A high-precision defect detection method and system for power distribution network micro-components

Abstract

The application discloses a kind of high-precision defect detection method and system for distribution network micro components, it is related to distribution network equipment defect detection technical field, including the composite radiation-electric energy conversion sensitive layer is attached to the surface of distribution network micro components;The composite radiation-electric energy conversion sensitive layer is used to receive environmental radiation and is converted into electric energy signal.The composite radiation-electric energy conversion sensitive layer is directly attached to the surface of distribution network micro components in the application, so that the component itself becomes sensor, such as wire clamp, insulator, etc.No need for artificial visual inspection, unmanned aerial vehicle inspection or external equipment intervention, such as infrared thermal imaging, realizes in-situ continuous monitoring in true sense.Array sensitive layer design further breaks through the limitation of external equipment view angle, and by comparing the spatial difference of each subunit electric energy signal, the defects of hidden parts such as the back of wire clamp and the inner side of insulator shed can be accurately positioned, and the positioning error is controlled within the size range of subunit, which fundamentally eliminates the sensing blind area.

Description

Technical Field

[0001] This invention relates to the field of defect detection technology for power distribution network equipment, and specifically to a high-precision defect detection method and system for small components in power distribution networks. Background Technology

[0002] As the final link in the power system, the reliability of the distribution network directly affects the quality of power supply to users. Small components such as clamps, insulators, tension clamps, suspension clamps, vibration dampers, parallel groove clamps, and terminals are critical connectors and supports in the distribution network, and their structural integrity is crucial for the safe operation of the grid. During long-term operation, these components are susceptible to mechanical stress, environmental corrosion, temperature changes, and other factors, leading to early defects such as micro-cracks, corrosion, and loosening. Failure to detect and address these defects in a timely manner can result in serious accidents such as line breaks and string slippage, causing widespread power outages. Currently, defect detection in small components of the distribution network mainly relies on manual visual inspection, drone inspections, infrared thermal imaging, and ultraviolet imaging. However, all of these technologies have a fundamental flaw: the detection methods passively depend on external detection equipment and cannot achieve continuous in-situ monitoring of the components themselves.

[0003] Specifically, manual visual inspection requires maintenance personnel to carry equipment close to live components for observation, which is not only inefficient but also limited by the angle and distance, making it difficult to detect tiny cracks and corrosion defects in hidden areas. Although drone inspection improves the detection range, it is essentially still an external imaging device taking periodic pictures of components. It is limited by weather conditions, battery life, and image resolution, making it impossible to achieve continuous monitoring around the clock, and there are blind spots in the inspection. Infrared thermal imaging and ultraviolet imaging can detect partial discharge and abnormal heating, but they are not sensitive to early structural defects such as microcracks and corrosion, and require dedicated imaging equipment, resulting in high costs per inspection.

[0004] The common characteristic of the aforementioned detection methods is that the detection is separated from the component itself—whether it's manual visual inspection, drone photography, or thermal imaging scanning, all rely on external equipment to detect the component, which is merely a passive object of observation and has no sensing capabilities. For a long time, it has been generally believed that the small size, complex installation environment, and lack of stable power supply conditions of power distribution network components make it impossible to achieve continuous in-situ monitoring of the components themselves. Instead, periodic inspections using external detection equipment are necessary. This has led to a long-standing technical dilemma in the field of "lagging detection, blind spots, and high costs," and no effective solution has been found to fundamentally address the industry pain point of the difficulty in quickly detecting early defects in small components. Summary of the Invention

[0005] The purpose of this invention is to provide a high-precision defect detection method and system for small components in power distribution networks, in order to address the shortcomings of the prior art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a high-precision defect detection method for micro-components in power distribution networks, comprising: S1. A composite radiation-to-electricity conversion sensitive layer is attached to the surface of a small component in a power distribution network; the composite radiation-to-electricity conversion sensitive layer is used to receive ambient radiation and convert it into an electrical signal, wherein the ambient radiation includes at least one of sunlight, infrared thermal radiation and ultraviolet light; S2. Utilize the radiation-electric energy conversion characteristics of the composite radiation-electric energy conversion sensitive layer under healthy conditions to establish a dynamic radiation-electric energy benchmark; S3. Monitor the power output of the composite radiation-to-electricity conversion sensitive layer in real time. When micro-cracks, corrosion or loosening occur in the micro-components of the power distribution network, the physical structural integrity of the composite radiation-to-electricity conversion sensitive layer is destroyed, resulting in a detectable sudden change in its radiation-to-electricity conversion efficiency, causing the real-time power signal to deviate from the dynamic reference. S4. Compare the real-time collected power signal with the dynamic benchmark, determine whether there are defects in the distribution network micro-components based on the comparison results, and identify the defect type when defects are found. S5. Output defect detection results.

[0007] In a preferred embodiment, in step S1, the composite radiation-electric energy conversion sensitive layer is a thin film structure with a thickness of 10 μm to 500 μm, and is made of at least two of photovoltaic conversion materials, infrared thermoelectric conversion materials, and ultraviolet photoelectric conversion materials.

[0008] In a preferred embodiment, establishing the radiation-electric energy mapping dynamic reference in S2 includes: Collect environmental radiation parameters, which include at least one of light intensity, spectral distribution, infrared radiation flux, and ultraviolet index; Based on historical electrical energy signals, environmental radiation parameters, component temperature, and radiation wavelength distribution, a radiation-electrical energy mapping function for the micro-component under healthy conditions is constructed using a machine learning regression model. =f(E, T, λ); in To predict electrical output, E is the environmental radiation parameter, T is the component temperature, and λ is the radiation wavelength distribution; The power prediction range under the current environmental conditions is calculated based on the mapping function, and the boundary of the prediction range is used as a dynamic threshold.

[0009] In a preferred embodiment, step S4, determining whether the power distribution network micro-component has a defect based on the comparison result, includes: The standardized deviation between the real-time acquired power signal and the dynamic benchmark prediction value is calculated. When the standardized deviation exceeds a preset threshold, it is determined that there is a defect. The standardized deviation ,in For real-time acquisition of electrical energy signals, The dynamic baseline forecast value is given by σ, where σ is the standard deviation of historical data. The defect type identification in S4 includes identifying the defect type based on the changing trend of the normalized deviation and the response differences of different band conversion units in the composite radiation-to-energy conversion sensitive layer, specifically including: When the standardization deviation shows a rapid abrupt change and the photovoltaic conversion unit response changes significantly, it is judged as a microcrack or delamination; When the standardization deviation shows a slow decreasing trend and the response of the thermoelectric conversion unit changes significantly, it is judged as corrosion; When the standardized deviation exhibits periodic fluctuations, it is considered loose.

[0010] In a preferred embodiment, in S1, the composite radiation-to-electricity conversion sensitive layer includes a plurality of radiation-to-electricity conversion units arranged in an array; In step S3, the power signals of each conversion unit are collected respectively; In step S4, the location of the defect is determined by comparing the differences in the electrical signals of each conversion unit.

[0011] In a preferred embodiment, a self-calibration step is also included: The self-calibration process is performed periodically. Under the condition that the environmental radiation is stable and measurable, the real-time electrical energy signal is compared with the theoretical calculation value to correct the model parameters of the dynamic reference and compensate for the aging effect of the composite radiation-electrical energy conversion sensitive layer.

[0012] In a preferred embodiment, an energy management step is also included: At least a portion of the electrical energy output from the composite radiation-to-electricity conversion sensitive layer is stored through an energy management circuit, which simultaneously powers the signal acquisition, defect identification, and wireless transmission modules, enabling passive, wiring-free, and long-term online monitoring.

[0013] In a preferred embodiment, the power distribution network micro-components include at least one of the following: clamps, insulators, tension clamps, suspension clamps, vibration dampers, parallel groove clamps, and terminals.

[0014] The present invention also provides a high-precision defect detection system for micro-components in power distribution networks, comprising: A composite radiation-to-electricity conversion sensitive layer is disposed on the surface of a micro component of a power distribution network to receive ambient radiation and convert it into an electrical signal. The ambient radiation includes at least one of sunlight, infrared thermal radiation, and ultraviolet light. When the micro component of the power distribution network develops microcracks, corrosion, or loosening, the physical structural integrity of the composite radiation-to-electricity conversion sensitive layer is compromised, resulting in a detectable abrupt change in its radiation-to-electricity conversion efficiency. The signal acquisition unit is electrically connected to the composite radiation-to-electricity conversion sensitive layer and is used to acquire the electrical signal. The dynamic benchmark modeling module is used to establish a dynamic benchmark of radiation-electrical energy mapping of the micro-components of the distribution network in a healthy state based on historical power signals and environmental radiation parameters, through regression analysis or machine learning methods. The defect identification module is used to compare the real-time collected power signals with the dynamic benchmark, determine whether there are defects in the micro components of the power distribution network based on the comparison results, and identify the type of defect when defects are found. The wireless communication unit is used to wirelessly transmit the defect identification results to the remote monitoring platform; An energy management unit is used to store at least a portion of the electrical energy output by the composite radiation-to-electricity conversion sensitive layer and to power the signal acquisition unit, dynamic reference modeling module, defect identification module, and wireless communication unit.

[0015] The technical effects and advantages provided by the present invention in the above technical solution are as follows: This invention directly attaches a composite radiation-to-electrical-energy conversion sensitive layer to the surface of tiny components in a power distribution network, turning components such as clamps and insulators into sensors themselves, fundamentally changing the traditional detection mode. It eliminates the need for external equipment such as manual visual inspection, drone inspection, or infrared thermal imaging, achieving true in-situ continuous monitoring. The arrayed sensitive layer design further overcomes the limitations of external equipment viewing angles. By comparing the spatial differences in electrical signals from each sub-unit, defects in hidden areas such as the back of clamps and the inner side of insulator skirts can be accurately located, with the positioning error controlled within the sub-unit size range, fundamentally eliminating blind spots.

[0016] This invention utilizes a continuously online sensitive layer to synchronize detection with component operation. When a defect occurs, the output electrical energy immediately deviates from the healthy baseline, enabling millisecond-level judgment and completely resolving the detection lag issue. The dynamic baseline modeling module establishes a radiation-electrical energy mapping function using a machine learning regression model, employing predicted interval boundaries as dynamic thresholds to compensate for changes in environmental parameters such as light intensity, spectral distribution, infrared radiation flux, ultraviolet index, and component temperature in real time. This maintains stable detection performance under various conditions, including sunny, cloudy, and rainy days.

[0017] This invention collects and stores the weak electrical energy output from the sensitive layer through an energy management unit, which powers the signal acquisition, defect identification, and wireless communication modules, enabling the detection system to be self-sufficient in energy. The design of this invention allows the detection system to be deployed on a large scale in various small components of the power distribution network, significantly reducing the overall cost. It fundamentally solves the core technical problem of separating detection from the component body, and provides a brand-new technical solution for the health status perception of small components in the power distribution network. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0019] Figure 1 This is a flowchart of the method of the present invention.

[0020] Figure 2 This is a system block diagram of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Example 1, please refer to Figure 1 As shown in this embodiment, a high-precision defect detection method for micro-components in power distribution networks includes: S1. A composite radiation-to-electricity conversion sensitive layer is attached to the surface of a small component in a power distribution network; the composite radiation-to-electricity conversion sensitive layer is used to receive ambient radiation and convert it into an electrical signal, wherein the ambient radiation includes at least one of sunlight, infrared thermal radiation and ultraviolet light; S2. Utilize the radiation-electric energy conversion characteristics of the composite radiation-electric energy conversion sensitive layer under healthy conditions to establish a dynamic radiation-electric energy benchmark; S3. Monitor the power output of the composite radiation-to-electricity conversion sensitive layer in real time. When micro-cracks, corrosion or loosening occur in the micro-components of the power distribution network, the physical structural integrity of the composite radiation-to-electricity conversion sensitive layer is destroyed, resulting in a detectable sudden change in its radiation-to-electricity conversion efficiency, causing the real-time power signal to deviate from the dynamic reference. S4. Compare the real-time collected power signal with the dynamic benchmark, determine whether there are defects in the distribution network micro-components based on the comparison results, and identify the defect type when defects are found. S5. Output defect detection results.

[0023] As described in S1-S5 above, by attaching the radiation-to-electrical-energy conversion sensitive layer to the surface of tiny components in the power distribution network, in-situ sensing of the components themselves is achieved, overcoming the technical limitations of traditional detection methods that rely on external detection equipment. This sensitive layer simultaneously performs the dual functions of defect detection and energy conversion, enabling the components to be self-powered and self-sensing, requiring no external power supply or wiring, and allowing for long-term continuous online monitoring. By establishing a dynamic benchmark for radiation-to-electrical-energy mapping, combined with multi-dimensional information such as environmental radiation parameters and component temperature, the system adaptively fits the energy output pattern under healthy conditions, effectively eliminating the interference of environmental changes on the detection results. By calculating the standardized deviation and identifying defect types based on the trend of deviation changes and the response differences of multi-band conversion units, accurate classification of defect types is achieved. The self-calibration module compensates for the aging effect of the sensitive layer, ensuring the reliability of detection during long-term service. The energy management unit enables the detection system to be energy-self-sufficient, and together with wireless communication, constitutes a completely passive, maintenance-free monitoring solution.

[0024] In one embodiment, step S1 specifically includes: S11. Clean the surface of small components of the power distribution network to remove oil, oxide layer and attachments, and ensure that the surface cleanliness reaches Sa2.5 level or above.

[0025] Cleaning methods include: wiping with anhydrous ethanol to remove organic contaminants; using plasma cleaning to remove the micro-oxidation layer; and using ultrasonic cleaning to remove micro-particulate impurities.

[0026] After cleaning, the surface roughness Ra is controlled within the range of 0.8μm to 1.6μm to enhance the adhesion strength of the sensitive layer.

[0027] S12. Preparation of composite radiation-electric energy conversion sensitive layer material: Mix at least two of the photovoltaic conversion material, infrared thermoelectric conversion material and ultraviolet photoelectric conversion material in a preset ratio.

[0028] Photovoltaic conversion materials use perovskite or dye-sensitized materials to convert sunlight or ultraviolet light into direct current electricity. Infrared thermoelectric conversion materials use bismuth telluride or lead telluride to convert infrared thermal radiation or the component's own thermal field into electrical energy. The ultraviolet photoelectric conversion materials use zinc oxide or gallium nitride to convert ultraviolet light into electrical energy. Material composites are manufactured using liquid-phase mixing or physical vapor deposition methods to ensure uniform dispersion of all components.

[0029] S13. Attach the composite sensitive layer material to the surface of the component: the attachment method is spraying or pasting.

[0030] The spraying method includes: using ultrasonic atomization spraying equipment to uniformly spray the composite sensitive layer material onto the surface of the component, with a spraying thickness of 10μm to 500μm, followed by heat curing treatment at 80℃ to 120℃ for 30 to 60 minutes.

[0031] The bonding method includes: pre-preparing the composite sensitive layer material into a thin film, adhering it to the component surface using thermally and electrically conductive adhesive, with an adhesive layer thickness ≤ 5 μm, and curing conditions of 24 hours at room temperature or 2 hours at 60°C. After adhesion, the sensitive layer and the component surface form a tight mechanical and thermal coupling.

[0032] S14. Conduct quality inspection on the attached sensitive layer: Use an optical microscope to check the uniformity of the sensitive layer to ensure there are no bubbles, cracks, or peeling; use the four-probe method to test the conductivity of the sensitive layer; use a UV-Vis spectrophotometer to test the absorption spectrum of the sensitive layer; and use a thermogravimetric analyzer to test the thermal stability of the sensitive layer. Only after passing the inspection can the layer be put into use.

[0033] As described in S11-S14 above, surface cleaning ensures a tight bond between the sensitive layer and the component, laying the foundation for stress transfer and heat conduction. The composite design of multiple materials enables the sensitive layer to respond simultaneously to radiation in different wavelength bands, improving energy harvesting efficiency and adaptability to various environmental conditions. The optimized design with a thickness range of 10μm to 500μm ensures that the sensitive layer has sufficient mechanical strength to withstand vibration and impact during long-term service, while also ensuring sufficient sensitivity to minute deformations (micrometer level). The application method, either spraying or bonding, can be flexibly selected according to the component shape and site conditions, suitable for components with different structural features such as wire clamps and insulators.

[0034] In one embodiment, establishing a dynamic baseline for radiation-electric energy mapping in step S2 specifically includes: S21. Construct a multi-dimensional environmental radiation monitoring network: Deploy environmental radiation monitoring units near small components of the power distribution network, including light intensity sensors (measurement range 0–2000 W / m², accuracy ±5%), spectrometers (measurement wavelength range 300–2500 nm, resolution 1 nm), infrared radiometers (measurement wavelength range 8–14 μm, sensitivity 0.1 W / m²), and ultraviolet index sensors (measurement wavelength range 280–400 nm, resolution 0.1). All sensors collect data synchronously at a sampling frequency ≥1 Hz, and the data is transmitted to the processing unit via wired or wireless means.

[0035] S22. Collect historical data under healthy conditions: During the initial installation of small components in the distribution network or after confirming a defect-free state, continuously collect power signals and environmental radiation parameters for no less than 30 days. Power signals include open-circuit voltage, short-circuit current, and maximum output power output from the sensitive layer. Environmental radiation parameters include illuminance. Spectral distribution S(λ), infrared radiation flux Ultraviolet Index (UVI) and component surface temperature (T). Data collection covered different weather conditions (sunny, cloudy, rainy) and different time periods (daytime, nighttime, dawn and dusk) to ensure the diversity and representativeness of the data sample.

[0036] S23. Data preprocessing: Clean the collected raw data and remove outliers caused by equipment failure or external interference; use sliding window filtering (window size 10-30 data points) to eliminate high-frequency noise; interpolate and fill in missing data using linear interpolation or cubic spline interpolation methods; align the preprocessed data according to timestamps to form a standardized time series dataset.

[0037] S24. Feature Engineering Construction: Extract the following features from the preprocessed data: (1) Electrical characteristics: output power P, voltage V, current I, calculate their statistics (mean, variance, maximum value, minimum value, coefficient of variation); (2) Radiation characteristics: total radiation intensity Spectral characteristic parameters (peak wavelength) Half-width at half maximum (FWHM), spectral barycenter Infrared radiation flux Ultraviolet Index (UVI); (3) Environmental characteristics: component temperature T, ambient temperature Relative humidity (RH); (4) Derived features: Radiation-to-electrical-energy conversion efficiency η= , where A is the effective area of ​​the sensitive layer; Normalized power = Power temperature coefficient α=( ) / ΔT.

[0038] S25. Constructing a Machine Learning Regression Model: Employing machine learning algorithms such as Support Vector Regression (SVR), Random Forest Regression (RFR), or Gradient Boosting Regression Tree (GBRT), construct the radiation-electrical energy mapping function. =f(E, T, λ).

[0039] The model input is the feature vector X=[ ,S(λ), [UVI, T, RH], the output is the predicted electrical energy. The model was trained using 80% of the samples as the training set and 20% as the validation set, with five-fold cross-validation used to evaluate model performance. Evaluation metrics included root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination (R²).

[0040] The specific training parameters of the machine learning regression model are as follows: the training sample size is no less than 1000 groups (covering health status data under different weather conditions and at different times), the number of iterations is set to 1000-2000, and the learning rate is 0.001-0.01; the kernel function of the support vector regression (SVR) model is the radial basis function (RBF), the penalty coefficient C is 1-10, and the gamma parameter is 0.1-1; the number of decision trees in the random forest regression (RFR) model is 100-200, and the maximum depth is set to 10-20 layers; the learning rate of the gradient boosting regression tree (GBRT) model is 0.01-0.1, the number of iterations is 500-1000, and the maximum tree depth is 5-10 layers. Through the above parameter settings, it is ensured that the root mean square error (RMSE) of the model prediction is ≤5%, the coefficient of determination R² is ≥0.95, and the prediction accuracy of the dynamic benchmark is improved.

[0041] S26. Dynamic Threshold Calculation: Based on the trained model, calculate the prediction interval at a confidence level of 1-α (α=0.05), using a 95% confidence level, based on the assumption of a normal distribution of the prediction residuals. The prediction interval is defined as [ -z·σ, +z·σ], where z is a standard normal distribution. Quantiles, where σ is the standard deviation of the prediction residuals. The upper and lower boundaries of the prediction interval are used as dynamic thresholds; when the real-time power signal exceeds these intervals, it is considered abnormal.

[0042] As described in S21-S26 above, the construction of a multi-dimensional environmental radiation monitoring network provides comprehensive input parameters for establishing a dynamic benchmark, accurately characterizing radiation environment changes under different weather conditions and time periods. Long-term data acquisition under healthy conditions ensures the model learns the normal power output patterns of components under various operating conditions. Multi-feature engineering design fully extracts key information related to radiation-to-power conversion, laying the foundation for accurate model prediction. Compared to traditional statistical models, machine learning regression models have stronger nonlinear fitting capabilities and generalization performance, effectively handling complex coupling relationships between multiple variables. Dynamic threshold settings based on prediction intervals consider the uncertainty of model predictions, effectively reducing false alarm rates while ensuring detection sensitivity.

[0043] In one embodiment, the real-time data acquisition in step S3 specifically includes: S31. Signal Acquisition Hardware Configuration: A low-power analog-to-digital converter (ADC) is used, electrically connected to the composite radiation-to-electrical-energy conversion sensitive layer. The sampling rate is set to 10–100 Hz, and the resolution is 16–24 bits to accurately capture minute changes in electrical signals. The signal conditioning circuit includes: a preamplifier (adjustable gain, ranging from 10 to 1000 times), a low-pass filter (cutoff frequency is half the sampling rate, designed using Butterworth or Chebyshev filters), and an anti-aliasing filter (cutoff frequency is the Nyquist frequency, attenuation ≥60 dB).

[0044] S32. Synchronization Acquisition Mechanism: Multi-sensor synchronization is achieved using GPS or IEEE 1588 Precise Time Protocol (PTP), with a time synchronization accuracy ≤1ms. Synchronously acquired data includes: real-time power signals output from the sensing layer. (Including voltage, current, and power), the environmental radiation monitoring unit outputs light intensity, spectral distribution, infrared radiation flux, ultraviolet index, as well as component temperature and ambient temperature and humidity. All data are timestamped and packaged for storage.

[0045] S33. Data Quality Monitoring: Real-time monitoring of signal quality, including signal-to-noise ratio (SNR), signal amplitude range, and data packet loss rate. When the SNR falls below a set threshold (e.g., 20dB) or the signal amplitude exceeds the expected range, an alarm is automatically triggered and the abnormal event is recorded. A circular buffer mechanism is used to store real-time data from the past 24 hours for subsequent analysis and anomaly tracing.

[0046] S34. Low Power Management: The sampling frequency is dynamically adjusted according to the ambient radiation intensity: a high sampling frequency (≥50Hz) is used during the day when radiation is strong to capture rapid changes; the sampling frequency is reduced (≤10Hz) at night when radiation is weak to save energy. When the system enters standby mode, the sampling frequency is reduced to the lowest level (1Hz), maintaining only basic status monitoring.

[0047] As described in S31-S34 above, the high-precision signal acquisition hardware configuration ensures the accurate capture of minute changes in electrical energy signals, providing a high-quality data foundation for subsequent defect identification. The synchronous acquisition mechanism guarantees the time consistency between the electrical energy signal and environmental radiation parameters, avoiding analysis errors caused by time shifts. Data quality monitoring can promptly detect abnormal data, preventing erroneous data from entering the defect identification process. The low-power management strategy dynamically adjusts the sampling frequency according to environmental conditions, minimizing system power consumption while ensuring detection performance. The detection principle of this invention is that when a component develops microcracks, corrosion, or loosening, the physical structural integrity of the sensitive layer is compromised, leading to a sudden change in conversion efficiency and causing the real-time electrical energy signal to deviate from the dynamic reference.

[0048] In one embodiment, defect identification in step S4 specifically includes: S41. Standardized Deviation Calculation: Based on the real-time collected electrical energy signal. and dynamic benchmark predictions Calculate the standardized deviation , where σ is the standard deviation of the historical data. Deviation It reflects the deviation of real-time electrical energy from a health benchmark. Dimensionless processing eliminates differences in the absolute value of electrical energy under different components and environmental conditions, facilitating unified threshold setting. The absolute value exceeds the preset threshold (Preset threshold) When the value is 2 to 3 (preferably 2.5), it is considered abnormal.

[0049] S42. Preliminary determination of anomaly type: Based on The sign and magnitude of the signal are used to initially determine the type of anomaly. >0 indicates that the real-time power is higher than the predicted value, which may correspond to the compaction of sensitive layers, short circuits, or abnormal environmental enhancements. A value less than 0 indicates that the real-time electrical energy is lower than the predicted value, which may correspond to cracking, peeling, corrosion, or blockage of the sensitive layer. The preliminary judgment result will serve as a reference for subsequent detailed classification.

[0050] S43. Multi-band Response Difference Analysis: The composite radiation-to-energy conversion sensitive layer contains conversion units of different bands (photovoltaic units, thermoelectric units, and ultraviolet units), and each unit has different response characteristics to radiation of different bands. The variation patterns of the output signal of each unit are analyzed: Photovoltaic unit output It is primarily sensitive to sunlight and ultraviolet light; Thermoelectric unit output It is primarily sensitive to infrared thermal radiation and component thermal fields; UV unit output It is primarily sensitive to ultraviolet light.

[0051] Calculate the rate of change of each unit's output relative to the historical baseline: Δ =( - 0) / 0, Δ =( - 0) / 0, Δ =( - 0) / 0, where the subscript 0 indicates the average value under healthy conditions.

[0052] S44. Deviation Trend Analysis: Perform time series analysis on δ to extract its trend characteristics. (1) Mutation detection: The CUSUM algorithm is used to detect mutation points of δ. When δ changes by more than 2σ in a short time (e.g., within 1 minute), it is marked as a rapid mutation.

[0053] (2) Trend analysis: Linear regression was used to fit the trend of δ in the most recent 24 hours. When the slope was significantly negative (p<0.05) and lasted for more than 7 days, it was marked as slow decay.

[0054] (3) Periodicity analysis: Fast Fourier Transform (FFT) is used to analyze the spectrum of δ to detect whether there are periodic fluctuation components. When the frequency corresponding to the main frequency is not related to external factors (such as load fluctuations and daily changes in ambient temperature), it is marked as periodic fluctuation.

[0055] S45. Comprehensive Defect Type Judgment: Based on the differences and deviation trends in multi-band response, a decision tree or support vector machine classification model is used to identify the defect type. (1) Criteria for determining microcracks or delamination: Rapid abrupt change in standardized deviation (CUSUM detection alarm), significant change in photovoltaic unit response (Δ >20%), with relatively small changes in the response of the thermoelectric unit and the ultraviolet unit (Δ). <10%, Δ <10%). Microcracks cause local fractures in the sensitive layer, interrupting the photovoltaic conversion path and causing a sharp drop in output power; delamination leads to an increase in interfacial thermal resistance, while the response of the thermoelectric unit changes relatively little.

[0056] (2) Corrosion judgment criteria: The standardized deviation shows a slow decay trend (the slope is significantly negative for more than 7 consecutive days), and the thermoelectric unit response changes significantly (Δ >15%), with relatively small changes in photovoltaic and ultraviolet units (Δ <10%, Δ <10%). The corrosion process occurs gradually, leading to the deterioration of the properties of the sensitive layer material, with a gradual decrease in thermal conductivity and thermoelectric conversion efficiency.

[0057] (3) Loosening judgment conditions: The standardized deviation shows periodic fluctuations (significant periodic components were detected by spectrum analysis), the fluctuation frequency is different from the frequency of load fluctuations or ambient temperature changes, and the responses of multiple band units all fluctuate but have different phases. Loosening causes the contact state between the sensitive layer and the component interface to change periodically, and the outputs of each band conversion unit fluctuate synchronously.

[0058] S46. Confidence Assessment of Determination Results: Based on the consistency of evidence from multiple sources, calculate the confidence score for the defect type determination: Confidence level = × Deviation significance score + ×Multi-band response consistency score+ × Trend feature clarity score; in, , , For the weighting coefficients, satisfying + + =1, determined based on experimental data (e.g.) =0.4, =0.3, =0.3).

[0059] The deviation significance score, multi-band response consistency score, and trend feature clarity score are all values ​​between 0 and 1. They are calculated by linear mapping or sigmoid function based on the comparison results of the actual measured values ​​of the corresponding features and the preset thresholds. That is, the scores of each item have been normalized to the [0,1] interval.

[0060] When the confidence level is ≥0.8, the judgment result is output directly; when the confidence level is ≤0.5 and <0.8, it is marked as pending confirmation and manual review is triggered; when the confidence level is <0.5, it is judged as data abnormality or edge case and enters the abnormality handling process.

[0061] As described in S41-S46 above, the calculation of standardized deviation eliminates dimensional differences under different component and environmental conditions, facilitating unified threshold setting and cross-component comparison. Multi-band response difference analysis fully utilizes the multi-material characteristics of the composite sensitive layer and the sensitivity differences of different band conversion units to different types of defects, providing key discriminative features for defect type identification. Various analysis techniques for deviation change trends (mutation detection, trend analysis, periodic analysis) can capture the temporal evolution characteristics of different defect patterns. A comprehensive judgment model based on decision trees or support vector machines can effectively integrate multi-source features to achieve high-precision defect type identification. The confidence assessment mechanism provides a reliability measure for the judgment results, facilitating different response strategies for maintenance personnel based on confidence levels.

[0062] In one embodiment, step S4 further includes defect location positioning, specifically including: S47. Arrayed Sensitive Layer Design: The composite radiation-to-electrical-energy conversion sensitive layer is divided into M×N independent radiation-to-electrical-energy conversion sub-units, which are distributed in a matrix array. The size of the sub-units is determined according to the required positioning accuracy, typically ranging from 2mm×2mm to 10mm×10mm. Isolation slots (0.1–0.5mm wide) are provided between adjacent sub-units to avoid electrical signal crosstalk. Each sub-unit has an independently led-out signal line, which is connected to the signal acquisition unit using a flexible circuit board or conductive adhesive.

[0063] Preferably, the M×N array size of the arrayed sensitive layer can be flexibly selected according to the size of the micro components of the power distribution network. Typical implementation schemes include 16×16 array (sub-unit size 5mm×5mm, suitable for small components such as clamps and terminals) and 32×32 array (sub-unit size 3mm×3mm, suitable for medium-sized components such as insulators and tension clamps). The total number of sub-units can be adjusted according to the surface area of ​​the component to ensure that the defect positioning accuracy is not lower than the sub-unit size and to meet the detection needs of different micro components. S48. Power Signal Acquisition for Each Sub-unit: The signal acquisition unit acquires the power signals of each sub-unit in a time-division multiplexing manner using a multiplexer switch, with a sampling rate of not less than 10Hz. The acquisition sequence adopts a line-by-line scanning or random skipping method to reduce signal interference between adjacent sub-units. The real-time power signal of each sub-unit is denoted as... (t), where i=1…M, j=1…N.

[0064] S49. Signal Difference Comparison and Localization: Calculate the deviation of each sub-unit from the healthy baseline. =( - 0) / ,in 0 represents the average output of the sub-unit in a healthy state. The historical standard deviation is used to generate a deviation heatmap. Sub-units significantly larger than the threshold ( > () marked as abnormal sub-units. The defect location is determined by analyzing the spatial distribution pattern of these abnormal sub-units. Single point of failure: Only a single sub-unit If the defect significantly exceeds the standard, and the adjacent sub-units are normal, it is determined to be a point defect (such as a microcrack point or a localized corrosion pit). The defect location is the coordinate of the sub-unit center. Regional anomalies: multiple consecutive adjacent sub-units If the defect significantly exceeds the standard and forms a connected region, it is determined to be a planar defect (such as delamination or large-area corrosion). The defect location is the geometric center of the region, and the defect range is determined by the boundary of the connected region. Linear anomalies: The anomalous sub-units are arranged linearly and are identified as linear defects (such as linear cracks). The defect location is the center of the line segment, and the defect direction is determined by linear fitting.

[0065] S410, Positioning Accuracy Verification: The positioning algorithm is calibrated and verified using simulated defects at known locations to ensure that the positioning error is less than or equal to the sub-unit size. For critical components or situations with low positioning reliability, a secondary, precise inspection can be triggered using a drone or handheld device.

[0066] As described in S47-S410 above, the arrayed sensitive layer design decomposes the detection area into multiple independent detection units, achieving precise spatial localization of defects by comparing the signal differences between each unit. The multiplexing acquisition scheme reduces hardware costs while ensuring positioning accuracy. The generated deviation heatmap visually displays the abnormal distribution, facilitating rapid defect location and extent assessment by maintenance personnel. Identification algorithms for different spatial distribution patterns (single point, region, line) can adapt to the positioning needs of different types of defects. A positioning accuracy verification mechanism ensures the reliability of the system output results.

[0067] In one embodiment, a self-calibration step is also included: S6. Periodic Self-Calibration Process: The system performs a self-calibration every 30 days of operation, or is triggered when environmental conditions change significantly (such as seasonal changes or extreme weather). The self-calibration process includes the following sub-steps: S61. Obtain reference conditions: Select a period of clear weather and stable radiation (hourly radiation intensity variation <10%) as the calibration window. Record the environmental radiation parameters within the calibration window. and component temperature .

[0068] S62. Theoretical Value Calculation: Based on the photoelectric conversion model of the composite radiation-electrical energy conversion sensitive layer, calculate the theoretical output electrical energy. =η0× ×A, where η0 is the initial conversion efficiency and A is the effective area of ​​the sensitive layer. The theoretical model considers the effect of temperature on the conversion efficiency: η(T) = η0 × [1 + ×( -T0)], where It represents the temperature coefficient (-0.2% / ℃ to -0.5% / ℃).

[0069] S63. Actual Value Measurement: Within the same calibration window, collect the actual output electrical energy of the sensitive layer. The average value over 10 consecutive minutes is used to eliminate instantaneous fluctuations.

[0070] S64. Aging Coefficient Calculation: Calculate the aging compensation coefficient. = / . This reflects the performance degradation of the sensitive layer due to long-term service; the normal range is 0.85–1.0. When… When the value is below 0.8, a warning for the lifespan of the sensitive layer is triggered.

[0071] S65, Dynamic Baseline Correction: Applying the aging compensation coefficient to the dynamic baseline model: = × Simultaneously update the standard deviation σ of historical data using the exponentially weighted moving average (EWMA) method: =β× +(1-β)× , where β is the smoothing factor (usually taken as 0.9). To calibrate the standard deviation of the measured data within the calibration window.

[0072] S66. Calibration Records and Traceability: Results of each self-calibration (including...) The system records data (temperature, environmental parameters, calibration time) to a cloud database, forming a performance degradation curve for the entire lifecycle of the sensitive layer. When an abnormal abrupt change occurs in the degradation curve, it automatically analyzes possible causes (such as environmental corrosion, mechanical damage, etc.) and prompts maintenance personnel to check.

[0073] As described in S61-S66 above, the periodic self-calibration mechanism effectively compensates for performance degradation and model drift caused by environmental changes during the long-term service of the sensitive layer, ensuring the stability and accuracy of the detection system throughout its entire lifecycle. The calibration method based on the theoretical model does not rely on external standard sources and is easy to implement in engineering. The calculation of the aging coefficient provides a quantitative basis for predicting the lifespan of the sensitive layer. The exponentially weighted moving average method balances the stability of historical data with the timeliness of new data, enabling the dynamic benchmark model to smoothly adapt to slow changes. The cloud storage and traceability of calibration records provide data support for the health management and replacement decisions of the sensitive layer.

[0074] In one embodiment, an energy management step is also included: S7. Energy Management Process: The electrical energy output from the composite radiation-to-electricity conversion sensitive layer is collected, stored, and distributed through the energy management circuit to achieve completely passive operation of the system. Specifically, this includes: S71. Energy Harvesting: Employs high-efficiency energy harvesting chips (such as LTC3108 and BQ25570) with an input voltage range of 0.1–5V, capable of extracting energy from milliwatt-level low-power electrical energy. The energy harvesting circuit includes: an input protection circuit (preventing overvoltage and reverse current), a maximum power point tracking (MPPT) circuit (adjusting the input impedance in real time to ensure the sensitive layer operates at its maximum power point), and a boost circuit (boosting the low voltage to the charging voltage of the energy storage device).

[0075] S72, Energy Storage: A hybrid energy storage solution is adopted, including: Supercapacitors (capacitance 0.1~1F, operating voltage 2.7~5.5V): used for short-term energy buffering and rapid response to load changes; Solid-state thin-film batteries (capacity 1-10mAh, operating voltage 3.0-4.2V): used for long-term energy storage and providing stable power supply.

[0076] The energy storage device is connected to the energy harvesting chip, and the charging current is controlled and protected to prevent overcharging and over-discharging. The energy storage status monitoring circuit monitors the voltage and capacity of the energy storage device in real time. When the stored energy is below the threshold, the system enters a low-power mode.

[0077] S73. Energy Distribution: The Power Management Unit (PMU) dynamically distributes electrical energy based on the power consumption characteristics and real-time requirements of each module. Signal acquisition unit: power consumption is about 0.5~2mW, priority power supply is given to ensure continuous data acquisition; Defect identification module: power consumption is about 1-5mW, power is supplied on demand, and it only starts when there is data to be collected; Wireless communication unit: power consumption is about 10-50mW (transient), adopts burst transmission mode, data is accumulated to a certain amount and then sent in batches, and is turned off immediately after the transmission is completed; Environmental radiation monitoring unit: power consumption is about 1-3mW, and it adopts an intermittent sampling mode (sampling for 1 minute, then sleeping for 5 minutes) to reduce average power consumption.

[0078] The PMU uses a power budget algorithm to ensure that the energy storage level is always maintained within a safe range. When insufficient power is predicted, it actively reduces the operating frequency of high-power modules or extends the sleep time.

[0079] S74. Wireless Transmission: Low-power wide-area network (LPWAN) technologies (such as LoRa and NB-IoT) are used to upload defect detection results to the remote monitoring platform. Data encapsulation includes: component ID, defect type, defect location, deviation, timestamp, confidence level, etc. The transmission strategy adopts adaptive data rate (ADR), which automatically adjusts the transmission power and data rate according to the signal strength to minimize energy consumption while ensuring transmission reliability. For critical alarms (such as microcracks and severe corrosion), high-priority transmission is triggered immediately; for routine status data, periodic batch transmission (such as once a day) is used.

[0080] S75. Energy Budget Management: An energy budget assessment is performed upon system startup, estimating the average daily collectable energy based on expected ambient radiation intensity and historical data. and daily energy consumption .when < When this happens, the system automatically enters energy-saving mode: reducing the sampling frequency (from 50Hz to 10Hz), extending the wireless transmission cycle (from 1 hour to 6 hours), and turning off non-core functions (such as real-time waveform display) until the energy storage level is restored.

[0081] As described in S71-S75 above, the application of energy harvesting circuitry and MPPT technology maximizes the efficiency of collecting weak electrical energy, ensuring continuous system operation under low radiation conditions. The hybrid energy storage scheme combines the high power density of supercapacitors and the high energy density of thin-film batteries, enabling it to handle sudden high power consumption demands while ensuring long-term stable power supply. The dynamic energy allocation strategy optimizes power usage based on the real-time needs of each module, improving energy utilization efficiency. The selection of low-power wide-area network (LPWAN) technology minimizes communication energy consumption while ensuring transmission distance and reliability. The energy budget management mechanism enables the system to be adaptive, automatically adjusting its operating strategy when environmental conditions change, ensuring long-term passive operation.

[0082] In one embodiment, the basis for determining the key threshold is verified through multiple sets of comparative experiments. All experimental samples are from the actual operation site of the power distribution network, and the experimental data are based on a unified data acquisition platform. The preprocessing process is consistent with the aforementioned embodiments of the present invention.

[0083] Verification experiments on composite radiation-to-electricity conversion sensitive layer thickness ranging from 10 μm to 500 μm Experimental objective: To determine the optimal range of sensitive layer thickness, balance mechanical strength and detection sensitivity, and ensure responsiveness to micron-level defects.

[0084] Experimental Samples: Composite radiation-to-electrical energy conversion sensitive layer samples with thicknesses of 5 μm, 10 μm, 50 μm, 100 μm, 200 μm, 300 μm, 500 μm, and 600 μm were prepared and attached to the surface of aluminum wire clamps. Simulated cracks with depths of 5 μm, 10 μm, 20 μm, 50 μm, and 100 μm were created on the surface of the wire clamps using a micron-level scratch instrument. The response sensitivity of sensitive layers of different thicknesses to cracks of different depths was tested.

[0085] Experimental steps: Under standard radiation conditions (light intensity 1000W / m², AM1.5 spectrum), measure the output power P0 of each sample in a defect-free state; after creating a simulated crack, measure the output power P1; calculate the power change rate ΔP=|P1-P0| / P0×100%; and statistically analyze the detection capabilities of samples of different thicknesses for various types of cracks.

[0086] The experimental results are shown in Table 1 below:

[0087] Table 1: Detection sensitivity of sensitive layers of different thicknesses to cracks of different depths (power change rate ΔP / %). Experimental conclusions: The thickness of the sensitive layer is negatively correlated with the detection sensitivity; the smaller the thickness, the more sensitive it is to microcracks, but the lower the mechanical strength. A thickness of 10 μm still shows an 8.5% power change rate for a 5 μm crack, which is effective for detection; however, a thickness of 500 μm only shows a 1.1% change for a 20 μm crack, making accurate detection difficult. While thicknesses below 10 μm offer higher sensitivity, the fabrication process is more complex, and the layer is prone to breakage during long-term service. Considering detection sensitivity, fabrication feasibility, and long-term reliability, the sensitive layer thickness range is determined to be 10 μm–500 μm, with an optimal range of 50 μm–200 μm.

[0088] Standardized deviation threshold =2~3 verification experiments; Experimental objective: To determine the standardized deviation threshold that distinguishes between normal fluctuations and defect anomalies, and to balance the false negative rate and false positive rate.

[0089] Experimental Samples: 50 defect-free power distribution network micro-components (clamps, insulators, etc.) were selected as normal samples, and their power signals and environmental radiation parameters were collected for 30 consecutive days; 30 components with real defects were selected (confirmed on-site: 10 micro-cracks, 10 corrosion, and 10 loosening), and their power signals before and after the defects were collected.

[0090] Experimental steps: Calculate the standardized deviation δ of each sample according to the method of this invention, and statistically analyze the distribution range of δ for normal samples and the distribution range of δ for defective samples; set different... Calculate the detection accuracy at each threshold (1.0, 1.5, 2.0, 2.5, 3.0, 3.5).

[0091] The experimental results are shown in Table 2 below:

[0092] Table 2: Differences Detection performance comparison under threshold Experimental conclusion: When At a resolution of 2.0, the overall accuracy is highest (92.2%), with a false positive rate of only 3.6% for normal samples and a defective sample detection rate of 95.8%. When the value is less than 2.0, the false alarm rate increases significantly; when... When the value is greater than 3.0, the detection rate decreases significantly. Considering that distribution network defect detection is more sensitive to missed detections (which could lead to safety accidents), a standardized deviation threshold was determined after comprehensive consideration. 2 to 3, preferred =2.5.

[0093] Verification experiment of defect type determination rules; Experimental objective: To verify the accuracy of the defect type determination rule based on the deviation change trend and multi-band response differences.

[0094] Experimental Samples: Forty small components of a power distribution network with real defects were selected, including 15 with microcracks, 15 with corrosion, and 10 with loosening (all confirmed by on-site investigation and laboratory analysis). All samples have been equipped with composite radiation-to-electricity conversion sensitive layers and are operating normally.

[0095] Experimental steps: Collect real-time electrical signals and environmental parameters of each sample according to the method of this invention, calculate the trend of the standardization deviation δ, analyze the response differences of multi-band conversion units (photovoltaic units, thermoelectric units); compare the automatic judgment results with the actual defect types, and statistically analyze the identification accuracy of each type of defect.

[0096] The experimental results are shown in Table 3 below:

[0097] Table 3: Statistics on the Accuracy of Defect Type Determination Analysis of misidentification cases: One microcrack sample was misidentified as loose because the crack was a progressive type, with the deviation showing a slow decay trend rather than a rapid abrupt change; one corrosion sample was misidentified as a microcrack because localized peeling in the corrosion area caused a rapid drop in electrical energy; one loose sample was misidentified as a microcrack because the gap between the sensitive layer and the component changed drastically after loosening. For these edge cases, a confidence assessment mechanism can be used to mark them as pending confirmation, triggering manual review.

[0098] Experimental conclusions: The defect type determination rule based on the deviation change trend and multi-band response differences achieved an overall accuracy of 92.5%, with a microcrack and corrosion identification accuracy of 93.3% and a loosening identification accuracy of 90.0%. It has a good ability to distinguish typical defect patterns, and edge cases can be compensated for by confidence assessment and manual verification mechanisms, meeting the needs of engineering applications.

[0099] Example 2, please refer to Figure 2 As shown in the figure, the high-precision defect detection system for micro-components in power distribution networks described in this embodiment includes: A composite radiation-to-electrical-energy conversion sensitive layer is disposed on the surface of a small component in a power distribution network. It receives ambient radiation and converts it into electrical signals. The ambient radiation includes at least one of sunlight, infrared thermal radiation, and ultraviolet light. This sensitive layer is a thin film structure with a thickness of 10 μm to 500 μm, made of at least two of photovoltaic conversion materials, infrared thermoelectric conversion materials, and ultraviolet photoelectric conversion materials. The sensitive layer contains at least two of these materials, forming conversion units for different wavelength bands. The sensitive layer can be designed in an array, containing multiple arrayed radiation-to-electrical-energy conversion sub-units for precise defect location. When microcracks, corrosion, or loosening occur in the small component of the power distribution network, the physical structural integrity of the sensitive layer is compromised, leading to a detectable abrupt change in its radiation-to-electrical-energy conversion efficiency.

[0100] A signal acquisition unit, electrically connected to the composite radiation-to-electricity conversion sensitive layer, is used to acquire the electrical energy signal. The signal acquisition unit includes: Multi-channel analog front end: includes a low-noise amplifier (adjustable gain from 10 to 1000 times), programmable filters (low-pass, high-pass, band-pass) and a 16 to 24-bit analog-to-digital converter (ADC) with an adjustable sampling rate from 10 to 100 Hz; Multiplexer switch: used for signal switching of array-type sensitive layers, supporting line-by-line scanning or random jump acquisition modes; Signal conditioning circuit: Provides sensor bias, reference voltage, and common-mode rejection to ensure the quality of the acquired signal; Data buffer: Adopts a dual-buffer design to support continuous acquisition and real-time processing.

[0101] An environmental radiation monitoring unit is used to collect environmental radiation parameters in real time. It includes: Light intensity sensor: Measurement range 0~2000W / m², accuracy ±5%, used to monitor sunlight intensity; Spectrum analyzer: measures wavelengths from 300 to 2500 nm with a resolution of 1 nm, used to obtain spectral distribution characteristics; Infrared radiometer: measures wavelength range of 8–14 μm, sensitivity of 0.1 W / m², used to monitor infrared thermal radiation flux; Ultraviolet index sensor: measures wavelength range 280–400 nm, resolution 0.1, used to monitor ultraviolet radiation intensity; Temperature sensor: Measurement range -40~125℃, accuracy ±0.5℃, used to monitor component surface temperature and ambient temperature; Humidity sensor: Measurement range 0~100%RH, accuracy ±3%RH, used to monitor ambient humidity.

[0102] The dynamic benchmark modeling module is used to establish a dynamic benchmark for the radiation-electrical energy mapping of micro-components in the distribution network under healthy conditions, based on historical power signals and environmental radiation parameters, through regression analysis or machine learning methods. The dynamic benchmark modeling module includes: Data preprocessing unit: performs data cleaning, filtering, interpolation, and standardization. Feature engineering unit: Extracts electrical energy features, radiation features, environmental features, and derived features from raw data; Model training unit: The radiation-electrical energy mapping function is trained using Support Vector Regression (SVR), Random Forest Regression (RFR), or Gradient Boosting Regression Tree (GBRT) algorithms. =f(E, T, λ); Model evaluation unit: Evaluates model performance through cross-validation and calculates metrics such as RMSE, MAE, and R². Dynamic threshold calculation unit: Calculates the confidence boundary of the prediction interval as the dynamic threshold, and outputs... , upper and lower bounds of the confidence interval.

[0103] The defect identification module compares real-time acquired power signals with a dynamic benchmark, determines the presence of defects in minute components of the distribution network based on the comparison results, and identifies the type of defect if present. The defect identification module includes: Deviation Calculation Unit: Calculates standardized deviation ; Anomaly detection unit: Judgment Does it exceed the preset threshold? Output an exception flag; Multi-band response analysis unit: Calculates the rate of change of output of each band conversion unit (photovoltaic unit, thermoelectric unit, ultraviolet unit); Trend Analysis Unit: The CUSUM algorithm is used to detect abrupt change points, linear regression is used to analyze the trend slope, and FFT is used to analyze the periodic components. Classifier unit: Employs decision tree or support vector machine models, integrates multi-band response differences and deviation trend features, and outputs defect types (microcracks, corrosion, loosening). Location unit: For array-type sensitive layers, the location coordinates and range of defects are output by comparing the differences in electrical signals of each sub-unit; Confidence assessment unit: Based on the consistency of multi-source evidence, outputs a confidence score.

[0104] The self-calibration module is used to periodically perform the self-calibration process. The self-calibration module includes: Calibration condition judgment unit: Monitors environmental conditions and determines whether the self-calibration requirements are met (sunny weather, stable radiation); Theoretical value calculation unit: calculates the theoretical output electrical energy based on the photoelectric conversion model; Actual value measurement unit: Collects the actual output electrical energy within the calibration window; Aging coefficient calculation unit: Calculation = / ; Model parameter correction unit: updates the parameters of the dynamic baseline model to compensate for the aging effect of the sensitive layer; Lifespan warning unit: when When the threshold is reached, a notification to replace the sensitive layer is triggered.

[0105] An energy management unit is used to store at least a portion of the electrical energy output from the composite radiation-to-electricity conversion sensitive layer and to power the signal acquisition unit, environmental radiation monitoring unit, dynamic benchmark modeling module, defect identification module, and wireless communication unit. The energy management unit includes: Energy harvesting circuitry: includes input protection, maximum power point tracking (MPPT), and boost circuitry; Energy storage devices: supercapacitors (short-term buffering) and solid-state thin-film batteries (long-term storage); Power Management Unit (PMU): Enables dynamic allocation of power supply and power consumption management for each module; Energy storage monitoring circuit: Real-time monitoring of energy storage voltage and capacity; Power budget module: Dynamically adjusts the system operation mode based on environmental radiation predictions.

[0106] A wireless communication unit is used to wirelessly transmit defect identification results to a remote monitoring platform. The wireless communication unit includes: Low-power wide-area network modules (LoRa or NB-IoT): Enable long-distance, low-power transmission; Data encapsulation unit: Packages the detection results (part ID, defect type, defect location, deviation, timestamp, confidence level) into structured data; Transmission strategy unit: Adaptive data rate (ADR) and burst transmission mode are used to optimize transmission energy consumption; Retransmission mechanism: Ensures reliable delivery of critical alarm data.

[0107] System workflow: Data input stage: The composite radiation-to-electricity conversion sensitive layer generates weak electrical energy under environmental radiation irradiation. The energy management unit collects and stores this electrical energy to power the system. The environmental radiation monitoring unit simultaneously collects parameters such as light intensity, spectral distribution, infrared radiation flux, and ultraviolet index.

[0108] Benchmark establishment phase: When the system is first deployed or periodically calibrated, historical power signals and environmental radiation parameters under healthy conditions are collected. The dynamic benchmark modeling module trains the radiation-power mapping function through machine learning to establish a dynamic benchmark.

[0109] Real-time monitoring phase: The signal acquisition unit acquires the output power of the sensitive layer in real time at a set sampling rate (10-100Hz), and the environmental radiation monitoring unit acquires the current environmental parameters simultaneously; the defect identification module compares the real-time power signal with the dynamic benchmark prediction value and calculates the standardized deviation δ.

[0110] Defect determination phase: When δ exceeds the preset threshold At that time, the defect identification module determines that a defect exists; further analysis of the variation trend of δ and the response differences of the multi-band conversion unit is performed, and the defect type (micro-crack, corrosion, loosening) is identified through the classification model; for the array-type sensitive layer, the defect location is located by comparing the signal differences of each sub-unit.

[0111] Self-calibration phase: The self-calibration module periodically (e.g., every 30 days) performs the self-calibration process and calculates the aging compensation coefficient. Correct the parameters of the dynamic benchmark model and compensate for the performance degradation of the sensitive layer.

[0112] Results output stage: Defect identification results (whether a defect exists, defect type, defect location) are uploaded to the remote monitoring platform via wireless communication unit; the energy management unit continuously supplies power to each module, achieving completely passive, wiring-free, and long-term online monitoring.

[0113] As described above, this system achieves in-situ continuous monitoring of the components themselves by attaching a composite radiation-to-electrical-energy conversion sensitive layer to the surface of tiny components in the power distribution network. This eliminates the need for external power supply and wiring, solving the problems of traditional detection methods such as reliance on external equipment, poor real-time performance, and high deployment costs. Dynamic benchmark modeling and environmental radiation compensation eliminate the interference of environmental changes on the detection results, improving detection accuracy. Standardized deviation calculation and multi-feature fusion analysis enable high-precision identification of defect types, providing richer information for operation and maintenance decisions. The array design enables precise location of defects, and self-calibration and energy management mechanisms ensure the system's long-term reliability and self-sufficiency.

[0114] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims. To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

Claims

1. A high-precision defect detection method for micro-components in power distribution networks, characterized in that, include: S1. Attach the composite radiation-to-electricity conversion sensitive layer to the surface of micro-components in the power distribution network; The composite radiation-to-electricity conversion sensitive layer is used to receive environmental radiation and convert it into electrical signals. The environmental radiation includes at least one of sunlight, infrared thermal radiation, and ultraviolet light. The composite radiation-to-electricity conversion sensitive layer is a thin film structure with a thickness of 10μm to 500μm, and is made of at least two of the following: photovoltaic conversion materials, infrared thermoelectric conversion materials, and ultraviolet photoelectric conversion materials. S2. Utilize the radiation-electric energy conversion characteristics of the composite radiation-electric energy conversion sensitive layer under healthy conditions to establish a dynamic radiation-electric energy benchmark; The establishment of the radiation-electric energy dynamic benchmark specifically includes: Collect environmental radiation parameters, which include at least one of light intensity, spectral distribution, infrared radiation flux, and ultraviolet index; Based on historical electrical energy signals, environmental radiation parameters, component temperature, and radiation wavelength distribution, a radiation-electrical energy mapping function for the micro-component under healthy conditions is constructed using a machine learning regression model. ； in To predict electrical energy output, E For environmental radiation parameters, T For component temperature, λ The radiation wavelength distribution; The power prediction range under the current environmental conditions is calculated based on the mapping function, and the boundary of the prediction range is used as a dynamic threshold. S3. Monitor the power output of the composite radiation-to-electricity conversion sensitive layer in real time. When micro-cracks, corrosion or loosening occur in the micro-components of the power distribution network, the physical structural integrity of the composite radiation-to-electricity conversion sensitive layer is destroyed, resulting in a detectable sudden change in its radiation-to-electricity conversion efficiency, causing the real-time power signal to deviate from the dynamic reference. S4. Compare the real-time collected power signal with the dynamic benchmark, determine whether there are defects in the distribution network micro-components based on the comparison results, and identify the defect type when defects are found. The step of determining whether the micro-components of the power distribution network have defects based on the comparison results includes: The standardized deviation between the real-time acquired power signal and the dynamic benchmark prediction value is calculated. When the standardized deviation exceeds a preset threshold, it is determined that there is a defect. The standardized deviation in For real-time acquisition of electrical energy signals, This is a dynamic benchmark prediction value. σ The standard deviation of historical data; The identification of defect types includes identifying defect types based on the changing trend of the normalized deviation and the response differences of different band conversion units in the composite radiation-to-electricity conversion sensitive layer, specifically including: When the standardization deviation shows a rapid abrupt change and the photovoltaic conversion unit response changes significantly, it is judged as a microcrack or delamination; When the standardization deviation shows a slow decreasing trend and the response of the thermoelectric conversion unit changes significantly, it is judged as corrosion; When the standardized deviation exhibits periodic fluctuations, it is considered loose. S5. Output defect detection results.

2. The high-precision defect detection method for micro-components in power distribution networks according to claim 1, characterized in that: In S1, the composite radiation-to-electricity conversion sensitive layer includes multiple radiation-to-electricity conversion units arranged in an array; In step S3, the power signals of each conversion unit are collected respectively; In step S4, the location of the defect is determined by comparing the differences in the electrical signals of each conversion unit.

3. The high-precision defect detection method for micro-components in power distribution networks according to claim 1, characterized in that: It also includes a self-calibration step: The self-calibration process is performed periodically. Under the condition that the environmental radiation is stable and measurable, the real-time electrical energy signal is compared with the theoretical calculation value to correct the model parameters of the dynamic reference and compensate for the aging effect of the composite radiation-electrical energy conversion sensitive layer.

4. The high-precision defect detection method for micro-components in power distribution networks according to claim 1, characterized in that: It also includes energy management steps: At least a portion of the electrical energy output from the composite radiation-to-electricity conversion sensitive layer is stored through an energy management circuit, which simultaneously powers the signal acquisition, defect identification, and wireless transmission modules, enabling passive, wiring-free, and long-term online monitoring.

5. The high-precision defect detection method for micro-components in power distribution networks according to claim 1, characterized in that: The power distribution network micro-components include at least one of the following: clamps, insulators, tension clamps, suspension clamps, vibration dampers, parallel groove clamps, and terminals.

6. A high-precision defect detection system for micro-components in power distribution networks, used to implement the high-precision defect detection method for micro-components in power distribution networks as described in any one of claims 1-5, characterized in that, include: A composite radiation-to-electricity conversion sensitive layer is disposed on the surface of a micro component of a power distribution network to receive ambient radiation and convert it into an electrical signal. The ambient radiation includes at least one of sunlight, infrared thermal radiation, and ultraviolet light. When the micro component of the power distribution network develops microcracks, corrosion, or loosening, the physical structural integrity of the composite radiation-to-electricity conversion sensitive layer is compromised, resulting in a detectable abrupt change in its radiation-to-electricity conversion efficiency. The signal acquisition unit is electrically connected to the composite radiation-to-electricity conversion sensitive layer and is used to acquire the electrical signal. The dynamic benchmark modeling module is used to establish a dynamic benchmark of radiation-electrical energy mapping of the micro-components of the distribution network in a healthy state based on historical power signals and environmental radiation parameters, through regression analysis or machine learning methods. The defect identification module is used to compare the real-time collected power signals with the dynamic benchmark, determine whether there are defects in the micro components of the power distribution network based on the comparison results, and identify the type of defect when defects are found. The wireless communication unit is used to wirelessly transmit the defect identification results to the remote monitoring platform; An energy management unit is used to store at least a portion of the electrical energy output by the composite radiation-to-electricity conversion sensitive layer and to power the signal acquisition unit, dynamic reference modeling module, defect identification module, and wireless communication unit.

Images