Power transmission line fault cause identification method based on fault recording and geographic meteorology

Abstract

The application discloses a power transmission line fault cause identification method based on fault recording and geographic meteorology, and the method comprises the following steps: acquiring power transmission line fault data, including fault recording data, tower geographic location data and all meteorological data of the area crossed by the line; processing and analyzing the fault recording data; calculating fault ranging results according to the fault recording data, determining the fault occurrence area in combination with the tower geographic location data, and acquiring the meteorological data of the fault occurrence area in combination with all meteorological data of the area crossed by the line; setting a fault cause label by calibrating the fault cause of the fault data, and predicting the fault cause label of the fault data which has not been calibrated; and establishing an integrated learner model, and completing model training and parameter optimization, wherein the model takes meteorological features, waveform morphological features and electrical quantity features of the fault data as input, and takes the fault cause label as output. The application can realize more intelligent power transmission line fault cause diagnosis.

Description

Technical Field

[0001] This invention belongs to the field of power system fault cause identification technology, and more specifically, relates to a method for identifying the cause of transmission line faults based on fault recording and geographical meteorology. Background Technology

[0002] Transmission lines, as a crucial component of the power system, play a vital role in ensuring power supply through their safe and stable operation. However, in actual operation, transmission line faults can occur due to natural environmental factors, equipment aging, and other reasons, thus affecting the normal operation of the power system. Currently, the handling of transmission line faults mainly focuses on the emergency response phase after the fault occurs, that is, analyzing the fault waveform data recorded by fault recorders to obtain information such as the location and type of the fault. However, existing technologies often cannot directly provide effective fault cause analysis. In practice, to determine the specific cause of a transmission line fault and take corresponding measures, on-site inspections by maintenance personnel are usually required. This process is often time-consuming and may affect subsequent emergency response decisions, such as whether to implement forced power restoration and how to arrange maintenance personnel for on-site inspection and repair work. In view of the above situation, it is particularly necessary to develop a method that can quickly and intelligently diagnose the causes of transmission line faults. Summary of the Invention

[0003] To address the aforementioned deficiencies or improvement needs of existing technologies, this invention proposes a method for identifying the causes of transmission line faults based on fault recording and geographic meteorological data. After a fault occurs in a transmission line, fault-related fault recordings and geographic meteorological data are collected and feature engineering is established. The model is then trained and predicted using ensemble learning, ultimately achieving the identification of the cause of the line fault.

[0004] To achieve the above objectives, according to a first aspect of the present invention, a method for identifying the causes of transmission line faults based on fault recording and geographical meteorological data is provided, the method comprising: Acquire transmission line fault data, including fault waveform data, tower geographical location data, and all meteorological data for the areas traversed by the line; Process and analyze fault waveform data; The fault location is calculated based on the fault recording data, the fault location is determined by combining the tower geographical location data, and the meteorological data of the fault location is obtained by combining all the meteorological data of the area traversed by the line. Extracting meteorological characteristics from meteorological data; Based on a one-dimensional convolutional neural network, waveform morphology features of fault recording data are extracted. Based on electrical quantity analysis, electrical quantity characteristics of fault recording data are extracted; Fault data is labeled with fault causes to set fault cause labels, and fault cause labels are predicted for fault data without labeled fault causes through an adaptive semi-supervised machine learning method. An integrated learner model was established and its training and parameter tuning were completed. The model takes the meteorological characteristics, waveform morphology characteristics and electrical quantity characteristics of the fault data as input and the fault cause label as output to identify the fault cause of the transmission line.

[0005] The above solution involves processing and analyzing fault recording data, including: After acquiring fault waveform data of transmission lines, if the waveform data model is complete, the parameters of the line and the corresponding analog and digital signals are parsed and extracted based on the power system transient data format standard and the model file. If the waveform data model is missing, fuzzy matching technology is used to find the corresponding data channel by the equipment name, and then the acquired data is parsed and extracted.

[0006] In the above scheme, the fault location result is calculated based on the fault recording data, the fault location area is determined by combining the tower geographical location data, and the meteorological data of the fault location area is obtained by combining all meteorological data of the area traversed by the line, including: Assume the spans between the poles of the faulty line are, starting from the measurement station, sequentially as follows: The fault location result is x Take the condition that is met pole tower i and i+ Latitude and longitude information of 1 and ; Meteorological data includes real-time meteorological data and meteorological bulletins. Real-time meteorological data is obtained using distance coordinates. Recent ground-based observation data, including temperature, relative humidity, wind speed, wind direction, and precipitation; meteorological bulletins are taken from the towers. i The weather conditions for the city or district in the most recent time period, including text-based weather descriptions.

[0007] The above scheme extracts meteorological characteristics from meteorological data, including: For real-time meteorological data, we directly take the numerical values ​​and standardize them to obtain numerical meteorological characteristics. For text-based weather descriptions in meteorological bulletins, numerical features of the text-based weather descriptions are obtained based on sentiment analysis and then merged with the numerical meteorological characteristics of the actual meteorological data to finally obtain the meteorological features of the meteorological data; among them, the sentiment analysis method includes the use of the sentiment analysis network SnowNLP.

[0008] In the above scheme, based on a one-dimensional convolutional neural network, waveform morphology features of fault recording data are extracted, including: The voltage and current waveforms of the fault phase are extracted from the fault recording data. By selecting the sampled values ​​of several cycles before and after the fault, they are merged to form a one-dimensional array to be analyzed. A one-dimensional convolutional neural network is used to extract features from the waveform data. The one-dimensional convolutional neural network includes an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer. It takes the one-dimensional array as input and the fault cause label as output. After model training, the output result of the second to last layer, i.e., the fully connected layer, is taken as the waveform morphology feature of the fault recording data for each fault recording data.

[0009] In the above scheme, based on the electrical quantity analysis method, the electrical quantity characteristics of the fault recording data are extracted, including: Electrical quantity characteristics include harmonic content, attenuated DC component, transition resistance, and measurement impedance angle. When extracting harmonic content, the 2nd to 25th harmonics of the fault phase voltage and current are calculated, and the harmonic order and content of the largest and second largest harmonics of voltage and current are taken as fault characteristics. When extracting attenuated DC component, the fault phase current is calculated, and the amplitude and time constant of the attenuated DC component of the current after the fault are taken as fault characteristics. When extracting transition resistance, the voltage and current magnitudes at the fault point are calculated using the transmission line equation, and the real part of the impedance value (voltage divided by current) is taken as the transition resistance characteristic. When extracting measurement impedance angle, the ratio of the measured voltage to the measured current at one end is calculated, and the phase angle of the ratio result is taken as the measurement impedance angle characteristic.

[0010] In the above scheme, the fault cause labels include: lightning strike, wildfire, wind deflection, foreign object, bird droppings flashover, other flashover, de-icing jump, insulator skirt bridging, sub-conductor breakage, complete breakage and others; An adaptive semi-supervised machine learning approach is used to predict fault cause labels for fault data with unlabeled fault causes, including: Let B be the number of fault data samples, and μ be the proportion of fault data samples without a defined fault cause. Then the objective function for fault data without a defined fault cause is: ,in, and They represent the first The output probabilities of fault data with unlabeled fault causes after weak and strong enhancement, respectively. Represents the cross-entropy loss function. To set a threshold, This is an indicator function; it is 1 if the condition is met, and 0 otherwise. use Replace global threshold The model prediction expectation for each cluster is obtained. The final adaptive threshold is obtained by integrating the global threshold and the local threshold. for: ,in CThis refers to the number of clusters, i.e., the number of fault cause labels. Labels indicating the cause of the fault. The global threshold is used; after obtaining the adaptive threshold, the fault cause label is predicted for the fault data samples without a defined fault cause based on the objective function of the fault data without a defined fault cause.

[0011] In the above scheme, the basic learner of the ensemble learner model consists of logistic regression and decision tree, and the basic learner is combined in a sequence ensemble manner; the hyperparameters of the ensemble learner model are optimized by using the swarm optimization algorithm symbiotic search.

[0012] According to a second aspect of the present invention, a computer device is provided, comprising: a processor and a memory, wherein the memory stores a program or instructions executable on the processor, and the program or instructions, when executed by the processor, implement the steps of the transmission line fault cause identification method based on fault recording and geographic meteorology as described in any one of the first aspects.

[0013] According to a third aspect of the present invention, a computer-readable storage medium is provided having a program or instructions stored thereon, wherein when the program or instructions are executed by a processor, the steps of the transmission line fault cause identification method based on fault recording and geographic meteorology as described in any one of the first aspects are implemented.

[0014] In summary, compared with the prior art, the above-described technical solutions conceived by this invention can achieve the following beneficial effects: (1) In the past, the identification of line fault causes generally only relied on meteorological bulletin information. However, in this invention, the fault distance measurement results and the geographical location information of the line towers are used to locate the data of the nearest ground observation station, thereby obtaining more accurate meteorological data about the fault point, such as rainfall, wind speed, temperature, etc.

[0015] (2) In the past, when analyzing the characteristics of meteorological bulletin data, it was only possible to make rigid classifications based on conditions such as sunny, rainy, and snowy. However, the sentiment analysis method used in this invention can abstract and refine the textual descriptions in meteorological textual information, and comprehensively provide more accurate meteorological bulletin data characteristic values.

[0016] (3) In addition to extracting external meteorological features and internal electrical quantity features, waveform curve feature extraction based on one-dimensional CNN can extract typical fault waveform features such as the impulse voltage of lightning strikes and the peak-valley depression features of wildfires, increasing the effective and usable features for cause identification.

[0017] (4) Previous fault cause identification techniques require the labeling of all fault samples, but there are actually a large number of sample data with unknown fault causes, making them unusable. This invention can automatically label samples with unknown causes through an adaptive semi-supervised learning method, which greatly increases the number of usable samples.

[0018] (5) Compared with other traditional machine learning models such as support vector machines, ensemble learning methods are more suitable for situations with complex feature spaces, large amounts of data, and a lot of noise. The feature engineering designed in this invention has a high dimension, and in actual faults, there are situations such as large measurement data errors and sample calibration errors. In such scenarios, the results provided by the ensemble method generally have better accuracy and robustness. Attached Figure Description

[0019] Figure 1 A schematic diagram of the overall process of a method for identifying the cause of transmission line faults based on fault recording and geographic meteorology, provided in an embodiment of the present invention; Figure 2 A partial screenshot of the data management interface provided in an embodiment of the present invention; Figure 3 This is a diagram illustrating the test results for fault cause identification provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the hardware structure of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. All other embodiments obtained by those skilled in the art based on the embodiments provided by this invention without inventive effort are within the scope of protection of this invention.

[0021] Obviously, the accompanying drawings described below are merely some examples or embodiments of the present invention. Those skilled in the art can apply the present invention to other similar scenarios based on these drawings without any inventive effort. Furthermore, it is understood that although the efforts made in this development process may be complex and lengthy, for those skilled in the art related to the content disclosed in this invention, modifications to design, manufacturing, or production based on the technical content disclosed in this invention are merely conventional technical means and should not be construed as insufficient disclosure of the present invention.

[0022] In this invention, the reference to "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described in this invention may be combined with other embodiments without conflict.

[0023] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "a," "an," "an," "the," and similar words used in this invention do not indicate quantity limitation and may indicate singular or plural. The terms "comprising," "including," "having," and any variations thereof used in this invention are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or modules (units) is not limited to the listed steps or units, but may also include steps or units not listed, or may include other steps or units inherent to these processes, methods, products, or devices. The terms "connected," "linked," "coupled," and similar words used in this invention are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "A plurality" in this invention refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships may exist; for example, "A and / or B" can represent: A alone, A and B simultaneously, and B alone. The character " / " generally indicates that the preceding and following objects have an "or" relationship. The terms "first," "second," and "third" used in this invention are merely to distinguish similar objects and do not represent a specific ordering of the objects.

[0024] This invention provides a method, computer equipment, and computer-readable storage medium for identifying the causes of transmission line faults based on fault waveform recordings and geographic meteorological data. The method includes the following steps: collecting data related to the line fault, such as fault waveform recordings, geographic location information, and meteorological data; parsing the fault waveform recording data and extracting sampled values; obtaining meteorological conditions in the fault location area based on the fault location and tower geographic information, including real-time data and bulletins; processing the textual descriptions in the meteorological bulletins using sentiment analysis to convert them into feature values ​​that can be used for analysis; extracting waveform morphology features from the fault waveform recordings using a one-dimensional convolutional neural network; extracting electrical quantity fault features, such as harmonic content, using electrical quantity analysis; calibrating fault samples using fault reports and semi-supervised machine learning; constructing an ensemble learning model and optimizing hyperparameters using a symbiotic biological search algorithm. This invention has significant advantages over traditional technologies, such as more accurate meteorological data analysis, effective extraction of waveform morphology features, and the ability to automatically calibrate faults with unknown causes. In this way, more intelligent fault diagnosis of transmission lines can be achieved, thereby improving the response efficiency and reliability of the power system.

[0025] The present invention provides a method for identifying the causes of transmission line faults based on fault recording and geographic meteorological data, such as... Figure 1 As shown, it includes: Step S1: Collect fault data. After a fault occurs on the transmission line, collect fault waveform data, tower geographical location data, and all meteorological data for the area traversed by the line.

[0026] Step S2: Process and parse fault waveform data. After obtaining the line fault waveform data, if the waveform data model is complete, the parameters of the line and the corresponding analog and digital signals are parsed and extracted based on the power system transient data format standard using the model file; if the waveform data model is missing, fuzzy matching technology is used to find the corresponding data channel by equipment name, and then the data is parsed and extracted.

[0027] Step S3: Obtain meteorological data for the fault location. Based on the distance measurement results calculated using the Takagi algorithm from the fault waveform data and the tower's geographical location data, calculate and obtain the latitude and longitude information of the towers near the fault location. Based on this latitude and longitude information, obtain the meteorological conditions of the fault location.

[0028] In some alternative implementations, the spans between the towers of the faulty line are assumed to be sequentially arranged from the measuring station as follows: The fault location result is x Take the condition that is met pole tower i and i+ Latitude and longitude information of 1 and Meteorological conditions include real-time meteorological data and meteorological bulletins. Real-time meteorological data is approximated using distance coordinates. Recent ground-based observation data, including temperature, relative humidity, wind speed, wind direction, and precipitation; meteorological bulletins are taken from the towers. i The weather conditions in the city or district for the most recent period, including some textual weather descriptions.

[0029] Step S4: Extract meteorological data features based on sentiment analysis. For numerical data such as temperature and wind speed from ground observation stations in the meteorological data, these can be directly selected as features for the fault sample dataset after standardization. For textual descriptions in meteorological bulletins, the descriptive text is first preprocessed with text cleaning and word segmentation. Then, a sentiment dictionary is constructed using weather-related words. Descriptions of good weather, such as "calm wind" and "no precipitation," are assigned higher scores (closer to 1), while descriptions of severe weather, such as "severe convective weather," are assigned lower scores (closer to 0). Specifically, the SnowNLP sentiment analysis network is trained using meteorological labeled data to obtain numerical features (between 0 and 1) of the meteorological text. Finally, these features are combined with numerical meteorological features such as temperature, relative humidity, wind speed, wind direction, and precipitation to form a feature engineering model of meteorological information from multiple dimensions.

[0030] Step S5: Extract morphological features of the recorded waveform using a one-dimensional convolutional neural network (CNN). The voltage and current waveforms of the fault phase are extracted from the fault recording. By selecting sampled values ​​from several cycles before and after the fault, a one-dimensional array to be analyzed is formed. The one-dimensional CNN is used to extract features from the waveform data. The established one-dimensional CNN model includes an input layer, convolutional layer, pooling layer, fully connected layer, and output layer. After model training, the output of the second-to-last layer (the fully connected layer) is taken as the morphological feature of the fault recording for each sample.

[0031] In some alternative solutions, the measured data of the fault phase voltage and current channels are extracted from the fault recording. Sample values ​​from the cycle before the fault and the three cycles after the fault are selected and merged into a one-dimensional array. After resampling, the number of sampling points for one power frequency cycle is unified to N, and the size of the one-dimensional array is then... .

[0032] Step S6: Extract electrical quantity characteristics from the fault recording based on electrical quantity analysis. The analyzed electrical quantity characteristics include harmonic content, attenuated DC component, transition resistance, and measurement impedance angle. When extracting harmonic content, the 2nd to 25th harmonics of the fault phase voltage and current are calculated, and the harmonic order and content rate of the largest and second largest harmonics of voltage and current are taken as fault characteristics. When extracting the attenuated DC component, the fault phase current is calculated, and the amplitude and time constant of the attenuated DC component of the current after the fault are taken as fault characteristics. When extracting transition resistance characteristics, the voltage and current magnitudes at the fault point are calculated using the transmission line equation, and the real part of the impedance value (voltage divided by current) is taken as the transition resistance characteristic value. When extracting the measurement impedance angle, the ratio of the measured voltage to the measured current at one end is calculated, and the phase angle of the ratio result is taken as the measurement impedance angle characteristic.

[0033] Step S7: Label the fault samples. For historical fault cases of AC lines, the causes of the faults are labeled by reviewing the fault inspection reports for subsequent model training. Some historical cases cannot be labeled using conventional methods due to missing inspection reports, and require adaptive semi-supervised machine learning to predict data labels.

[0034] In some optional implementations, the classification labels for fault causes include: lightning strike, wildfire, wind deflection, foreign object, pollution flashover (bird droppings), pollution flashover (other), de-icing jump, insulator skirt bridging, sub-conductor breakage, complete breakage, and others. When performing adaptive semi-supervised machine learning label prediction, let the number of historical dataset samples be B, and the proportion of unlabeled data samples be μ. Then, the objective function for the unlabeled data is: ,in, and These represent the output probabilities of unlabeled data after weak and strong augmentation, respectively. Represents the cross-entropy loss function. To set a threshold; use Replace global threshold The model prediction expectation for each cluster is obtained. The final adaptive threshold is obtained by integrating the global threshold and the local threshold. for: ,in C The number of classification clusters is denoted by 1; after obtaining the adaptive threshold, label prediction is performed on the entire historical dataset of unlabeled data according to the objective function of the unlabeled data.

[0035] Step S8: Establish the ensemble learner model and complete model training and parameter tuning. The previous steps established a fault sample dataset. Each sample in the dataset contains features such as meteorological characteristics, waveform morphology features, and electrical quantity features, as well as a labeled fault cause. The basic learners in the ensemble learner consist of logistic regression and decision trees, combined using a sequence ensemble (Boosting) approach. Hyperparameters of the ensemble learning model, such as the learning rate, the number of basic learners, and the maximum tree depth, are optimized using the swarm optimization algorithm Symbiotic Biology Search (SOS). When a real-time fault occurs, the fault data features are extracted and input into the trained model, and the system outputs the fault cause identification result.

[0036] In some optional implementations, samples that were mispredicted by the previous base learner during sequence ensemble are weighted and added to the sample set to train the next base learner. The objective function of the SOS algorithm is the prediction accuracy on the test set. After iterative optimization through three stages—mutualism, symbiosis, and parasitism—it seeks the optimal combination of hyperparameters for the ensemble learning model.

[0037] Specifically, in a specific embodiment of the present invention: Step S1: Collect fault data. After a fault occurs on the transmission line, collect fault waveform data, tower geographical location data, and all meteorological data for the area traversed by the line.

[0038] Specifically, historical fault case data of power grid transmission lines in recent years were compiled and collected, including fault recording or protection recording data, geographical and meteorological data, and fault inspection analysis reports. A total of 7019 cases were collected, with 6272 cases having known causes: 1828 cases of lightning strikes, 1751 cases of foreign objects, 930 cases of wind deflection, 906 cases of flashover (bird droppings), 73 cases of flashover (other), 230 cases of wildfires, 260 cases of ice breakage, 54 cases of insulator skirt bridging, 67 cases of conductor breakage, 43 cases of complete breakage, and 130 cases of other causes (such as secondary circuit faults); 707 cases had unknown causes. Screenshots of the data management interface are shown below. Figure 2 As shown.

[0039] Step S2: Process and parse the fault waveform data. After acquiring the line fault waveform data, if the waveform data model is complete, the parameters of the line and the corresponding analog and digital input channels are parsed and extracted based on the power system transient data format standard using the model file. If the waveform data model is missing, fuzzy matching technology is used to find the corresponding data channel by device name, and then the data is parsed and extracted. For example, extracting the voltage and current data of the fault phase.

[0040] Specifically, power system fault recording files contain up to four related standard files: title file (.HDR), configuration file (.CFG), data file (.DAT), and information file (.INF). The recording sampling data is parsed and extracted according to the national standard GB / T 14598.24-2017 Measuring relays and protection devices Part 24: General format for power system transient data exchange (COMTRADE).

[0041] Step S3: Obtain meteorological data for the fault location. Based on the ranging results calculated from the fault recording data and the tower's geographical location data, calculate and obtain the latitude and longitude information of the towers near the fault location. Based on this latitude and longitude information, obtain the meteorological conditions of the fault location.

[0042] Specifically, a certain 220kV line is 29.763km long with a total of 73 towers. The spans between the towers, starting from the measurement station, are as follows: The fault location result is 24.2 km, which meets the conditions. Latitude and longitude information of towers 69 and 70 and Meteorological conditions include real-time meteorological data and meteorological bulletins. Real-time meteorological data is approximated using distance coordinates. Recent ground-based observation data, including air temperature (°C), relative humidity (%), wind speed (m / s), wind direction (°), and precipitation (mm), are as follows: The weather bulletin uses the most recent weather conditions for the area where tower No. 69 is located, stating: "The Meteorological Bureau issues an orange alert for thunderstorms and a yellow alert for strong winds. Severe convective weather is expected, with localized hail and short-term heavy precipitation. Gusts of 8-10 levels, locally exceeding level 11, are expected during thunderstorms."

[0043] Step S4: Extract meteorological data features based on sentiment analysis. For numerical data such as temperature and wind speed from ground observation stations in the meteorological data, these can be directly selected as features of the fault sample dataset after standardization. For textual descriptions in meteorological bulletins, the descriptive text is first preprocessed with text cleaning and word segmentation. Then, a sentiment dictionary is constructed using weather-related words. Descriptions of good weather, such as "calm wind" and "no precipitation," are assigned higher scores (closer to 1), while descriptions of severe weather, such as "severe convective weather," are assigned lower scores (closer to 0). In this embodiment, the SnowNLP sentiment analysis network is trained using meteorological labeled data to obtain numerical features of the meteorological text (between 0 and 1). In this case, the score is 0.91. Finally, these features are combined with numerical meteorological features such as temperature, relative humidity, wind speed, wind direction, and precipitation to form a feature engineering representation of meteorological information from multiple dimensions. .

[0044] Step S5: Extract morphological features of the recorded waveform using a one-dimensional convolutional neural network (CNN). The voltage and current waveforms of the fault phase are extracted from the fault recording. By selecting sampled values ​​from several cycles before and after the fault, a one-dimensional array to be analyzed is formed. The one-dimensional CNN is used to extract features from the waveform data. The established one-dimensional CNN model includes an input layer, convolutional layer, pooling layer, fully connected layer, and output layer. After model training, the output of the second-to-last layer (the fully connected layer) is taken as the morphological feature of the fault recording for each sample.

[0045] Specifically, the measured data of the fault phase voltage and current channels are extracted from the fault recording. Sample values ​​from the cycle before the fault and the three cycles after the fault are selected and merged into a one-dimensional array. After resampling, the number of sampling points for one power frequency cycle is unified to N, and the size of the one-dimensional array is... Considering the number of samples, feature dimensions, and number of labels, the designed one-dimensional CNN convolutional layer has 64 filters, a filter width of 5, an activation function of ReLU, and a pooling window size of 2. The fully connected layer contains 32 neurons, and the activation function is also ReLU. The output of the fully connected layer is the extracted morphological features of the recorded waveform.

[0046] Step S6: Extract electrical quantity characteristics from the fault recording based on electrical quantity analysis. The analyzed electrical quantity characteristics include harmonic content, attenuated DC component, transition resistance, and measurement impedance angle. When extracting harmonic content, the 2nd to 25th harmonics of the fault phase voltage and current are calculated, and the harmonic order and content rate of the largest and second largest harmonics of voltage and current are taken as fault characteristics. When extracting the attenuated DC component, the fault phase current is calculated, and the amplitude and time constant of the attenuated DC component of the current after the fault are taken as fault characteristics. When extracting transition resistance characteristics, the voltage and current magnitudes at the fault point are calculated using the transmission line equation, and the real part of the impedance value (voltage divided by current) is taken as the transition resistance characteristic value. When extracting the measurement impedance angle, the ratio of the measured voltage to the measured current at one end is calculated, and the phase angle of the ratio result is taken as the measurement impedance angle characteristic.

[0047] Step S7: Label the fault samples. For historical fault cases of AC lines, the causes of the faults are labeled by reviewing the fault inspection reports for subsequent model training. Some historical cases cannot be labeled using conventional methods due to missing inspection reports, and require adaptive semi-supervised machine learning to predict data labels.

[0048] Specifically, the classification labels for fault causes include: lightning strike, wildfire, wind deflection, foreign object, pollution flashover (bird droppings), pollution flashover (other), de-icing jump, insulator skirt bridging, sub-conductor breakage, complete breakage, and others. When performing adaptive semi-supervised machine learning label prediction, let the number of historical dataset samples be B, and the proportion of unlabeled data samples be μ. Then, the objective function for the unlabeled data is: ,in, and These represent the output probabilities of unlabeled data after weak and strong augmentation, respectively. Represents the cross-entropy loss function. Set a threshold for the third; use Replace global threshold The model prediction expectation for each cluster is obtained. The final adaptive threshold is obtained by integrating the global threshold and the local threshold. for: ,in C The number of classification clusters is denoted by 1; after obtaining the adaptive threshold, label prediction is performed on the entire historical dataset of unlabeled data according to the objective function of the unlabeled data.

[0049] Step S8: Establish the ensemble learner model and complete model training and parameter tuning. The previous steps established a fault sample dataset. Each sample in the dataset contains features such as meteorological characteristics, waveform morphology features, and electrical quantity features, as well as a labeled fault cause. The basic learners in the ensemble learner consist of logistic regression and decision trees, combined using a sequence ensemble (Boosting) approach. Hyperparameters of the ensemble learning model, such as the learning rate, the number of basic learners, and the maximum tree depth, are optimized using the swarm optimization algorithm Symbiotic Biology Search (SOS). When a real-time fault occurs, the fault data features are extracted and input into the trained model, and the system outputs the fault cause identification result.

[0050] In some optional implementations, samples that were mispredicted by the previous basic learner during sequence ensemble are weighted and added to the sample set to train the next basic learner. The objective function of the SOS algorithm is the prediction accuracy on the test set. After iterative optimization through three stages—mutualism, symbiosis, and parasitism—the optimal combination of hyperparameters for the ensemble learning model is sought. The test set consists of 100 cases, including 20 cases of lightning strikes, 20 cases of foreign objects, 13 cases of wildfires, 10 cases of wind deflection, 7 cases of flashover (bird droppings), 4 cases of flashover (other), 7 cases of de-icing jumps, 5 cases of insulator skirt bridging, 5 cases of complete line breakage, 5 cases of line sub-conductor breakage, and 4 other cases. After training and testing using the method provided in this invention, 90 fault causes were accurately identified, as shown in the results below. Figure 3As shown, the blue, green, and yellow bar charts above each fault cause represent the number of test cases, the number of correct predictions, and the number of incorrect predictions of that type.

[0051] It should be noted that the steps shown in the above process or in the flowchart of the accompanying figures can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0052] In addition, combined Figure 1 The transmission line fault cause identification method based on fault recording and geographic meteorology described in this embodiment of the invention can be implemented by a computer device. Figure 4 This is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. Figure 4 As shown, the device may include a processor 301 and a memory 302 storing computer program instructions.

[0053] Specifically, the processor 301 may include a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of the present invention.

[0054] Memory 302 may include a large-capacity memory for data or instructions. For example, and not limitingly, memory 302 may include a hard disk drive (HDD), a floppy disk drive, a solid-state drive (SSD), flash memory, an optical disk drive, a magneto-optical disk drive, magnetic tape, or a Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, memory 302 may include removable or non-removable (or fixed) media. Where appropriate, memory 302 may be internal or external to a data processing device. In a particular embodiment, memory 302 is non-volatile memory. In a particular embodiment, memory 302 includes read-only memory (ROM) and random access memory (RAM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable read-only memory (PROM), an erasable read-only memory (EPROM), an electrically erasable read-only memory (EEPROM), an electrically alterable read-only memory (EAROM), or flash memory, or a combination of two or more of these. Where appropriate, the RAM can be Static Random-Access Memory (SRAM) or Dynamic Random-Access Memory (DRAM). DRAM can be Fast Page Mode Dynamic Random-Access Memory (FPMDRAM), Extended Data Out Dynamic Random-Access Memory (EDODRAM), Synchronous Dynamic Random-Access Memory (SDRAM), etc.

[0055] The memory 302 can be used to store or cache various data files that need to be processed and / or communicated, as well as possible computer program instructions executed by the processor 301.

[0056] The processor 301 reads and executes computer program instructions stored in the memory 302 to implement any of the transmission line fault cause identification methods based on fault recording and geographical meteorology in the above embodiments.

[0057] In some embodiments, the computer device may further include a communication interface 303 and a bus 300. For example, Figure 4 As shown, the processor 301, memory 302, and communication interface 303 are connected through bus 300 and complete communication with each other.

[0058] The communication interface 303 is used to enable communication between the various modules, devices, units, and / or equipment in the embodiments of the present invention. The communication interface 303 can also enable data communication with other components such as external devices, image / data acquisition devices, databases, external storage, and image / data processing workstations.

[0059] Bus 300 includes hardware, software, or both, that couples components of a computer device together. Bus 300 includes, but is not limited to, at least one of the following: data bus, address bus, control bus, expansion bus, and local bus. For example, and not as a limitation, bus 300 may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an InfiniBand interconnect, a Low Pin Count (LPC) bus, a memory bus, a MicroChannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local Bus (VLB) bus, or other suitable buses, or a combination of two or more of these. Where appropriate, bus 300 may include one or more buses. Although specific buses are described and illustrated in embodiments of the invention, the invention contemplates any suitable bus or interconnect.

[0060] The computer device can execute the transmission line fault cause identification method based on fault recording and geographic meteorology in this embodiment of the invention, thereby achieving a combination of... Figure 1 This paper describes a method for identifying the causes of transmission line faults based on fault recording and geographic meteorology.

[0061] Furthermore, in conjunction with the transmission line fault cause identification method based on fault recording and geographical meteorology in the above embodiments, this invention can be implemented using a computer-readable storage medium. This computer-readable storage medium stores computer program instructions; when executed by a processor, these computer program instructions implement any of the transmission line fault cause identification methods based on fault recording and geographical meteorology in the above embodiments.

[0062] It should be noted that the technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments have been described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification. In addition, depending on the implementation needs, the various steps / components described in this invention can be broken down into more steps / components, or two or more steps / components or parts of steps / components can be combined into new steps / components to achieve the purpose of this invention.

[0063] It will be readily understood by those skilled in the art that the above-described embodiments merely illustrate several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this patent should be determined by the appended claims.

Claims

1. A power transmission line fault cause identification method based on fault recording and geographic meteorology, characterized in that, The method includes: Acquire transmission line fault data, including fault waveform data, tower geographical location data, and all meteorological data for the areas traversed by the line; Process and analyze fault waveform data; The fault location is calculated based on the fault recording data, the fault location is determined by combining the tower geographical location data, and the meteorological data of the fault location is obtained by combining all the meteorological data of the area traversed by the line. Extracting meteorological characteristics from meteorological data; Based on a one-dimensional convolutional neural network, waveform morphology features of fault recording data are extracted. Based on electrical quantity analysis, electrical quantity characteristics of fault recording data are extracted; Fault data is labeled with fault causes to set fault cause labels, and fault cause labels are predicted for fault data without labeled fault causes through an adaptive semi-supervised machine learning method. An integrated learner model was established and model training and parameter tuning were completed. The model takes the meteorological characteristics, waveform morphology characteristics and electrical quantity characteristics of fault data as input and the fault cause label as output to identify the fault cause of transmission lines. This includes processing and analyzing fault recording data, including: After acquiring fault waveform data of transmission lines, if the waveform data model is complete, the parameters of the line and the corresponding analog and digital signals are parsed and extracted based on the power system transient data format standard using the model file; if the waveform data model is missing, fuzzy matching technology is used to find the corresponding data channel by equipment name, and then the acquired data is parsed and extracted. An adaptive semi-supervised machine learning approach is used to predict fault cause labels for fault data with unlabeled fault causes, including: Let B be the number of fault data samples, and μ be the proportion of fault data samples without a defined fault cause. Then the objective function for fault data without a defined fault cause is: ,in, and They represent the first The output probabilities of fault data with unlabeled fault causes after weak and strong enhancement, respectively. Represents the cross-entropy loss function. To set a threshold, This is an indicator function; it is 1 if the condition is met, and 0 otherwise. use Replace global threshold The model prediction expectation for each cluster is obtained. The final adaptive threshold is obtained by integrating the global threshold and the local threshold. for: ,in C This refers to the number of clusters, i.e., the number of fault cause labels. Labels indicating the cause of the fault. The global threshold is used; after obtaining the adaptive threshold, the fault cause label is predicted for the fault data samples without a defined fault cause based on the objective function of the fault data without a defined fault cause.

2. The method for identifying the cause of transmission line faults based on fault recording and geographic meteorology as described in claim 1, characterized in that, The fault location is calculated based on the fault recording data, and the fault location area is determined by combining the tower geographical location data. Then, meteorological data for the fault location area is obtained by combining all meteorological data across the area traversed by the line, including: Assume the spans between the poles of the faulty line are, starting from the measurement station, sequentially as follows: The fault location result is x Take the condition that is met pole tower i and i+ Latitude and longitude information of 1 and ; Meteorological data includes real-time meteorological data and meteorological bulletins. Real-time meteorological data is obtained using distance coordinates. Recent ground-based observation data, including temperature, relative humidity, wind speed, wind direction, and precipitation; meteorological bulletins are taken from the towers. i The weather conditions for the city or district in the most recent time period, including text-based weather descriptions.

3. The method for identifying the cause of transmission line faults based on fault recording and geographic meteorology as described in claim 2, characterized in that, Extracting meteorological characteristics from meteorological data, including: For real-time meteorological data, we directly take the numerical values ​​and standardize them to obtain numerical meteorological characteristics. For text-based weather descriptions in meteorological bulletins, numerical features of the text-based weather descriptions are obtained based on sentiment analysis and then merged with the numerical meteorological characteristics of the actual meteorological data to finally obtain the meteorological features of the meteorological data; among them, the sentiment analysis method includes the use of the sentiment analysis network SnowNLP.

4. The method for identifying the cause of transmission line faults based on fault recording and geographic meteorology as described in claim 1, characterized in that, Based on a one-dimensional convolutional neural network, waveform morphology features of fault recording data are extracted, including: The voltage and current waveforms of the fault phase are extracted from the fault recording data. By selecting the sampled values ​​of several cycles before and after the fault, they are merged to form a one-dimensional array to be analyzed. A one-dimensional convolutional neural network is used to extract features from the waveform data. The one-dimensional convolutional neural network includes an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer. It takes the one-dimensional array as input and the fault cause label as output. After model training, the output result of the second to last layer, i.e., the fully connected layer, is taken as the waveform morphology feature of the fault recording data for each fault recording data.

5. The method for identifying the cause of transmission line faults based on fault recording and geographic meteorology as described in claim 1, characterized in that, Based on electrical quantity analysis, electrical quantity characteristics of fault recording data are extracted, including: Electrical quantity characteristics include harmonic content, attenuated DC component, transition resistance, and measurement impedance angle. When extracting harmonic content, the 2nd to 25th harmonics of the fault phase voltage and current are calculated, and the harmonic order and content of the largest and second largest harmonics of voltage and current are taken as fault characteristics. When extracting attenuated DC component, the fault phase current is calculated, and the amplitude and time constant of the attenuated DC component of the current after the fault are taken as fault characteristics. When extracting transition resistance, the voltage and current magnitudes at the fault point are calculated using the transmission line equation, and the real part of the impedance value (voltage divided by current) is taken as the transition resistance characteristic. When extracting measurement impedance angle, the ratio of the measured voltage to the measured current at one end is calculated, and the phase angle of the ratio result is taken as the measurement impedance angle characteristic.

6. The method for identifying the cause of transmission line faults based on fault recording and geographic meteorology as described in claim 1, characterized in that, Fault cause labels include: lightning strike, wildfire, wind deflection, foreign object, bird droppings flashover, other flashover, de-icing jump, insulator skirt bridging, sub-conductor breakage, complete breakage, and others.

7. The method for identifying the cause of transmission line faults based on fault recording and geographic meteorology as described in claim 1, characterized in that, The basic learner of the ensemble learner model consists of logistic regression and decision tree, and the basic learner is combined using a sequence ensemble approach. The hyperparameters of the ensemble learner model are optimized using the symbiotic search algorithm of swarm optimization.

8. A computer device, characterized in that, include: The processor and memory, wherein the memory stores a program or instructions that can run on the processor, and when the program or instructions are executed by the processor, implement the steps of the transmission line fault cause identification method based on fault recording and geographic meteorology as described in any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that, It stores programs or instructions, which, when executed by a processor, implement the steps of the transmission line fault cause identification method based on fault recording and geographical meteorology as described in any one of claims 1 to 7.

Images