fault early warning. There is an urgent need for an electrical connection overheating fault early warning method that takes into account thermal accumulation effect modeling, environmental adaptive adjustment, and multi-dimensional accurate judgment to address the various deficiencies of existing technologies and improve the intelligent early warning level of power equipment overheating faults. Summary of the Invention

[0006] The main objective of this invention is to provide a method for early warning of electrical connection overheating faults based on infrared thermal imaging.

[0007] Another objective of this invention is to provide an electrical connection overheating fault early warning device based on infrared thermal imaging.

[0008] The third objective of this invention is to provide an electronic device.

[0009] A fourth objective of this invention is to provide a non-transitory computer-readable storage medium.

[0010] To achieve the above objectives, a first aspect of the present invention provides a method for early warning of electrical connection overheating faults based on infrared thermal imaging, comprising:

[0011] The infrared thermal image sequence of the electrical connection area is acquired by an infrared thermal imaging device, and the infrared thermal image sequence contains temperature information at multiple consecutive time points. Based on the infrared thermal image sequence, electrical connection points in each frame of the image are identified and located, and the corresponding temperature values ​​are extracted. After associating the timestamps, time-series temperature data of each connection point is formed. Based on the time-series temperature data, numerical integration coupled with a heat conduction theoretical model is used, combined with ambient temperature compensation, to calculate the heat accumulation effect index of each electrical connection point, and the index is normalized to quantify the degree of heat energy accumulation at the electrical connection point during operation. The heat accumulation effect index is compared with the warning threshold output by the dynamic threshold model, and the change trend of the heat accumulation effect index is combined to perform multi-condition fusion judgment to determine whether the warning triggering conditions are met. The dynamic threshold model is pre-trained based on historical overheating fault data and normal operation data. When the warning triggering conditions are met, a warning signal is generated and output. The warning signal includes the identifier of the overheated connection point, the current thermal accumulation effect index value, and the recommended handling measures, thus completing the warning triggering and signal output for the overheating fault of the electrical connection point.

[0012] Optionally, an infrared thermal image sequence of the electrical connection area is acquired using an infrared thermal imaging device, including: Based on the real-time calculated trend of the heat accumulation effect index and combined with the data fed back from the ambient temperature sensor, the image acquisition frequency of the infrared thermal imaging equipment is dynamically adjusted. When the thermal accumulation effect index shows a rapid upward trend, the acquisition time interval is automatically shortened to increase the data acquisition density. The acquired infrared thermal images are subjected to denoising and enhancement processing to improve image clarity and the accuracy of temperature measurement.

[0013] Optionally, based on the infrared thermal image sequence, electrical connection points in each frame are identified and located, and their corresponding temperature values ​​are extracted. After associating timestamps, time-series temperature data for each connection point is generated, including: A multi-scale image segmentation algorithm is used to process each frame of infrared thermal image to divide candidate connection point regions; The candidate connection point regions are matched with pre-stored connection point geometric feature templates to determine the precise location of each electrical connection point in the image; Based on the positioning results, the temperature value is read from the corresponding location, wherein the temperature value is selected from at least one of the average temperature, the maximum temperature, or the temperature distribution gradient. The extracted temperature values ​​are associated with timestamps to form high-resolution time-series temperature data for each connection point.

[0014] Optionally, based on the time-series temperature data, numerical integration coupled with a heat conduction theoretical model, combined with ambient temperature compensation, is used to calculate the heat accumulation effect index of each electrical connection point, including: Numerical integration is performed on the time-series temperature data of each electrical connection point to obtain the cumulative amount of temperature change over time; Based on the heat conduction theory model, the instantaneous heat flux density and heat capacity changes are solved by combining the cumulative quantity; Dynamically select the integration time window based on the equipment's operating history; An ambient temperature compensation factor is introduced to correct the integral results. The corrected integral results are then normalized to output a normalized thermal accumulation effect index.

[0015] Optionally, the pre-training process of the dynamic threshold model includes: Historical overheating fault data and normal operation data were collected as a training sample set; A deep learning neural network is constructed using the heat accumulation effect index, ambient temperature and equipment current load in the training sample set as input features and the corresponding warning threshold as the output target. By iteratively learning the training sample set through a neural network, a dynamic threshold model is generated, enabling the model to output multi-level early warning thresholds under complex working conditions.

[0016] Optionally, the heat accumulation effect index is compared with the warning threshold output by the dynamic threshold model, and a multi-condition fusion judgment is performed based on the changing trend of the heat accumulation effect index to determine whether the warning triggering conditions are met, including: The current heat accumulation effect index is compared with the warning threshold output by the dynamic threshold model to determine whether the heat accumulation effect index exceeds the warning threshold. Calculate the first and second derivatives of the thermal accumulation effect index and analyze the stability of its changing trend; Based on the comparison results and trend stability analysis results, a decision tree algorithm is used to perform multi-condition fusion judgment; When the heat accumulation effect index exceeds the warning threshold and continues for a predetermined period of time, or when the trend of change meets the preset instability conditions, the warning trigger condition is determined to be met.

[0017] Optionally, when the warning triggering conditions are met, a warning signal is generated and output, including: The warning level is determined based on the magnitude and duration by which the current heat accumulation effect index exceeds the warning threshold; Generate a structured early warning message containing the spatial coordinates of the overheated connection point, historical change curves of the heat accumulation effect index, current environmental conditions, and early warning level; Structured early warning messages are pushed to the cloud monitoring platform or local display terminal in real time via wireless communication protocols, simultaneously triggering audible and visual alarm devices to complete the on-site output of early warning signals.

[0018] To achieve the above objectives, a second aspect of the present invention provides an electrical connection overheating fault early warning device based on infrared thermal imaging, comprising: An image acquisition unit is used to acquire a sequence of infrared thermal images of an electrical connection area using an infrared thermal imaging device. The sequence of infrared thermal images contains temperature information at multiple consecutive time points. The time-series temperature extraction unit is used to identify and locate electrical connection points in each frame of the infrared thermal image sequence and extract the corresponding temperature values, and form time-series temperature data for each connection point after associating timestamps. The heat accumulation index calculation unit is used to calculate the heat accumulation effect index of each electrical connection point based on the time series temperature data, using numerical integration and coupled with the heat conduction theoretical model, combined with ambient temperature compensation, and to normalize the index to quantify the degree of heat energy accumulation of the electrical connection point during operation. The fault early warning and discrimination unit is used to compare the thermal accumulation effect index with the early warning threshold output by the dynamic threshold model, and to perform multi-condition fusion judgment based on the changing trend of the thermal accumulation effect index to determine whether the early warning triggering conditions are met. The dynamic threshold model is pre-trained based on historical overheating fault data and normal operation data. The early warning signal output unit is used to generate and output an early warning signal when the early warning triggering conditions are met. The early warning signal includes the identifier of the overheated connection point, the current thermal accumulation effect index value, and the recommended handling measures, thereby completing the early warning triggering and signal output for the overheating fault of the electrical connection point.

[0019] Regarding the apparatus in the above embodiments, the specific manner in which each unit performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0020] To achieve the above objectives, a third aspect of this application provides an electronic device, including a processor and a memory; wherein the processor reads executable program code stored in the memory to run a program corresponding to the executable program code, for implementing an electrical connection overheating fault early warning method based on infrared thermal imaging as described in the first aspect embodiment.

[0021] To achieve the above objectives, a fourth aspect of this application provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements a method for early warning of electrical connection overheating faults based on infrared thermal imaging as described in the first aspect embodiment.

[0022] The embodiments of the present invention have the following beneficial effects: 1. Based on quantitative analysis of heat accumulation effect, the predictability and accuracy of overheat warning are significantly improved. This application constructs a time-series temperature integral calculation model based on heat conduction theory, and obtains a heat accumulation effect index by normalization combined with an environmental compensation mechanism. This index can accurately characterize the heat accumulation state of electrical connection points during long-term operation, breaking through the limitations of traditional technologies that rely solely on instantaneous temperature or a single temperature change rate. It can identify hidden overheating hazards in advance and effectively avoid problems such as delayed warnings and missed detections caused by ignoring the heat accumulation process.

[0023] 2. Employing a deep learning dynamic threshold model significantly enhances the adaptive capability for complex operating conditions. This application integrates multi-dimensional features such as heat accumulation index, ambient temperature, and equipment load to construct a neural network early warning model. The model is trained using massive amounts of historical normal operation data and fault data, enabling it to adaptively and dynamically update the early warning threshold based on changes in on-site operating conditions. This overcomes the technical shortcomings of traditional fixed thresholds, which cannot adapt to environmental fluctuations and dynamic load changes, ensuring the stability and reliability of early warning results under diverse operating scenarios.

[0024] 3. Based on high-precision image recognition technology, the integrity and accuracy of monitoring data are ensured. This application combines multi-scale image segmentation and geometric feature template matching algorithms to achieve precise positioning of electrical connection points. It can effectively remove interference from complex monitoring backgrounds, efficiently extract multi-dimensional temperature features of the target area, and construct a high-resolution time-series temperature dataset. This solves the problems of target positioning deviation and single temperature data in traditional temperature measurement methods, providing solid data support for subsequent thermal state analysis and fault diagnosis.

[0025] 4. Construct a closed-loop processing framework to ensure long-term stable system performance. This application integrates a complete technical system encompassing data preprocessing, feature index calculation, intelligent early warning judgment, data storage backtracking, and model iterative optimization. It improves the quality of raw data through filtering and noise reduction, anomaly removal, optimizes storage resources using time-series data compression technology, and continuously iterates and optimizes model accuracy based on historical operating data. This effectively solves the problems of incomplete data processing, model aging, and long-term performance degradation associated with traditional technologies, significantly improving the robustness and engineering applicability of the entire early warning solution. Attached Figure Description

[0026] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 A flowchart illustrating an electrical connection overheating fault early warning method based on infrared thermal imaging, provided as an embodiment of the present invention; Figure 2 A schematic diagram of the overall process of an electrical connection overheating fault early warning method based on infrared thermal imaging provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the temperature data extraction and preprocessing process provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the numerical calculation process for the thermal accumulation effect index provided in an embodiment of the present invention; Figure 5 This is a schematic diagram of the dynamic threshold model training and multi-level early warning judgment process provided in an embodiment of the present invention; Figure 6 This is a structural diagram of an electrical connection overheating fault early warning device based on infrared thermal imaging, provided in an embodiment of the present invention. Detailed Implementation

[0027] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0028] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. 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 should fall within the scope of protection of the present invention.

[0029] The following description, with reference to the accompanying drawings, describes an embodiment of the present invention of an electrical connection overheating fault early warning method and apparatus based on infrared thermal imaging.

[0030] Example 1 This invention provides a method for early warning of electrical connection overheating faults based on infrared thermal imaging. Figure 1 This is a schematic flowchart illustrating an electrical connection overheating fault early warning method based on infrared thermal imaging, provided in an embodiment of the present invention. Figure 1 As shown, the method includes the following steps: Step S1: Acquire an infrared thermal image sequence of the electrical connection area using an infrared thermal imaging device. The infrared thermal image sequence contains temperature information at multiple consecutive time points.

[0031] Combination Figure 2 Topmost process node Figure 3 The first image input node is used to carry out this step. In this embodiment of the application, the deployment of infrared thermal imaging equipment is completed first. Infrared thermal imagers are installed at fixed points in all key electrical connection areas in the substation and power distribution system. The monitoring points cover all electrical connection points such as disconnector contacts, circuit breaker terminals, and busbar joints, with no monitoring blind spots.

[0032] The infrared thermal imagers selected in this application are uniformly configured with an imaging resolution of 640×512 pixels, an optical lens focal length of 25mm, and a temperature measurement range covering -20℃ to 300℃, meeting the temperature measurement requirements of high-voltage electrical equipment under all operating conditions. The thermal imagers have built-in industrial Ethernet and 4G dual communication links, which can transmit the acquired images to the background data processing server in real time. The server hardware configuration includes a Xeon E5 processor, 64GB of RAM, 10TB of time-series storage media, and runs on the CentOS operating system to support batch image processing.

[0033] Image sequence acquisition adopts a dynamic adaptive acquisition frequency mechanism. The basic acquisition time interval is set to 10 seconds, and the acquisition interval can be dynamically adjusted within the range of 0.5 seconds to 30 seconds according to the working conditions. The frequency adjustment logic relies on two sets of real-time monitoring parameters: one is the change trend of the heat accumulation effect index calculated and output in real time, and the other is the ambient temperature value collected by the temperature sensor around the equipment.

[0034] In the actual operation logic of this application embodiment, if the instantaneous change rate of the thermal accumulation effect index is detected to be higher than 0.5℃ / min, or the ambient temperature fluctuation exceeds 5℃ within 5 minutes, it will automatically switch to high-frequency acquisition mode, shorten the image acquisition interval to 0.5 seconds, and acquire image frames at high density to capture details of rapid temperature changes; if the change rate of the thermal accumulation effect index is consistently lower than 0.1℃ / min for 30 consecutive minutes and the ambient temperature does not fluctuate significantly, the acquisition frequency will be automatically reduced and the interval will be lengthened to 30 seconds to save server computing power and storage resources.

[0035] After acquiring the raw infrared thermal image, the image preprocessing workflow is started, corresponding to... Figure 3The second process node involves preprocessing in two stages: the first stage uses a 3×3 window mid-range filtering algorithm to remove random salt-and-pepper noise generated during image acquisition and transmission; the second stage uses a contrast-limited adaptive histogram equalization algorithm to stretch the local temperature contrast of the image, making the temperature distribution boundaries in the electrical connection point area clearer. Simultaneously, the system periodically performs online calibration of the thermal imager using a blackbody radiation reference source, controlling the temperature measurement error within 1℃, ensuring the clarity of the infrared thermal images and the accuracy of temperature measurement throughout the process. Finally, it outputs a continuous sequence of infrared thermal images with precise timestamps and a resolution of 640×512, which is then transferred to the S2 execution stage.

[0036] Step S2: Based on the infrared thermal image sequence, identify and locate the electrical connection points in each frame of the image and extract the corresponding temperature values. After associating the timestamps, form the time series temperature data of each connection point.

[0037] The complete execution process for this step corresponds to the attached document. Figure 3 The entire process involves receiving pre-processed infrared thermal images output by S1 frame by frame, and performing eight continuous processes: electrical connection point identification, template matching and positioning, temperature value extraction, timestamp association, data quality control, ambient temperature compensation, raw data time-series storage, and data integrity verification.

[0038] The first step is to identify the electrical connection points, corresponding to... Figure 3 In the third step, this embodiment employs a specially trained U-Net convolutional neural network for region segmentation. This network comprises a dual structure of encoder and decoder. The encoder extracts image temperature grayscale features through multi-layer convolution and pooling operations, while the decoder restores pixel spatial details through upsampling and skip connections. During model training, over 5000 labeled thermal images of electrical joints in substations are input, with a learning rate of 0.001 and a training iteration period of 100 epochs. The final test set segmentation and recognition accuracy reaches 96.5%. After inputting a single frame of infrared thermal image into the network, the segmentation mask for each suspected electrical connection point is output. The segmentation confidence threshold is set to 0.92, and only target regions with a confidence level higher than the threshold are retained as segmentation seed points.

[0039] The second step involves template matching for precise localization, corresponding to... Figure 3 The fourth node pre-stores various geometric feature templates specific to electrical connection points, including circular features of bolt joints, rectangular terminal features, and long strip features of busbar overlaps. An improved normalized cross-correlation algorithm is used to calculate the similarity between the image segmentation region and the template. The similarity judgment threshold is set to 0.85. Only regions with scores exceeding the threshold are judged as real electrical connection points, eliminating false segmentation regions caused by noise.

[0040] The third step completes the temperature value extraction, corresponding to... Figure 3The fifth node reads three types of core temperature features from the pixel area locked by template matching: the average temperature of the area, the highest temperature of the area, and the pixel temperature distribution gradient, with units of ℃, ℃, and ℃ / cm, respectively. The three types of temperature values ​​are synchronously cached in the memory buffer.

[0041] The fourth step is to associate the temperature data with the timestamp, corresponding to... Figure 3 The sixth node binds the precise time stamp of the image acquisition to each group of extracted temperature values, establishing a single-connection-point independent original time-series dataset to achieve a one-to-one correspondence between temperature values ​​and acquisition time.

[0042] The fifth step involves performing data quality control procedures, corresponding to... Figure 3 The seventh step involves multi-level noise reduction and anomaly removal processing of the raw time-series temperature data with timestamps. First, a Kalman filter algorithm is used to smooth short-term random temperature fluctuations. The Kalman filter comprises two core formulas: a state prediction equation and a state update equation. State prediction equation:

[0043] State update equation:

[0044] in, A represents the prior state estimate at the current moment, and A is the state transition matrix. This is an estimate of the optimal state at the previous time step. For external control input items; This is the optimal posterior state estimate for this moment. It is the Kalman filter gain. These are actual temperature observations measured by an infrared thermal imager. This is the observation transformation matrix. The embodiments of this application optimize the filtering parameters based on the characteristics of the temperature sensor, and the process noise covariance... The observation noise covariance R is fixed at 0.01.

[0045] Outlier removal uses a modified 3σ criterion, and the criterion formula is as follows:

[0046] In the formula The average temperature over 30 consecutive sampling sliding windows. The standard deviation of temperature within the window is used; if the output outlier=1, the sampling point is determined to be an outlier, and the normal sampling values ​​before and after the point are automatically used for linear interpolation to replace it, thus eliminating temperature jump interference.

[0047] The sixth step is to perform ambient temperature compensation correction, corresponding to... Figure 3 The formula for calculating the basic compensation at the eighth node is as follows:

[0048] In the formula It is an infrared thermal imager that reads the raw temperature. The correction factor is used to match the metal material of the connection point. To synchronously collect ambient temperature, The preset standard reference temperature; in the preferred embodiment of this application, a load current variation correction term is further introduced, and the correction formula is extended as follows:

[0049] in, This is the load change correction factor. To measure the real-time rate of change of the equipment load current, two sets of correction coefficients will be dynamically and adaptively adjusted according to the equipment model and historical operating data, completely eliminating the temperature measurement offset caused by environmental temperature changes and load fluctuations, and obtaining the corrected true temperature reading.

[0050] The seventh step is to store all the calibrated temperature time-series data into the InfluxDB time-series database. Figure 3 The ninth node uses a three-layer independent measurement storage structure: `temperature_raw` stores uncorrected raw temperature data, `indicator_calculated` stores subsequently calculated derived indicators, and `alert_log` stores all alert event records. The database uses the Gorilla time-series compression algorithm to store data, with the core XOR operation formula being: Differential compression is applied to continuous time-series values, achieving a temperature data compression ratio of over 10:1, thus saving server storage space.

[0051] The eighth step is to perform data integrity verification, corresponding to... Figure 4 The tenth node verifies the temporal continuity and valid range of values ​​of the time series dataset in batches, removes abnormal data segments with missing or out-of-order data, and finally outputs a complete, reliable, high-resolution time series temperature dataset with unique connection point identifiers, which is then transmitted to the S3 heat accumulation effect calculation step.

[0052] Step S3: Based on the time series temperature data, numerical integration is used and coupled with the heat conduction theoretical model, combined with ambient temperature compensation, to calculate the heat accumulation effect index of each electrical connection point, and the index is normalized to quantify the degree of heat energy accumulation of the electrical connection point during operation.

[0053] The complete calculation logic for this step is attached. Figure 4All process nodes receive the corrected time-series temperature data output by S2 and sequentially execute nine calculation steps: numerical integration calculation, solving the two-dimensional unsteady heat conduction model, dynamically selecting the integration time window, secondary ambient temperature compensation, assigning the material thermal diffusivity, solving the finite difference instantaneous heat flux density, evaluating the heat capacity change, index normalization, and outputting the heat accumulation effect index. The entire process relies on heat transfer numerical algorithms to quantify the total accumulated heat energy at the connection point.

[0054] The first step is to perform numerical integration calculations, corresponding to Figure 4 The second point is that this application defines the thermal accumulation effect index using an integral-based formula:

[0055] in, The index represents the cumulative heat effect at time t, measured in °C. min; For electrical connection points at time The infrared measured temperature; The ambient temperature was collected synchronously at the same time. It is a time-varying thermal conductivity coefficient, the value of which is determined by the metal material of the connection point and the contact and tightening state of the bolts. This marks the start time of the monitoring and scoring.

[0056] To adapt to discrete numerical computation in computers, the system uses the composite Simpson's rule to perform discrete solutions of definite integrals, defining the integration interval... Divide the data into n equal sub-intervals, with each sub-interval lasting for a specified duration. The formula for calculating the discretized integral is:

[0057] In the formula The Simpson weighting coefficients are assigned according to the following rules: interval endpoints. or hour, When i is an odd number When i is an even number This numerical integration scheme effectively reduces the time consumed by server floating-point operations while ensuring computational accuracy.

[0058] The second step is to couple a two-dimensional unsteady heat conduction theoretical model, corresponding to... Figure 4 At the third node, the partial differential equation for heat conduction is:

[0059] in, The thermal diffusivity of the connection point material. , Let be the second-order partial derivatives of temperature in the x and y directions in two-dimensional space, respectively. The partial differential equation is transformed into an iteratively computable difference equation using the finite difference method:

[0060] In the formula, Represents spatial grid coordinates The temperature value at discrete time n, The step size of the two-dimensional spatial grid. The time step for numerical calculation is defined as follows: In the 500kV substation implementation scenario, the spatial step is 0.5cm, the time step is fixed at 1 second, and the thermal diffusivity of the contact copper material is... Set as .

[0061] The third step is to dynamically select the integration time window, corresponding to... Figure 4 The fourth node retrieves historical runtime sequence data of the device stored in InfluxDB, automatically matches and adapts the integration duration, and sets the window range from 1 minute to 60 minutes. For devices that have been running stably for a long time, a long window of 60 minutes is selected, and for devices with frequent temperature fluctuations and early signs of failure, a medium window of 15 to 30 minutes is selected. For short-term instantaneous temperature rise conditions, a short window of 1 minute is enabled.

[0062] The fourth step introduces a secondary environmental temperature compensation factor, corresponding to Figure 4 The fifth step involves making an environmental temperature change correction on the intermediate results of the integral calculation to offset the interference of external environmental heat such as day-night temperature difference, ventilation, and sunshine on the heat energy accumulation index, ensuring that the index only reflects the heat accumulation of the electrical connection point itself.

[0063] The fifth step is to configure the material's thermal diffusivity parameters, corresponding to... Figure 4 In the sixth node, the system automatically matches the thermal diffusivity value of the corresponding metal conductor based on the equipment material ledger stored in the database, and substitutes it into the thermal conduction difference equation for iterative solution.

[0064] Step 6: Calculate the instantaneous heat flux density distribution, corresponding to Figure 4 The seventh node outputs the heat flux density distribution value in the two-dimensional plane of the connection point based on the finite difference iteration results, distinguishing between the normal operating baseline heat flux density (500W / m²) and the abnormal overheating heat flux density (1500W / m² and above), to assist in the subsequent early warning trend judgment.

[0065] Step 7: Assess the change in heat capacity at the connection point, corresponding to... Figure 4 The eighth node combines temperature time-series changes and material specific heat capacity parameters to simultaneously output the heat capacity change curve as temperature increases, thus improving the comprehensive quantitative basis for heat energy accumulation.

[0066] The eighth step is to perform indicator normalization processing, corresponding to... Figure 5The ninth node unifies the original integration results output by different integration windows and equipment of different materials to the same dimensional standard, eliminating numerical deviations caused by equipment model and monitoring duration.

[0067] Step 9 outputs the normalized heat accumulation effect index ,correspond Figure 5 At the tenth node, the indicator value, first derivative, and second derivative are simultaneously packaged and transmitted to the S4 dynamic threshold and early warning judgment step. The discrete calculation formulas for the first and second derivatives are as follows: First derivative (rate of change) of the heat accumulation index:

[0068] Second derivative of the heat accumulation index (acceleration of change):

[0069] Step S4: Compare the heat accumulation effect index with the warning threshold output by the dynamic threshold model, and combine the changing trend of the heat accumulation effect index to perform multi-condition fusion judgment to determine whether the warning triggering conditions are met. The dynamic threshold model is pre-trained based on historical overheating fault data and normal operation data.

[0070] This step involves a complete process of model training, threshold updating, and multi-condition fusion judgment, as detailed in the attached document. Figure 5 All process nodes receive the heat accumulation effect index output by S3. Ambient temperature The system uses three sets of input features: equipment load current I. It involves eight processes: deep learning model training, model periodic automatic update, multi-level dynamic threshold output, indicator trend derivative analysis, C4.5 decision tree multi-condition fusion, early warning trigger judgment, three-level early warning level classification, and early warning decision output.

[0071] The first step is to input the features into the deep learning neural network, corresponding to... Figure 5 In the first process node, this embodiment constructs a three-layer fully connected neural network with hidden layers as the basic model for dynamic thresholding. The network hierarchy is fixed: the input layer has 3 neurons, each corresponding to a thermal accumulation effect index. Ambient temperature The system has three types of input features: device current load, device current load, and device current load. The first hidden layer has 64 neurons, the second hidden layer has 32 neurons, and the third hidden layer has 16 neurons. The output layer has 3 neurons, corresponding to the first, second, and third level warning dynamic thresholds, respectively.

[0072] All hidden layers use the ReLU activation function, the function expression of which is: The output layer is paired with the Softmax activation function to map the network output values ​​to the probability distributions corresponding to the three warning levels. The Softmax calculation formula is as follows:

[0073] During the model training phase, the mean squared error (MSE) is used as the loss function. The loss calculation formula is as follows:

[0074] in, This represents the total number of training samples in a single iteration. For the sample's true warning label, The neural network model is used to predict the output threshold; the Adam algorithm is selected as the model optimizer, with a fixed learning rate of 0.001 and a complete training iteration cycle of 500 epochs.

[0075] The second step is to configure the model training dataset and online performance monitoring mechanism, corresponding to... Figure 5 The branch parallel nodes consist of two types of data in the training sample set: time-series data of normal operating conditions of equipment that has been running stably for a long time, and full-process feature data of electrical connection overheating faults that have occurred in history. The samples cover a variety of complex operating conditions with different loads, day and night temperature changes, and different equipment models, ensuring the generalization ability of the model. The model performance monitoring module is started in parallel to continuously collect real-time early warning results and manually review fault records, calculate the model early warning accuracy periodically, and automatically fine-tune the network weights to complete parameter optimization.

[0076] The third step is to set up the automatic model update logic, corresponding to... Figure 5 The main process threshold is dynamically updated. The neural network model is retrained every 24 hours. The training set automatically extracts the complete equipment runtime sequence data of the most recent 30 days from the database to adapt to changes in operating conditions caused by equipment aging and seasonal temperature changes, so as to realize real-time adaptive updates of the warning threshold.

[0077] The fourth step outputs multi-level dynamic early warning thresholds, corresponding to... Figure 5 Multi-level threshold output nodes, the updated neural network is based on real-time input. , I. It dynamically outputs three sets of independent threshold values: Level 1 warning threshold, Level 2 warning threshold, and Level 3 warning threshold. The threshold values ​​fluctuate in real time according to the operating conditions, which is different from the traditional fixed threshold warning scheme.

[0078] The fifth step is to conduct a trend analysis of the changes in heat accumulation indicators, corresponding to... Figure 5 The trend analysis node retrieves the first and second derivatives of the indicators output by S3, which respectively represent the real-time rate of increase and the acceleration of rate change of the heat accumulation indicator, and distinguish three types of trend characteristics: steady and slow temperature rise, continuous and accelerating temperature rise, and instantaneous and abrupt temperature rise.

[0079] The sixth step uses a C4.5 decision tree to complete the multi-condition fusion judgment, corresponding to... Figure 5 Multi-condition fusion judgment nodes, and feature selection in decision trees is completed based on the information gain ratio formula:

[0080] In the formula, S is the set of feature samples currently involved in the judgment, and A is a single-class input feature attribute. The decision tree incorporates all core judgment features: the absolute value of the thermal accumulation effect index, the duration of the positive first derivative of the index, the ambient temperature compensation factor, the rate of change of the load current, and the magnitude of the temperature gradient exceeding the standard. It also integrates the multi-level dynamic thresholds output by the neural network, and distinguishes between normal small fluctuations in equipment temperature and true overheating fault precursors through multi-dimensional cross-judgment, which greatly reduces the probability of false alarms and false alarms caused by single threshold comparison.

[0081] Step 7: Set standardized early warning triggering rules, corresponding to Figure 5 The warning trigger condition node is determined to meet the warning trigger standard if either of the following two conditions is met: First, the current value of the heat accumulation effect index exceeds the dynamic threshold of the corresponding working condition, and the exceeding state continues for a preset time period; Second, the first derivative of the heat accumulation index remains positive for a long time, the second derivative remains greater than 0, and the temperature rise trend continues to be unstable. Even if the index has not yet completely exceeded the threshold, the warning is determined to be triggered in advance.

[0082] The eighth step is to divide the warning levels into three standardized levels, corresponding to... Figure 2 The warning level classification is based on two dimensions: the magnitude of the indicator exceeding the standard and the duration of the exceedance. Level 1 warning is when the indicator exceeds the corresponding dynamic threshold by less than 10% and the duration is less than 5 minutes, and only a text notification is pushed to the software. Level 2 warning is when the indicator exceeds the threshold by 10% to 30% and the duration is 5 to 15 minutes, and the on-site audible and visual warning device is activated simultaneously. Level 3 warning is when the indicator exceeds the threshold by more than 30% or the exceedance duration is more than 15 minutes, which is determined to be a major overheating defect, and an emergency alarm is immediately pushed and the message is directly pushed to the maintenance personnel's terminal.

[0083] The ninth step outputs the final early warning decision result, corresponding to Figure 6 The early warning decision output node packages and transmits the determined early warning level, triggering characteristics, and duration of exceeding the limit to the S5 early warning signal generation and output step.

[0084] Step S5: When the warning triggering conditions are met, a warning signal is generated and output. The warning signal includes the identifier of the overheated connection point, the current thermal accumulation effect index value, and the recommended handling measures, thus completing the warning triggering and signal output for the overheating fault of the electrical connection point.

[0085] When the S4 output determines that the warning triggering condition is met, the entire process of generating structured warning messages, dual-channel redundant transmission, on-site audible and visual alarms, platform data synchronization, and event archiving is initiated, with complete logical continuity. Figure 6 The last two process nodes.

[0086] The first step is to generate a standardized JSON-formatted structured early warning message. The message contains all the core fields: a unique electrical connection point device identifier, two-dimensional spatial coordinates of the connection point, sampling curve data of the thermal accumulation effect index for the most recent 60 minutes (sampling interval of 1 minute), real-time ambient temperature, real-time equipment load current, real-time value of the normalized thermal accumulation effect index, three-level early warning level identifier, and corresponding recommended maintenance and operation measures. The curve field fully stores the entire process of time-series index changes, making it easy for maintenance personnel to trace the temperature rise process.

[0087] The second step is to build a dual-channel redundant transmission architecture to ensure that early warnings are not lost in the event of a network outage. The main transmission channel adopts the MQTT communication protocol, and the format of the early warning messages fully complies with the ISO / IEC20922 industry standard, pushing the structured JSON early warning messages to the cloud remote monitoring platform in real time. The backup transmission channel adopts the ModbusTCP industrial communication protocol, synchronously sending the same early warning data packets to the local monitoring and operation terminal of the substation, realizing the synchronous push of early warnings from both the cloud and local systems.

[0088] The third step involves linking the on-site audible and visual alarm hardware devices and adjusting the alarm output mode according to the warning levels defined by S4: Level 1 warnings do not trigger on-site audible and visual devices, but only push notifications online; Level 2 warnings activate low-frequency flashing and medium-decibel alarm sounds; Level 3 warnings activate 2Hz high-frequency flashing and 85dB high-decibel continuous alarm sounds to intuitively remind on-site personnel to handle overheating defects.

[0089] The fourth step is to complete the standardized data integration across multiple platforms. Based on a microservice architecture, it provides a standardized RESTful API interface with communication protocols compatible with HTTP / HTTPS. The data exchange encoding uniformly adopts the UTF-8 character set. The API interface includes four core functions: device registration, real-time data reporting, device status query, and remote algorithm parameter distribution. At the same time, it has a built-in power industry protocol conversion gateway, which is compatible with the IEC61850 substation communication protocol and the DNP3 distributed network protocol. It automatically completes the data semantic mapping and format conversion between different monitoring systems, and can seamlessly connect to existing substation automation systems and plant energy management platforms. It automatically pushes early warning briefings to the platform and supports maintenance personnel to remotely retrieve complete time-series temperature and indicator calculation raw data.

[0090] The fifth step is to complete the full archiving and storage of the early warning event. The entire process data of this early warning is stored in the alert_log storage space of the InfluxDB time series database. The archived content includes: the original infrared thermal image frame, the preprocessed and corrected temperature sequence, the complete calculation process of the thermal accumulation effect index, the output results of the neural network dynamic threshold, the decision tree multi-condition judgment features, the structured early warning message, the alarm trigger time, and the operation and maintenance handling records. The background automatically generates equipment operation health analysis reports on a regular basis. The reports include full-cycle early warning statistics, equipment heat defect distribution, and system algorithm operation performance indicators, providing complete data support for operation and maintenance personnel to conduct long-term equipment status assessments and subsequent model iteration optimizations.

[0091] If maintenance personnel conduct on-site inspections based on the early warning prompts and confirm that there are overheating defects such as loose or oxidized electrical connection points and complete the repairs, the system will continuously collect temperature time-series data after the repairs are completed, automatically update the equipment's historical operating dataset, participate in the daily automatic retraining of the subsequent neural network model, continuously optimize the accuracy of dynamic threshold judgment, and form a complete closed-loop early warning system of monitoring-calculation-early warning-handling-model optimization.

[0092] Example 2 This invention provides an electrical connection overheating fault early warning device based on infrared thermal imaging. ​ This is a schematic diagram of an electrical connection overheating fault early warning device based on infrared thermal imaging, provided as an embodiment of the present invention. ​ As shown, the device includes: The image acquisition unit 100 is used to acquire an infrared thermal image sequence of the electrical connection area through an infrared thermal imaging device, the infrared thermal image sequence containing temperature information at multiple consecutive time points; The time-series temperature extraction unit 200 is used to identify and locate electrical connection points in each frame of the infrared thermal image sequence and extract the corresponding temperature values, and form time-series temperature data for each connection point after associating timestamps. The heat accumulation index calculation unit 300 is used to calculate the heat accumulation effect index of each electrical connection point based on the time series temperature data, using numerical integration and coupled with the heat conduction theoretical model, combined with the ambient temperature compensation, and to normalize the index to quantify the degree of heat energy accumulation of the electrical connection point during operation. The fault early warning discrimination unit 400 is used to compare the heat accumulation effect index with the early warning threshold output by the dynamic threshold model, and to perform multi-condition fusion judgment based on the changing trend of the heat accumulation effect index to determine whether the early warning triggering conditions are met. The dynamic threshold model is pre-trained based on historical overheating fault data and normal operation data. The warning signal output unit 500 is used to generate and output a warning signal when the warning triggering conditions are met. The warning signal includes the identifier of the overheated connection point, the current thermal accumulation effect index value, and the recommended handling measures, thereby completing the warning triggering and signal output for the overheating fault of the electrical connection point.

[0093] Regarding the apparatus in the above embodiments, the specific manner in which each unit performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0094] Example 3 To implement the methods of the above embodiments, the present invention also provides an electronic device, which includes a memory and a processor; wherein the processor reads executable program code stored in the memory to run a program corresponding to the executable program code, so as to implement the various steps of the methods described above.

[0095] Example 4 To implement the above embodiments, this application also proposes a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the method described in the foregoing embodiments.

[0096] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0097] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0098] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

Claims

1. A method for early warning of electrical connection overheating faults based on infrared thermal imaging, characterized in that, include: The infrared thermal image sequence of the electrical connection area is acquired by an infrared thermal imaging device, and the infrared thermal image sequence contains temperature information at multiple consecutive time points. Based on the infrared thermal image sequence, electrical connection points in each frame of the image are identified and located, and the corresponding temperature values ​​are extracted. After associating the timestamps, time-series temperature data of each connection point is formed. Based on the time-series temperature data, numerical integration coupled with a heat conduction theoretical model is used, combined with ambient temperature compensation, to calculate the heat accumulation effect index of each electrical connection point, and the index is normalized to quantify the degree of heat energy accumulation at the electrical connection point during operation. The heat accumulation effect index is compared with the warning threshold output by the dynamic threshold model, and the change trend of the heat accumulation effect index is combined to perform multi-condition fusion judgment to determine whether the warning triggering conditions are met. The dynamic threshold model is pre-trained based on historical overheating fault data and normal operation data. When the warning triggering conditions are met, a warning signal is generated and output. The warning signal includes the identifier of the overheated connection point, the current thermal accumulation effect index value, and the recommended handling measures, thus completing the warning triggering and signal output for the overheating fault of the electrical connection point.

2. The method according to claim 1, characterized in that, A sequence of infrared thermal images of electrical connection areas is acquired using infrared thermal imaging equipment, including: Based on the real-time calculated trend of the heat accumulation effect index and combined with the data fed back from the ambient temperature sensor, the image acquisition frequency of the infrared thermal imaging equipment is dynamically adjusted. When the thermal accumulation effect index shows a rapid upward trend, the acquisition time interval is automatically shortened to increase the data acquisition density. The acquired infrared thermal images are subjected to denoising and enhancement processing to improve image clarity and the accuracy of temperature measurement.

3. The method according to claim 2, characterized in that, Based on the infrared thermal image sequence, electrical connection points in each frame are identified and located, and their corresponding temperature values ​​are extracted. After associating timestamps, time-series temperature data for each connection point is generated, including: A multi-scale image segmentation algorithm is used to process each frame of infrared thermal image to divide candidate connection point regions; The candidate connection point regions are matched with pre-stored connection point geometric feature templates to determine the precise location of each electrical connection point in the image; Based on the positioning results, the temperature value is read from the corresponding location, wherein the temperature value is selected from at least one of the average temperature, the maximum temperature, or the temperature distribution gradient. The extracted temperature values ​​are associated with timestamps to form high-resolution time-series temperature data for each connection point.

4. The method according to claim 3, characterized in that, Based on the aforementioned time-series temperature data, numerical integration coupled with a heat conduction theoretical model, combined with ambient temperature compensation, is used to calculate the heat accumulation effect index of each electrical connection point, including: Numerical integration is performed on the time-series temperature data of each electrical connection point to obtain the cumulative amount of temperature change over time; Based on the heat conduction theory model, the instantaneous heat flux density and heat capacity changes are solved by combining the cumulative quantity; Dynamically select the integration time window based on the equipment's operating history; An ambient temperature compensation factor is introduced to correct the integral results. The corrected integral results are then normalized to output a normalized thermal accumulation effect index.

5. The method according to claim 4, characterized in that, The pre-training process of the dynamic threshold model includes: Historical overheating fault data and normal operation data were collected as a training sample set; A deep learning neural network is constructed using the heat accumulation effect index, ambient temperature and equipment current load in the training sample set as input features and the corresponding warning threshold as the output target. By iteratively learning the training sample set through a neural network, a dynamic threshold model is generated, enabling the model to output multi-level early warning thresholds under complex working conditions.

6. The method according to claim 5, characterized in that, The heat accumulation effect index is compared with the warning threshold output by the dynamic threshold model, and a multi-condition fusion judgment is performed based on the changing trend of the heat accumulation effect index to determine whether the current warning triggering conditions are met, including: The current heat accumulation effect index is compared with the warning threshold output by the dynamic threshold model to determine whether the heat accumulation effect index exceeds the warning threshold. Calculate the first and second derivatives of the thermal accumulation effect index and analyze the stability of its changing trend; Based on the comparison results and trend stability analysis results, a decision tree algorithm is used to perform multi-condition fusion judgment; When the heat accumulation effect index exceeds the warning threshold and continues for a predetermined period of time, or when the trend of change meets the preset instability conditions, the warning trigger condition is determined to be met.

7. The method according to claim 6, characterized in that, When the warning triggering conditions are met, a warning signal is generated and output, including: The warning level is determined based on the magnitude and duration by which the current heat accumulation effect index exceeds the warning threshold; Generate a structured early warning message containing the spatial coordinates of the overheated connection point, historical change curves of the heat accumulation effect index, current environmental conditions, and early warning level; Structured early warning messages are pushed to the cloud monitoring platform or local display terminal in real time via wireless communication protocols, simultaneously triggering audible and visual alarm devices to complete the on-site output of early warning signals.

8. An electrical connection overheating fault early warning device based on infrared thermal imaging, characterized in that, include: An image acquisition unit is used to acquire a sequence of infrared thermal images of an electrical connection area using an infrared thermal imaging device. The sequence of infrared thermal images contains temperature information at multiple consecutive time points. The time-series temperature extraction unit is used to identify and locate electrical connection points in each frame of the infrared thermal image sequence and extract the corresponding temperature values, and form time-series temperature data for each connection point after associating timestamps. The heat accumulation index calculation unit is used to calculate the heat accumulation effect index of each electrical connection point based on the time series temperature data, using numerical integration and coupled with the heat conduction theoretical model, combined with ambient temperature compensation, and to normalize the index to quantify the degree of heat energy accumulation of the electrical connection point during operation. The fault early warning and discrimination unit is used to compare the thermal accumulation effect index with the early warning threshold output by the dynamic threshold model, and to perform multi-condition fusion judgment based on the changing trend of the thermal accumulation effect index to determine whether the early warning triggering conditions are met. The dynamic threshold model is pre-trained based on historical overheating fault data and normal operation data. The early warning signal output unit is used to generate and output an early warning signal when the early warning triggering conditions are met. The early warning signal includes the identifier of the overheated connection point, the current thermal accumulation effect index value, and the recommended handling measures, thereby completing the early warning triggering and signal output for the overheating fault of the electrical connection point.

9. An electronic device, characterized in that, Including processor and memory; The processor reads executable program code stored in the memory to run a program corresponding to the executable program code, so as to implement the method as described in any one of claims 1-7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-7.

Images