A power distribution network fault identification method and system based on multi-source data

Abstract

A method and system for identifying distribution network faults based on multi-source data is disclosed. This method involves acquiring weather warning level data, querying to obtain a matching data acquisition pattern, collecting feature data according to the matching pattern, normalizing the feature data, calculating feature impact data and dynamic weight coefficients, calculating line risk values ​​based on the feature impact data and dynamic weight coefficients, comparing the line risk values ​​with thresholds to obtain the distribution network risk status level, and finally sending distribution network warning information to the operation and maintenance monitoring platform based on the distribution network risk status level. This invention achieves distribution network fault identification by obtaining feature data through matching data acquisition patterns, calculating feature impact data, dynamic weight coefficients, and line risk values, and comparing line risk values ​​with thresholds. This invention can facilitate efficient scheduling of operation and maintenance resources and reduce distribution network fault losses.

Description

Technical Field

[0001] This invention belongs to the field of power distribution network fault identification technology, and in particular relates to a power distribution network fault identification method and system based on multi-source data. Background Technology

[0002] my country's coastal areas frequently experience extreme weather events such as typhoons and torrential rains, which often impact power distribution and can even cause large-scale outages. As a crucial link in power transmission, the stable operation of the power distribution network directly affects residents' lives and businesses' production.

[0003] Traditional fault identification often relies on manual inspections and user reports before or after a fault occurs. This is inefficient and cannot identify fault risks or phenomena in a timely and effective manner before or during typhoons or rainstorms, which can have a certain impact on people's lives or work.

[0004] Existing technologies, such as the patent document with publication number CN118468216A, provide a solution that determines the first fault identification result of the target distribution network based on electrical measurement data. Simultaneously, a second fault identification result is obtained through inspection image data of the target distribution network. The comprehensive identification result obtained by fusing the two results determines the safety hazard identification result of the distribution network. Another example is the patent document with publication number CN104280668A, which obtains real-time waveforms from the distribution network transmission lines, matches them with waveforms in a waveform library for coarse fault type identification, and then further matches the real-time waveforms with waveforms in the waveform library by extracting edge feature points to determine the specific fault type of the real-time waveform. Yet another example is the patent document with publication number CN113625108A, which uses a random forest scheme composed of multiple decision trees. Utilizing machine learning, it trains the random forest based on the feature vectors of the faulty lines to identify fault categories. While there are some solutions for integrating multiple data sources in existing technologies, they often only rely on electrical data for judgment, taking into account fewer factors and failing to comprehensively consider complex environmental factors. They lack the ability to identify and predict fault points in the distribution network based on the environment, and thus cannot meet the requirements for the safe and stable operation of the distribution network.

[0005] Effective technical solutions are urgently needed to address the above problems. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a method and system for identifying distribution network faults based on multi-source data. The method involves acquiring weather warning level data and then querying to obtain a matching data acquisition pattern. Feature data is then collected according to the matching pattern. After normalizing the feature data, feature impact data and dynamic weight coefficients are calculated. Line risk values ​​are calculated based on the feature impact data and dynamic weight coefficients. The line risk values ​​are then compared to thresholds to obtain the distribution network risk status level. Finally, the distribution network warning information is sent to the operation and maintenance monitoring platform based on the distribution network risk status level. This method achieves distribution network fault identification by obtaining feature data through matching data acquisition patterns, calculating feature impact data, dynamic weight coefficients, and line risk values, and comparing line risk values ​​to thresholds.

[0007] The present invention adopts the following technical solution.

[0008] This invention proposes a method for distribution network fault identification based on multi-source data, comprising: Obtain weather warning level data, and query the preset warning collection mode mapping table based on the weather warning level data to obtain the matching data collection mode; According to the matching data acquisition mode, feature data is collected in the power distribution network line by preset sensors. The feature data includes environmental feature data, equipment feature data and hidden danger feature data. Based on the normalized feature data, feature impact data and dynamic weight coefficients are calculated; the feature impact data includes environmental impact data, equipment impact data, and hazard impact data. The line risk value is obtained by weighted summation based on the characteristic impact data and the dynamic weight coefficient. The risk value of the line is compared with the preset risk assessment threshold of the distribution network to obtain the risk level of the distribution network. Based on the risk level of the distribution network, the corresponding early warning information of the distribution network will be sent to the operation and maintenance monitoring platform.

[0009] More preferably, the step of obtaining the weather warning level is as follows: The real-time monitored wind force level and rainfall amount are mapped to risk score values ​​respectively; Assign weights to the risk score values ​​corresponding to the wind force level and the rainfall, and calculate the comprehensive risk score; The final weather warning level is determined by comparing the comprehensive risk score with a preset threshold range.

[0010] More preferably, the method for calculating the feature influence data is as follows: Environmental impact data = wind speed characteristic coefficient × normalized wind speed data + rainfall characteristic coefficient × normalized rainfall data + correction characteristic coefficient × normalized wind speed data × normalized rainfall data; Equipment impact data = Line current trend slope characteristic coefficient × Normalized value of line current trend slope + Tower tilt angle change characteristic coefficient × Normalized value of tower tilt angle change coefficient + Line temperature characteristic coefficient × Normalized value of line temperature. Hazard impact data = Line wind resistance level characteristic coefficient / Normalized value of line wind resistance level data + Annual condition characteristic coefficient × (Line operating years data / Standard operating years data) All of the aforementioned characteristic coefficients are obtained through a preset process.

[0011] More preferably, the dynamic weighting coefficient is calculated as follows: The environmental impact data raised to the power of k is used as the environmental impact conversion value, the equipment impact data raised to the power of k is used as the equipment impact conversion value, and the hidden danger impact data raised to the power of k is used as the hidden danger impact conversion value. The environmental impact conversion value, the equipment impact conversion value, and the hazard impact conversion value are normalized to obtain the dynamic weight coefficients corresponding to the environmental impact data, the equipment impact data, and the hazard impact data.

[0012] More preferably, the step of comparing the line risk value with a preset distribution network risk status assessment threshold to obtain the distribution network risk status level specifically includes: The first threshold and the second threshold are extracted based on the preset risk assessment threshold of the distribution network, and the first threshold is greater than the second threshold; Compare the line risk value with the first threshold and the second threshold; If the line risk value is less than or equal to the second threshold, the risk level of the distribution network is low risk. If the line risk value is greater than the second threshold and less than or equal to the first threshold, the risk level of the distribution network is medium risk. If the risk value of the line is greater than the first threshold, the risk level of the distribution network is high.

[0013] More preferably, the step of sending the corresponding distribution network early warning information to the operation and maintenance monitoring platform according to the distribution network risk status level includes the following steps: If the risk level of the distribution network is low, then the current situation will remain unchanged; If the risk level of the distribution network is medium risk, a medium-level early warning and the risk location will be sent to the operation and maintenance monitoring platform. If the risk level of the distribution network is high, a high-level warning, risk location, and maintenance priority will be sent to the operation and maintenance monitoring platform.

[0014] More preferably, the environmental feature data is collected through the following steps: Acquire environmental feature data of the current location and a preset number of environmental feature data of the same category within a preset distance; Determine the status data of the environmental characteristics obtained at the current location, including whether it is normal or abnormal; If the status data is normal, proceed to the next step; If the status data is abnormal, obtain the effective accuracy of the sensor corresponding to other environmental feature data within the preset distance; The weight data corresponding to other environmental feature data is obtained based on the effective accuracy. The corrected values ​​for environmental characteristic data are obtained by weighting other environmental characteristic data and corresponding weight data; Change the abnormal environmental feature data of the current location to the corrected environmental feature data value.

[0015] This invention also proposes a distribution network fault identification system based on multi-source data, including a matching data acquisition mode determination module, a feature data acquisition module, a line risk value calculation module, a distribution network risk status level acquisition module, and an execution module: The matching data acquisition mode determination module acquires weather warning level data, queries a preset warning acquisition mode mapping table based on the weather warning level data, and obtains the matching data acquisition mode. The feature data acquisition module collects feature data in the power distribution network line through preset sensors according to the matching data acquisition mode. The feature data includes environmental feature data, equipment feature data, and potential hazard feature data. The line risk value calculation module calculates feature impact data and dynamic weight coefficients based on the normalized feature data; the feature impact data includes environmental impact data, equipment impact data, and hidden danger impact data; and obtains the line risk value by weighted summation based on the feature impact data and the dynamic weight coefficients. The distribution network risk status level acquisition module compares the line risk value with a preset distribution network risk status assessment threshold to obtain the distribution network risk status level. The execution module sends the corresponding early warning information of the distribution network to the operation and maintenance monitoring platform according to the risk status level of the distribution network.

[0016] The present invention also proposes a terminal, including a processor and a storage medium: The storage medium is used to store instructions; The processor is configured to operate according to the instructions to perform the steps of the method according to any of the foregoing.

[0017] The present invention also proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the methods described above.

[0018] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention solves the problems of low efficiency and slow response of traditional methods: by using a weather warning-linked dynamic data acquisition mode, it enables early identification and warning of fault risks without relying on manual inspections; 2. This invention solves the problem of existing solutions considering only one factor: it integrates multi-source data from the environment, equipment, and potential hazards, and combines dynamic weight adjustment to improve the comprehensiveness and accuracy of fault identification; 3. This invention enables tiered early warning and intelligent operation and maintenance: it pushes targeted early warning information according to risk level, and combines multi-risk point path planning to help efficiently schedule operation and maintenance resources and reduce distribution network fault losses. Attached Figure Description

[0019] Figure 1 This is a flowchart of a method for identifying distribution network faults based on multi-source data according to the present invention; Figure 2 This is a flowchart of a power distribution network fault identification method based on multi-source data provided in an embodiment of the present invention; Figure 3 A flowchart illustrating the acquisition and matching data collection mode of a power distribution network fault identification method based on multi-source data, provided in an embodiment of the present invention; Figure 4 This is a flowchart illustrating the acquisition of characteristic influence data and dynamic weight coefficients in a distribution network fault identification method based on multi-source data, as provided in an embodiment of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of this invention. The embodiments described in this application are merely some embodiments of this invention, and not all embodiments. Based on the spirit of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of this invention.

[0021] The present invention proposes the following technical solution: like Figure 1 As shown, this invention proposes a method for identifying distribution network faults based on multi-source data, specifically including the following steps: Obtain weather warning level data, and query the preset warning collection mode mapping table based on the weather warning level data to obtain the matching data collection mode; The steps for obtaining the weather warning level are as follows: The real-time monitored wind force level and rainfall amount are mapped to risk score values ​​respectively; Assign weights to the risk score values ​​corresponding to the wind force level and the rainfall, and calculate the comprehensive risk score; The final weather warning level is determined by comparing the comprehensive risk score with a preset threshold range.

[0022] According to the matching data acquisition mode, feature data is collected in the power distribution network line by preset sensors. The feature data includes environmental feature data, equipment feature data and hidden danger feature data. Based on the normalized feature data, feature impact data and dynamic weight coefficients are calculated; the feature impact data includes environmental impact data, equipment impact data, and hazard impact data. The method for calculating the feature influence data is as follows: Environmental impact data = wind speed characteristic coefficient × normalized wind speed data + rainfall characteristic coefficient × normalized rainfall data + correction characteristic coefficient × normalized wind speed data × normalized rainfall data; Equipment impact data = Line current trend slope characteristic coefficient × Normalized value of line current trend slope + Tower tilt angle change characteristic coefficient × Normalized value of tower tilt angle change coefficient + Line temperature characteristic coefficient × Normalized value of line temperature. Hazard impact data = Line wind resistance level characteristic coefficient / Normalized value of line wind resistance level data + Annual condition characteristic coefficient × (Line operating years data / Standard operating years data) All of the aforementioned characteristic coefficients are obtained through a preset process.

[0023] The calculation method for the dynamic weighting coefficient is as follows: The environmental impact data raised to the power of k is used as the environmental impact conversion value, the equipment impact data raised to the power of k is used as the equipment impact conversion value, and the hidden danger impact data raised to the power of k is used as the hidden danger impact conversion value. The environmental impact conversion value, the equipment impact conversion value, and the hazard impact conversion value are normalized to obtain the dynamic weight coefficients corresponding to the environmental impact data, the equipment impact data, and the hazard impact data.

[0024] The line risk value is obtained by weighted summation based on the characteristic impact data and the dynamic weight coefficient. The risk value of the line is compared with the preset risk assessment threshold of the distribution network to obtain the risk level of the distribution network. The first threshold and the second threshold are extracted based on the preset risk assessment threshold of the distribution network, and the first threshold is greater than the second threshold; Compare the line risk value with the first threshold and the second threshold; If the line risk value is less than or equal to the second threshold, the risk level of the distribution network is low risk. If the line risk value is greater than the second threshold and less than or equal to the first threshold, the risk level of the distribution network is medium risk. If the risk value of the line is greater than the first threshold, the risk level of the distribution network is high.

[0025] Based on the risk level of the distribution network, the corresponding early warning information of the distribution network will be sent to the operation and maintenance monitoring platform.

[0026] If the risk level of the distribution network is low, then the current situation will remain unchanged; If the risk level of the distribution network is medium risk, a medium-level early warning and the risk location will be sent to the operation and maintenance monitoring platform. If the risk level of the distribution network is high, a high-level warning, risk location, and maintenance priority will be sent to the operation and maintenance monitoring platform.

[0027] The environmental feature data collection steps are as follows: Acquire environmental feature data of the current location and a preset number of environmental feature data of the same category within a preset distance; Determine the status data of the environmental characteristics obtained at the current location, including whether it is normal or abnormal; If the status data is normal, proceed to the next step; If the status data is abnormal, obtain the effective accuracy of the sensor corresponding to other environmental feature data within the preset distance; The weight data corresponding to other environmental feature data is obtained based on the effective accuracy. The corrected values ​​for environmental characteristic data are obtained by weighting other environmental characteristic data and corresponding weight data; Change the abnormal environmental feature data of the current location to the corrected environmental feature data value.

[0028] This invention also proposes a distribution network fault identification system based on multi-source data, including a matching data acquisition mode determination module, a feature data acquisition module, a line risk value calculation module, a distribution network risk status level acquisition module, and an execution module: The matching data acquisition mode determination module acquires weather warning level data, queries a preset warning acquisition mode mapping table based on the weather warning level data, and obtains the matching data acquisition mode. The feature data acquisition module collects feature data in the power distribution network line through preset sensors according to the matching data acquisition mode. The feature data includes environmental feature data, equipment feature data, and potential hazard feature data. The line risk value calculation module calculates feature impact data and dynamic weight coefficients based on the normalized feature data; the feature impact data includes environmental impact data, equipment impact data, and hidden danger impact data; and obtains the line risk value by weighted summation based on the feature impact data and the dynamic weight coefficients. The distribution network risk status level acquisition module compares the line risk value with a preset distribution network risk status assessment threshold to obtain the distribution network risk status level. The execution module sends the corresponding early warning information of the distribution network to the operation and maintenance monitoring platform according to the risk status level of the distribution network.

[0029] The present invention also proposes a terminal, including a processor and a storage medium: The storage medium is used to store instructions; The processor is configured to operate according to the instructions to perform the steps of the method according to any of the foregoing.

[0030] The present invention also proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the methods described above.

[0031] Example 1 Please refer to Figure 2 , Figure 2 This is a flowchart of a distribution network fault identification method based on multi-source data according to some embodiments of this application. This multi-source data-based distribution network fault identification method is used in terminal devices, such as computers and mobile phones. The multi-source data-based distribution network fault identification method includes the following steps: S11. Obtain weather warning level data, and obtain the matching data collection mode by querying the preset warning collection mode mapping table based on the weather warning level data. S12. Collect characteristic data in the power distribution network line through preset sensors according to the matching data acquisition mode, including environmental characteristic data, equipment characteristic data and hidden danger characteristic data; S13. After normalizing the feature data, calculate the feature influence data and dynamic weight coefficients. S14. The line risk value is calculated using the weighted summation method based on the characteristic impact data and dynamic weight coefficients. S15. Compare the line risk value with the preset distribution network risk status assessment threshold to obtain the distribution network risk status level; S16. Send the corresponding early warning information of the distribution network to the operation and maintenance monitoring platform according to the risk status level of the distribution network.

[0032] It should be noted that weather warning level data reflects the potential harm, urgency, and development trend of weather. It can be a warning level for a future period obtained from meteorological data, or a warning level for the current weather assessed through real-time weather data. In this embodiment, the latter is used. The weather warning level is calculated based on wind speed and rainfall using a preset warning level assessment model, which is trained on historical wind speed and rainfall data. In the weather warning level data, Level 1 is the highest, and so on. The required data collection frequency for the distribution network varies under different weather warning levels. Therefore, the data collection mode needs to be determined based on the weather warning level. After collecting characteristic data according to the determined data collection mode, normalization processing is performed to calculate characteristic impact data and dynamic weight coefficients. Then, the line risk value is calculated based on the characteristic impact data and dynamic weight coefficients. After threshold comparison, the power grid risk status level is obtained. Finally, the corresponding distribution network warning information is sent to the operation and maintenance monitoring platform.

[0033] Please refer to Figure 3 , Figure 3 This is a flowchart illustrating the process of obtaining a matching data acquisition mode in a distribution network fault identification method based on multi-source data, as provided in this application embodiment. According to this embodiment, the step of obtaining weather warning level data and a data acquisition mode, and obtaining a matching data acquisition mode by querying a preset warning acquisition mode mapping table based on the weather warning level data, specifically includes: S21. Obtain weather warning level data, including Level 1 warning, Level 2 warning, or Level 3 warning; As an example of this invention, the real-time monitored wind speed and rainfall are mapped to risk score values, with higher scores indicating greater risk. These risk score values ​​are preset; for example, the risk score value corresponding to the wind speed level could be: Level 1-3 (light breeze): 0-20 points; Level 4-5 (gentle breeze): 21-40 points; Level 6-7 (strong winds): 41-60 points; Level 8-9 (Gale): 61-80 points; Level 10-11 (Gale): 81-95 points; ≥Category 12 (Hurricane): 96-100 points; The risk score corresponding to rainfall amount can be: <10 mm / h (light rain): 0-20 minutes; 10-25 mm / h (moderate rain): 21-40 minutes; 25-50 mm / h (heavy rain): 41-60 minutes; 50-100 mm / h (heavy rain): 61-80 minutes; 100-250 mm / h (heavy rainstorm): 81-95 minutes; 250 mm / h (Extremely heavy rain): 96-100 points.

[0034] Weights are assigned to wind speed and rainfall to reflect the varying importance of different factors on the power distribution network in different regions or scenarios. A weighted sum of wind speed and rainfall is calculated as a comprehensive risk score. Based on the calculated comprehensive risk score and compared with a preset threshold range, the final weather warning data is determined, as an example of this invention: Level 3 Warning (Low Risk): 0 ≤ Comprehensive Risk Score < 40; Level 2 Warning (Medium Risk): 40 ≤ Comprehensive Risk Score < 70; Level 1 Warning (High Risk): Comprehensive risk score ≥ 70; Among these methods, historical meteorological data and records of actual power distribution network faults or their impact can be used to optimize the critical points, weights, and level determination thresholds of the scoring mapping through machine learning algorithms (such as logistic regression, support vector machines, or decision trees). Alternatively, the thresholds for critical points, weights, and level determination can be empirically set based on experimental performance to maximize the match between the warning level and the actual historical risk situation.

[0035] S22. Based on the weather warning level data, query the preset warning collection mode mapping table to obtain the matching data collection mode.

[0036] It should be noted that a Level 1 warning indicates severe weather conditions and the greatest potential impact on the power distribution network, followed by a Level 2 warning, with the Level 3 warning having the least impact. Data acquisition modes refer to different frequency modes used when collecting data, including Level 1, Level 2, and Level 3 acquisition frequency modes, with Level 1 having the highest acquisition frequency, and so on. A preset warning acquisition mode mapping table shows the correspondence between weather warning levels and data acquisition modes. This preset warning acquisition mode mapping table is developed based on actual needs and historical data analysis. In this embodiment, taking the collection of wind speed, rainfall, and tower tilt angle variation coefficient as examples, in the Level 1 acquisition frequency mode, wind speed (1Hz), rainfall (10 min / time), and tower tilt angle variation coefficient (5 min / time) are collected.

[0037] According to an embodiment of the present invention, the step of collecting feature data through a preset sensor according to a matching data acquisition mode includes environmental feature data, equipment feature data, and potential hazard feature data, specifically including: Acquire characteristic data, including environmental characteristic data, equipment characteristic data, and potential hazard characteristic data; Environmental characteristic data includes rainfall data and wind speed data; equipment characteristic data includes line current trend slope, tower tilt angle variation coefficient and line temperature; and hidden danger characteristic data includes line wind resistance level data, line service life data and standard service life data.

[0038] It should be noted that environmental characteristic data, equipment characteristic data, and potential hazard characteristic data can all cause faults in the distribution network and collectively affect its status. Therefore, to more comprehensively assess and identify fault points, analysis can be performed using the above data. The line current trend slope is an equipment characteristic parameter reflecting the degree of change in line current over time, used to determine whether abnormal current fluctuations exist in the line. It can be calculated using the sliding window analysis method. The tower tilt angle variation coefficient is a parameter that measures how quickly the tower's tilt changes over time, reflecting the tower's stability. Line wind resistance rating data indicates the level of wind resistance the line can withstand; the higher the wind resistance rating, the stronger the wind resistance capability.

[0039] Please refer to Figure 4 , Figure 4 This is a flowchart illustrating the process of obtaining characteristic impact data and dynamic weight coefficients in a distribution network fault identification method based on multi-source data, as provided in this application embodiment. According to this embodiment, the step of calculating the characteristic impact data and dynamic weight coefficients after normalizing the characteristic data specifically includes: S31. After normalizing the environmental characteristic data, equipment characteristic data, and hazard characteristic data, calculate the environmental impact data, equipment impact data, and hazard impact data. S32. After performing power transformations on the environmental impact data, equipment impact data, and hazard impact data respectively, obtain the environmental impact transformation value, equipment impact transformation value, and hazard impact transformation value; S33. Calculate the dynamic weighting coefficients based on the environmental impact conversion value, equipment impact conversion value, and hidden danger impact conversion value.

[0040] It should be noted that normalizing environmental characteristic data, equipment characteristic data, and hazard characteristic data is a crucial step in ensuring the fair and effective weighting of each feature during multi-source feature fusion. Environmental impact data, equipment impact data, and hazard impact data are calculated based on the normalized environmental characteristic data, equipment characteristic data, and hazard characteristic data, respectively, reflecting the degree of influence of these data on the power distribution network status. The environmental impact data can be calculated based on normalized rainfall and wind speed data, specifically: Environmental impact data = Wind speed characteristic coefficient × Normalized wind speed value + Rainfall. Characteristic coefficient × normalized value of rainfall data + corrected characteristic coefficient × normalized value of wind speed data × normalized value of rainfall data; Equipment impact data = characteristic coefficient of line current trend slope × normalized value of line current trend slope + characteristic coefficient of tower tilt angle change × normalized value of tower tilt angle change coefficient + characteristic coefficient of line temperature × normalized value of line temperature; Hazard impact data = characteristic coefficient of line wind resistance level / normalized value of line wind resistance level data + characteristic coefficient of age status × (line operating years data / standard operating years data); The characteristic coefficients mentioned above are preset after training based on historical data; Environmental impact data, equipment impact data, and hazard impact data... Loudness data consists of numbers between 0 and 1, representing standardized data. In this embodiment, a larger value among these three data points indicates a greater degree of influence, and thus amplifies the leading advantage. Therefore, a power-law transformation method is used to obtain the environmental impact transformation value, equipment impact transformation value, and hazard impact transformation value. In this embodiment, the k-th power of the environmental impact data is used as the environmental impact transformation value, the k-th power of the equipment impact data is used as the equipment impact transformation value, and the k-th power of the hazard impact data is used as the hazard impact transformation value. The value of k ranges from [1.2, 2] and can be selected according to actual needs. The dynamic weighting coefficient is obtained by calculating the environmental impact conversion value, equipment impact conversion value, and hazard impact conversion value. In this embodiment, the formula for calculating the dynamic weighting coefficient corresponding to the environmental impact data is: Environmental impact weighting coefficient = Environmental impact conversion value / (Environmental impact conversion value + Equipment impact conversion value + Hazard impact conversion value); Similarly, the dynamic weighting coefficient for equipment impact = Equipment impact conversion value / (Environmental impact conversion value + Equipment impact conversion value + Hazard impact conversion value); The dynamic weighting coefficient for hazard impact = Equipment impact conversion value / (Environmental impact conversion value + Equipment impact conversion value + Hazard impact conversion value).

[0041] In a preferred embodiment of the present invention, the calculation process of the characteristic coefficients includes: Historical data is collected to construct a sample library containing various features such as environment, equipment, and potential hazards, along with corresponding power grid state labels. Subsequently, a deep learning model integrating an attention mechanism (such as a neural network with an attention layer) is built and trained. This model automatically learns from the data and outputs an importance score for each feature's impact on the state through its attention weight layer, using this as the initial value for each feature coefficient (such as wind speed coefficient, tower tilt angle change coefficient, etc.). For further optimization, principal component analysis and other methods are used periodically to reduce the dimensionality of the accumulated feature data, identifying the core features that contribute the most to power grid state fluctuations. Finally, combining the initial weights provided by the attention mechanism with the key features identified by dimensionality reduction analysis, a set of optimized feature coefficient preset values ​​for different typical scenarios (such as deserts, coastal areas, and urban areas) is generated and fixed. In practical applications, the system calls the corresponding coefficient set for calculation based on the real-time scenario and can be continuously updated through online learning, thereby achieving dynamic and intelligent adjustment of feature weights. For example, in desert scenarios with high wind erosion, wind speed characteristics, pole tilt change characteristics, and wind resistance characteristics are enhanced, with their corresponding characteristic coefficients preset to 0.7, 0.8, and 0.8, respectively, while the corresponding rainfall characteristics and line temperature change characteristics are weakened.

[0042] Furthermore, those skilled in the art should understand that the method for calculating feature coefficients is not limited to the deep learning model based on the attention mechanism described above. Other methods can also be used, such as entropy weighting, analytic hierarchy process, or feature importance assessment based on tree models such as random forest and XGBoost, which can achieve similar technical objectives and are all within the scope of protection of this invention.

[0043] According to an embodiment of the present invention, the step of calculating the line risk value based on the characteristic influence data and dynamic weight coefficients using a weighted summation method specifically includes: The environmental impact data, equipment impact data, hazard impact data, and dynamic weighting coefficients are input into a preset line risk assessment model for processing to obtain the line risk value.

[0044] It should be noted that, in this embodiment, the formula for calculating the line risk value is: ; in, This represents the risk value of the line. For environmental impact data, For equipment impact data, For data on the impact of potential hazards, , , These are the dynamic weighting coefficients corresponding to environmental impact data, equipment impact data, and hazard impact data, respectively.

[0045] According to an embodiment of the present invention, the step of comparing the line risk value with a preset distribution network risk status assessment threshold to obtain the distribution network risk status level specifically includes: The first threshold and the second threshold are extracted based on the preset risk assessment threshold of the distribution network, and the first threshold is greater than the second threshold; Compare the line risk value with the first threshold and the second threshold; If the line risk value is less than or equal to the second threshold, the risk level of the distribution network is low risk. If the line risk value is greater than the second threshold and less than or equal to the first threshold, the risk level of the distribution network is medium risk. If the risk value of the line is greater than the first threshold, the risk level of the distribution network is high.

[0046] It should be noted that the higher the line risk value, the greater the possibility of a fault in the distribution network line. Therefore, the risk level of the distribution network can be assessed by comparing thresholds. In this embodiment, the preset distribution network risk assessment threshold can be obtained after training based on user needs and historical data.

[0047] In a preferred embodiment of the present invention, the potential hazards caused by different risk values ​​can be obtained through analysis of historical data. Power grid management or operation units can customize distribution network risk assessment thresholds based on line management requirements. In this embodiment, the preset distribution network risk assessment thresholds are set as follows: [0, 0.6] indicates a low risk level, (0.6, 0.8] indicates a medium risk level, and (0.8, 1] indicates a high risk level. When the line risk value is 0.87, the distribution network risk level is considered high.

[0048] According to an embodiment of the present invention, the step of sending the corresponding distribution network early warning information to the operation and maintenance monitoring platform based on the distribution network risk status level specifically includes: If the risk level of the distribution network is low, then the current situation will remain unchanged; If the risk level of the distribution network is medium risk, a medium-level early warning and the risk location will be sent to the operation and maintenance monitoring platform. If the risk level of the distribution network is high, a high-level warning, risk location, and maintenance priority will be sent to the operation and maintenance monitoring platform.

[0049] It should be noted that low risk indicates a low probability of failure, so the current situation can be maintained; medium risk indicates a possibility of failure, and a medium-level warning and risk location need to be sent to the operation and maintenance monitoring platform. The risk location is determined based on the location of the sensor at that point; high risk indicates a high risk that urgently needs to be addressed, and the maintenance priority is obtained through preset analysis methods.

[0050] In this embodiment, if the system receives multiple power grid risk status levels at the same time, it will sort them in the order of high risk and medium risk. After sorting, the high-risk ones will be prioritized for maintenance. If there are multiple high-risk ones, they will be sorted again according to maintenance priority among the high-risk levels, and the ones with higher priority will be maintained first.

[0051] It is worth mentioning that when collecting environmental feature data from the feature data, it also includes: Acquire environmental feature data of the current location and a preset number of environmental feature data of the same category within a preset distance; The status data of the current location's environmental characteristics, including whether it is normal or abnormal, is obtained through data consistency judgment; If the status data is normal, proceed to the next step; If the status data is abnormal, obtain the effective accuracy of the sensor corresponding to other environmental feature data within the preset distance; The weight data corresponding to other environmental feature data is obtained based on the effective accuracy. The corrected values ​​for environmental characteristic data are obtained by weighting other environmental characteristic data and corresponding weight data; Change the abnormal environmental feature data of the current location to the corrected environmental feature data value.

[0052] It should be noted that during typhoons and heavy rainstorms, due to extreme weather conditions and the sensor's own performance, the sensor may sometimes collect abnormal data. Therefore, to better ensure the accuracy of the collected data, when collecting environmental feature data, after obtaining the current environmental feature data, a preset number of environmental feature data of the same category within a preset distance are also collected. In this embodiment, taking wind speed as an example, after collecting the wind speed data at point A, wind speed data from two other wind speed sensors within 200 meters are also collected simultaneously. If all the collected data are normal, the next step is executed; if the status data is abnormal, it indicates the presence of abnormal data, and the possibility of simultaneous sensor failure is extremely small.

[0053] In a preferred embodiment of the present invention, it is assumed that at most one of the three sensors may exhibit an anomaly, and a consistency judgment method is used to determine whether the data is normal or abnormal. The consistency judgment method is as follows: First, a preliminary screening is performed based on the physical principles of wind speed and sensor parameters, eliminating data that does not conform to these principles and marking it as abnormal data. Examples include wind speed values ​​exceeding the sensor's range (e.g., 0-60 m / s); wind speed values ​​showing negative values ​​or obviously unreasonable abrupt changes; and wind speed change rates exceeding physical limits (e.g., ±20 m / s / s) between adjacent time points.

[0054] Secondly, consistency is carefully judged. Through "pairwise comparison + majority principle", if two of the three data are very close (the difference is less than the set threshold, such as 1m / s) and one is significantly deviated (the difference is greater than the threshold), then the significantly deviated one is an outlier.

[0055] When the results of the coarse screening and the fine consistency evaluation of physical rules are inconsistent, the following priority shall be applied: Physical rule priority principle: If a physical rule determines that a certain data is abnormal, even if the consistency detail judgment considers the data to be normal, the physical rule shall prevail and the data shall be marked as abnormal. Additional principle: If no anomalies are found during the initial screening by physical rules, but anomalies are identified during the consistency review, then the data is marked as an anomaly. Double confirmation mechanism: If no abnormalities are found in both the coarse screening of physical rules and the fine consistency judgment, then the data is normal; Unable to determine: If the differences between any of the three data points are too large to be determined by the majority rule, they are marked as "pending confirmation" and require manual intervention or further analysis in conjunction with historical data.

[0056] The above measures can accurately locate the current position status data. In this embodiment, the effective accuracy rate refers to the average accuracy rate of the sensor over the past 4 months. Let the historical accuracy rates of the sensor corresponding to the two normal values ​​be a1 and a2 (accuracy range 0-1), then the corresponding weights are w1=a1 / (a1+a2) and w2=a2 / (a1+a2), and the sum of the weights of w1 and w2 is 1; the environmental feature data correction value x'=w1x1+w2x2 (x1 and x2 are the sensor acquisition values ​​corresponding to the two normal values).

[0057] Through the above four-step logic, the rigid constraints of physical rules are guaranteed, the advantages of multi-sensor redundancy are utilized, and a clear conflict resolution mechanism is established to ensure the accuracy and reliability of data quality judgment.

[0058] It is worth mentioning that it also includes: If the distribution network risk level is medium or high at multiple locations simultaneously on the operation and maintenance monitoring platform; Data categorizing risk locations includes priority protection areas, densely populated residential areas, industrial and commercial areas, or other non-critical areas; The processing priority is then ranked in the following order: priority protection area, residential area, industrial and commercial area, and other non-critical area. The optimal path is determined using a pre-defined ant colony algorithm, and the path planning results are dynamically adjusted based on real-time updated traffic data.

[0059] It should be noted that in this embodiment, the priority protection area includes key public service units such as hospitals, nursing homes, and emergency command centers; the residential area includes areas affecting 400 or more households; the industrial and commercial area includes industrial parks and commercial areas; and other non-critical areas include scattered residents or non-critical facilities. The preset ant colony algorithm is trained based on the road conditions in this area.

[0060] This invention combines a dynamic matching data acquisition mode for weather warning levels. It collects multi-dimensional feature data of the environment, equipment, and potential hazards through preset sensors. After normalization processing, it calculates the feature impact data and dynamic weight coefficients, and then obtains the line risk value through a weighted summation method. The risk level is then classified by comparing it with a preset threshold and the corresponding warning is sent to the operation and maintenance monitoring platform. Through dynamic fusion of multi-source data, dynamic weight adjustment, and a hierarchical warning mechanism, it achieves accurate identification and timely warning of distribution network fault risks, solving the problems of single-factor consideration and delayed response in existing technologies.

[0061] Example 2 This invention also proposes a distribution network fault identification system based on multi-source data, including a matching data acquisition mode determination module, a feature data acquisition module, a line risk value calculation module, a distribution network risk status level acquisition module, and an execution module: The matching data acquisition mode determination module acquires weather warning level data, queries a preset warning acquisition mode mapping table based on the weather warning level data, and obtains the matching data acquisition mode. The feature data acquisition module collects feature data in the power distribution network line through preset sensors according to the matching data acquisition mode. The feature data includes environmental feature data, equipment feature data, and potential hazard feature data. The line risk value calculation module calculates feature impact data and dynamic weight coefficients based on the normalized feature data; the feature impact data includes environmental impact data, equipment impact data, and hidden danger impact data; and obtains the line risk value by weighted summation based on the feature impact data and the dynamic weight coefficients. The distribution network risk status level acquisition module compares the line risk value with a preset distribution network risk status assessment threshold to obtain the distribution network risk status level. The execution module sends the corresponding early warning information of the distribution network to the operation and maintenance monitoring platform according to the risk status level of the distribution network.

[0062] Example 3 The present invention also proposes a terminal, including a processor and a storage medium: The storage medium is used to store instructions; The processor is configured to operate according to the instructions to perform the steps of the method according to any one of Embodiment 1.

[0063] Example 4 The present invention also proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in any one of the embodiments.

[0064] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. A power distribution network fault identification method based on multi-source data, characterized in that, include: Obtain weather warning level data, and query the preset warning collection mode mapping table based on the weather warning level data to obtain the matching data collection mode; According to the matching data acquisition mode, feature data is collected in the power distribution network line by preset sensors. The feature data includes environmental feature data, equipment feature data and hidden danger feature data. Based on the normalized feature data, feature impact data and dynamic weight coefficients are calculated; the feature impact data includes environmental impact data, equipment impact data, and hazard impact data. The line risk value is obtained by weighted summation based on the characteristic impact data and the dynamic weight coefficient. The risk value of the line is compared with the preset risk assessment threshold of the distribution network to obtain the risk level of the distribution network. Based on the risk level of the distribution network, the corresponding early warning information of the distribution network will be sent to the operation and maintenance monitoring platform.

2. The method for distribution network fault identification based on multi-source data according to claim 1, characterized in that: The steps for obtaining the weather warning level are as follows: The real-time monitored wind force level and rainfall amount are mapped to risk score values ​​respectively; Assign weights to the risk score values ​​corresponding to the wind force level and the rainfall, and calculate the comprehensive risk score; The final weather warning level is determined by comparing the comprehensive risk score with a preset threshold range.

3. The method for distribution network fault identification based on multi-source data according to claim 1, characterized in that: The method for calculating the feature influence data is as follows: Environmental impact data = wind speed characteristic coefficient × normalized wind speed data + rainfall characteristic coefficient × normalized rainfall data + correction characteristic coefficient × normalized wind speed data × normalized rainfall data; Equipment impact data = Line current trend slope characteristic coefficient × Normalized value of line current trend slope + Tower tilt angle change characteristic coefficient × Normalized value of tower tilt angle change coefficient + Line temperature characteristic coefficient × Normalized value of line temperature. Hazard impact data = Line wind resistance level characteristic coefficient / Normalized value of line wind resistance level data + Annual condition characteristic coefficient × (Line operating years data / Standard operating years data) All of the aforementioned characteristic coefficients are obtained through a preset process.

4. The method for distribution network fault identification based on multi-source data according to claim 1, characterized in that: The calculation method for the dynamic weighting coefficient is as follows: The environmental impact data raised to the power of k is used as the environmental impact conversion value, the equipment impact data raised to the power of k is used as the equipment impact conversion value, and the hidden danger impact data raised to the power of k is used as the hidden danger impact conversion value. The environmental impact conversion value, the equipment impact conversion value, and the hazard impact conversion value are normalized to obtain the dynamic weight coefficients corresponding to the environmental impact data, the equipment impact data, and the hazard impact data.

5. The method for distribution network fault identification based on multi-source data according to claim 1, characterized in that: The step of comparing the line risk value with a preset distribution network risk status assessment threshold to obtain the distribution network risk status level is as follows: The first threshold and the second threshold are extracted based on the preset risk assessment threshold of the distribution network, and the first threshold is greater than the second threshold; Compare the line risk value with the first threshold and the second threshold; If the line risk value is less than or equal to the second threshold, the risk level of the distribution network is low risk. If the line risk value is greater than the second threshold and less than or equal to the first threshold, the risk level of the distribution network is medium risk. If the risk value of the line is greater than the first threshold, the risk level of the distribution network is high.

6. The method for distribution network fault identification based on multi-source data according to claim 1, characterized in that: The specific steps for sending the corresponding distribution network early warning information to the operation and maintenance monitoring platform based on the distribution network risk status level are as follows: If the risk level of the distribution network is low, then the current situation will remain unchanged; If the risk level of the distribution network is medium risk, a medium-level early warning and the risk location will be sent to the operation and maintenance monitoring platform. If the risk level of the distribution network is high, a high-level warning, risk location, and maintenance priority will be sent to the operation and maintenance monitoring platform.

7. The method for distribution network fault identification based on multi-source data according to claim 1, characterized in that: The environmental feature data collection steps are as follows: Acquire environmental feature data of the current location and a preset number of environmental feature data of the same category within a preset distance; Determine the status data of the environmental characteristics obtained at the current location, including whether it is normal or abnormal; If the status data is normal, proceed to the next step; If the status data is abnormal, obtain the effective accuracy of the sensor corresponding to other environmental feature data within the preset distance; The weight data corresponding to other environmental feature data is obtained based on the effective accuracy. The corrected values ​​for environmental characteristic data are obtained by weighting other environmental characteristic data and corresponding weight data; Change the abnormal environmental feature data of the current location to the corrected environmental feature data value.

8. A distribution network fault identification system based on multi-source data using the method described in any one of claims 1-7, comprising a matching data acquisition mode determination module, a feature data acquisition module, a line risk value calculation module, a distribution network risk status level acquisition module, and an execution module, characterized in that: The matching data acquisition mode determination module acquires weather warning level data, queries a preset warning acquisition mode mapping table based on the weather warning level data, and obtains the matching data acquisition mode. The feature data acquisition module collects feature data in the power distribution network line through preset sensors according to the matching data acquisition mode. The feature data includes environmental feature data, equipment feature data, and potential hazard feature data. The line risk value calculation module calculates feature impact data and dynamic weight coefficients based on the normalized feature data; the feature impact data includes environmental impact data, equipment impact data, and hidden danger impact data; and obtains the line risk value by weighted summation based on the feature impact data and the dynamic weight coefficients. The distribution network risk status level acquisition module compares the line risk value with a preset distribution network risk status assessment threshold to obtain the distribution network risk status level. The execution module sends the corresponding early warning information of the distribution network to the operation and maintenance monitoring platform according to the risk status level of the distribution network.

9. A terminal, comprising a processor and a storage medium; characterized in that: The storage medium is used to store instructions; The processor is configured to operate according to the instructions to perform the steps of the method according to any one of claims 1-7.

10. A computer readable storage medium having stored thereon a computer program, characterized in that, When executed by a processor, the program implements the steps of the method according to any one of claims 1-7.

Images