Distributed intelligent line selection depth evaluation method for ground fault of power distribution network

Abstract

The invention relates to a power distribution network ground fault distributed intelligent line selection depth evaluation method, and belongs to the technical field of power distribution network fault detection and analysis. The method comprises the following steps: collecting power distribution network line data, distributed power supply data, novel power system data, environment data and edge node data, identifying the ground fault condition of each line in the power distribution network, judging the detection requirement of sharing ground fault characteristic data by edge nodes after topology change, and determining the ground fault characteristic data according to the detection requirement. And calculating a ground fault feature data accuracy degree detection coefficient, constructing a data accuracy degree calibration model, further calibrating the ground fault feature data accuracy degree detection coefficient, and generating a distributed intelligent line selection data report. The problems of fault current distortion, difference between edge node data sampling frequency and precision and low feature data precision influenced by topological change caused by distributed power supply access in a traditional method are solved, and the intelligent degree of a power distribution network ground fault distributed intelligent line selection process is remarkably enhanced.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of power distribution network fault detection and analysis, and particularly relates to a power distribution network grounding fault distributed intelligent line selection and depth evaluation method. BACKGROUND

[0002] In a power system, as a key link connecting users and the main network, the safe and stable operation of the power distribution network is directly related to the normal order of social production and life. However, the power distribution network has a complex structure, numerous branches, and is often in an outdoor environment, which is prone to single-phase grounding faults caused by lightning strikes, equipment aging, external damage and other factors. If the fault cannot be quickly and accurately selected and eliminated, it may lead to the expansion of the fault, causing phase-to-phase short circuit, equipment burning and even power outage accidents, causing huge economic losses. Traditional grounding fault line selection methods mostly rely on centralized monitoring devices to analyze the electrical quantity signals collected at the bus of the substation. However, the power distribution network has complex conditions such as asymmetric line parameters, weak fault signals, and arc grounding, and the centralized method is easily affected by signal attenuation and interference, resulting in low accuracy of distributed line selection for power distribution network grounding faults. Therefore, it is urgent to establish a scientific depth evaluation method to quantify its performance under different network structures and fault types, provide a basis for technical optimization and engineering selection, and become a research hotspot in the field of power distribution network automation.

[0003] The traditional power distribution network grounding fault distributed intelligent line selection technology mainly adopts the way of edge node on-site sensing analysis and regional cooperation for distributed intelligent line selection. However, in the process of distributed intelligent line selection based on the above method, the access of distributed power supply may cause distortion of the grounding fault current. At the same time, the existing technology is difficult to comprehensively monitor the sampling frequency and accuracy of the edge node shared grounding fault characteristic data, causing the problem of inconsistent sampling frequency and low accuracy of the edge node shared grounding fault characteristic data. Due to the output of distributed power supply, there are factors affecting the operation state of the power distribution network and related environmental factors, which may cause changes in the edge node shared grounding fault characteristic data, resulting in low accuracy of the edge node shared grounding fault characteristic data, and further affecting the accuracy of the power distribution network grounding fault distributed intelligent line selection. Therefore, the power distribution network grounding fault distributed intelligent line selection depth evaluation method is proposed. SUMMARY

[0004] The purpose of the present application is to provide a power distribution network grounding fault distributed intelligent line selection depth evaluation method, which solves the problem of traditional methods caused by the access of distributed power supply, such as fault current distortion, edge node data sampling frequency and accuracy difference, and low accuracy of characteristic data affected by topology change, and significantly enhances the intelligent degree of the power distribution network grounding fault distributed intelligent line selection process.

[0005] To achieve the above objectives, the technical solution of the present invention is: a distributed intelligent fault location depth assessment method for grounding faults in distribution networks, comprising the following steps:

[0006] S1. Collect data on distribution network lines, distributed power sources, new power systems, environmental data, and edge node data;

[0007] S2. Based on the collected distribution network line data, and combined with the weighted summation method, calculate the ground fault coefficient of each line in the distribution network and identify the ground fault situation of each line in the distribution network.

[0008] S3, combining decision tree algorithm, edge computing technology, distributed communication protocol, time synchronization technology, and random forest algorithm, determines the detection requirements for shared grounding fault feature data of edge nodes after topology changes. The steps include:

[0009] S31. Integrating new power system data and edge node data, calculate the system clock synchronization error and sampling period deviation of each edge node in the distribution network. Using the decision tree algorithm, construct a data sampling frequency analysis model, output the sampling frequency evaluation coefficient of ground fault feature data, and obtain the sampling frequency of shared ground fault feature data of edge nodes.

[0010] S32. Based on the sampling frequency of the shared grounding fault characteristic data of edge nodes, combined with acquisition equipment, edge computing technology, distributed communication protocol and time synchronization technology, acquire the shared grounding fault characteristic data of edge nodes before and after topology change and the grounding fault current data of distribution network before and after topology change;

[0011] S33. Using the random forest algorithm, a ground fault current distortion detection model for the distribution network is constructed, and the ground fault current distortion coefficient is output to correct the ground fault current data of the distribution network. Combined with the weighted average method, the edge node sharing consistency is calculated to determine the detection requirements of the edge node sharing ground fault feature data after the topology change.

[0012] S4. Using a combination of multiple linear regression and weighted summation methods, calculate the accuracy detection coefficient of ground fault feature data. Then, employ a neural network algorithm to obtain the accuracy calibration coefficient of the ground fault feature data. The calibration of the accuracy detection coefficient of the ground fault feature data includes the following steps:

[0013] S41. Using edge node data and a multiple linear regression algorithm, a data accuracy analysis model is constructed to calculate the accuracy of the shared grounding fault characteristic data of edge nodes;

[0014] S42. Based on the accuracy of the shared ground fault feature data of the edge nodes, and combined with the consistency between the corrected distribution network ground fault current data and the edge node sharing, the weighted summation method is used to calculate the detection coefficient of the accuracy of the ground fault feature data.

[0015] S43. Based on distributed power source data and new power system data, calculate the active power fluctuation and reactive power fluctuation of distributed power sources and the proportion of distributed power source access capacity respectively. Combine environmental data and neural network algorithms to construct a data accuracy calibration model, output the accuracy calibration coefficient of ground fault feature data, and calibrate the accuracy detection coefficient of ground fault feature data.

[0016] S5. Based on the accuracy of the data, calibrate the output of the model, evaluate the accuracy of the distributed intelligent fault location for grounding faults in the distribution network adapted to the new power system, and generate a distributed intelligent fault location data report by using edge-side data fusion and standardized protocol technology.

[0017] Furthermore, the process of collecting the distribution network line data, distributed power source data, new power system data, environmental data, and edge node data includes:

[0018] Different types of data acquisition devices are deployed, combined with data entry technology, to collect data on power distribution lines, distributed power sources, new power systems, environmental data, and edge nodes. The data acquisition devices include broadband zero-sequence current transformers, zero-sequence voltage transformers, high-precision transient waveform recording devices, broadband current transformers, broadband voltage transformers, voltage sensors, multi-functional power meters, front-end sensing devices, edge computing gateways, BeiDou dual-mode time synchronization devices, temperature and humidity sensors, tipping bucket rain gauges, cup-type anemometers, soil resistivity meters, electromagnetic interference testers, signal analyzers, sampling controllers, power quality analyzers, and ultra-high frequency current sensors.

[0019] The distribution network line data includes the zero-sequence current, zero-sequence current waveform and its harmonic content, zero-sequence voltage, transient current and transient voltage of each line in the distribution network; the distributed power source data includes the grid connection point voltage, active power, reactive power, total capacity and current limiting threshold of the inverter output current of the distributed power source; the new power system data includes the total load and reference clock of the new power system; the environmental data includes the real-time temperature, real-time humidity, rainfall intensity, wind speed, soil resistivity, high-frequency electromagnetic interference intensity and communication channel noise of the distribution network ground fault environment; the edge node data includes the sampling frequency and sampling clock of each edge node in the distribution network, as well as the bit depth, range and cutoff frequency of the AD converter;

[0020] The collected distribution network line data, distributed power source data, new power system data, environmental data, and edge node data are cleaned and standardized. Timestamps are assigned to each of these data sets. By adjusting the timestamps, the collection time of the distribution network line data, distributed power source data, new power system data, environmental data, and edge node data is synchronized. The preprocessed distribution network line data, distributed power source data, new power system data, environmental data, and edge node data are then integrated to generate a distributed intelligent fault location dataset for distribution network grounding faults.

[0021] Furthermore, the process of calculating the ground fault coefficient of each line in the distribution network and identifying the ground fault status of each line in the distribution network includes:

[0022] Using data from the distribution network lines, calculate the average zero-sequence current, average zero-sequence current harmonic content, average zero-sequence voltage, average transient current, and average transient voltage for each line in the distribution network.

[0023] By measuring the zero-sequence current, its waveform and harmonic content, and the proportions of zero-sequence voltage, transient current, and transient voltage in the average zero-sequence current, average zero-sequence current harmonic content, average zero-sequence voltage, average transient current, and average transient voltage of each line in the distribution network, the abnormal values ​​of zero-sequence current, zero-sequence current harmonic content, zero-sequence voltage, transient current, and transient voltage of each line in the distribution network are obtained.

[0024] The criteria for determining ground faults for each line in the distribution network are set, and the weights are assigned to the abnormal values ​​of zero-sequence current, zero-sequence current harmonic content, zero-sequence voltage, transient current, and transient voltage for each line in the distribution network. The ground fault coefficient of each line in the distribution network is calculated by combining the weighted summation method.

[0025] Based on the grounding fault coefficient of each line in the distribution network, the grounding fault status of each line in the distribution network is identified. The identification of the grounding fault status of each line in the distribution network includes the occurrence of grounding faults and the absence of grounding faults. The lines in the distribution network that have grounding faults are called grounding fault lines, and the lines in the distribution network that have not grounding faults are called non-grounding fault lines.

[0026] Furthermore, the process of obtaining the sampling frequency of the shared ground fault characteristic data of the edge nodes includes:

[0027] The system clock synchronization error of each edge node in the distribution network is calculated by using the difference between the total load of the new power system and the sampling clock of each edge node in the distribution network.

[0028] The sampling period deviation of each edge node in the distribution network is calculated by the absolute value of the difference between the sampling clocks of each edge node in the distribution network, and the system clock synchronization error and sampling period deviation of each edge node in the distribution network are integrated into the distributed intelligent fault location dataset of the distribution network.

[0029] The sampling frequency of each edge node in the distribution network and the system clock synchronization error and sampling period deviation of each edge node in the distributed intelligent fault location dataset of the distribution network are extracted. The extracted data is divided into the first training set and the first test set.

[0030] Using the decision tree algorithm, the first training set data is set as input, and the sampling frequency evaluation coefficient of the ground fault feature data is set as output. The nonlinear relationship between each item of the first training set data and the sampling frequency evaluation coefficient of the ground fault feature data is learned, and the data sampling frequency analysis model is trained.

[0031] The first test set data is input into the data sampling frequency analysis model. The SGD optimizer is used to adjust the parameters of the data sampling frequency analysis model, optimize the performance of the data sampling frequency analysis model, and obtain the final data sampling frequency analysis model.

[0032] Based on the sampling frequency of each edge node in the current distribution network and the system clock synchronization error and sampling period deviation of each edge node in the distribution network, the corresponding ground fault characteristic data sampling frequency evaluation coefficient is output.

[0033] Based on the evaluation coefficient of the ground fault characteristic data sampling frequency, the corresponding sampling frequency is assigned to the edge nodes to share the ground fault characteristic data.

[0034] Furthermore, the process of acquiring the edge node shared ground fault characteristic data and distribution network ground fault current data before and after the topology change includes:

[0035] The shared ground fault characteristic data of edge nodes before and after the topology change includes the zero-sequence current, voltage amplitude, high-frequency component of transient current and harmonic content of the ground fault line before and after the topology change, as well as the phase difference of the zero-sequence current with the unground fault line; the ground fault current data of the distribution network includes the real-time ground fault current of each line in the distribution network and its fundamental amplitude, harmonic amplitude and ultra-high frequency component.

[0036] Based on the sampling frequency of the shared ground fault characteristic data of edge nodes, the zero-sequence current, voltage amplitude, high-frequency component of transient current, harmonic content and zero-sequence current phase of the ground fault line before and after the topology change, the zero-sequence current phase of the unground fault line, and the real-time ground fault current and its ultra-high frequency component of each line in the distribution network are collected using the acquisition equipment.

[0037] The zero-sequence current phase difference between the ground fault line and the unground fault line before and after the topology change is obtained by using the absolute value of the difference between the zero-sequence current phase of the ground fault line before the topology change and the absolute value of the difference between the zero-sequence current phase of the ground fault line and the unground fault line after the topology change.

[0038] Using edge computing technology and the FFT algorithm, the fundamental and harmonic amplitudes of the real-time ground fault current of each line in the distribution network are extracted, and the shared ground fault feature data of the edge nodes before and after the topology change and the ground fault current data of the distribution network are obtained.

[0039] Based on a distributed communication protocol, the communication list is automatically updated when the topology changes. Combined with time synchronization technology, the edge nodes share ground fault characteristic data and distribution network ground fault current data before and after the topology change.

[0040] The ground fault current data of the distribution network is integrated into the distributed intelligent fault location dataset of the distribution network.

[0041] Furthermore, the process of correcting the distribution network ground fault current data and determining the detection requirements for shared ground fault characteristic data of edge nodes after topology changes includes:

[0042] The zero-sequence current, zero-sequence current waveform and harmonic content, and ground fault current data of each line in the distribution network are extracted from the distributed intelligent line selection dataset of the distribution network ground fault. The extracted data are divided into a second training set and a second test set.

[0043] Using the random forest algorithm, the second training set data is used as input and the ground fault current distortion coefficient is used as output to learn the nonlinear relationship between the second training set data and the ground fault current distortion coefficient, thereby training the distribution network ground fault current distortion detection model.

[0044] The second test set data is input into the distribution network ground fault current distortion detection model. The Adam optimizer is used to adjust the parameters of the distribution network ground fault current distortion detection model, optimize the performance of the distribution network ground fault current distortion detection model, and obtain the final distribution network ground fault current distortion detection model. Combining the zero-sequence current and zero-sequence current waveform and its harmonic content of each line in the current distribution network, as well as the distribution network ground fault current data, the corresponding ground fault current distortion coefficient is output.

[0045] Based on the ground fault current distortion coefficient, the real-time ground fault current of each line in the distribution network is calibrated to obtain the real-time ground fault current of each line in the distribution network after calibration, as well as its fundamental amplitude, harmonic amplitude and ultra-high frequency component. Using the real-time ground fault current of each line in the distribution network after calibration, as well as its fundamental amplitude, harmonic amplitude and ultra-high frequency component, the ground fault current data of the distribution network is updated to obtain the calibrated ground fault current data of the distribution network.

[0046] By combining the edge node shared grounding fault characteristic data before and after topology change with the weighted average method, the edge node shared consistency degree is calculated;

[0047] Based on the edge node sharing consistency, the detection requirement of edge node shared grounding fault feature data after topology change is determined. The detection requirement of edge node shared grounding fault feature data after topology change includes whether there is a detection requirement or not.

[0048] Furthermore, the process of constructing a data accuracy analysis model and calculating the accuracy of shared grounding fault characteristic data of edge nodes includes:

[0049] Extract the bit depth, range, and cutoff frequency of the AD converters of each edge node in the distribution network from the distributed intelligent fault location dataset of the distribution network grounding fault, and convert the extracted data into a third training set and a third test set;

[0050] The multivariate linear regression algorithm is used, with the third training set data as input and the accuracy of the edge node shared grounding fault feature data as output. The linear relationship between the accuracy of each item of the third training set data and the accuracy of the edge node shared grounding fault feature data is learned, and the data accuracy analysis model is trained.

[0051] The third test set data is input into the data accuracy analysis model. The regression coefficients and intercept terms of the data accuracy analysis model are adjusted to optimize the performance of the data accuracy analysis model and obtain the final data accuracy analysis model. Combined with the bit depth, range, and cutoff frequency of the AD converters of each edge node in the current distribution network, the accuracy of the corresponding edge node shared grounding fault characteristic data is output.

[0052] Furthermore, the process of calculating the accuracy detection coefficient of ground fault characteristic data includes:

[0053] Weights are assigned to the accuracy of the shared ground fault characteristic data of edge nodes, the corrected distribution network ground fault current data, and the shared consistency of edge nodes, respectively.

[0054] The weighted summation method is used to calculate the detection coefficient for the accuracy of ground fault characteristic data. The calculation formula is as follows:

[0055] τ0=w1σ1+w2t2+w3t3+w4t4+w5t5+w6α

[0056] Wherein, τ0 is the detection coefficient for the accuracy of ground fault feature data, σ1, t2, t3, t4, t5 and α are the accuracy of the ground fault feature data shared by edge nodes, the real-time ground fault current and its fundamental amplitude, harmonic amplitude and ultra-high frequency component of each line in the calibrated distribution network and the edge node sharing consistency, respectively, and w1, w2, w3, w4, w5 and w6 are the weights of the accuracy of the ground fault feature data shared by edge nodes, the real-time ground fault current and its fundamental amplitude, harmonic amplitude and ultra-high frequency component of each line in the calibrated distribution network and the edge node sharing consistency, respectively.

[0057] Furthermore, the process of constructing a data accuracy calibration model, outputting ground fault feature data accuracy calibration coefficients, and then calibrating the ground fault feature data accuracy detection coefficients includes:

[0058] Select time t1 and time t2, and extract the active power and reactive power of the distributed power source corresponding to time t1 and time t2 respectively from the distributed power source data;

[0059] Based on the active and reactive power of the distributed generation at times t1 and t2 respectively, the active power fluctuation and reactive power fluctuation of the distributed generation are calculated respectively; the calculation process is as follows:

[0060]

[0061] Where, Δp y and Δp w These represent the active power fluctuation and reactive power fluctuation of the distributed power source, respectively; p y1 and p y2 Let p be the active power of the distributed generation at times t1 and t2, respectively; w1 and p w2 Let t1 and t2 be the reactive power of the distributed generation corresponding to time t1 and t2 respectively; t1 and t2 are selected times.

[0062] The proportion of distributed power generation capacity in the total load of the new power system is used to calculate the proportion of distributed power generation access capacity. The calculated active power fluctuation and reactive power fluctuation of distributed power generation, as well as the proportion of distributed power generation access capacity, are then integrated into the distributed intelligent fault location dataset for distribution network grounding faults.

[0063] Environmental data, active power fluctuation and reactive power fluctuation of distributed power sources, the proportion of distributed power source access capacity, and the grid connection point voltage and current limiting threshold of the inverter output current of distributed power sources are extracted from the distributed intelligent fault location dataset of the distribution network. The extracted data is divided into the fourth training set and the fourth test set.

[0064] A neural network algorithm is used, with the fourth training set data as input and the ground fault feature data accuracy calibration coefficient as output. The nonlinear relationship between each of the fourth training set data and the ground fault feature data accuracy calibration coefficient is learned to train the data accuracy calibration model.

[0065] The fourth test set data is input into the data accuracy calibration model, the parameters of the data accuracy calibration model are adjusted, the performance of the data accuracy calibration model is optimized, and the final data accuracy calibration model is obtained.

[0066] Based on current environmental data, the active and reactive power fluctuations of distributed power sources, the proportion of distributed power source access capacity, and the grid connection point voltage and current limiting threshold of the inverter output current of distributed power sources, the corresponding ground fault characteristic data accuracy calibration coefficient is output.

[0067] Based on the accuracy calibration coefficient of ground fault feature data, the accuracy detection coefficient of ground fault feature data is calibrated to obtain the calibrated accuracy detection coefficient of ground fault feature data.

[0068] Furthermore, the process of assessing the accuracy of distributed intelligent fault location for grounding faults in distribution networks adapted to new power systems and generating distributed intelligent fault location data reports includes:

[0069] The accuracy of the model's output is evaluated by calibrating the model's output based on the accuracy of the data, thus assessing the accuracy of distributed intelligent fault location for distribution network grounding faults adapted to the new power system.

[0070] By employing edge-side data fusion and standardized protocol technology, this method integrates the grounding fault status of each line in the distribution network, the detection requirements of edge nodes sharing grounding fault characteristic data after topology changes, and the accuracy of distributed intelligent line selection for grounding faults in the distribution network adapted to the new power system, and generates distributed intelligent line selection data reports.

[0071] Compared to existing technologies, this invention has the following advantages: The distributed intelligent line selection depth assessment method for distribution network grounding faults in this invention, compared to traditional methods, closely integrates edge computing technology, distributed communication protocols, time synchronization technology, decision tree algorithms, random forest algorithms, grounding fault current distortion detection technology, and accuracy analysis and calibration technology with modern information technology. This accurately captures multi-dimensional data such as distribution network line data, distributed power source data, new power system data, environmental data, and edge node data, achieving precise identification of grounding fault lines in the distribution network. Furthermore, by constructing a data sampling frequency analysis model and grounding... The fault current distortion detection model and data accuracy analysis model enable effective monitoring of fault characteristic data sampling frequency, current distortion, and data accuracy. This solves the problems in traditional methods, such as fault current distortion caused by distributed power source access, differences in data sampling frequency and accuracy at edge nodes, and the impact of topology changes on the accuracy of distributed intelligent fault location in distribution networks. This ensures that the method in this invention can refine the dynamic monitoring standards of the distributed intelligent fault location depth assessment method for distribution network grounding faults within a more precise range, making the monitored data more accurate indicators under the same conditions. The development and application of this method significantly enhances the intelligence level in the distributed intelligent fault location depth assessment process for distribution network grounding faults. Attached Figure Description

[0072] Figure 1 This is a flowchart of the distributed intelligent fault location depth assessment method for grounding faults in distribution networks according to the present invention. Detailed Implementation

[0073] The technical solution of the present invention will now be described in detail with reference to the accompanying drawings.

[0074] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0075] like Figure 1 As shown, this invention provides a distributed intelligent fault location depth assessment method for grounding faults in distribution networks, comprising the following steps:

[0076] S1. Collect data on distribution network lines, distributed power sources, new power systems, environmental data, and edge nodes to provide a data basis for subsequent steps.

[0077] S2. Based on the collected distribution network line data, combined with the weighted summation method, the grounding fault coefficient of each line in the distribution network is calculated, thereby identifying the grounding fault status of each line in the distribution network, providing a basis for generating distributed intelligent line selection data reports.

[0078] S3, combining decision tree algorithm, edge computing technology, distributed communication protocol, time synchronization technology, and random forest algorithm, determines the detection requirements for shared grounding fault feature data of edge nodes after topology changes. The steps include:

[0079] S31. Integrating new power system data and edge node data, calculate the system clock synchronization error and sampling period deviation of each edge node in the distribution network. Use decision tree algorithm to construct a data sampling frequency analysis model, output the sampling frequency evaluation coefficient of ground fault feature data, and then obtain the sampling frequency of shared ground fault feature data of edge nodes. This provides a basis for the acquisition of shared ground fault feature data of edge nodes before and after subsequent topology changes, as well as ground fault current data of distribution network.

[0080] S32. Based on the sampling frequency of the shared grounding fault characteristic data of edge nodes, combined with acquisition equipment, edge computing technology, distributed communication protocol and time synchronization technology, acquire the shared grounding fault characteristic data of edge nodes before and after topology change and the grounding fault current data of distribution network before and after topology change;

[0081] S33. Using the random forest algorithm, a ground fault current distortion detection model for the distribution network is constructed, and the ground fault current distortion coefficient is output. Then, the ground fault current data of the distribution network is corrected. Combined with the weighted average method, the edge node sharing consistency is calculated, and the detection requirements of the edge node shared ground fault feature data after the topology change are determined, which provides a basis for generating distributed intelligent line selection data reports.

[0082] S4. Using a combination of multiple linear regression and weighted summation methods, calculate the accuracy detection coefficient of ground fault feature data. Then, employ a neural network algorithm to obtain the accuracy calibration coefficient of the ground fault feature data, and finally calibrate the accuracy detection coefficient of the ground fault feature data. The steps include:

[0083] S41. Using edge node data and a multiple linear regression algorithm, a data accuracy analysis model is constructed to calculate the accuracy of the shared grounding fault characteristic data of edge nodes;

[0084] S42. Based on the accuracy of the shared ground fault feature data of the edge nodes, and combined with the consistency between the corrected distribution network ground fault current data and the edge node sharing, the weighted summation method is used to calculate the detection coefficient of the accuracy of the ground fault feature data.

[0085] S43. Based on distributed power source data and new power system data, the active power fluctuation and reactive power fluctuation of distributed power sources, as well as the proportion of distributed power source access capacity, are calculated respectively. Combining environmental data and neural network algorithms, a data accuracy calibration model is constructed, and the ground fault feature data accuracy calibration coefficient is output. Then, the ground fault feature data accuracy detection coefficient is calibrated, which improves the accuracy of the ground fault feature data accuracy detection coefficient and provides a basis for generating distributed intelligent line selection data reports.

[0086] S5. Based on the accuracy of the data, calibrate the output of the model, evaluate the accuracy of the distributed intelligent fault location for grounding faults in the distribution network adapted to the new power system, and generate a distributed intelligent fault location data report by using edge-side data fusion and standardized protocol technology.

[0087] In step S1, the process of collecting data from distribution network lines, distributed power sources, new power systems, environmental data, and edge nodes includes:

[0088] Deploy different types of data acquisition equipment, combined with data entry technology, to collect data on power distribution lines, distributed power sources, new power systems, environmental data, and edge nodes. The data acquisition equipment includes broadband zero-sequence current transformers, zero-sequence voltage transformers, high-precision transient waveform recording devices, broadband current transformers, broadband voltage transformers, voltage sensors, multi-functional power meters, front-end sensing devices, edge computing gateways, BeiDou dual-mode time synchronization devices, temperature and humidity sensors, tipping bucket rain gauges, cup-type anemometers, soil resistivity meters, electromagnetic interference testers, signal analyzers, sampling controllers, power quality analyzers, and ultra-high frequency current sensors.

[0089] Distribution network line data includes zero-sequence current, zero-sequence current waveform and its harmonic content, zero-sequence voltage, transient current and transient voltage of each line in the distribution network; distributed generation data includes grid connection point voltage, active power, reactive power, total capacity and current limiting threshold of inverter output current of distributed generation; new power system data includes total load and reference clock of new power system; environmental data includes real-time temperature, real-time humidity, rainfall intensity, wind speed, soil resistivity, high-frequency electromagnetic interference intensity and communication channel noise of the distribution network ground fault environment; edge node data includes sampling frequency and sampling clock of each edge node in the distribution network, as well as bit depth, range and cutoff frequency of AD converter;

[0090] Specifically, using broadband zero-sequence current transformers, the zero-sequence current, waveform, and harmonic content of each line in the distribution network are collected; using zero-sequence voltage transformers, the zero-sequence voltage of each line in the distribution network is collected; combining high-precision transient waveform recording devices, broadband current transformers, and broadband voltage transformers, the transient current and transient voltage of each line in the distribution network are collected; using voltage sensors and multi-functional power meters, the grid connection point voltage, active power, and reactive power of distributed power sources are collected; using data entry technology, the total capacity of distributed power sources and the current limiting threshold of their inverter output current are extracted from the energy management system database; and utilizing front-end sensing equipment and edge computing networks... The system collects the total load of the new power system; it uses a Beidou dual-mode time synchronization device to collect the reference clock of the new power system; it integrates temperature and humidity sensors, tipping bucket rain gauges, cup-type anemometers, soil resistivity meters, electromagnetic interference testers, and signal analyzers to collect real-time temperature, real-time humidity, rainfall intensity, wind speed, soil resistivity, high-frequency electromagnetic interference intensity, and communication channel noise in the distribution network grounding fault environment; it uses a sampling controller to collect the sampling frequency and sampling clock of each edge node in the distribution network; and it uses data entry technology to obtain the bit depth, range, and cutoff frequency of the corresponding AD converters used from the edge node parameter database.

[0091] The collected distribution network line data, distributed power source data, new power system data, environmental data, and edge node data are cleaned and standardized. Timestamps are assigned to each of these data sets. By adjusting the timestamps, the collection time of the distribution network line data, distributed power source data, new power system data, environmental data, and edge node data is synchronized. The preprocessed distribution network line data, distributed power source data, new power system data, environmental data, and edge node data are then integrated to generate a distributed intelligent fault location dataset for distribution network grounding faults.

[0092] Step S2, which involves calculating the ground fault coefficient of each line in the distribution network and identifying the ground fault status of each line in the distribution network, includes:

[0093] Using data from the distribution network lines, calculate the average zero-sequence current, average zero-sequence current harmonic content, average zero-sequence voltage, average transient current, and average transient voltage for each line in the distribution network.

[0094] By measuring the zero-sequence current, its waveform and harmonic content, and the proportions of zero-sequence voltage, transient current, and transient voltage in the average zero-sequence current, average zero-sequence current harmonic content, average zero-sequence voltage, average transient current, and average transient voltage of each line in the distribution network, the abnormal values ​​of zero-sequence current, zero-sequence current harmonic content, zero-sequence voltage, transient current, and transient voltage of each line in the distribution network are obtained.

[0095] The criteria for determining ground faults for each line in the distribution network are established. Weights are then assigned to the abnormal values ​​of zero-sequence current, zero-sequence current harmonic content, zero-sequence voltage, transient current, and transient voltage for each line in the distribution network. Using a weighted summation method, the ground fault coefficient for each line in the distribution network is calculated. The specific calculation process includes:

[0096] g d = a1x1 + a2x2 + a3x3 + a4x4 + a5x5

[0097] Among them, g d Let x1, x2, x3, x4, and x5 be the grounding fault coefficients of each line in the distribution network, respectively, and let a1, a2, a3, a4, and a5 be the weights of the zero-sequence current anomaly, zero-sequence current harmonic content anomaly, zero-sequence voltage anomaly, transient current anomaly, and transient voltage anomaly of each line in the distribution network.

[0098] Based on the grounding fault coefficient of each line in the distribution network, the grounding fault status of each line in the distribution network is identified. The identification of the grounding fault status of each line in the distribution network includes the occurrence of grounding faults and the absence of grounding faults. The lines in the distribution network that have grounding faults are called grounding fault lines, and the lines in the distribution network that have not grounding faults are called non-grounding fault lines.

[0099] The specific identification process is as follows: when the ground fault coefficient of each line in the distribution network is less than 0.5, no ground fault has occurred in the corresponding distribution network line; when the ground fault coefficient of each line in the distribution network is equal to or greater than 0.5, a ground fault has occurred in the corresponding distribution network line.

[0100] In step S3, the process of obtaining the sampling frequency of the shared ground fault characteristic data of the edge nodes includes:

[0101] The system clock synchronization error of each edge node in the distribution network is calculated by using the difference between the total load of the new power system and the sampling clock of each edge node in the distribution network.

[0102] The sampling period deviation of each edge node in the distribution network is calculated by the absolute value of the difference between the sampling clocks of each edge node in the distribution network, and the system clock synchronization error and sampling period deviation of each edge node in the distribution network are integrated into the distributed intelligent fault location dataset of the distribution network.

[0103] The sampling frequency of each edge node in the distribution network and the system clock synchronization error and sampling period deviation of each edge node in the distributed intelligent fault location dataset of the distribution network are extracted. The extracted data are divided into the first training set and the first test set in a ratio of 7:3.

[0104] Using the decision tree algorithm, the first training set data is set as input, and the sampling frequency evaluation coefficient of the ground fault feature data is set as output. The sampling frequency of each edge node in the distribution network and the nonlinear relationship between the system clock synchronization error and sampling period deviation of each edge node in the distribution network and the sampling frequency evaluation coefficient of the ground fault feature data are learned to train the data sampling frequency analysis model.

[0105] The first test set data is input into the data sampling frequency analysis model. The SGD optimizer is used to adjust the parameters of the data sampling frequency analysis model, optimize the performance of the data sampling frequency analysis model, and obtain the final data sampling frequency analysis model.

[0106] Based on the sampling frequency of each edge node in the current distribution network and the system clock synchronization error and sampling period deviation of each edge node in the distribution network, the corresponding ground fault characteristic data sampling frequency evaluation coefficient is output.

[0107] Based on the evaluation coefficient of the ground fault characteristic data sampling frequency, the corresponding sampling frequency is assigned to the edge nodes to share the ground fault characteristic data.

[0108] In step S3, the process of acquiring the shared ground fault characteristic data of edge nodes before and after the topology change, as well as the ground fault current data of the distribution network, includes:

[0109] The edge node shared ground fault characteristic data before and after topology change includes the zero-sequence current, voltage amplitude, high-frequency component of transient current and harmonic content of the ground fault line before and after topology change, as well as the phase difference of the zero-sequence current with the unground fault line; the distribution network ground fault current data includes the real-time ground fault current of each line in the distribution network and its fundamental amplitude, harmonic amplitude and ultra-high frequency component.

[0110] Based on the sampling frequency of the shared ground fault characteristic data of edge nodes, the zero-sequence current, voltage amplitude, high-frequency component of transient current, harmonic content and zero-sequence current phase of the ground fault line before and after the topology change, the zero-sequence current phase of the unground fault line, and the real-time ground fault current and its ultra-high frequency component of each line in the distribution network are collected using the acquisition equipment.

[0111] The specific data acquisition process includes: combining broadband zero-sequence current transformers, broadband voltage transformers, broadband current transformers, and power quality analyzers to acquire the zero-sequence current, voltage amplitude, transient current high-frequency components, and harmonic content of the ground fault line; using broadband zero-sequence current transformers to acquire the zero-sequence current phase of the ground fault line and the unground fault line; and using broadband zero-sequence current transformers and ultra-high frequency current sensors to acquire the real-time ground fault current and its ultra-high frequency components of each line in the distribution network.

[0112] The zero-sequence current phase difference between the ground fault line and the unground fault line before and after the topology change is obtained by using the absolute value of the difference between the zero-sequence current phase of the ground fault line before the topology change and the absolute value of the difference between the zero-sequence current phase of the ground fault line and the unground fault line after the topology change.

[0113] Using edge computing technology and the FFT algorithm, the fundamental and harmonic amplitudes of the real-time ground fault current of each line in the distribution network are extracted, and the shared ground fault feature data of the edge nodes before and after the topology change and the ground fault current data of the distribution network are obtained.

[0114] Based on a distributed communication protocol, the communication list is automatically updated when the topology changes, ensuring that the edge nodes before and after the topology change share ground fault feature data and the distribution network ground fault current data match the current network structure. Combined with time synchronization technology, the time synchronization of the edge nodes sharing ground fault feature data and the distribution network ground fault current data before and after the topology change is achieved.

[0115] The ground fault current data of the distribution network is integrated into the distributed intelligent fault location dataset of the distribution network.

[0116] Step S3, the process of correcting the distribution network ground fault current data and determining the detection requirements for shared ground fault characteristic data of edge nodes after topology changes, includes:

[0117] The zero-sequence current, zero-sequence current waveform and harmonic content, and ground fault current data of each line in the distribution network are extracted from the distributed intelligent line selection dataset of the distribution network ground fault. The extracted data is divided into a second training set and a second test set, wherein the ratio of the second training set data to the second test set data is 8:2.

[0118] Using the random forest algorithm, the second training set data is used as input and the ground fault current distortion coefficient is used as output to learn the zero-sequence current, zero-sequence current waveform and harmonic content of each line in the distribution network, as well as the nonlinear relationship between the distribution network ground fault current data and the ground fault current distortion coefficient, and to train the distribution network ground fault current distortion detection model.

[0119] The second test set data is input into the distribution network ground fault current distortion detection model. The Adam optimizer is used to adjust the parameters of the distribution network ground fault current distortion detection model, optimize the performance of the distribution network ground fault current distortion detection model, and obtain the final distribution network ground fault current distortion detection model. Combining the zero-sequence current and zero-sequence current waveform and its harmonic content of each line in the current distribution network, as well as the distribution network ground fault current data, the corresponding ground fault current distortion coefficient is output.

[0120] Based on the ground fault current distortion coefficient, the real-time ground fault current of each line in the distribution network is calibrated to obtain the real-time ground fault current of each line in the distribution network after calibration, as well as its fundamental amplitude, harmonic amplitude and ultra-high frequency component. Using the real-time ground fault current of each line in the distribution network after calibration, as well as its fundamental amplitude, harmonic amplitude and ultra-high frequency component, the ground fault current data of the distribution network is updated to obtain the calibrated ground fault current data of the distribution network.

[0121] By combining the edge node shared grounding fault characteristic data before and after the topology change with the weighted average method, the edge node shared compliance degree is calculated. The calculation process includes:

[0122]

[0123] Where α is the edge node shared consistency; u1, u2, u3, u4, u5 and u6 are the real-time ground fault current, zero-sequence current, voltage amplitude, high-frequency component and harmonic content of transient current, and phase difference of zero-sequence current with ungrounded fault lines of each line in the calibrated distribution network, respectively; b1, b2, b3, b4, b5 and b6 are the weights of the real-time ground fault current, zero-sequence current, voltage amplitude, high-frequency component and harmonic content of transient current, and phase difference of zero-sequence current with ungrounded fault lines of each line in the calibrated distribution network, respectively.

[0124] Based on the edge node sharing consistency, the detection requirement for edge node shared grounding fault characteristic data after topology change is determined. This requirement includes whether or not the edge node shared grounding fault characteristic data has a detection requirement. Specifically, the determination process is as follows: when the edge node sharing consistency is below 85%, there is no detection requirement for the edge node shared grounding fault characteristic data after topology change; when the edge node sharing consistency is equal to or higher than 85%, there is a detection requirement for the edge node shared grounding fault characteristic data after topology change.

[0125] In step S4, the process of constructing a data accuracy analysis model and calculating the accuracy of the shared grounding fault characteristic data of edge nodes includes:

[0126] The bit depth, range, and cutoff frequency of the AD converters of each edge node in the distribution network are extracted from the distributed intelligent fault location dataset of the distribution network. The extracted data are then converted into a third training set and a third test set, with the ratio of the third training set data to the third test set data being 6:4.

[0127] Using a multiple linear regression algorithm, the third training set data is used as input, and the accuracy of the shared ground fault feature data of edge nodes is used as output. The linear relationship between the bit depth, range, and cutoff frequency of the AD converter of each edge node in the distribution network and the accuracy of the shared ground fault feature data of edge nodes is learned, and the data accuracy analysis model is trained.

[0128] The third test set data is input into the data accuracy analysis model. The regression coefficients and intercept terms of the data accuracy analysis model are adjusted to optimize the performance of the data accuracy analysis model and obtain the final data accuracy analysis model. Combined with the bit depth, range, and cutoff frequency of the AD converters of each edge node in the current distribution network, the accuracy of the corresponding edge node shared grounding fault characteristic data is output.

[0129] The expression for the above data accuracy analysis model is:

[0130] σ1=β0+β1ad1+β2ad2+β3ad3+∈

[0131] Where σ1 represents the accuracy of the shared ground fault characteristic data of the edge nodes; ad1, ad2, and ad3 represent the number of bits, range, and cutoff frequency of the AD converters of each edge node in the distribution network, respectively; β1, β2, and β3 represent the regression coefficients of the number of bits, range, and cutoff frequency of the AD converters of each edge node in the distribution network, respectively; and β0 and ∈ represent the intercept term and error term of the data accuracy analysis model, respectively.

[0132] Step S4, the process of calculating the accuracy detection coefficient of ground fault characteristic data includes:

[0133] Weights are assigned to the accuracy of the shared ground fault characteristic data of edge nodes, the corrected distribution network ground fault current data, and the shared consistency of edge nodes, respectively.

[0134] The weighted summation method is used to calculate the detection coefficient for the accuracy of ground fault characteristic data. The calculation formula is as follows:

[0135] τ0=w1σ1+w2t2+w3t3+w4t4+w5t5+w6α

[0136] Wherein, τ0 is the detection coefficient for the accuracy of ground fault feature data, σ1, t2, t3, t4, t5 and α are the accuracy of the ground fault feature data shared by edge nodes, the real-time ground fault current and its fundamental amplitude, harmonic amplitude and ultra-high frequency component of each line in the calibrated distribution network and the edge node sharing consistency, respectively, and w1, w2, w3, w4, w5 and w6 are the weights of the accuracy of the ground fault feature data shared by edge nodes, the real-time ground fault current and its fundamental amplitude, harmonic amplitude and ultra-high frequency component of each line in the calibrated distribution network and the edge node sharing consistency, respectively.

[0137] In step S4, the process of constructing a data accuracy calibration model, outputting ground fault feature data accuracy calibration coefficients, and then calibrating the ground fault feature data accuracy detection coefficients includes:

[0138] Select time t1 and time t2, and extract the active power and reactive power of the distributed power source corresponding to time t1 and time t2 respectively from the distributed power source data;

[0139] Based on the active and reactive power of the distributed generation at times t1 and t2 respectively, the active power fluctuation and reactive power fluctuation of the distributed generation are calculated respectively; the calculation process is as follows:

[0140]

[0141] Where, Δp y and Δp w These represent the active power fluctuation and reactive power fluctuation of the distributed power source, respectively; p y1 and p y2 Let p be the active power of the distributed generation at times t1 and t2, respectively; w1 and p w2 Let t1 and t2 be the reactive power of the distributed generation corresponding to time t1 and t2 respectively; t1 and t2 are selected times.

[0142] The proportion of distributed power generation capacity in the total load of the new power system is used to calculate the proportion of distributed power generation access capacity. The calculated active power fluctuation and reactive power fluctuation of distributed power generation, as well as the proportion of distributed power generation access capacity, are then integrated into the distributed intelligent fault location dataset for distribution network grounding faults.

[0143] Environmental data, active power fluctuation and reactive power fluctuation of distributed power sources, the proportion of distributed power source access capacity, and the grid connection point voltage and current limiting threshold of the inverter output current of distributed power sources are extracted from the distributed intelligent fault location dataset of the distribution network. The extracted data is divided into a fourth training set and a fourth test set, with the ratio of the fourth training set data to the fourth test set data being 7:3.

[0144] A neural network algorithm is used, with the fourth training set data as input and the ground fault feature data accuracy calibration coefficient as output. The nonlinear relationship between each of the fourth training set data and the ground fault feature data accuracy calibration coefficient is learned to train the data accuracy calibration model.

[0145] The fourth test set data is input into the data accuracy calibration model, the parameters of the data accuracy calibration model are adjusted, the performance of the data accuracy calibration model is optimized, and the final data accuracy calibration model is obtained.

[0146] Based on current environmental data, the active and reactive power fluctuations of distributed power sources, the proportion of distributed power source access capacity, and the grid connection point voltage and current limiting threshold of the inverter output current of distributed power sources, the corresponding ground fault characteristic data accuracy calibration coefficient is output.

[0147] Based on the accuracy calibration coefficient of ground fault feature data, the accuracy detection coefficient of ground fault feature data is calibrated to obtain the calibrated accuracy detection coefficient of ground fault feature data.

[0148] The specific calibration process includes: when the ground fault characteristic data accuracy calibration coefficient satisfies 0.4 ≤ γ ≤ 0.6, no calibration is performed on the ground fault characteristic data accuracy detection coefficient; when the ground fault characteristic data accuracy calibration coefficient satisfies 0 ≤ γ < 0.4, calibration is performed based on τ t = (1-γ)×τ0 formula, to calibrate the detection coefficient of ground fault characteristic data accuracy; when the ground fault characteristic data accuracy calibration coefficient satisfies 0.6<γ≤1, according to τ t =γ×τ0 formula, used to calibrate the detection coefficient for the accuracy of ground fault characteristic data, where τ tγ is the accuracy detection coefficient of the calibrated ground fault feature data, τ0 is the accuracy detection coefficient of the ground fault feature data output by the data accuracy calibration model.

[0149] Step S5, which assesses the accuracy of distributed intelligent fault location for grounding faults in distribution networks adapted to the new power system and generates a distributed intelligent fault location data report, includes:

[0150] The accuracy of the model's output is evaluated based on the data accuracy, assessing the accuracy of distributed intelligent fault location for grounding faults in distribution networks adapted to the new power system. The specific evaluation process includes:

[0151] When the accuracy detection coefficient of the calibrated ground fault characteristic data satisfies 0≤τ t When the accuracy coefficient is less than 0.5, the distributed intelligent fault location for grounding faults in the distribution network adapted to the new power system has low accuracy; when the accuracy detection coefficient of the calibrated grounding fault characteristic data satisfies 0.5≤τ t When the accuracy is ≤0.7, the distributed intelligent fault location for grounding faults in the distribution network adapted to the new power system is of medium accuracy; when the accuracy detection coefficient of the calibrated grounding fault characteristic data satisfies 0.7<τ t When the value is ≤1, the distributed intelligent fault location for grounding faults in the distribution network, which is adapted to the new power system, achieves a high degree of accuracy.

[0152] By employing edge-side data fusion and standardized protocol technology, this method integrates the grounding fault status of each line in the distribution network, the detection requirements of edge nodes sharing grounding fault characteristic data after topology changes, and the accuracy of distributed intelligent line selection for grounding faults in the distribution network adapted to the new power system, and generates distributed intelligent line selection data reports.

[0153] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A distributed intelligent fault location depth assessment method for grounding faults in distribution networks, characterized in that, Includes the following steps: S1. Collect data on distribution network lines, distributed power sources, new power systems, environmental data, and edge node data; S2. Based on the collected distribution network line data, and combined with the weighted summation method, calculate the ground fault coefficient of each line in the distribution network and identify the ground fault situation of each line in the distribution network. S3, combining decision tree algorithm, edge computing technology, distributed communication protocol, time synchronization technology, and random forest algorithm, determines the detection requirements for shared grounding fault feature data of edge nodes after topology changes. The steps include: S31. Integrating new power system data and edge node data, calculate the system clock synchronization error and sampling period deviation of each edge node in the distribution network. Using the decision tree algorithm, construct a data sampling frequency analysis model, output the sampling frequency evaluation coefficient of ground fault feature data, and obtain the sampling frequency of shared ground fault feature data of edge nodes. S32. Based on the sampling frequency of the shared grounding fault characteristic data of edge nodes, combined with acquisition equipment, edge computing technology, distributed communication protocol and time synchronization technology, acquire the shared grounding fault characteristic data of edge nodes before and after topology change and the grounding fault current data of distribution network before and after topology change; S33. Using the random forest algorithm, a ground fault current distortion detection model for the distribution network is constructed, and the ground fault current distortion coefficient is output to correct the ground fault current data of the distribution network. Combined with the weighted average method, the edge node sharing consistency is calculated to determine the detection requirements of the edge node sharing ground fault feature data after the topology change. S4. Using a combination of multiple linear regression and weighted summation methods, calculate the accuracy detection coefficient of ground fault feature data. Then, employ a neural network algorithm to obtain the accuracy calibration coefficient of the ground fault feature data. The calibration of the accuracy detection coefficient of the ground fault feature data includes the following steps: S41. Using edge node data and a multiple linear regression algorithm, a data accuracy analysis model is constructed to calculate the accuracy of the shared grounding fault characteristic data of edge nodes; S42. Based on the accuracy of the shared ground fault feature data of the edge nodes, and combined with the consistency between the corrected distribution network ground fault current data and the edge node sharing, the weighted summation method is used to calculate the detection coefficient of the accuracy of the ground fault feature data. S43. Based on distributed power source data and new power system data, calculate the active power fluctuation and reactive power fluctuation of distributed power sources and the proportion of distributed power source access capacity respectively. Combine environmental data and neural network algorithms to construct a data accuracy calibration model, output the accuracy calibration coefficient of ground fault feature data, and calibrate the accuracy detection coefficient of ground fault feature data. S5. Based on the accuracy of the data, calibrate the output of the model, evaluate the accuracy of the distributed intelligent fault location for grounding faults in the distribution network adapted to the new power system, and generate a distributed intelligent fault location data report by using edge-side data fusion and standardized protocol technology.

2. The distributed intelligent fault location depth assessment method for distribution network grounding faults according to claim 1, characterized in that, The process of collecting data from power distribution network lines, distributed power sources, new power systems, environmental data, and edge nodes includes: Different types of data acquisition devices are deployed, combined with data entry technology, to collect data on power distribution lines, distributed power sources, new power systems, environmental data, and edge nodes. The data acquisition devices include broadband zero-sequence current transformers, zero-sequence voltage transformers, high-precision transient waveform recording devices, broadband current transformers, broadband voltage transformers, voltage sensors, multi-functional power meters, front-end sensing devices, edge computing gateways, BeiDou dual-mode time synchronization devices, temperature and humidity sensors, tipping bucket rain gauges, cup-type anemometers, soil resistivity meters, electromagnetic interference testers, signal analyzers, sampling controllers, power quality analyzers, and ultra-high frequency current sensors. The distribution network line data includes the zero-sequence current, zero-sequence current waveform and its harmonic content, zero-sequence voltage, transient current and transient voltage of each line in the distribution network; the distributed power source data includes the grid connection point voltage, active power, reactive power, total capacity and current limiting threshold of the inverter output current of the distributed power source; the new power system data includes the total load and reference clock of the new power system; the environmental data includes the real-time temperature, real-time humidity, rainfall intensity, wind speed, soil resistivity, high-frequency electromagnetic interference intensity and communication channel noise of the distribution network ground fault environment; the edge node data includes the sampling frequency and sampling clock of each edge node in the distribution network, as well as the bit depth, range and cutoff frequency of the AD converter; The collected distribution network line data, distributed power source data, new power system data, environmental data, and edge node data are cleaned and standardized. Timestamps are assigned to each of these data sets. By adjusting the timestamps, the collection time of the distribution network line data, distributed power source data, new power system data, environmental data, and edge node data is synchronized. The preprocessed distribution network line data, distributed power source data, new power system data, environmental data, and edge node data are then integrated to generate a distributed intelligent fault location dataset for distribution network grounding faults.

3. The distributed intelligent fault location depth assessment method for distribution network grounding faults according to claim 2, characterized in that, The process of calculating the ground fault coefficient of each line in the distribution network and identifying the ground fault status of each line in the distribution network includes: Using data from the distribution network lines, calculate the average zero-sequence current, average zero-sequence current harmonic content, average zero-sequence voltage, average transient current, and average transient voltage for each line in the distribution network. By measuring the zero-sequence current, its waveform and harmonic content, and the proportions of zero-sequence voltage, transient current, and transient voltage in the average zero-sequence current, average zero-sequence current harmonic content, average zero-sequence voltage, average transient current, and average transient voltage of each line in the distribution network, the abnormal values ​​of zero-sequence current, zero-sequence current harmonic content, zero-sequence voltage, transient current, and transient voltage of each line in the distribution network are obtained. The criteria for determining ground faults for each line in the distribution network are set, and the weights are assigned to the abnormal values ​​of zero-sequence current, zero-sequence current harmonic content, zero-sequence voltage, transient current, and transient voltage for each line in the distribution network. The ground fault coefficient of each line in the distribution network is calculated by combining the weighted summation method. Based on the grounding fault coefficient of each line in the distribution network, the grounding fault status of each line in the distribution network is identified. The identification of the grounding fault status of each line in the distribution network includes the occurrence of grounding faults and the absence of grounding faults. The lines in the distribution network that have grounding faults are called grounding fault lines, and the lines in the distribution network that have not grounding faults are called non-grounding fault lines.

4. The distributed intelligent fault location depth assessment method for distribution network grounding faults according to claim 3, characterized in that, The process of obtaining the sampling frequency of the shared ground fault characteristic data of the edge nodes includes: The system clock synchronization error of each edge node in the distribution network is calculated by using the difference between the total load of the new power system and the sampling clock of each edge node in the distribution network. The sampling period deviation of each edge node in the distribution network is calculated by the absolute value of the difference between the sampling clocks of each edge node in the distribution network, and the system clock synchronization error and sampling period deviation of each edge node in the distribution network are integrated into the distributed intelligent fault location dataset of the distribution network. The sampling frequency of each edge node in the distribution network and the system clock synchronization error and sampling period deviation of each edge node in the distributed intelligent fault location dataset of the distribution network are extracted. The extracted data is divided into the first training set and the first test set. Using the decision tree algorithm, the first training set data is set as input, and the sampling frequency evaluation coefficient of the ground fault feature data is set as output. The nonlinear relationship between each item of the first training set data and the sampling frequency evaluation coefficient of the ground fault feature data is learned, and the data sampling frequency analysis model is trained. The first test set data is input into the data sampling frequency analysis model. The SGD optimizer is used to adjust the parameters of the data sampling frequency analysis model, optimize the performance of the data sampling frequency analysis model, and obtain the final data sampling frequency analysis model. Based on the sampling frequency of each edge node in the current distribution network and the system clock synchronization error and sampling period deviation of each edge node in the distribution network, the corresponding ground fault characteristic data sampling frequency evaluation coefficient is output. Based on the evaluation coefficient of the ground fault characteristic data sampling frequency, the corresponding sampling frequency is assigned to the edge nodes to share the ground fault characteristic data.

5. The distributed intelligent fault location depth assessment method for distribution network grounding faults according to claim 4, characterized in that, The process of acquiring the shared ground fault characteristic data of edge nodes before and after the topology change, as well as the ground fault current data of the distribution network, includes: The shared ground fault characteristic data of edge nodes before and after the topology change includes the zero-sequence current, voltage amplitude, high-frequency component of transient current and harmonic content of the ground fault line before and after the topology change, as well as the phase difference of the zero-sequence current with the unground fault line; the ground fault current data of the distribution network includes the real-time ground fault current of each line in the distribution network and its fundamental amplitude, harmonic amplitude and ultra-high frequency component. Based on the sampling frequency of the shared ground fault characteristic data of edge nodes, the zero-sequence current, voltage amplitude, high-frequency component of transient current, harmonic content and zero-sequence current phase of the ground fault line before and after the topology change, the zero-sequence current phase of the unground fault line, and the real-time ground fault current and its ultra-high frequency component of each line in the distribution network are collected using the acquisition equipment. The zero-sequence current phase difference between the ground fault line and the unground fault line before and after the topology change is obtained by using the absolute value of the difference between the zero-sequence current phase of the ground fault line before the topology change and the absolute value of the difference between the zero-sequence current phase of the ground fault line and the unground fault line after the topology change. Using edge computing technology and the FFT algorithm, the fundamental and harmonic amplitudes of the real-time ground fault current of each line in the distribution network are extracted, and the shared ground fault feature data of the edge nodes before and after the topology change and the ground fault current data of the distribution network are obtained. Based on a distributed communication protocol, the communication list is automatically updated when the topology changes. Combined with time synchronization technology, the edge nodes share ground fault characteristic data and distribution network ground fault current data before and after the topology change. The ground fault current data of the distribution network is integrated into the distributed intelligent fault location dataset of the distribution network.

6. The distributed intelligent fault location depth assessment method for distribution network grounding faults according to claim 5, characterized in that, The process of correcting the distribution network ground fault current data and determining the detection requirements for shared ground fault characteristic data of edge nodes after topology changes includes: The zero-sequence current, zero-sequence current waveform and harmonic content, and ground fault current data of each line in the distribution network are extracted from the distributed intelligent line selection dataset of the distribution network ground fault. The extracted data are divided into a second training set and a second test set. Using the random forest algorithm, the second training set data is used as input and the ground fault current distortion coefficient is used as output to learn the nonlinear relationship between the second training set data and the ground fault current distortion coefficient, thereby training the distribution network ground fault current distortion detection model. The second test set data is input into the distribution network ground fault current distortion detection model. The Adam optimizer is used to adjust the parameters of the distribution network ground fault current distortion detection model, optimize the performance of the distribution network ground fault current distortion detection model, and obtain the final distribution network ground fault current distortion detection model. Combining the zero-sequence current and zero-sequence current waveform and its harmonic content of each line in the current distribution network, as well as the distribution network ground fault current data, the corresponding ground fault current distortion coefficient is output. Based on the ground fault current distortion coefficient, the real-time ground fault current of each line in the distribution network is calibrated to obtain the real-time ground fault current of each line in the distribution network after calibration, as well as its fundamental amplitude, harmonic amplitude and ultra-high frequency component. Using the real-time ground fault current of each line in the distribution network after calibration, as well as its fundamental amplitude, harmonic amplitude and ultra-high frequency component, the ground fault current data of the distribution network is updated to obtain the calibrated ground fault current data of the distribution network. By combining the edge node shared grounding fault characteristic data before and after topology change with the weighted average method, the edge node shared consistency degree is calculated; Based on the edge node sharing consistency, the detection requirement of edge node shared grounding fault feature data after topology change is determined. The detection requirement of edge node shared grounding fault feature data after topology change includes whether there is a detection requirement or not.

7. The distributed intelligent fault location depth assessment method for distribution network grounding faults according to claim 6, characterized in that, The process of constructing a data accuracy analysis model and calculating the accuracy of shared grounding fault characteristic data of edge nodes includes: Extract the bit depth, range, and cutoff frequency of the AD converters of each edge node in the distribution network from the distributed intelligent fault location dataset of the distribution network grounding fault, and convert the extracted data into a third training set and a third test set; The multivariate linear regression algorithm is used, with the third training set data as input and the accuracy of the edge node shared grounding fault feature data as output. The linear relationship between the accuracy of each item of the third training set data and the accuracy of the edge node shared grounding fault feature data is learned, and the data accuracy analysis model is trained. The third test set data is input into the data accuracy analysis model. The regression coefficients and intercept terms of the data accuracy analysis model are adjusted to optimize the performance of the data accuracy analysis model and obtain the final data accuracy analysis model. Combined with the bit depth, range, and cutoff frequency of the AD converters of each edge node in the current distribution network, the accuracy of the corresponding edge node shared grounding fault characteristic data is output.

8. The distributed intelligent fault location depth assessment method for distribution network grounding faults according to claim 7, characterized in that, The process of calculating the accuracy detection coefficient of ground fault characteristic data includes: Weights are assigned to the accuracy of the shared ground fault characteristic data of edge nodes, the corrected distribution network ground fault current data, and the shared consistency of edge nodes, respectively. The weighted summation method is used to calculate the detection coefficient for the accuracy of ground fault characteristic data. The calculation formula is as follows: τ0=w1σ1+w2t2+w3t3+w4t4+w5t5+w6α Wherein, τ0 is the detection coefficient for the accuracy of ground fault feature data, σ1, t2, t3, t4, t5 and α are the accuracy of the ground fault feature data shared by edge nodes, the real-time ground fault current and its fundamental amplitude, harmonic amplitude and ultra-high frequency component of each line in the calibrated distribution network and the edge node sharing consistency, respectively, and w1, w2, w3, w4, w5 and w6 are the weights of the accuracy of the ground fault feature data shared by edge nodes, the real-time ground fault current and its fundamental amplitude, harmonic amplitude and ultra-high frequency component of each line in the calibrated distribution network and the edge node sharing consistency, respectively.

9. The distributed intelligent fault location depth assessment method for distribution network grounding faults according to claim 8, characterized in that, The process of constructing a data accuracy calibration model, outputting ground fault feature data accuracy calibration coefficients, and then calibrating the ground fault feature data accuracy detection coefficients includes: Select time t1 and time t2, and extract the active power and reactive power of the distributed power source corresponding to time t1 and time t2 respectively from the distributed power source data; Based on the active and reactive power of the distributed generation at times t1 and t2 respectively, the active power fluctuation and reactive power fluctuation of the distributed generation are calculated respectively; the calculation process is as follows: Where, Δp y and Δp w These represent the active power fluctuation and reactive power fluctuation of the distributed power source, respectively; p y1 and p y2 Let p be the active power of the distributed generation at times t1 and t2, respectively; w1 and p w2 Let t1 and t2 be the reactive power of the distributed generation corresponding to time t1 and t2 respectively; t1 and t2 are selected times. The proportion of distributed power generation capacity in the total load of the new power system is used to calculate the proportion of distributed power generation access capacity. The calculated active power fluctuation and reactive power fluctuation of distributed power generation, as well as the proportion of distributed power generation access capacity, are then integrated into the distributed intelligent fault location dataset for distribution network grounding faults. Environmental data, active power fluctuation and reactive power fluctuation of distributed power sources, the proportion of distributed power source access capacity, and the grid connection point voltage and current limiting threshold of the inverter output current of distributed power sources are extracted from the distributed intelligent fault location dataset of the distribution network. The extracted data is divided into the fourth training set and the fourth test set. A neural network algorithm is used, with the fourth training set data as input and the ground fault feature data accuracy calibration coefficient as output. The nonlinear relationship between each of the fourth training set data and the ground fault feature data accuracy calibration coefficient is learned to train the data accuracy calibration model. The fourth test set data is input into the data accuracy calibration model, the parameters of the data accuracy calibration model are adjusted, the performance of the data accuracy calibration model is optimized, and the final data accuracy calibration model is obtained. Based on current environmental data, the active and reactive power fluctuations of distributed power sources, the proportion of distributed power source access capacity, and the grid connection point voltage and current limiting threshold of the inverter output current of distributed power sources, the corresponding ground fault characteristic data accuracy calibration coefficient is output. Based on the accuracy calibration coefficient of ground fault feature data, the accuracy detection coefficient of ground fault feature data is calibrated to obtain the calibrated accuracy detection coefficient of ground fault feature data.

10. The distributed intelligent fault location depth assessment method for distribution network grounding faults according to claim 9, characterized in that, The process of assessing the accuracy of distributed intelligent fault location for grounding faults in distribution networks adapted to new power systems and generating distributed intelligent fault location data reports includes: The accuracy of the model's output is evaluated by calibrating the model's output based on the accuracy of the data, thus assessing the accuracy of distributed intelligent fault location for distribution network grounding faults adapted to the new power system. By employing edge-side data fusion and standardized protocol technology, this method integrates the grounding fault status of each line in the distribution network, the detection requirements of edge nodes sharing grounding fault characteristic data after topology changes, and the accuracy of distributed intelligent line selection for grounding faults in the distribution network adapted to the new power system, and generates distributed intelligent line selection data reports.

Images