A Fault Detection Method for High-Resolution Infrared Thermal Imaging Arrays

Abstract

The present invention discloses a fault detection method for a high-resolution infrared thermal imaging array, which relates to the technical field of power equipment monitoring. The fault detection method for the high-resolution infrared thermal imaging array obtains a thermal imaging image of the surface of the power equipment through a high-resolution infrared thermal imaging device, dynamically corrects the temperature field by combining environmental parameters and load current, and generates a temperature matrix free of interference. A graph neural network is used to model the heat conduction relationship between devices, and a U-Net segmentation network is used to detect temperature abnormal regions, generating an abnormal region mask and temperature spatial features. The dynamic time warping algorithm is used to analyze the temporal correlation between temperature fluctuations and electrical parameters to distinguish continuous faults from instantaneous interference. Based on few-shot transfer learning, combined with a pre-trained feature extractor and an attention mechanism, fault type classification is performed, and the fault type and confidence level are output. The efficient and accurate detection of power equipment faults is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of power equipment monitoring, and particularly to a fault detection method for a high-resolution infrared thermal imaging array. Background Art

[0002] With the continuous growth of energy demand, the reliability and safety of power equipment have become increasingly important. As a key device in power transmission and distribution, the operating state of a transformer directly affects the stability and safety of the power system. Therefore, fault detection and early warning of transformers play a crucial role in the power industry. Especially in the rapid development of smart grids and new energy fields, real-time monitoring and fault diagnosis of transformers are particularly urgent. Due to its non-contact, high-sensitivity, and ability to real-time monitor the surface temperature change of equipment, high-resolution infrared thermal imaging technology has become an important tool in fault diagnosis. However, in complex power equipment and environmental conditions, how to accurately and quickly detect faults and determine the fault type is still a challenge faced by current technologies.

[0003] Currently, traditional fault detection methods mainly rely on sensor data collection and empirical models. These methods usually have the following disadvantages: First, sensor-based monitoring methods are difficult to comprehensively capture the thermal characteristics of equipment and are easily affected by environmental noise; Second, existing fault diagnosis technologies mostly rely on single temperature or electrical parameters, lacking comprehensive analysis of multi-dimensional data such as temperature, load, and environment, resulting in limited accuracy of detection results; In addition, data processing and feature extraction in traditional methods mostly rely on manual design, lacking flexibility and self-adaptability. Although deep learning and image segmentation technologies have been proposed to improve the accuracy of fault detection, in the real-time monitoring of power equipment, how to effectively fuse information from different data sources and perform accurate classification is still an urgent problem to be solved. Summary of the Invention

[0004] Aiming at the deficiencies of the prior art, the present invention provides a fault detection method for a high-resolution infrared thermal imaging array, which solves the problems in the above background art.

[0005] To achieve the above objectives, the present invention is realized through the following technical solutions: A fault detection method for a high-resolution infrared thermal imaging array, comprising the following steps: S1. Obtain the thermal imaging image of the transformer surface through a high-resolution infrared thermal imaging device, and dynamically correct the infrared temperature field in combination with environmental parameters and the transformer load current to generate a temperature matrix free of interference; S2. According to the temperature matrix and the physical topology map of the battery pack, model the thermal conduction relationship between batteries through a graph neural network, and combine the U-Net segmentation network to detect the temperature anomaly area, identify the anomaly area that conforms to the thermal conduction law, and generate an anomaly area mask and temperature spatial features under spatial structure constraints; S3. Extract the temperature fluctuation time series data of the anomaly area mask, and synchronously match the real-time current and voltage changes of the battery pack. Quantify the correlation between temperature and electrical parameters through the dynamic time warping algorithm, distinguish continuous faults from instantaneous interference, and output the fault detection index and evolution mode label; S4. According to the anomaly area mask, the fault detection index and the evolution mode label, classify the battery fault types through few-shot transfer learning. Through a pre-trained feature extractor, fine-tune the classifier in combination with few-shot battery fault data, and combine the attention mechanism to weight and fuse the temperature spatial features and the evolution mode label, and output the fault type and its confidence level.

[0006] Further, the specific process of dynamically correcting the infrared temperature field is as follows: Establish a battery load current-temperature characteristic curve through the battery load current and voltage data, in combination with the environmental temperature and the heat dissipation wind speed; Dynamically correct the infrared temperature field through a Kalman filter to eliminate the temperature error caused by environmental factors and generate infrared temperature field data free of interference; Calculate and correct the temperature field data to eliminate the temperature rise effect caused by normal charge and discharge conditions, and dynamically correct the infrared temperature field.

[0007] Further, the specific process of generating a temperature matrix free of interference is as follows: Conduct a multiple regression analysis on the corrected temperature field data and environmental factors; Based on the battery load current-temperature characteristic curve, eliminate the temperature rise effect under normal conditions; Further optimize the temperature data to remove the interference caused by environmental noise and generate a temperature matrix free of interference.

[0008] Furthermore, according to the temperature matrix and the physical topology of the battery pack, the graph neural network is used to model the heat conduction relationship between batteries, and the U-Net segmentation network is used to detect the temperature abnormal areas. The specific process of identifying abnormal areas that conform to the law of heat conduction is as follows: According to the temperature matrix and the physical topology of the battery pack, a graph neural network model of heat conduction between batteries is established to simulate the thermal coupling relationship between each single battery in the battery pack, and the heat flow path between batteries is determined based on the battery position and heat conduction characteristics; the battery pack topology map is combined with the temperature field data, and the heat conduction law between batteries is learned through the graph neural network to optimize the spatial distribution prediction of the temperature field; based on the U-Net segmentation network, the abnormal areas in the temperature field are identified, and combined with the heat conduction law between batteries, those areas with abnormal temperature and drastic fluctuations on the heat conduction path are preferentially identified, especially the temperature difference mutation area between adjacent batteries; the temperature abnormal area is output through the detection results of the temperature abnormal area, and the specific location of the area is calibrated.

[0009] Furthermore, the specific process of generating abnormal area masks and temperature spatial features under spatial structure constraints is as follows: based on the output of the U-Net segmentation network, the spatial structure mask of the temperature abnormal area is generated by combining it with the topological map of the battery pack. The mask marks all abnormal areas where the temperature exceeds the threshold, and further eliminates noise through image post-processing; the spatial features of the abnormal area are extracted, including area, shape factor, and boundary complexity, which are used to describe the geometric characteristics of the abnormal area; based on the spatial distribution characteristics of the temperature field, the spatial temperature features related to the fault are further extracted, including temperature gradient and temperature fluctuation amplitude.

[0010] Furthermore, the correlation between temperature and electrical parameters is quantified by a dynamic time warping algorithm, and the specific process of distinguishing between persistent faults and transient interference is as follows: extracting the time series data of temperature fluctuations from the abnormal area mask, and obtaining the temperature fluctuation characteristics of each abnormal area within a certain time interval by performing time series analysis on the dynamic changes of the temperature abnormal area; synchronously collecting the electrical parameters of the battery pack including real-time current and voltage data, and aligning the electrical parameters with the temperature fluctuation time series data to ensure the time consistency of the temperature data and the electrical data; calculating the time series correlation coefficient of the temperature fluctuation data and the electrical parameters through the dynamic time warping algorithm, and quantifying the time delay relationship between the temperature fluctuation and the electrical parameters; judging the fault type based on the analysis results of the dynamic time warping algorithm and the relationship between the temperature fluctuation and the electrical parameters; if the temperature fluctuation and the electrical parameter changes are synchronized and continuous, it is determined to be a persistent fault; if the temperature fluctuation and the electrical parameter changes are short-lived and irregular, it is determined to be a fluctuation caused by transient interference.

[0011] Further, the specific process of outputting the fault detection index and the evolution mode label is as follows: Calculate the temporal correlation coefficient between the temperature fluctuation data and the electrical parameters, and generate the fault detection index in combination with the operating state of the battery pack; Generate the fault evolution mode label by the evolution mode classification algorithm according to the temporal characteristics of the temperature fluctuation and the change trend of the electrical data, and classify the fault into a continuous type or an intermittent type.

[0012] Further, the specific process of classifying the battery fault types by small-sample transfer learning according to the abnormal area mask, the fault detection index, and the evolution mode label is as follows: Use the abnormal area mask, the fault detection index, and the evolution mode label as input features to form a multi-dimensional fault feature vector; By the method of small-sample transfer learning, transfer the already trained network model and fine-tune it for a specific battery type, and map the features of different battery faults to the corresponding class labels through fine-tuning.

[0013] Further, the specific process of fine-tuning the classifier by a pre-trained feature extractor, combining small-sample battery fault data, and weighted fusing the temperature space feature and the evolution mode label by the attention mechanism to output the fault type and its confidence is as follows: Use the pre-trained feature extractor to extract features from the abnormal area mask, the fault detection index, and the evolution mode label to obtain a high-dimensional feature vector containing the temperature space feature and the evolution mode label; Combine the small-sample battery fault data to fine-tune the classifier, and adjust the network parameters to make the classifier adapt to the feature mapping of the target battery fault; Apply the attention mechanism to weighted fuse the temperature space feature and the evolution mode label to strengthen the contribution of key features in classification; Use the weighted fused feature vector to classify the fault type by the fine-tuned classifier, calculate the fault confidence according to the classification result, and finally output the battery fault type and its confidence.

[0014] The present invention has the following beneficial effects:

[0015] (1) A fault detection method for a high-resolution infrared thermal imaging array, which obtains the thermal imaging image of the transformer surface through a high-resolution infrared thermal imaging device, dynamically corrects the infrared temperature field in combination with the environmental parameters and the transformer load current, generates a temperature matrix free of interference, can effectively remove the interference caused by environmental factors or load current fluctuations, provides more accurate and stable temperature data, and ensures the accuracy of subsequent fault detection. By combining the temperature matrix and the physical topology map of the battery pack, using a graph neural network to model the heat conduction relationship between batteries, the fault detection can more accurately identify the abnormal area according to the heat conduction law, generate an abnormal area mask and a temperature space feature under spatial structure constraints, and effectively improve the reliability and accuracy of fault detection.

[0016] (2) A fault detection method for a high-resolution infrared thermal imaging array. By extracting the temperature fluctuation time-series data of the abnormal area mask and synchronously matching the real-time current and voltage changes of the battery pack, the dynamic time warping algorithm is used to quantify the correlation between the temperature fluctuation and the electrical parameters, and the fault detection index and the evolution pattern label are output, quantifying the correlation between the temperature fluctuation and the electrical parameters, and timely reflecting the severity and change trend of the fault through the fault detection index, providing an important basis for subsequent analysis. Classify the battery fault types through small-sample transfer learning, and fine-tune the classifier in combination with a pre-trained feature extractor, so that the system can use the transfer learning method to improve the classification performance under limited battery fault data, and achieve accurate identification of specific battery faults; by combining the attention mechanism to weighted-fuse the temperature spatial features and the evolution pattern label, it is possible to increase the attention to key features, optimize the fault classification model, improve the accuracy of fault type recognition, and output the corresponding confidence level, providing more reliable fault diagnosis support for engineers.

[0017] Of course, it is not necessary for any product implementing the present invention to achieve all the above-mentioned advantages simultaneously. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a flowchart of a fault detection method for a high-resolution infrared thermal imaging array of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] The embodiment of the present application solves the problems in traditional fault detection such as environmental interference, data inconsistency, and insufficient accuracy of fault type discrimination through a fault detection method for a high-resolution infrared thermal imaging array. Specifically, the present invention accurately identifies the abnormal temperature area on the surface of the transformer or battery pack by dynamically correcting the infrared temperature field, combining the physical topology map of the battery pack, and efficient heat conduction modeling technology, effectively avoiding the detection error caused by external factors in the traditional method. Quantify the correlation between temperature and electrical parameters through the dynamic time warping algorithm, and optimize the classifier in combination with small-sample transfer learning and attention mechanism, greatly improving the accuracy of fault type recognition, reducing the dependence on a large amount of labeled data, and further enhancing the intelligent level and practicality of fault detection.

[0020] The general idea of the solution in the embodiment of the present application is as follows:

[0021] Obtain the thermal imaging image of the surface of the transformer through a high-resolution infrared thermal imaging device, and dynamically correct the infrared temperature field in combination with the environmental parameters and the load current of the transformer to generate a temperature matrix free of interference.

[0022] Based on the temperature matrix and the physical topology diagram of the battery pack, model the heat conduction relationship between batteries through a graph neural network, combine with the U-Net segmentation network to detect temperature anomaly regions, identify anomaly regions that conform to the heat conduction law, and generate an anomaly region mask and temperature spatial features under spatial structure constraints.

[0023] Extract the temperature fluctuation time series data of the anomaly region mask, synchronously match the real-time current and voltage changes of the battery pack, quantify the correlation between temperature and electrical parameters through the dynamic time warping algorithm, distinguish continuous faults from instantaneous interference, and output a fault detection index and an evolution mode label.

[0024] Based on the anomaly region mask, the fault detection index and the evolution mode label, classify the battery fault types through few-shot transfer learning. Through a pre-trained feature extractor, fine-tune the classifier in combination with few-shot battery fault data, and combine the attention mechanism to weight and fuse the temperature spatial features and the evolution mode label, and output the fault type and its confidence level.

[0025] Please refer to Figure 1 For this, an embodiment of the present invention provides a technical solution: a fault detection method for a high-resolution infrared thermal imaging array, including the following steps: S1. Obtain a thermal imaging image of the surface of the transformer through a high-resolution infrared thermal imaging device, and combine the environmental parameters and the transformer load current to dynamically correct the infrared temperature field and generate a temperature matrix free of interference; S2. Based on the temperature matrix and the physical topology diagram of the battery pack, model the heat conduction relationship between batteries through a graph neural network, combine with the U-Net segmentation network to detect temperature anomaly regions, identify anomaly regions that conform to the heat conduction law, and generate an anomaly region mask and temperature spatial features under spatial structure constraints; S3. Extract the temperature fluctuation time series data of the anomaly region mask, synchronously match the real-time current and voltage changes of the battery pack, quantify the correlation between temperature and electrical parameters through the dynamic time warping algorithm, distinguish continuous faults from instantaneous interference, and output a fault detection index and an evolution mode label; S4. Based on the anomaly region mask, the fault detection index and the evolution mode label, classify the battery fault types through few-shot transfer learning. Through a pre-trained feature extractor, fine-tune the classifier in combination with few-shot battery fault data, and combine the attention mechanism to weight and fuse the temperature spatial features and the evolution mode label, and output the fault type and its confidence level.

[0026] In this implementation scheme, step S1: Obtain the thermal imaging image of the transformer surface through a high-resolution infrared thermal imaging device, and correct the interference of temperature data in real time with the help of environmental factors and load current data. This can eliminate the influence of factors such as temperature fluctuations and external environment changes, ensure that the generated temperature matrix reflects the real transformer temperature situation, and provide reliable basic data for subsequent fault detection. Step S2: Use the temperature matrix and the physical layout diagram of the battery pack, and use a graph neural network (GNN) to model the heat conduction relationship between the batteries. Then, combine the U-Net segmentation network to accurately locate and detect the temperature anomaly area. Through the U-Net segmentation network, it is possible to identify the area that conforms to the heat conduction law, and generate a "mask" to indicate the location of the anomaly area, and further extract the temperature spatial characteristics of this area. This process helps to accurately calibrate the area where the fault occurs and provides key information for subsequent diagnosis. Step S3: By extracting the temperature fluctuation time series data and comparing it with the real-time current and voltage changes of the battery pack, further analyze the correlation between temperature and electrical parameters. Through the dynamic time warping (DTW) algorithm, it is possible to quantify the changes in temperature fluctuations and electrical parameters, so as to distinguish continuous faults from instantaneous interference. In addition, this step also helps to distinguish different types of faults by outputting the fault detection index and the evolution mode label, enhancing the accuracy of the detection system. Step S4: By using the small sample transfer learning technology, combined with the existing small amount of battery fault data and the pre-trained classifier model, accurately classify the battery fault types. The attention mechanism further helps to focus on the most recognizable information in the temperature spatial characteristics and the evolution mode label, improving the classification accuracy. The final output is the fault type and its confidence value, indicating the degree of certainty of the fault type. High-resolution infrared thermal imaging device: This device presents the surface temperature of an object in the form of a thermal map by sensing the infrared rays radiated from the object surface. It is widely used in fault detection to monitor the surface temperature distribution of equipment and help discover thermal anomaly areas. Graph neural network: A graph neural network is a deep learning model for processing graph-structured data, which can capture the complex relationships between nodes (such as battery cells) in the graph. In the present invention, GNN is used to establish a heat conduction relationship model between the batteries. U-Net is a convolutional neural network (CNN) commonly used in medical image segmentation. It classifies images at the pixel level and divides the image into different regions. In the present invention, U-Net is used to identify the area with temperature anomalies in the transformer or battery pack, so as to locate the fault source. Dynamic time warping DTW is an algorithm for calculating the similarity between two time series. By dynamically adjusting the data points on the time axis, it finds the best matching method between the two time series. This algorithm is used to quantify the relationship between temperature and electrical parameters (such as current and voltage) and distinguish continuous faults from instantaneous interference. Small sample transfer learning is a transfer learning technique that aims to transfer the model trained on large-scale data to small sample data and perform fine-tuning.It helps improve the model's performance in new fields in the case of scarce data. In the present invention, it is used to accurately classify battery fault types through a small amount of battery fault data. The attention mechanism simulates the visual attention process of the human brain, enabling the network to automatically focus on key information in the input data. In the present invention, the attention mechanism helps the model assign appropriate weights to the temperature spatial features and evolution pattern labels, improving the accuracy of fault classification.

[0027] Specifically, the specific process of dynamically correcting the infrared temperature field is as follows: By using the battery load current and voltage data, combined with the ambient temperature and heat dissipation wind speed, establish a battery load current-temperature characteristic curve; Dynamically correct the infrared temperature field through a Kalman filter to eliminate the temperature error caused by environmental factors and generate interference-free infrared temperature field data; Calculate and correct the temperature field data to eliminate the temperature rise effect caused by normal charge and discharge conditions and dynamically correct the infrared temperature field.

[0028] In this embodiment, first, through the battery load current ( ) and battery voltage ( ) data, the influence of the battery on temperature during operation can be deduced. Since the battery generates heat during charge and discharge, this heat will increase with the increase of the battery load, thus affecting the infrared measured temperature. The ambient temperature is also an important factor in temperature field correction. It will be superimposed on the battery temperature, affecting the finally measured temperature field. To model this relationship, we use a temperature characteristic curve, that is, the relationship between the battery load current and voltage and the battery temperature. This characteristic curve is usually obtained through experiments and can predict temperature changes. To eliminate the temperature error caused by environmental factors (such as external climate conditions), we adopt a Kalman filter to dynamically correct the temperature field. The core role of the Kalman filter is to reduce the influence of noise by combining prediction and measurement data, gradually correct the error of the temperature data, and thus generate more accurate temperature field data. In this process, we use the Kalman filter to adjust the measured value at the current moment and combine parameters such as battery load and voltage to continuously optimize the prediction of the temperature field, finally eliminating the error and generating interference-free infrared temperature data. The battery generates a temperature rise during normal charge and discharge, and this temperature rise is a natural reaction of the battery load current. In infrared thermography, the normal temperature rise may affect fault detection, so it is necessary to correct the temperature data. Through the battery load current and temperature characteristic curve, calculate the temperature rise generated during normal charge and discharge and remove it from the actually measured temperature field data, thus eliminating the temperature deviation generated during normal operation. The corrected temperature data can better reflect the true working state of the battery and reduce the interference of errors. The dynamic temperature correction formula, set the correction formula for the infrared temperature field as: ; where: : Corrected temperature field data. : Infrared temperature field measurement value. : Battery load current. : Battery voltage. : Ambient temperature. : Corrected temperature at the previous moment. : Model parameters. Parameter explanation: : Influence coefficient of infrared temperature measurement value on the corrected temperature field, used to adjust the influence of temperature measurement error. : Influence coefficients of battery load current and voltage on temperature change, indicating the contribution of load current and voltage to battery temperature. and : Respectively represent the response coefficients of load current and voltage to temperature, which can be adjusted according to the working characteristics of the battery. : Influence coefficient of ambient temperature on the temperature field, reflecting the interference of external ambient temperature change on battery temperature. : Correction coefficient between the corrected temperature at the previous moment and the current measured temperature, ensuring that the correction process takes into account the continuity of the time series and the smoothness of historical data.

[0029] Specifically, the specific process of generating the interference-free temperature matrix is as follows: Perform multiple regression analysis on the corrected temperature field data and environmental factors; Based on the battery load current-temperature characteristic curve, eliminate the temperature rise effect under normal working conditions; Further optimize the temperature data to remove the interference caused by environmental noise and generate the interference-free temperature matrix.

[0030] In this implementation plan, multiple regression analysis of the corrected temperature data and environmental factors: We have completed the dynamic correction of the temperature field through the Kalman filter and generated relatively accurate temperature data (i.e., ). However, the temperature data is still affected by environmental factors (such as external air temperature, humidity, wind speed, etc.) and battery load current. Therefore, we further process the corrected temperature data through multiple regression analysis to eliminate these external interferences. Application of the battery load current-temperature characteristic curve: Before performing multiple regression analysis, we first need to consider the relationship between battery load current and battery temperature. During the operation of the battery, additional heat will be generated due to the increase in load current, and this temperature rise is inevitable under normal working conditions. We need to use the battery load current-temperature characteristic curve to quantify this influence and thus eliminate it in the analysis. Removal of environmental noise and optimization of temperature data: To further optimize the temperature data and remove the interference caused by environmental noise, we introduce environmental factors (such as ambient temperature , humidity , wind speed ) Related items. Through the regression model, we can calculate the relationships with these external factors and eliminate these interference factors from the temperature data, finally generating a temperature matrix free of interference. Generate the temperature matrix free of interference as follows: ; Explanation of the formula: : Temperature matrix after removing interference, : Corrected temperature data. : Values of the battery load current-temperature characteristic curve, representing the relationship between the battery load current and temperature. : Ambient temperature. : Ambient humidity. : Wind speed. : Ambient noise interference factor, Other environment-related variables. : Coefficients of the regression model, representing the influence degrees of different factors on the temperature data. Optimize these coefficients through training data. : Influence factor representing ambient noise, corresponding to different interference factors. Relationship between the temperature data and the battery load current ( ): Relationship between the corrected temperature data and the battery load current. Load current will directly affect the temperature, so we use to represent the temperature influence of the load current and incorporate it through the parameter . Through the optimized regression coefficients , we eliminate the influence of the battery load current from the temperature data, ensuring that only the part related to ambient interference is retained. Removal of ambient factor interference ( ): Ambient temperature , humidity and wind speed will all interfere with the temperature data. Through the coefficients , and in the regression model, we can precisely remove the influence of these factors on the temperature data and obtain the temperature data free of interference. Removal of ambient noise ( ): In addition to the above factors, there may be other ambient noises (electromagnetic interference, light changes) affecting the temperature measurement. These interferences are represented by the ambient noise factor . The coefficient represents the influence degrees of different noise factors. Through weighted summation, these noises are eliminated from the temperature data, and finally a temperature matrix free of interference is obtained.

[0031] Specifically, according to the temperature matrix and the physical topology diagram of the battery pack, the heat conduction relationship between batteries is modeled by a graph neural network, and the U-Net segmentation network is combined to detect the temperature anomaly region. The specific process of identifying the anomaly region that conforms to the heat conduction law is as follows: According to the temperature matrix and the physical topology diagram of the battery pack, a graph neural network model for heat conduction between batteries is established to simulate the thermal coupling relationship between individual batteries in the battery pack, and the heat flow path between batteries is determined based on the battery position and heat conduction characteristics; The topology diagram of the battery pack is combined with the temperature field data, and the heat conduction law between batteries is learned through the graph neural network to optimize the prediction of the spatial distribution of the temperature field; Based on the U-Net segmentation network, the anomaly region in the temperature field is identified. Combining the heat conduction law between batteries, the regions where temperature anomalies and severe fluctuations occur on the heat conduction path are preferentially identified, especially the regions where the temperature difference between adjacent batteries changes suddenly; Through the detection results of the temperature anomaly region, the temperature anomaly region is output, and the specific position of the region is calibrated.

[0032] In this implementation, the modeling of the thermal conduction relationship between cells: Based on the physical topology diagram of the battery pack, the thermal conduction relationship between cells is established. The battery pack consists of multiple single cells, and there is a thermal coupling effect between the cells. To accurately simulate this thermal conduction relationship, we use a Graph Neural Network (GNN) for modeling. In the graph neural network, the cells are represented as nodes in the graph, and the edges between the nodes represent the thermal conduction paths between the cells. We define the weights of these edges according to the positions, thermal properties, and connection methods of the cells, thereby reflecting the thermal conduction characteristics between the cells. The graph neural network learns the thermal conduction law: The graph neural network optimizes the prediction of the spatial distribution of heat flow by iteratively calculating the message passing between nodes. Through the training process, we can obtain the temperature information of each node (cell), thereby estimating the temperature distribution of the entire battery pack. The goal of this step is to effectively learn the thermal conduction law through the graph neural network and optimize the temperature prediction. Identification of abnormal regions: After completing the temperature prediction of the battery pack, we need to detect the abnormal regions in the temperature field based on the U-Net segmentation network. U-Net is a neural network commonly used in image segmentation tasks, especially suitable for the segmentation of structured data. Through the U-Net network, we can identify the temperature abnormal regions from the temperature field. The input of U-Net is a temperature matrix, which passes the temperature information in the temperature field to the convolutional layer, extracts features step by step through the encoder, then restores the spatial resolution step by step through the decoder, and finally outputs the segmented abnormal regions. Prioritize the identification of abnormal regions on the thermal conduction path: Since the thermal conduction between cells is regular, when detecting abnormal regions, we will prioritize the identification of those regions that are on the heat flow path and have abnormal temperatures and significant fluctuations. Especially the regions where the temperature difference between adjacent cells changes suddenly, which is usually a precursor to a fault. By combining the topological structure of the cells and the thermal conduction law, U-Net can label these abnormal regions and provide specific location information for further analysis and repair. Generate the final output of the temperature abnormal region: Through the segmentation results of U-Net, we can output a temperature matrix containing the abnormal region and calibrate the specific location of the abnormal region. The output of the abnormal region includes not only the temperature abnormality but also the corresponding spatial position and the change information of the thermal conduction path.

[0033] Specifically, the specific process of generating the abnormal area mask and temperature spatial features under the spatial structure constraint is as follows: According to the output of the U-Net segmentation network, by combining it with the topology map of the battery pack, a spatial structure mask of the temperature abnormal area is generated. The mask marks all abnormal areas where the temperature exceeds the threshold, and further eliminates noise through image post-processing; extract the spatial features of the abnormal area, including area, shape factor, and boundary complexity, to describe the geometric characteristics of the abnormal area; based on the spatial distribution characteristics of the temperature field, further extract the spatial temperature features related to the fault, including temperature gradient and temperature fluctuation amplitude.

[0034] In this implementation, a spatial structure mask is generated: The temperature field of the battery pack is processed by a U-Net segmentation network. The U-Net extracts features at different scales through convolutional operations and fuses low-level features with high-level features through skip connections, enabling accurate identification of temperature anomaly regions. The output of the network is a binary image indicating whether the temperature exceeds a predetermined threshold. On this basis, combined with the topology map of the battery pack, the positional relationship between the temperature anomaly region and the battery pack can be combined to generate a temperature anomaly region mask with spatial structure constraints. The role of this mask is to mark all regions where the temperature exceeds the set threshold, identifying parts of the battery or battery pack that may have faults. Image post-processing techniques such as morphological operations (erosion, dilation, etc.) can further clean up noise and correct the shape of the marked regions to make them more accurate and avoid misidentification or overmarking. Extract spatial features of the anomaly region: After generating the mask, a series of spatial geometric features can be extracted from the temperature anomaly region, which can help us further analyze the morphology and size of the anomaly region. These spatial features include: Area: The total area of the anomaly region, which can be used to quantify the scale of the region. Usually, the area is calculated by counting the number of pixels in the mask or by the spatial positions of all pixels within the region. Shape factor: Used to describe the shape of the anomaly region. The shape factor can be measured by a series of parameters (such as aspect ratio, circularity, etc.) to determine whether the morphology of a region is regular or irregular. For example, the shape factor can be calculated based on the ratio of the perimeter and area of the region. The more regular the shape, the closer its factor value is to that of a circle. Boundary complexity: Measures the complexity of the boundary of the anomaly region. Usually, it is determined by calculating the curvature of the boundary of the anomaly region or the ratio of the boundary length of a polygon to its area. The more complex the boundary, the more irregular the fault region may be, indicating that there may be more serious temperature anomalies or battery faults. Extract spatial temperature features related to faults: Further, combined with the spatial distribution characteristics of the temperature field, spatial temperature features related to battery faults can be extracted, which can more accurately help diagnose the cause of the fault. Temperature gradient: The temperature gradient refers to the rate of temperature change in space, that is, the amount of temperature change per unit distance. In a battery pack, the change in temperature gradient is often related to uneven heat conduction or fault regions. By calculating the temperature difference between adjacent pixels, the distribution of the temperature gradient can be obtained, and then it can be judged whether the temperature change conforms to the expected heat conduction law. Temperature fluctuation amplitude: The temperature fluctuation amplitude describes the degree of temperature fluctuation in the time or space dimension. A large temperature fluctuation amplitude in the anomaly region may mean thermal runaway or battery faults. By calculating the temperature fluctuation amplitude within the temperature anomaly region, it can be further evaluated whether the region is in an abnormal state and combined with electrical parameters to analyze the nature of the fault.

[0035] Specifically, the specific process of quantifying the correlation between temperature and electrical parameters through the dynamic time warping algorithm and distinguishing continuous faults from instantaneous interference is as follows: Extract the time series data of temperature fluctuations from the abnormal area mask. Through time series analysis of the dynamic changes in the temperature abnormal area, obtain the temperature fluctuation characteristics of each abnormal area within a certain time interval; Synchronously collect the electrical parameters of the battery pack, including real-time current and voltage data, and align the electrical parameters with the time series data of temperature fluctuations to ensure the time series consistency of temperature data and electrical data; Through the dynamic time warping algorithm, calculate the time series correlation coefficient between the temperature fluctuation data and the electrical parameters, and quantify the time delay relationship between the temperature fluctuation and the electrical parameters; According to the analysis results of the dynamic time warping algorithm, combined with the relationship between temperature fluctuations and electrical parameters, judge the type of fault. If the temperature fluctuations are synchronized and continuous with the changes in electrical parameters, it is determined as a continuous fault. If the temperature fluctuations are short-term and irregular with the changes in electrical parameters, it is determined as a fluctuation caused by instantaneous interference.

[0036] In this implementation, extract the time-series data of temperature fluctuations from the abnormal area mask: Extract the time-series data of temperature fluctuations from the temperature abnormal area mask generated by the U-Net segmentation network. Each temperature abnormal area has its specific temperature change pattern, usually manifested as fluctuations over time. By continuously monitoring the temperature abnormal area, time-series data of temperature fluctuations can be formed. This time-series data contains the fluctuations of temperature at different time points and is an important basis for analyzing the fault modes of the battery pack. Synchronously collect electrical parameter data: Synchronously collect the electrical parameter data of the battery pack, such as real-time current and voltage, with the temperature fluctuation data. Electrical parameters have a significant impact on the operating state and thermal performance of the battery pack. Therefore, it is crucial to compare and analyze the electrical data with the temperature fluctuation data. In practical applications, the collection of electrical parameters (such as current and voltage) and temperature data usually has different time granularities or frequencies. To ensure the temporal consistency of the temperature fluctuation data and the electrical data, we need to align the two sets of data in time. A common approach is to use interpolation methods or select a unified time interval to align the two types of data to the same time step. Application of the Dynamic Time Warping (DTW) algorithm: The Dynamic Time Warping (DTW) algorithm is used to calculate the similarity between two time series. Especially when there are time offsets or misalignments in the time series, DTW can effectively align these series and find the optimal matching path. In this process, the DTW algorithm is used to quantify the correlation between the time-series data of temperature fluctuations and the time-series data of electrical parameters. The core purpose of DTW calculation is to measure the time delay relationship between temperature fluctuations and electrical parameter changes, that is, the degree of their time alignment. DTW finds the optimal matching path between them by minimizing the distance between the two time series (achieved through dynamic programming). Determine the type of fault: Persistent faults and transient disturbances: Based on the correlation coefficient calculated by the DTW algorithm, we can obtain the time delay relationship between temperature fluctuations and electrical parameters. If the temperature fluctuations are synchronized and persistent with the electrical parameter changes (that is, the two are well-aligned in time and this correlation remains unchanged over a long period), it can be judged as a persistent fault. Persistent faults may be caused by internal problems of the battery (such as internal short circuit or thermal runaway), manifested as continuous and regular temperature fluctuations. If the temperature fluctuations are short-lived and irregular with the electrical parameter changes (that is, the correlation between the two in time is low and there are obvious fluctuations), it can be judged as fluctuations caused by transient disturbances. Transient disturbances are usually caused by external environmental factors or occasional events (such as electrical noise, transient load changes), and the temperature fluctuations are manifested as sudden and short-lived fluctuations.

[0037] Specifically, the specific process of outputting the fault detection index and the evolution mode label is as follows: Calculate the temporal correlation coefficient of the temperature fluctuation data and the electrical parameters, and combine the operating state of the battery pack to generate the fault detection index; According to the temporal characteristics of the temperature fluctuation and the change trend of the electrical data, generate the fault evolution mode label through the evolution mode classification algorithm, and classify the fault into a continuous type or an intermittent type.

[0038] In this implementation, calculate the temporal correlation coefficient between the temperature fluctuation and the electrical parameters: Through the dynamic time warping (DTW) algorithm or other time series analysis methods, calculate the temporal correlation coefficient between the temperature fluctuation and the electrical parameters. This coefficient reflects the synchronization between the temperature fluctuation and the changes in electrical parameters (such as current and voltage). The formula is as follows: ; where: represents the temporal correlation coefficient. represents the time point at which the th measurement value of the temperature fluctuation data. and are the mean and standard deviation of the temperature fluctuation data at the time point respectively. represents the time point at which the th measurement value of the electrical parameter (current, voltage). and are the mean and standard deviation of the electrical data at the time point respectively. This correlation coefficient is used to quantify the synchronization degree between the temperature fluctuation and the change of electrical parameters, so as to help identify whether there is a fault. Generate the fault detection index: By calculating the temporal correlation coefficient between the temperature fluctuation and the electrical parameter, and combining the operating state of the battery pack (such as charging, discharging, standby, etc.), the fault detection index can be generated. This index reflects the severity and possibility of the fault occurrence. The weighted comprehensive method can be used to combine the temporal correlation coefficient with the weight of the operating state of the battery pack to generate the comprehensive fault detection index. The calculation formula of the fault detection index is as follows: ; where: represents the fault detection index. represents the temporal correlation coefficient. is a quantitative index of the current operating state of the battery pack, with a value range between 0 and 1, reflecting the current working mode of the battery. and is the weighting coefficient that controls the influence degree of the timing correlation coefficient and the battery pack state on the fault detection index. The higher the value of the fault detection index, the greater the possibility of a fault; conversely, it indicates a smaller possibility of a fault. By analyzing the timing characteristics of temperature fluctuations and the change trend of electrical parameters, combined with the evolutionary pattern classification algorithm (support vector machine), faults can be classified into different evolutionary patterns. Generally, the evolutionary patterns of faults can be divided into continuous faults and intermittent faults. Continuous fault: It refers to the situation where the temperature fluctuations and the changes in electrical parameters show a continuous trend, and the fault condition will not recover in a short time. Intermittent fault: It refers to the situation where the temperature fluctuations and the changes in electrical parameters show periodic or sporadic fluctuations, and the fault state appears and disappears sometimes. The generation of the evolutionary pattern label is based on the timing characteristics of temperature fluctuations (such as fluctuation amplitude, periodicity) and the change trend of electrical data. The fault pattern can be judged through the timing classification algorithm, and an evolutionary pattern label is assigned to each fault event.

[0039] Specifically, according to the abnormal area mask, the fault detection index, and the evolutionary pattern label, the specific process of classifying battery fault types through few-shot transfer learning is as follows: The abnormal area mask, the fault detection index, and the evolutionary pattern label are used as input features to form a multi-dimensional fault feature vector; through the method of few-shot transfer learning, the trained network model is transferred and fine-tuned for a specific battery type, and the features of different battery faults are mapped to the corresponding class labels through fine-tuning.

[0040] In this implementation, a multi-dimensional fault feature vector is generated: The abnormal region mask, fault detection index, and evolution pattern label are used as input features, and through feature fusion, an information vector containing multiple dimensions is formed. Each input feature provides information on different aspects of battery faults: Abnormal region mask: Represents the spatial distribution of the temperature abnormal region, providing information on the spatial location and shape of the fault region. Fault detection index: Quantifies the temporal correlation between temperature fluctuations and electrical parameters, reflecting the severity and likelihood of the fault. Evolution pattern label: Describes the evolution pattern of the fault, distinguishing between continuous faults and intermittent faults. After combining these three types of feature data, a multi-dimensional fault feature vector can be obtained. This feature vector includes temperature spatial features, temporal features, and fault evolution pattern features, providing rich input information for subsequent classification tasks. Few-shot transfer learning: Since the actual battery fault dataset often has insufficient sample numbers, few-shot transfer learning can be used to classify fault types. Few-shot transfer learning improves classification performance by transferring a network model that has been trained on a large-scale dataset to a new task with less data. In this stage, the process is as follows: Transfer the pre-trained model: Select a network model that has been trained on a similar task. Usually, these models have good learning ability and feature extraction ability on large-scale datasets (for example, using convolutional neural networks (CNNs) or graph neural networks (GNNs), etc.). These pre-trained models have learned general feature representations, so they can be used for the battery fault classification task. Fine-tuning: Apply the pre-trained model to the new battery fault classification task through fine-tuning. The purpose of fine-tuning is to adjust the model's parameters according to the new few-shot data to make it adapt to specific battery fault types. This is usually done by performing local training on a small amount of battery fault data. The network adjusts its weights to map the features of different battery faults to the corresponding class labels. During the fine-tuning process, the key is to retain the general feature extraction layer in the pre-trained model and perform targeted training on the last few layers to make the model adapt to the new fault classification task. Classification process: After fine-tuning through few-shot transfer learning, the model can classify battery fault types based on the input fault feature vector (containing information such as abnormal region mask, fault detection index, evolution pattern label, etc.). By using transfer learning, the model can achieve good performance on a small amount of training data, effectively solving the data scarcity problem in battery fault classification. During the classification process, the model maps the input feature vector to the class label of the fault type according to the learned relationship between fault features and labels. For example, possible fault types include short circuit faults, overheat faults, internal damage, etc.Finally, the model outputs a prediction result of the fault type, identifying whether the battery has a fault and determining the specific type of the fault. Through such classification, the operating state of the battery can be diagnosed more accurately, helping to take maintenance measures in a timely manner.

[0041] Specifically, through a pre-trained feature extractor, combining a small sample of battery fault data to fine-tune the classifier, and combining the attention mechanism to weighted fuse the temperature spatial features and the evolution pattern labels, the specific process of outputting the fault type and its confidence is as follows: The pre-trained feature extractor extracts features from the abnormal region mask, the fault detection index, and the evolution pattern labels to obtain a high-dimensional feature vector containing the temperature spatial features and the evolution pattern labels; Combining a small sample of battery fault data to fine-tune the classifier, and adjusting the network parameters to make the classifier adapt to the feature mapping of the target battery fault; Applying the attention mechanism to weighted fuse the temperature spatial features and the evolution pattern labels to strengthen the contribution of key features in classification; Using the fine-tuned classifier to classify the fault type through the weighted fused feature vector, calculating the fault confidence according to the classification result, and finally outputting the battery fault type and its confidence.

[0042] In this implementation, in the first step of this process, a pre-trained feature extractor is used to extract features from the input data. The goal of the feature extractor is to extract useful feature information from the abnormal region mask, fault detection index, and evolution pattern label to obtain a high-dimensional feature vector, which mainly includes the following aspects: Abnormal region mask: Output by the U-Net segmentation network, representing the spatial distribution of the temperature abnormal region and identifying the battery regions where faults may occur. Fault detection index: The temporal correlation coefficient between the temperature fluctuation data and electrical parameters calculated by the dynamic time warping algorithm, reflecting the temporal characteristics of battery faults. Evolution pattern label: Indicating the evolution trend of the fault and distinguishing between continuous and intermittent fault modes. Through the pre-trained feature extractor (usually a deep neural network, such as a convolutional neural network (CNN) or a graph neural network (GNN)), multi-level high-dimensional features are extracted from the above input data. These features include the temperature spatial distribution features and fault evolution patterns, which can provide sufficient information for subsequent classification tasks. Fine-tuning the classifier: After obtaining the high-dimensional feature vector, the next step is to fine-tune the classifier using a small sample of battery fault data. The purpose of the fine-tuning process is to apply the feature extraction ability of the pre-trained model to a specific battery fault classification task and adjust the classifier through a small amount of target battery fault data. Few-shot learning: Since it is usually difficult to obtain a large number of samples of battery fault data, during the fine-tuning process, the classifier can be made to adapt to the battery fault features contained in a small amount of data through the trained network parameters. Goal: Through the fine-tuning process, adjust the weights and biases of the network so that the classifier can accurately map the battery fault data to the corresponding fault types. The fine-tuning process is usually carried out based on the pre-trained model, maintaining the generalization ability of the feature extraction part, and at the same time refining the classification layer through a small amount of data to ensure that the classifier can accurately identify different battery faults. Applying the attention mechanism for weighted fusion: To enhance the model's attention to key features, the attention mechanism is applied to weightedly fuse the temperature spatial features and evolution pattern labels. The attention mechanism can dynamically assign different weights to different features according to the relevance of the input features, thereby better highlighting the role of key features in classification. Temperature spatial features: Describing the temperature distribution in the battery region, including the heat conduction relationship, temperature gradient, and temperature fluctuation amplitude between batteries. Evolution pattern label: Describing the evolution process of the fault, indicating whether the fault is a continuous fault or an intermittent fault. Through the attention mechanism, the model will automatically learn which features are more important for fault classification and increase their weights accordingly. This weighted fusion method helps to improve the accuracy of classification, especially when the input features have different importance. Fault type classification and confidence calculation: In the weighted fusion feature vector after applying the attention mechanism, the fine-tuned classifier is used to classify the battery fault types. The classifier will classify according to the fault features learned during training and map each sample to the corresponding fault type.In addition, the classifier also calculates the fault confidence based on the input feature vector. Confidence refers to the degree of trust of the model in its classification result, which is usually obtained by calculating the probability values of each fault type. For example, the model may output a probability value of 0.85 for a fault type of "overheating fault", which means the model has 85% confidence that the battery has an overheating fault. Finally, the system outputs the battery fault type and confidence, which provides a clear and reliable basis for the diagnosis of the battery state. High-confidence predictions usually represent a higher severity of battery faults, while low confidence may indicate a greater uncertainty of the fault type.

[0043] In summary, the present application has at least the following effects:

[0044] The present invention provides a fault detection method for a high-resolution infrared thermal imaging array, aiming to improve the fault detection ability of devices such as transformers or battery packs. The thermal imaging image of the device surface is obtained by a high-resolution infrared thermal imaging device, and the temperature field is dynamically corrected by combining environmental parameters and load current to generate a temperature matrix free of interference. The thermal conduction relationship between devices is modeled by a graph neural network, and the temperature anomaly region is detected by a U-Net segmentation network to generate an anomaly region mask and temperature spatial features under spatial structure constraints. Further, the time series data of temperature fluctuations in the anomaly region is extracted, and combined with the real-time electrical parameters of the battery pack, the temporal correlation between temperature and electrical parameters is analyzed by the dynamic time warping algorithm to distinguish persistent faults from instantaneous interference, and a fault detection index and an evolution pattern label are generated. Finally, through the small-sample transfer learning method, the classifier is fine-tuned by combining a pre-trained feature extractor, and the temperature spatial features and evolution pattern labels are weighted and fused by the attention mechanism to classify the battery fault type and output the fault type and its confidence. By comprehensively using high-resolution infrared imaging, graph neural networks, dynamic time warping, transfer learning, and attention mechanisms, the method realizes efficient and accurate detection of faults and has broad application prospects.

[0045] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0046] The present invention is described with reference to the flowcharts and / or block diagrams of systems, apparatus (systems), and computer program products according to embodiments of the invention. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and combinations of flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to the processors of general purpose computers, special purpose computers, embedded processors, or other programmable data processing devices to produce a machine, such that the instructions executed by the processors of the computer or other programmable data processing devices produce means for implementing the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or multiple blocks.

[0047] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a particular manner, such that the instructions stored in the computer-readable memory produce a manufactured article including instruction means that implement the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or multiple blocks.

[0048] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or multiple blocks.

[0049] Although the preferred embodiments of the present invention have been described, additional changes and modifications can be made to these embodiments by those skilled in the art once they learn the basic creative concepts. Therefore, the appended claims are intended to be construed to include the preferred embodiments as well as all changes and modifications that fall within the scope of the present invention.

[0050] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these modifications and variations.

Claims

1. A high-resolution infrared thermal imaging display fault detection method, characterized in that: The following steps are involved: S1. Obtain thermal imaging images of the transformer surface through high-resolution infrared thermal imaging equipment, and dynamically correct the infrared temperature field in combination with environmental parameters and transformer load current to generate a temperature matrix to eliminate interference; S2. Based on the temperature matrix and the physical topology of the battery pack, the heat conduction relationship between batteries is modeled through the graph neural network, and the temperature abnormality area is detected in combination with the U-Net segmentation network, the abnormal area that conforms to the heat conduction law is identified, and the abnormal area mask and temperature space features under the spatial structure constraint are generated; S3. Extract the temperature fluctuation time series data of the abnormal area mask, and synchronously match the real-time current and voltage changes of the battery pack. Quantify the correlation between temperature and electrical parameters through the dynamic time warping algorithm, distinguish between continuous faults and transient interference, and output the fault detection index and evolution mode label; S4. Classify the battery fault type through small sample transfer learning based on the abnormal area mask, fault detection index and evolution mode label. Fine-tune the classifier with the pre-trained feature extractor and small sample battery fault data, and combine the attention mechanism to weightedly fuse the temperature space features and evolution mode labels to output the fault type and its confidence. The heat conduction relationship between batteries is modeled through graph neural networks, and the abnormal temperature areas are detected in combination with the U-Net segmentation network. The specific process of identifying abnormal areas that conform to the heat conduction law is as follows: Based on the temperature matrix and the physical topology of the battery pack, a graph neural network model of heat conduction between batteries is established to simulate the thermal coupling relationship between the individual batteries in the battery pack, and determine the heat flow path between batteries based on the battery position and thermal conduction characteristics; Combine the battery pack topology with the temperature field data, learn the heat conduction law between batteries through the graph neural network, and optimize the spatial distribution prediction of the temperature field; Based on the U-Net segmentation network, the abnormal areas in the temperature field are identified. Combined with the heat conduction law between batteries, the areas with abnormal and drastic temperature fluctuations on the heat conduction path and the areas with sudden temperature changes between adjacent batteries are identified. Through the detection results of the temperature abnormal area, the temperature abnormal area is output and the specific location of the area is marked; The specific process of outputting the fault detection index and evolution mode label is as follows: Generate a fault detection index based on the calculated time series correlation coefficient of temperature fluctuation data and electrical parameters combined with the operating status of the battery pack; According to the timing characteristics of temperature fluctuations and the changing trend of electrical data, the fault evolution pattern label is generated through the evolution pattern classification algorithm, and the fault is divided into continuous or intermittent type.

2. A high-resolution infrared thermal imaging display fault detection method according to claim 1, characterized in that: The specific process of dynamic correction of infrared temperature field is as follows: The battery load current-temperature characteristic curve is established by combining the battery load current and voltage data with the ambient temperature and cooling wind speed; Dynamically correct the infrared temperature field through the Kalman filter to eliminate the temperature error caused by environmental factors and generate interference-free infrared temperature field data; Calculate and correct the temperature field data, eliminate the temperature rise effect caused by normal charging and discharging conditions, and dynamically correct the infrared temperature field.

3. A high-resolution infrared thermal imaging display fault detection method according to claim 2, characterized in that: The specific process of generating the interference-free temperature matrix is ​​as follows: The corrected temperature field data and environmental factors were subjected to multiple regression analysis; Based on the battery load current-temperature characteristic curve, the temperature rise effect under normal working conditions is eliminated; The temperature data is further optimized to remove the interference caused by environmental noise and generate a de-interference temperature matrix.

4. The method for fault detection of a high-resolution infrared thermal imaging display according to claim 3, characterized in that: The specific process of generating abnormal area masks and temperature spatial features under spatial structure constraints is as follows: Based on the output of the U-Net segmentation network, the spatial structure mask of the temperature abnormality area is generated by combining it with the topological map of the battery pack. The mask marks all abnormal areas where the temperature exceeds the threshold, and further eliminates noise through image post-processing; Extract the spatial features of the abnormal region, including area, shape factor, and boundary complexity, to describe the geometric characteristics of the abnormal region; Based on the spatial distribution characteristics of the temperature field, the spatial temperature characteristics related to the fault are further extracted, including the temperature gradient and the temperature fluctuation amplitude.

5. A method for fault detection of a high-resolution infrared thermal imaging display according to claim 4, characterized in that: The dynamic time warping algorithm is used to quantify the correlation between temperature and electrical parameters and distinguish between sustained faults and transient disturbances as follows: Extract the time series data of temperature fluctuation from the abnormal area mask, and obtain the temperature fluctuation characteristics of each abnormal area within a certain time interval by performing time series analysis on the dynamic changes of the temperature abnormal area; Synchronously collect the electrical parameters of the battery pack, including real-time current and voltage data, and align the electrical parameters with the temperature fluctuation timing data to ensure the timing consistency of the temperature data and the electrical data; The dynamic time warping algorithm is used to calculate the time series correlation coefficient between temperature fluctuation data and electrical parameters, and to quantify the time delay relationship between temperature fluctuation and electrical parameters. According to the analysis results of the dynamic time warping algorithm and the relationship between temperature fluctuations and electrical parameters, the fault type is determined. If the temperature fluctuations and electrical parameter changes are synchronized and continuous, it is determined to be a continuous fault. If the temperature fluctuations and electrical parameter changes are short-lived and irregular, it is determined to be a fluctuation caused by instantaneous interference.

6. A method for fault detection of a high-resolution infrared thermal imaging array according to claim 5, characterized in that: The specific process of classifying battery fault types through small sample transfer learning based on abnormal area masks, fault detection indexes, and evolution mode labels is as follows: The abnormal region mask, fault detection index and evolution mode label are used as input features to form a multi-dimensional fault feature vector; Through the small sample transfer learning method, the trained network model is migrated and fine-tuned for specific battery types. Through fine-tuning, the characteristics of different battery failures are mapped to corresponding category labels.

7. A method for fault detection of a high-resolution infrared thermal imaging array according to claim 6, characterized in that: The specific process of outputting the fault type and its confidence level by using the pre-trained feature extractor, fine-tuning the classifier with small sample battery failure data, and weighted fusion of temperature space features and evolution mode labels with the attention mechanism is as follows: The pre-trained feature extractor is used to extract features of abnormal area masks, fault detection indexes, and evolution mode labels, and a high-dimensional feature vector containing temperature spatial features and evolution mode labels is obtained. Combined with small sample battery failure data, the classifier is fine-tuned and the network parameters are adjusted to make the classifier adapt to the feature mapping of the target battery failure. The attention mechanism is used to perform weighted fusion of temperature space features and evolution mode labels to enhance the contribution of key features in classification. The fine-tuned classifier is used to classify the fault type through the weighted fused feature vector, and the fault confidence is calculated based on the classification result, and finally the battery fault type and its confidence are output.

Images