Power transmission line hidden danger prediction method and device based on big data analysis

Abstract

The invention discloses a power transmission line hidden danger prediction method and device based on big data analysis, and relates to the technical field of power grid safety. According to the main technical scheme, the method comprises the steps of collecting multi-source heterogeneous data associated with operation of a power transmission line, then preprocessing the data, and obtaining first data which converts time sequence data of different data sources into a time sequence structure of a unified time scale in a data processing category conforming to wavelet transform and a convolutional neural network; the second data is used for pre-labeling images and texts in the non-structural data of different data sources; processing the first data and the second data by applying cooperation of wavelet transform and a convolutional neural network so as to extract structured time sequence features, unstructured image features and text features related to the hidden danger of the power transmission line; and finally, inputting the feature data into a preset hidden danger prediction model of a Bayesian network for processing, and outputting existence probabilities of different preset hidden dangers so as to finally obtain a prediction result of the hidden dangers of the power transmission line.

Description

Technical Field

[0001] This application relates to the field of power grid safety technology, and in particular to a method and device for predicting potential hazards in transmission lines based on big data analysis. Background Technology

[0002] As the "blood vessels" of the power grid, the hidden dangers of transmission lines are directly related to the stable operation of the power system and may even lead to serious consequences such as large-scale power outages. Effectively identifying, investigating and managing these hidden dangers is a core link in ensuring the safety of the power grid.

[0003] Currently, there are various hidden dangers in power transmission lines, which can be divided into three main categories from the perspective of influencing factors: natural, equipment, and human factors. Among them, hidden dangers caused by natural factors include lightning strikes causing line breakage, ice galloping, bird damage causing short circuits, tree interference causing discharge, etc. Hidden dangers caused by equipment include broken conductor strands, aging insulators, corrosion of hardware, and tilting of towers, etc. Hidden dangers caused by external factors include illegal construction excavation, hanging foreign objects, etc.

[0004] Traditional manual inspections suffer from low efficiency and incomplete coverage. They are now gradually shifting towards intelligent monitoring. So, how to combine big data analysis with intelligent technology to predict potential hazards in power transmission lines in response to diverse types of hazards is an urgent problem to be solved. Summary of the Invention

[0005] This application provides a method and device for predicting hidden dangers in power transmission lines based on big data analysis. The main purpose is to use multi-source heterogeneous data, and to extract features related to hidden dangers in power transmission lines by using wavelet transform and convolutional neural network collaboration. Then, Bayesian network is used to perform inference and prediction, thus forming a complete link of "feature extraction → feature fusion → probabilistic inference". This provides an intelligent solution for predicting hidden dangers in power transmission lines by making full use of big data to cope with diverse types of hidden dangers.

[0006] To achieve the above objectives, this application mainly provides the following technical solutions: The first aspect of this application provides a method for predicting potential hazards in power transmission lines based on big data analysis. This method includes: Collect multi-source heterogeneous data related to the operation of transmission lines. The data dimensions included in the multi-source heterogeneous data include at least environmental meteorological data, measurement data, equipment ledger data, historical fault data, remote signaling data, and inspection data. Preprocessing is performed on the multi-source heterogeneous data to obtain first data and second data. The first data is data that converts time-series data from different data sources into time-series structure with a unified time scale. The second data is data that pre-labels images and text in unstructured data from different data sources. The first data is processed using a preset wavelet transform algorithm to extract structured time-series features associated with the analysis of potential hazards in transmission lines, which are then used as the first feature data. The second data is processed using a pre-set feature extraction model to extract unstructured image features and text features associated with the analysis of potential hazards in transmission lines. The unstructured image features correspond to the second feature data, and the text features correspond to the third feature data. The pre-set feature extraction model is pre-trained using a convolutional neural network. The first feature data, the second feature data, and the third feature data are input into the pre-set hidden danger prediction model for processing, and the existence probability corresponding to different pre-set hidden dangers is output to obtain the prediction result of the hidden danger of the transmission line. The pre-set hidden danger prediction model is pre-trained using a Bayesian network.

[0007] The second aspect of this application provides a transmission line hazard prediction device based on big data analysis, the device comprising: The acquisition unit is used to acquire multi-source heterogeneous data related to the operation of transmission lines. The data dimensions included in the multi-source heterogeneous data include at least environmental meteorological data, measurement data, equipment ledger data, historical fault data, remote signaling data, and inspection data. The first processing unit is used to perform preprocessing on the multi-source heterogeneous data to obtain first data and second data. The first data is data that converts time-series data from different data sources into time-series structure data with a unified time scale, and the second data is data that pre-labels images and text in unstructured data from different data sources. The second processing unit is used to process the first data using a preset wavelet transform algorithm to extract structured time-series features associated with the analysis of potential hazards in transmission lines, as the first feature data. The third processing unit is used to process the second data using a preset feature extraction model to extract unstructured image features and text features associated with the analysis of potential hazards in transmission lines. The unstructured image features correspond to the second feature data, and the text features correspond to the third feature data. The preset feature extraction model is pre-trained using a convolutional neural network. The fourth processing unit is used to input the first feature data, the second feature data and the third feature data into the preset hidden danger prediction model for processing, output the existence probability corresponding to different preset hidden dangers, and obtain the prediction result of the hidden danger of the transmission line. The preset hidden danger prediction model is pre-trained using a Bayesian network.

[0008] A third aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method for predicting potential hazards in power transmission lines based on big data analysis as described above.

[0009] A fourth aspect of this application provides an electronic device, the device including at least one processor, and at least one memory and bus connected to the processor; The processor and the memory communicate with each other via the bus. The processor is used to call program instructions in the memory to execute the above-described method for predicting potential hazards in power transmission lines based on big data analysis.

[0010] By employing the above-described technical solution, the technical solution provided in this application has at least the following advantages: This application provides a method and apparatus for predicting potential hazards in power transmission lines based on big data analysis. It collects multi-source heterogeneous data related to the operation of power transmission lines, including at least environmental meteorological data, measurement data, equipment ledger data, historical fault data, remote signaling data, and inspection data. This data is then preprocessed, and in accordance with the data processing principles of wavelet transform and convolutional neural networks, the following results are obtained: first data is obtained by converting time-series data from different data sources into a time-series structure with a unified time scale; second data is obtained by pre-labeling images and text from unstructured data from different data sources. The first data is then processed using a pre-set wavelet transform algorithm to extract structured time-series features, and the second data is processed using a pre-set feature extraction model trained by a convolutional neural network to extract unstructured image and text features. Finally, these feature data are input into a pre-set hazard prediction model pre-trained by a Bayesian network for processing, outputting the probability of existence for different pre-set hazards, thus obtaining the prediction result for potential hazards in the power transmission lines.

[0011] Compared to existing technologies for predicting potential hazards, this application uses multi-source heterogeneous data and employs wavelet transform and convolutional neural network collaboration to extract features related to potential hazards in power transmission lines. Then, a Bayesian network is used to perform inference and prediction, thus forming a complete chain of "feature extraction → feature fusion → probabilistic inference". This solution addresses diverse hazard types, makes full use of big data, and provides an intelligent solution for predicting potential hazards in power transmission lines.

[0012] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0013] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of this application. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 A flowchart illustrating a method for predicting potential hazards in power transmission lines based on big data analysis, provided in this application embodiment; Figure 2 A block diagram illustrating the composition of a power transmission line hazard prediction device based on big data analysis, provided in this application embodiment; Figure 3 A block diagram of another power transmission line hazard prediction device based on big data analysis provided in this application embodiment. Detailed Implementation

[0014] Exemplary embodiments of the present application will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present application are shown in the drawings, it should be understood that the present application may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this application will be thorough and complete, and will fully convey the scope of the present application to those skilled in the art.

[0015] This application provides a method for predicting potential hazards in power transmission lines based on big data analysis, such as... Figure 1 As shown, the following specific steps are provided in this embodiment of the invention: 101. Collect multi-source heterogeneous data related to the operation of transmission lines. The data dimensions included in the multi-source heterogeneous data shall include at least environmental meteorological data, measurement data, equipment ledger data, historical fault data, remote signaling data, and inspection data.

[0016] Among these, environmental meteorological data, such as time-series data on temperature, humidity, wind speed, wind direction, rainfall, lightning, and icing thickness, can be used to identify risks such as lightning strikes, icing, and wind deflection; measurement data, such as parameters including voltage, current, and line load, reflects the operating status of transmission lines and can be used to analyze potential hazards such as overload and temperature rise; equipment ledger data, such as tower type, conductor type, installation time, and historical maintenance records, provides background information on infrastructure for hazard correlation modeling; and historical fault data, such as fault time, cause, and impact range, helps train hazard prediction models. Generally, the collected environmental meteorological data, measurement data, equipment ledger data, and historical fault data are usually structured data.

[0017] Remote signaling data refers to switch signal data collected through remote monitoring technology that reflects the "state changes" of transmission lines and related equipment. It is mainly used for real-time monitoring of whether the operating status of equipment has changed and is one of the core data types of the Supervisory Control And Data Acquisition (SCADA) system for power systems. The collected remote signaling data is typically structured data in power systems.

[0018] The inspection data includes, for example, drone inspection information and manual inspection information. Drone inspection information is used to identify visual hazards such as hanging foreign objects, damaged insulators, loose conductors, and equipment corrosion. Manual inspection information consists of text data containing descriptions of potential hazards, such as tower tilting, foundation settlement, and construction within power line corridors. The collected inspection data typically includes unstructured data, such as image data and text data.

[0019] 102. Perform preprocessing on multi-source heterogeneous data to obtain first data and second data. The first data is data that converts time-series data from different data sources into time-series structure with a unified time scale. The second data is data that pre-labels images and text in unstructured data from different data sources.

[0020] This step involves preprocessing the collected multi-source heterogeneous data, and examples include the following: Outlier removal includes removing data that exceeds the physically reasonable range, such as current exceeding the conductor's maximum current carrying capacity or significantly abnormal temperature. Missing value handling includes methods such as interpolation, mean filling, or deletion, depending on the specific data type. Format and unit standardization are also important, such as converting temperature to degrees Celsius and wind speed to m / s.

[0021] Temporal reconstruction and alignment, such as performing time window sliding, resampling, and time alignment operations on time-series meteorological and measurement data to form a unified temporal data structure; image and text pre-annotation, such as manually or semi-automatically annotating potential hazards on UAV data and manual inspection text, such as marking the location of insulator damage or keywords such as "loose tower base" in the text.

[0022] The main purpose of this preprocessing step is to extract valuable and high-quality data from multi-source heterogeneous data for subsequent prediction of transmission line hazards, and to perform feature extraction operations for the "wavelet transform algorithm" and "feature extraction model trained on convolutional neural network" used in the subsequent embodiments of this application. This step performs preprocessing operations on multi-source heterogeneous data to obtain two types of data. To distinguish them, the embodiments of this application use the terms "first" and "second" to identify them. The first data is the data in which time series data from different data sources are converted into time series structure with a unified time scale, and the second data is the data in which images and text in unstructured data from different data sources are pre-annotated.

[0023] This step involves preprocessing the multi-source heterogeneous data based on its inherent characteristics and the feature extraction methods to be used (such as wavelet transform algorithm and feature extraction model trained based on convolutional neural network). Under the condition that these two requirements are met, this step is implemented to obtain the first and second data. The specific implementation steps of the preprocessing are not specifically limited here, as long as they meet the above two requirements.

[0024] 103. The first data is processed using a preset wavelet transform algorithm to extract the structured time-series features associated with the analysis of potential hazards in transmission lines, which are then used as the first feature data.

[0025] Specifically, in conjunction with the need to extract features using the "wavelet transform algorithm," preprocessing is performed on the multi-source heterogeneous data (such as environmental meteorological data, measurement data, equipment ledger data, historical fault data, remote signaling data, and inspection data) collected as illustrated in Example 1 of this application: Example 1: Wavelet Transform is a time-frequency analysis tool whose core advantages are: it can analyze signals simultaneously in the time and frequency domains, and is particularly good at capturing transient features (such as abrupt changes, shocks, and local fluctuations) in non-stationary signals (signals that change over time); it can separate effective features from noise in signals through multi-scale decomposition (such as filtering out interference and extracting key fluctuation information); it is applicable to one-dimensional signals (such as time series) and can also be extended to the decomposition and feature extraction of two-dimensional signals (such as images).

[0026] (1) Environmental meteorological data (such as wind speed, wind direction, temperature, rainfall, ice thickness, etc.) are essentially non-stationary one-dimensional signals that change over time; (2) Measurement data of transmission lines (such as line current, voltage, power, tower vibration acceleration, etc.) are electrical signals or physical vibration signals, which have obvious non-stationarity. The above (1)-(2) are within the scope of wavelet transform processing. Therefore, in the embodiments of this application, (1)-(2) are preprocessed in 102 to obtain the first data, that is, the time series data from different data sources are converted into time series structure data with a unified time scale.

[0027] Accordingly, a pre-set wavelet transform algorithm is used to process the first data to extract structured time-series features associated with the analysis of potential transmission line hazards, such as, but not limited to, current, voltage, conductor temperature, wind speed, wind direction, and rainfall. Multi-scale features extracted through wavelet transform, such as energy concentration and frequency change rate, are used to reflect risk signs such as short-term overload, icing, wind deflection, and lightning strikes. It is evident that wavelet transform focuses on the "dynamic change patterns" of time-series data, supporting the prediction of time-correlated risks such as icing and lightning strikes.

[0028] 104. The second data is processed using a pre-set feature extraction model to extract unstructured image features and text features associated with the analysis of potential hazards in transmission lines. The unstructured image features correspond to the second feature data, and the text features correspond to the third feature data. The pre-set feature extraction model is pre-trained using a convolutional neural network.

[0029] Specifically, in conjunction with the requirement of using a "feature extraction model trained based on a convolutional neural network", preprocessing is performed on the multi-source heterogeneous data (such as environmental meteorological data, measurement data, equipment ledger data, historical fault data, remote signaling data, and inspection data) collected as listed in example 101. The embodiments of this application are illustrated in the following example 2: Example 2: Convolutional neural networks are good at processing grid-structured data (such as the pixel grid of a two-dimensional image and the word vector matrix after text conversion). They automatically extract local features from the data (such as the edges and textures of an image, and the phrases and semantic combinations of a text) through convolution operations.

[0030] (1) The drone aerial images, manually taken equipment photos, and infrared thermal images in the inspection data are typical unstructured two-dimensional image data, which are fully compatible with the processing scope of convolutional neural networks; (2) The manually recorded text in the inspection data (such as "Three rust spots appeared on the base of tower A, with a diameter of about 5cm" and "The distance between conductor B and the tree is less than 1 meter") are unstructured text data, and features can be extracted through text convolutional neural networks. The above (1)-(2) are compatible with the processing scope of convolutional neural networks. Therefore, in the embodiment of this application, (1)-(2) are preprocessed in 102 to obtain the second data, which is the data in which images and texts from unstructured data from different data sources are pre-annotated.

[0031] Accordingly, a pre-built feature extraction model trained by a convolutional neural network is used to process the second data to extract unstructured temporal features associated with the analysis of potential hazards in transmission lines, such as, but not limited to, crack texture, insulator failure probability, and conductor posture changes. These features are standardized and encoded into discrete state variables such as "damage level," "icing level," and "type of suspended object." The extracted text features, such as, but not limited to, "tower base cracking," "abnormal noise," and "visible deformation," are converted into Boolean or rank variables to represent potential hazards perceived by humans. It is evident that the convolutional neural network (including text processing) focuses on the "spatial / semantic features" of unstructured data, supporting the identification of risks requiring visual / textual descriptions, such as equipment aging and suspended foreign objects.

[0032] 105. Input the first feature data, the second feature data, and the third feature data into the pre-set hidden danger prediction model for processing, output the existence probability corresponding to different pre-set hidden dangers, and obtain the prediction results of the hidden dangers of the transmission line. The pre-set hidden danger prediction model is pre-trained using a Bayesian network.

[0033] It should be noted that, in order to distinguish the different types of features extracted in this application embodiment, the words "first", "second" and "third" are used for identification. The structured time-series features associated with the analysis of hidden dangers in transmission lines are used as the first feature data, the unstructured image features are used as the second feature data, and the text features are used as the third feature data.

[0034] The potential hazards provided in this application embodiment may include, but are not limited to, the following: (1) Icing risk: refers to the hidden dangers caused by icing on transmission lines (such as conductor overload, excessive stress on towers, etc.). (2) Lightning strike risk: refers to the faults caused by lightning strikes on the line (such as insulator breakdown, conductor burnout, etc.). (3) Foreign object hanging risk: refers to the hidden dangers such as short circuits and discharges caused by foreign objects such as kites and plastic sheets hanging on the conductors. (4) Equipment aging risk: refers to the risk of performance degradation or failure of line equipment (such as insulators, tower components, etc.) due to aging. (5) Tower tilting risk caused by geological disasters: specifically refers to the hidden danger of tower tilting caused by geological disasters (such as landslides, subsidence).

[0035] In the prediction of potential hazards in power transmission lines based on big data analysis, Bayesian networks are a powerful probabilistic graphical model that can integrate structured temporal features extracted by wavelet transforms and unstructured image / text features extracted by convolutional neural networks to perform probabilistic inference on pre-defined hazards. Pre-defined hazards may include, but are not limited to, icing risks, lightning strike risks, foreign object hanging risks, equipment aging risks, and tower tilting risks caused by geological disasters.

[0036] As shown in sections 101-105 above, the assessment results based on the power grid's operational status, combined with real-world factors such as equipment condition, historical data, and environmental factors, can generate targeted operation and maintenance (O&M) recommendations. These recommendations cover fault prevention measures, equipment maintenance plans, and resource allocation schemes, helping O&M personnel take proactive action, optimize O&M strategies, improve power grid reliability and operational efficiency, and ensure the safe and stable operation of the power grid. This series of intelligent processes achieves efficient transformation from data to decision-making, providing comprehensive business support for power grid O&M and promoting the intelligent upgrade of power grid O&M.

[0037] In some modified embodiments, preprocessing is performed on multi-source heterogeneous data to obtain first data. The embodiments of this application provide the following detailed implementation steps: A1. Identify the first data source with time-series data from the multi-source heterogeneous data.

[0038] As explained in Example 1 of 103, the purpose of obtaining the first data is to perform feature extraction operations using the wavelet transform algorithm. For this purpose, and in conjunction with the data processing scope of wavelet transform, this step is to determine the data source with time-series data from the multi-source heterogeneous data, such as (1) environmental meteorological data and (2) transmission line measurement data shown in Example 1 of 103.

[0039] A2. In time series data collected from different primary data sources, timestamps from different primary data sources that meet the preset range of numerical differences are mapped to the same time base through time alignment to complete the time alignment process.

[0040] This application employs time alignment to eliminate "minor misalignments" in timestamps. Timestamps from different data sources may have slight differences (e.g., environmental meteorological data is "08:00:00", while measurement data is "08:00:05"). In such cases, time alignment is necessary to map them to the same time base.

[0041] This application embodiment can, but is not limited to, use the timestamp of a certain data source as a reference to match records from other data sources to the nearest reference time point (such as aligning "08:00:05" to "08:00:00"), thereby ensuring that different types of data at the same "actual moment" (such as wind speed and current at that moment) can be associated, avoiding "data mismatch" caused by a difference of a few seconds in timestamps.

[0042] A3. After completing the time alignment process, adjust the acquisition frequency of different first data sources to the same target frequency to achieve resampling.

[0043] Even with timestamp alignment, different data sources may still have different collection frequencies (e.g., environmental meteorological data every 10 minutes, and measurement data every 5 minutes). Resampling aims to standardize the data to a preset time interval (called the "target frequency"), and methods include: Downsampling: High-frequency data (e.g., 1 data point every 5 minutes) are merged into low-frequency data (e.g., 1 data point every 10 minutes). Common aggregation methods include mean and maximum value (e.g., taking the mean of every two 5-minute measurements as 10-minute data).

[0044] Upsampling: Low-frequency data (e.g., 1 record every 10 minutes) is interpolated to high-frequency data (e.g., 1 record every 5 minutes). Common methods include linear interpolation and filling in the preceding and following values ​​(e.g., using two adjacent 10-minute environmental meteorological data to estimate the value of the middle 5 minutes).

[0045] After resampling, data from all data sources will be arranged into equally spaced time sequences according to the "target frequency" (e.g., one record every 10 minutes, with timestamps of 08:00:00, 08:10:00, 08:20:00, etc.).

[0046] A4. After time alignment and resampling, the data corresponding to different first data sources are processed according to the preset time window to obtain multiple target time series data, which are used as the first data.

[0047] After alignment and resampling, the data becomes a structured sequence with equal time intervals (e.g., one record every 10 minutes, containing fields such as wind speed and current). However, time series processing requires "input containing historical information" (e.g., data from the current moment plus the previous four moments), rather than a single isolated point in time. In this case, the sliding time window serves the purpose of: Set a fixed-length window (e.g., window size = 5), slide the window along the time axis, and pack N consecutive time points into a "sample".

[0048] For example, when the window is slid to "08:40:00", the sample contains data from 5 time points: "08:00, 08:10, 08:20, 08:30, and 08:40", which are used to predict the state at subsequent times (such as the current at 08:50).

[0049] As shown in A1-A4 above, the embodiments of this application employ temporal reconstruction and alignment to process environmental meteorological data and measurement data with temporal sequence characteristics, thereby creating a unified temporal data structure and using time window sliding to capture temporal dependencies, thus obtaining high-quality target temporal data as the first data for subsequent processing by wavelet transform algorithm.

[0050] In some modified embodiments, preprocessing is performed on multi-source heterogeneous data to obtain second data. The embodiments of this application provide the following detailed implementation steps: B1. Identify a second data source with unstructured data from multi-source heterogeneous data, including image data and text data.

[0051] As explained in Example 2 of 104, the purpose of obtaining the second data is to perform feature extraction operations using a feature extraction model trained by a convolutional neural network. For this purpose, and in conjunction with the data processing scope of convolutional neural networks, this step is to determine the data source with unstructured data from multi-source heterogeneous data, such as the inspection data (1) image and (2) text shown in Example 2 of 104.

[0052] B2. Mark the potential hazard areas associated with power transmission lines in the image data.

[0053] This step can be done manually or semi-automatically. For semi-automatic annotation, a target detection algorithm (such as YOLO) can be used to initially identify potential hazard areas in the image (such as "suspected insulator damage"), and then manual review and correction can be performed to reduce the workload of purely manual annotation. After annotation, the image data can be converted into a structured form of "image + hazard label" (such as "insulator image + damage = 1 / normal = 0").

[0054] B3. Mark keywords related to power transmission lines in the text data.

[0055] This step involves identifying terms related to potential hazards in the text (such as "loose tower base", "ice accumulation on conductors", "lightning burns", and "loose bolts") and labeling them (e.g., "loose tower base" belongs to "risk of tower tilting caused by geological disasters", and "lightning burns" belongs to "lightning risk"). Key information in the text (time, tower number, hazard type, and location) is then extracted into tables or labels.

[0056] B4. Use the unstructured data from the second data source that has been annotated as the second data.

[0057] In this embodiment of the application, the significance of pre-annotation is to transform unstructured image and text data into "hazard features" that the model can understand (such as damaged area features in images and hazard keyword vectors in text), so that it can be integrated with structured data such as environmental meteorology and measurement, providing multi-dimensional support for predicting risks such as icing, lightning strikes, and geological disasters (such as combining icing annotations of UAV images with temperature and humidity data of environmental meteorology, which can more accurately predict icing risks).

[0058] In some modified embodiments, a preset wavelet transform algorithm is used to process the first data to extract structured time-series features associated with the analysis of potential transmission line hazards, including: C1. Based on the first data corresponding to the time sequence signal, the time sequence signal is denoised and optimized, and normalized and mapped to the specified interval.

[0059] This step is signal preprocessing, which performs noise reduction optimization on the original signal after time alignment (such as current data sampled once every 10 minutes) (supplementing (1) residual noise processing after outlier removal, such as using soft thresholding to filter high-frequency noise), normalization (mapping signals of different dimensions (such as wind speed m / s, temperature ℃) to the [0,1] interval to avoid numerical differences affecting the decomposition effect).

[0060] C2. By selecting a wavelet basis suitable for the signal characteristics of the transmission line, the decomposition scale is set, including at least: a first scale and a second scale. The first scale is used to capture high-frequency details, and the second scale is used to extract low-frequency trends. Among them, the high-frequency details include at least the current spike during lightning strikes, and the low-frequency trends include at least the slow decrease in conductor temperature caused by icing.

[0061] This step involves wavelet basis and scale selection. Select a wavelet basis suitable for the characteristics of the transmission line signal (such as the db4 wavelet, which is sensitive to abrupt signals) and set the decomposition scale (such as 3-5 layers): low scale (such as 1-2 layers) captures high-frequency details (such as current spikes during lightning strikes), and high scale (such as 4-5 layers) extracts low-frequency trends (such as the slow temperature drop trend of conductors caused by icing).

[0062] C3. Calculate the features of each decomposition result obtained according to the decomposition scale to realize multi-scale feature extraction. The multi-scale features include at least energy features, mutation features and frequency changes.

[0063] This step calculates the characteristics of each layer's decomposition results: energy characteristics (the proportion of energy at a certain scale, reflecting the signal strength in that frequency band, such as the energy of conductor vibration concentrated in the low frequency band when icing occurs); abrupt change characteristics (the difference in coefficients between adjacent moments, such as the abrupt change in current signal during a lightning strike); and frequency change rate (frequency drift at different scales, such as the change in conductor sway frequency during wind deflection).

[0064] C4. Redundant features are removed, and multi-scale features are fused with the original time-series features corresponding to the time-series signal to form structured time-series features.

[0065] This step involves feature selection and fusion, eliminating redundant features (such as repetition frequency features with correlation > 0.9), and fusing multi-scale features with original time-series features (such as current peaks) to form structured time-series features.

[0066] In some modified embodiments, a preset feature extraction model is used to process the second data to extract unstructured image features and text features associated with the analysis of potential hazards in transmission lines. The embodiments of this application provide the following detailed implementation steps: (1) Image feature extraction based on convolutional neural networks, including the following: (1.1) Image enhancement: perform data enhancement on the pre-labeled image (such as rotation, scaling, brightness adjustment) to solve the problem of sample imbalance (such as few damaged samples); based on the bounding box, crop out the region of interest (ROI, such as the insulator region) to reduce background interference.

[0067] (1.2) Convolutional layer feature extraction: Local features are extracted by sliding through multiple convolutional kernels (such as 3×3, 5×5): shallow convolution (layers 1-2) captures edges and textures (such as the line features of cracks); deep convolution (layers 3-5) fuses local features to generate higher-order features (such as the overall shape of "insulator damage").

[0068] (1.3) Feature transformation and quantization: The high-dimensional feature vector (e.g., 1024-dimensional) output by the convolutional layer is compressed through the fully connected layer and classified in combination with the labels (e.g., “damaged”, “normal”) in (5), and transformed into discrete variables such as “damage level (1-5)” and “ice thickness (mm)”, which are convenient for fusion with time-series features.

[0069] (2) Text feature extraction based on (TextCNN / BERT), including the following: (2.1) Text preprocessing: segment the pre-annotated text into words (e.g., split “loose tower foundation with risk of collapse” into “tower foundation / loose / existing / collapse / risk”), construct a professional dictionary based on the keywords in (5) (e.g., “tower tilt”, “wire wear”), and filter out irrelevant words (e.g., “inspection time”).

[0070] (2.2) Semantic feature extraction: TextCNN: captures local semantics (such as the combination of "tower base + loose") through convolutional kernels and outputs feature vectors representing "tower stability risk"; BERT: understands ambiguity in context (such as "abnormal" means fault in "current abnormal" and environmental meteorological risk in "weather abnormal") and outputs more accurate semantic labels.

[0071] (2.3) Feature quantization: convert semantic features into Boolean variables (e.g., "there is a loose base = 1") or grade variables (e.g., "risk level: high / medium / low"), and unify the format with image features and time series features.

[0072] In some modified embodiments, the first feature data, the second feature data, and the third feature data are input into a pre-set hazard prediction model for processing, and the existence probability corresponding to different pre-set hazards is output. The embodiments of this application provide the following detailed implementation steps: D1 abstracts the first feature data, the second feature data, and the third feature data into state variable nodes in a Bayesian network.

[0073] D2 constructs causal paths based on the power grid operation logic and the hidden danger development mechanism. The causal paths are used to connect the causal logic between state variable nodes.

[0074] Based on the evidence gleaned from the first, second, and third feature data, and combined with the causal path, D3 outputs the probability of existence for different pre-existing hidden dangers.

[0075] Bayesian networks are graphical models based on probabilistic reasoning. Their core advantage lies in quantifying causal relationships between variables and handling uncertainty, making them highly suitable for the scenario of "multiple factors intertwined and dynamic risks" in power transmission line hazard prediction. Combining the aforementioned three characteristics, its specific implementation logic can be unfolded in four levels: "node design—causal modeling—reasoning mechanism—decision implementation." The following is a detailed analysis: First, state variable nodes: abstraction and mapping of multi-source features. The core of a Bayesian network is "nodes" and "edges," where nodes represent random variables, corresponding to the three types of features extracted above; edges represent the dependencies between variables, and the following mapping logic is provided as an example: Table 1 Transforming high-dimensional, heterogeneous features into network-recognizable "state variables" lays the foundation for subsequent causal modeling. For example, "icing level = 2" can be used as a node to directly link downstream nodes such as "abnormal conductor tension," thus mapping features to risks.

[0076] Second, causal path construction: reconstructing the physical and logical links of hazard development. The "edges" of a Bayesian network are used to characterize the causal relationships between nodes. Their construction is based on power grid operation patterns, hazard evolution mechanisms, and expert experience. An example analysis is as follows: (1) "Rainfall + sudden drop in temperature → increased risk of icing → abnormal conductor tension"; Causal logic: Rainfall (structured temporal characteristics) and low temperature (structured temporal characteristics) are necessary conditions for icing formation (physical laws); increased icing thickness leads to increased conductor tension (mechanical principles), which can be verified through measurement data (such as tension sensors).

[0077] Feature association: the antecedent nodes come from structured time-series features, and the result nodes are associated with measurement data anomalies, forming a link of "environmental factors → intermediate risks → equipment status".

[0078] (2) "Equipment aging → insulator damage → increased risk of discharge"; Causal logic: Equipment aging ("service life" in the ledger data + "aging texture" in the image features) will lead to a decrease in the mechanical strength of the insulator (the law of equipment degradation); damaged insulators may cause partial discharge (electrical principle), which can be corroborated by remote signaling data (such as discharge alarm).

[0079] Feature association: Integrating ledger data (structured), image features (unstructured), and remote information data to form a link of "equipment status → component failure → risk event".

[0080] (3) "Tower foundation settlement → tower tilting → uneven conductor tension → structural hidden dangers"; Causal logic: Tower base settlement (text feature "tower base loosening" + image feature "tower verticality deviation") will lead to tower tilting (geometric relationship); tilting causes the conductor tension distribution to become unbalanced (mechanical effect), which will eventually lead to structural failure (such as tower collapse).

[0081] Feature association: Integrate text descriptions, image visual features, and measurement data (such as wire tension sensors) to form a link of "basic problems → structural deformation → safety risks".

[0082] By connecting dispersed multi-source features through causal paths, the network can not only "predict risks" but also "explain the causes of risks," providing a traceable basis for operation and maintenance strategies.

[0083] Third, the predictive inference mechanism: dynamic probability updates based on evidence. The essence of the Bayesian network's inference process is "calculating the posterior probability of the target node based on evidence," and its specific implementation steps and advantages are as follows: (1) Selection of evidence nodes. Evidence refers to "known observation data," which comes from three types of features after preprocessing, such as: Structured temporal characteristics: "Temperature will continue to drop in the next 6 hours" (environmental meteorological trend), "rapid changes in conductor tension" (measurement data); Unstructured image features: "Ice level = 2" (discrete state output by convolutional neural network); Textual feature: "High probability of insulator damage" (semantic tag).

[0084] (2) The posterior probability is calculated based on Bayes' theorem (P(A|B) = P(B|A)P(A) / P(B)), combined with the prior probability of the network (set by historical data or expert experience), to calculate the probability of the risk event. For example: "P(line icing instability|slight icing has occurred + wind speed > 10m / s + temperature < 0℃) = 72%": This means that when "slight icing" (image feature), "strong wind" (structured time series feature), and "low temperature" (structured time series feature) occur simultaneously, the probability of the line becoming unstable due to icing is 72%.

[0085] Calculation logic: The prior probability comes from historical icing accident data, the conditional probability reflects the degree of impact of "icing + strong wind + low temperature" on instability, and finally the posterior probability is obtained through network topology propagation.

[0086] (3) Dynamic update mechanism: When new evidence is input (such as real-time measurement data, new inspection images), the network will automatically update the probability distribution of all relevant nodes. For example: initial evidence: "Icing level = 1" + "Temperature = -2℃", calculate "Icing instability probability = 30%"; New evidence: "Wind speed suddenly increased to 15 m / s" (real-time environmental meteorological data), and the network recalculated "probability of icing instability = 65%", realizing dynamic risk assessment.

[0087] Fourth, decision support and closed-loop mechanism: From risk prediction to operation and maintenance actions, the output of Bayesian networks ultimately serves intelligent operation and maintenance decision-making, specifically reflected in: (1) Risk threshold alarm: The system presets risk thresholds (such as “probability of ice instability > 70%” to trigger an emergency alarm). When the reasoning result exceeds the threshold, it automatically pushes operation and maintenance suggestions (such as “melt ice immediately” or “suspend load transmission”).

[0088] (2) Targeted operation and maintenance scheduling, locating the source of risk based on causal paths, for example: If the main evidence for "insulation failure probability = 83%" is "equipment service life exceeding 10 years", then priority should be given to scheduling inspections of old lines; if "high risk of tower tilting" stems from "tower foundation settlement", then foundation reinforcement projects should be prioritized.

[0089] (3) Closed-loop optimization: The results of operation and maintenance actions (such as "the icing level drops to 0 after melting ice") will feed back into the network, update the node probability parameters, and form a closed loop of "data collection → preprocessing → feature extraction → inference prediction → operation and maintenance actions → data feedback" to continuously improve the prediction accuracy.

[0090] The first four points above demonstrate that Bayesian networks, through the entire process of "node abstraction - causal modeling - probabilistic reasoning - decision support," perfectly meet the needs of predicting potential hazards in power transmission lines.

[0091] Furthermore, as a response to the above Figure 1 To implement the method shown, this application provides a transmission line hazard prediction device based on big data analysis. This device embodiment corresponds to the aforementioned method embodiment. For ease of reading, this device embodiment will not repeat the details of the aforementioned method embodiment, but it should be understood that the device in this embodiment can implement all the contents of the aforementioned method embodiment. This device is applied to provide an intelligent solution for predicting transmission line hazards, specifically as follows... Figure 2 As shown, the device includes: The acquisition unit 21 is used to acquire multi-source heterogeneous data related to the operation of transmission lines. The data dimensions included in the multi-source heterogeneous data include at least environmental meteorological data, measurement data, equipment ledger data, historical fault data, remote signaling data, and inspection data. The first processing unit 22 is used to perform preprocessing on the multi-source heterogeneous data to obtain first data and second data. The first data is data that converts time-series data from different data sources into time-series structure data with a unified time scale, and the second data is data that pre-labels images and text in unstructured data from different data sources. The second processing unit 23 is used to process the first data using a preset wavelet transform algorithm to extract structured time-series features associated with the analysis of potential hazards in transmission lines, as the first feature data. The third processing unit 24 is used to process the second data using a preset feature extraction model to extract unstructured image features and text features associated with the analysis of potential hazards in transmission lines. The unstructured image features correspond to the second feature data, and the text features correspond to the third feature data. The preset feature extraction model is pre-trained using a convolutional neural network. The fourth processing unit 25 is used to input the first feature data, the second feature data and the third feature data into the preset hidden danger prediction model for processing, output the existence probability corresponding to different preset hidden dangers, and obtain the prediction result of the hidden danger of the transmission line. The preset hidden danger prediction model is pre-trained using a Bayesian network.

[0092] Furthermore, such as Figure 3 As shown, the first processing unit 22 includes: The first determining module 221 is used to determine a first data source with time-series data from the multi-source heterogeneous data; The first processing module 222 is used to map the timestamps of different first data sources that meet the preset range of numerical differences to the same time base through time alignment in time-series data collected from different first data sources, thereby completing time alignment processing. The second processing module 223 is used to adjust the acquisition frequency of different first data sources to the same target frequency after completing the time alignment processing, so as to achieve resampling. The third processing module 224 is used to process the data corresponding to different first data sources after time alignment and resampling according to a preset time window to obtain multiple target time series data as the first data.

[0093] Furthermore, such as Figure 3 As shown, the first processing unit 22 further includes: The second determining module 225 is used to determine a second data source with unstructured data from the multi-source heterogeneous data, wherein the unstructured data includes image data and text data; The annotation module 226 is used to annotate potential hazard areas associated with power transmission lines in the image data; The annotation module 226 is also used to annotate keywords associated with transmission lines in the text data; The third determining module 227 is used to take the unstructured data after annotation processing in the second data source as the second data.

[0094] Furthermore, the second processing unit 23 is specifically used for: Based on the first data corresponding to the time series signal, the time series signal is denoised and optimized, and normalized and mapped to a specified interval. By selecting a wavelet basis suitable for the signal characteristics of the transmission line, a decomposition scale is set, including at least a first scale and a second scale. The first scale is used to capture high-frequency details, and the second scale is used to extract low-frequency trends. The high-frequency details include at least the current spikes during lightning strikes, and the low-frequency trends include at least the slow temperature decrease trend of the conductor caused by icing. Based on the decomposition scale, the features of each layer decomposition result are calculated to realize multi-scale feature extraction. The multi-scale features include at least energy features, mutation features, and frequency changes. Redundant features are removed, and the multi-scale features are fused with the original time-series features corresponding to the time-series signal to form structured time-series features.

[0095] Furthermore, the fourth processing unit 25 is specifically used for: The first feature data, the second feature data, and the third feature data are abstracted as state variable nodes in a Bayesian network; A causal path is constructed based on the power grid operation logic and the hidden danger development mechanism. The causal path is used to connect the causal logic existing between the state variable nodes. Based on the evidence extracted from the first feature data, the second feature data, and the third feature data, and combined with the causal path, the probability of existence corresponding to different pre-set hidden dangers is output.

[0096] Furthermore, the pre-existing hazards include at least: icing risk, lightning strike risk, foreign object hanging risk, equipment aging risk, and tower tilting risk caused by geological disasters.

[0097] As described above, the transmission line hidden danger prediction device based on big data analysis includes a processor and a memory. The above-mentioned acquisition unit, first processing unit, second processing unit, third processing unit and fourth processing unit are all stored in the memory as program units. The processor executes the above-mentioned program units stored in the memory to realize the corresponding functions.

[0098] The processor contains a kernel, which retrieves the corresponding program units from memory. One or more kernels can be configured. By adjusting kernel parameters, multi-source heterogeneous data is used, and wavelet transform and convolutional neural network collaboration are employed to extract features related to potential transmission line hazards. Then, Bayesian networks are used to perform inference and prediction, thus forming a complete chain of "feature extraction → feature fusion → probabilistic inference." This approach addresses diverse hazard types, fully utilizes big data, and provides an intelligent solution for predicting potential transmission line hazards.

[0099] This application provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the above-described method for predicting potential hazards in power transmission lines based on big data analysis.

[0100] This application provides an electronic device, which includes at least one processor, at least one memory and a bus connected to the processor; wherein the processor and the memory communicate with each other through the bus; the processor is used to call program instructions in the memory to execute the transmission line hidden danger prediction method based on big data analysis as described above.

[0101] This application also provides a computer program product that, when executed on a data processing device, is suitable for performing the initialization steps of a method for predicting potential hazards in transmission lines based on big data analysis.

[0102] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0103] In a typical configuration, the device includes one or more processors (CPUs), memory, and a bus. The device may also include input / output interfaces, network interfaces, etc.

[0104] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, like read-only memory (ROM) or flash RAM, and memory includes at least one memory chip. Memory is an example of computer-readable media.

[0105] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0106] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0107] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied 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.

[0108] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for predicting potential hazards in power transmission lines based on big data analysis, characterized in that, The method includes: Collect multi-source heterogeneous data related to the operation of transmission lines. The data dimensions included in the multi-source heterogeneous data include at least environmental meteorological data, measurement data, equipment ledger data, historical fault data, remote signaling data, and inspection data. Preprocessing is performed on the multi-source heterogeneous data to obtain first data and second data. The first data is data that converts time-series data from different data sources into time-series structure with a unified time scale. The second data is data that pre-labels images and text in unstructured data from different data sources. The first data is processed using a preset wavelet transform algorithm to extract structured time-series features associated with the analysis of potential hazards in transmission lines, which are then used as the first feature data. The second data is processed using a pre-set feature extraction model to extract unstructured image features and text features associated with the analysis of potential hazards in transmission lines. The unstructured image features correspond to the second feature data, and the text features correspond to the third feature data. The pre-set feature extraction model is pre-trained using a convolutional neural network. The first feature data, the second feature data, and the third feature data are input into the pre-set hidden danger prediction model for processing, and the existence probability corresponding to different pre-set hidden dangers is output to obtain the prediction result of the hidden danger of the transmission line. The pre-set hidden danger prediction model is pre-trained using a Bayesian network.

2. The method according to claim 1, characterized in that, Preprocessing is performed on the multi-source heterogeneous data to obtain first data, including: From the multi-source heterogeneous data, determine the first data source with time-series data; In the time series data collected from different first data sources, for timestamps of different first data sources that meet the preset range of numerical differences, time alignment is performed and mapped to the same time base to complete the time alignment process; After completing the time alignment process, the sampling frequency of different first data sources is adjusted to the same target frequency to achieve resampling; After time alignment and resampling, the data corresponding to different first data sources are processed according to a preset time window to obtain multiple target time series data, which are used as the first data.

3. The method according to claim 1, characterized in that, Preprocessing is performed on the multi-source heterogeneous data to obtain second data, including: From the multi-source heterogeneous data, a second data source with unstructured data is determined, the unstructured data including image data and text data; Mark the potential hazard areas associated with the power transmission lines in the image data; The text data is annotated with keywords associated with power transmission lines; The unstructured data after annotation processing in the second data source is used as the second data.

4. The method according to any one of claims 1 to 3, characterized in that, The first data is processed using a pre-set wavelet transform algorithm to extract structured time-series features associated with the analysis of potential transmission line hazards, including: Based on the first data corresponding to the time series signal, the time series signal is denoised and optimized, and normalized and mapped to a specified interval. By selecting a wavelet basis suitable for the signal characteristics of the transmission line, a decomposition scale is set, including at least a first scale and a second scale. The first scale is used to capture high-frequency details, and the second scale is used to extract low-frequency trends. The high-frequency details include at least the current spikes during lightning strikes, and the low-frequency trends include at least the slow temperature decrease trend of the conductor caused by icing. Based on the decomposition scale, the features of each layer decomposition result are calculated to realize multi-scale feature extraction. The multi-scale features include at least energy features, mutation features, and frequency changes. Redundant features are removed, and the multi-scale features are fused with the original time-series features corresponding to the time-series signal to form structured time-series features.

5. The method according to any one of claims 1 to 3, characterized in that, The step of inputting the first feature data, the second feature data, and the third feature data into a preset hazard prediction model for processing, and outputting the existence probability corresponding to different preset hazards, includes: The first feature data, the second feature data, and the third feature data are abstracted as state variable nodes in a Bayesian network; A causal path is constructed based on the power grid operation logic and the hidden danger development mechanism. The causal path is used to connect the causal logic existing between the state variable nodes. Based on the evidence extracted from the first feature data, the second feature data, and the third feature data, and combined with the causal path, the probability of existence corresponding to different pre-set hidden dangers is output.

6. The method according to any one of claims 1 to 3, characterized in that, The pre-existing hazards include at least: icing risk, lightning strike risk, foreign object hanging risk, equipment aging risk, and tower tilting risk caused by geological disasters.

7. A transmission line hazard prediction device based on big data analysis, characterized in that, The device includes: The acquisition unit is used to acquire multi-source heterogeneous data related to the operation of transmission lines. The data dimensions included in the multi-source heterogeneous data include at least environmental meteorological data, measurement data, equipment ledger data, historical fault data, remote signaling data, and inspection data. The first processing unit is used to perform preprocessing on the multi-source heterogeneous data to obtain first data and second data. The first data is data that converts time-series data from different data sources into time-series structure data with a unified time scale, and the second data is data that pre-labels images and text in unstructured data from different data sources. The second processing unit is used to process the first data using a preset wavelet transform algorithm to extract structured time-series features associated with the analysis of potential hazards in transmission lines, as the first feature data. The third processing unit is used to process the second data using a preset feature extraction model to extract unstructured image features and text features associated with the analysis of potential hazards in transmission lines. The unstructured image features correspond to the second feature data, and the text features correspond to the third feature data. The preset feature extraction model is pre-trained using a convolutional neural network. The fourth processing unit is used to input the first feature data, the second feature data and the third feature data into the preset hidden danger prediction model for processing, output the existence probability corresponding to different preset hidden dangers, and obtain the prediction result of the hidden danger of the transmission line. The preset hidden danger prediction model is pre-trained using a Bayesian network.

8. The apparatus according to claim 7, characterized in that, The first processing unit includes: The first determining module is used to determine a first data source with time-series data from the multi-source heterogeneous data; The first processing module is used to map the timestamps of different first data sources that meet the preset range of numerical differences to the same time base through time alignment in time-series data collected from different first data sources, thereby completing the time alignment process. The second processing module is used to adjust the acquisition frequency of different first data sources to the same target frequency after completing the time alignment processing, so as to achieve resampling. The third processing module is used to process the data corresponding to different first data sources after time alignment and resampling according to a preset time window to obtain multiple target time series data as the first data.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the method for predicting potential hazards in transmission lines based on big data analysis as described in any one of claims 1-6.

10. An electronic device, characterized in that, The device includes at least one processor, and at least one memory and bus connected to the processor; The processor and the memory communicate with each other via the bus. The processor is used to call program instructions in the memory to execute the transmission line hidden danger prediction method based on big data analysis as described in any one of claims 1-6.

Images