Single-phase grounding fault locating method for distribution network with high proportion of new energy access

Abstract

This application discloses a method for locating single-phase grounding faults in distribution networks with a high proportion of renewable energy integration, relating to the field of distribution network technology. The method includes dividing the distribution network into multiple feeder monitoring sections, acquiring active power data and zero-sequence differential current data of each feeder monitoring section within a preset time period, constructing a first operational dataset based on the active power data and zero-sequence differential current data, calculating the first operational dataset using a preset ellipsoidal algorithm to construct a dynamic correlation boundary model, processing the first operational dataset based on the model parameters of the dynamic correlation boundary model to obtain the boundary deviation index of each feeder monitoring section at the current time, and determining that a single-phase grounding fault has occurred in the current feeder monitoring section when the boundary deviation index meets preset conditions. This application achieves the technical effect of accurately locating the fault location while adapting to photovoltaic power fluctuations, dynamically adjusting protection boundaries, and reducing costs.

Description

Technical Field

[0001] This application relates to the field of power distribution network technology, specifically to a method for locating single-phase grounding faults in power distribution networks with a high proportion of renewable energy integration. Background Technology

[0002] In existing neutral-point non-effectively grounded distribution networks, single-phase ground fault currents are mainly composed of line-to-ground capacitance currents, with relatively small amplitudes. With the high proportion of distributed photovoltaic (PV) systems integrated, PV inverters inject zero-sequence currents (i.e., background noise) into the grid that dynamically varies with active power output. Traditional ground fault protection typically uses fixed zero-sequence current settings: if the setting is too high, it cannot detect single-phase ground faults with high transition resistance when PV output is low (insufficient sensitivity); if the setting is too low, it is easily affected by background noise and malfunctions during periods of high PV power generation (insufficient reliability). Although high-precision methods such as traveling wave ranging can solve this problem, they have extremely high requirements for sampling equipment and communication channels, resulting in high retrofit costs and making them difficult to implement in large-scale distribution networks.

[0003] Existing technologies lack a low-cost fault location method that can adapt to photovoltaic power fluctuations and dynamically adjust protection boundaries. Summary of the Invention

[0004] To address the technical problems in related technologies, such as inaccurate fault location when using a fixed zero-sequence current setting value and high cost of high-precision methods such as traveling wave ranging, this application provides a method for locating single-phase grounding faults in distribution networks with a high proportion of new energy access.

[0005] The specific technical solution adopted is as follows: The distribution network is divided into multiple feeder monitoring sections, and active power data and zero-sequence differential current data of each feeder monitoring section are obtained within a preset time period. Each feeder monitoring section includes multiple distributed photovoltaic inverters. Based on active power data and line zero-sequence differential current data, a first operational dataset is constructed; By using a preset ellipsoid algorithm, the first running dataset is calculated to construct a dynamic associated boundary model. The boundary of the dynamic associated boundary model changes in real time with the photovoltaic power. Based on the model parameters of the dynamic correlation boundary model, the first running dataset at the current moment is calculated and processed to obtain the boundary deviation index of each feeder monitoring segment at the current moment. If the boundary deviation index meets the preset conditions, it is determined that a single-phase grounding fault has occurred in the current feeder monitoring section.

[0006] In one possible implementation of this application, before acquiring the active power data and zero-sequence differential current data of each feeder monitoring section within a preset time period, the method further includes: Acquire the uploaded power data of each distributed photovoltaic inverter in each feeder monitoring section within a preset time period, the zero-sequence current value at the first-end switch, and the zero-sequence current value at the end of the first-end switch; For any feeder monitoring section, the sum of each uploaded power data is calculated to obtain the active power data, and the line zero-sequence differential current data is calculated based on the difference between the zero-sequence current value at the beginning and the zero-sequence current value at the end.

[0007] In one possible implementation of this application, a first operational dataset is constructed based on active power data and line zero-sequence differential current data, including: The active power data is normalized to obtain normalized power data; The zero-sequence differential current data of the line is normalized to obtain normalized differential current data; The normalized power data and the normalized differential current data are combined to construct the first data pair; Data cleaning is performed on each of the first data pairs to obtain the first running dataset.

[0008] In one possible implementation of this application, data cleaning processing is performed on each first data pair to obtain a first running dataset, including: Obtain the zero-sequence voltage value of the bus at each moment within a preset time period; If the zero-sequence voltage value of the busbar is greater than the preset safe voltage threshold, the feeder monitoring section at the current moment is in an abnormal operating state. The first data pair at the current moment is then discarded, resulting in the first operating dataset. In one possible implementation of this application, a dynamic correlation boundary model is constructed by calculating the first operating dataset using a preset ellipsoid algorithm, including: By using a pre-defined ellipsoid algorithm, iterative calculations are performed on the first running dataset to determine the center vector and shape matrix of the ellipsoid model; A dynamic associated boundary model is constructed based on the center vector and the shape matrix.

[0009] In one possible implementation of this application, a preset ellipsoid algorithm is used to iteratively calculate the first running dataset to determine the center vector and shape matrix of the ellipsoid model, including: Construct the objective function for the ellipsoid model; By using a pre-defined ellipsoid algorithm, the objective function is minimized to obtain candidate center vectors and candidate shape matrices. Based on the candidate center vector and the candidate shape matrix, the Mahalanobis distance between each data point in the first running dataset and the candidate center vector is calculated. When the Mahalanobis distance meets the preset constraints, the center vector and shape matrix of the ellipsoid model are generated.

[0010] In one possible implementation of this application, the model parameters include a center vector and a shape matrix. Based on the model parameters of the dynamic associated boundary model, the first running dataset at the current time is calculated and processed to obtain the boundary deviation index of each feeder monitoring segment at the current time, including: Using the Mahalanobis distance formula, the center vector, shape matrix, and the corresponding data pairs of the first running dataset at the current moment are calculated and processed to obtain the boundary deviation index of each feeder monitoring segment at the current moment. The boundary deviation index is used to characterize the degree of deviation between the corresponding data pairs of the first running dataset at the current moment and the dynamic associated boundary model.

[0011] In one possible implementation of this application, before calculating and processing the first running dataset at the current time based on the model parameters of the dynamic correlation boundary model to obtain the boundary deviation index of each feeder monitoring segment at the current time, the method further includes: Compare the normalized differential current data with the preset current threshold; If the normalized differential current data is less than the preset current threshold, the normalized differential current data of the feeder monitoring section is determined to be normal measurement noise, and the boundary deviation index is not calculated. If the normalized differential current data is greater than or equal to the preset current threshold, it is determined that the normalized differential current data of the feeder monitoring section belongs to abnormal measurement noise caused by fault or large power fluctuation, and the boundary deviation index calculation process is executed.

[0012] In one possible implementation of this application, determining that a single-phase ground fault has occurred in the current feeder monitoring section when the boundary deviation index meets preset conditions includes: Compare the boundary deviation index with the preset anomaly detection threshold; If the boundary deviation index is less than or equal to the preset anomaly judgment threshold, then the current feeder monitoring section is within the normal background fluctuation range and no single-phase grounding fault has occurred. If the boundary deviation index is greater than the preset anomaly judgment threshold, then the boundary deviation index meets the preset conditions. If the boundary deviation index meets the preset conditions within multiple consecutive preset sampling periods, it is determined that a single-phase grounding fault has occurred in the current feeder monitoring section.

[0013] In one possible implementation of this application, after determining that a single-phase ground fault has occurred in the current feeder monitoring section when the boundary deviation index meets a preset condition, the method further includes: If multiple feeder monitoring sections are detected to have single-phase grounding faults, the feeder monitoring section with the largest boundary deviation index is selected as the source of the single-phase grounding fault. Based on the source of the fault, fault location information is sent to the dispatch interface of the distribution network.

[0014] This application has, but is not limited to, the following technical effects: By dividing the distribution network into multiple feeder monitoring sections, active power data and zero-sequence differential current data of each feeder monitoring section within a preset time period are obtained, and a first operating dataset is constructed. Then, a preset ellipsoidal algorithm is used to calculate the first operating dataset to construct a dynamic correlation boundary model representing the correlation between power and current. Subsequently, the model parameters of the dynamic correlation boundary model are used to calculate and process the first operating dataset at the current moment to obtain the boundary deviation index of each feeder monitoring section at the current moment. When the boundary deviation index meets the preset conditions, it is determined that a single-phase grounding fault has occurred in the current feeder monitoring section. In this application, active power data and line zero-sequence differential current data are acquired, and these two types of data are processed by a preset ellipsoidal algorithm to generate a dynamic correlation boundary model. Since the boundary of the dynamic correlation boundary model changes in real time with the photovoltaic power, the boundary can automatically tighten when the photovoltaic output is low, which can identify weak high-resistance grounding faults; when the photovoltaic output is high, the boundary can automatically widen to effectively accommodate background noise fluctuations and prevent false triggering. Then, the fault status of each feeder monitoring section is determined by the boundary deviation index. Thus, the fault location can be accurately located while adapting to photovoltaic power fluctuations, dynamically adjusting the protection boundary, and reducing costs. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating the first embodiment of the method for locating single-phase grounding faults in a distribution network with a high proportion of renewable energy access in this application. Figure 2 This is a schematic diagram of the device structure of the hardware operating environment involved in the embodiments of this application. Detailed Implementation

[0016] It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit this application.

[0017] This application provides a method for locating single-phase grounding faults in distribution networks with a high proportion of renewable energy integration. In the first embodiment of this method, referring to... Figure 1 The methods include: Step S10: Divide the distribution network into multiple feeder monitoring sections, and obtain the active power data and zero-sequence differential current data of each feeder monitoring section within a preset time period. The feeder monitoring section includes multiple distributed photovoltaic inverters.

[0018] As an example, the method for locating single-phase grounding faults in distribution networks with a high proportion of renewable energy access can also be applied to the single-phase grounding fault location system in distribution networks with a high proportion of renewable energy access. Since the location of single-phase grounding faults in distribution networks is carried out on a line section basis, and the zero-sequence background current injected by distributed photovoltaic inverters will be superimposed on the line current flowing through the section, in order to accurately assess the background noise level in a specific section, it is necessary to first determine the spatial location of the interference source.

[0019] Specifically, based on the static topology information maintained by the distribution automation master station, the system first divides the entire distribution feeder into several independent logical units, defined as feeder monitoring sections, using the automated sectionalizing switches as boundaries. ,in, A unique segment index across the entire network ( N represents the number of feeder monitoring sections. For each distributed photovoltaic inverter, based on its grid connection point (PCC) location in the electrical topology, its feeder monitoring section is identified, and a section inverter set containing all distributed photovoltaic inverters within the power supply range of that section is established, denoted as . : in, Represents the first in the set Taiwanese inverter, This represents the total number of inverters connected within the feeder monitoring section. If there is no photovoltaic power connected within this section, then... This is an empty set, and once established, the mapping relationship is only updated when the distribution network topology changes (such as power transfer operations).

[0020] As an example, the preset time period can be data from the past 10 days up to the current moment, or it can be 14 days; there is no specific limitation.

[0021] As an example, active power data can be the sum of the active power values ​​of all photovoltaic inverters in the feeder monitoring section at the same time.

[0022] As an example, line zero-sequence differential current data can be the difference between the zero-sequence current flowing in at the beginning and the zero-sequence current flowing out at the end of the feeder monitoring section.

[0023] Before step S10, the following are included: The system acquires the uploaded power data of each distributed photovoltaic inverter in each feeder monitoring section within a preset time period, the zero-sequence current value at the first-end switch, and the zero-sequence current value at the end of the first-end switch.

[0024] As an example, because feeder terminals installed at switches typically provide high-frequency current sampling at the second level, while photovoltaic inverters or smart terminals in distribution areas typically only support data uploads at the minute level (e.g., 15 minutes), there is a natural frequency mismatch on the timeline. To achieve simultaneous... To construct synchronized feature data, this step employs a latest value retention strategy with time constraints.

[0025] Specifically, the current real-time detection time is set as For any inverter in the feeder monitoring section Obtain the inverter in the first... The timestamp of the most recent successful data upload before the specified time is denoted as time. (satisfy Simultaneously, read the upload power data at that moment. And calculate the time difference between the current time and the data upload time. The following judgment and processing will be performed: like less than the preset data validity period If (for example, 5 minutes), then the photovoltaic power data is deemed valid, and the system uses it directly. As of the present moment Upload power data.

[0026] like Greater than or equal to If the photovoltaic power data is invalid (possibly due to communication interruption or equipment failure), the system automatically switches to conservative mode to prevent misjudgments caused by data lag (e.g., using high power data from 15 minutes ago to assess the current low power condition). This forces the inverter's current power estimate to be set to the conservative low power value for that period (or 20% of its rated power) to ensure the model boundary covers the worst-case scenario and prevents malfunctioning protection systems, thus obtaining the uploaded power data for each time period. If the value is greater than or equal to the preset long-term failure threshold (e.g., 24 hours), the photovoltaic data is determined to be missing for a long period of time, and the calculation will be performed again after communication is restored.

[0027] As an example, data is collected from the feeder monitoring section via the feeder terminal. The first-end switch is at the Zero-sequence current value at the beginning of time and the end switch in the Zero-sequence current value at the end of time (For the terminal branch section, ).

[0028] For any feeder monitoring section, the sum of each uploaded power data is calculated to obtain the active power data, and the line zero-sequence differential current data is calculated based on the difference between the zero-sequence current value at the beginning and the zero-sequence current value at the end.

[0029] As an example, this step is used to calculate active power data characterizing the excitation intensity and line zero-sequence differential current data characterizing the response level. The active power data quantifies the sum of active power injected into the grid by all inverters in the feeder monitoring section at the current moment, for the first... Active power data at any time The calculation method can be: in For the first Upload power data at any given time. If the set If empty, then .

[0030] As an example, the zero-sequence differential current data of the line reflects the total zero-sequence current leaking to the ground within that section through ground capacitance or fault points, specifically for the first... At any given time, the zero-sequence differential current data of the line The calculation method can be: in, This represents the zero-sequence current value at the beginning. This represents the zero-sequence current value at the end.

[0031] Step S20: Construct the first operational dataset based on active power data and line zero-sequence differential current data.

[0032] As an example, after normalizing the active power data and the line zero-sequence differential current data respectively, the two types of data are combined to obtain data pairs containing the two types of data. The data pairs at each time point are integrated to construct the first running dataset.

[0033] Step S20 includes: The active power data is normalized to obtain normalized power data.

[0034] As an example, the normalization method could be min-max normalization to normalize power data. The calculation method can be: in, It represents the sum of the rated capacities of all inverters within the feeder monitoring section. This represents active power data.

[0035] The zero-sequence differential current data of the line is normalized to obtain normalized differential current data.

[0036] The normalized power data and the normalized differential current data are combined to construct the first data pair.

[0037] As an example, normalized differential flow data can also be calculated using the method described above. The calculation method can be: in, This is the maximum permissible zero-sequence unbalanced current within the monitoring section of the feeder (set according to grid connection standards, such as 5% of the rated current). This represents the zero-sequence differential current data of the line.

[0038] As an example, normalized power data is combined with normalized differential current data to generate the first... The feeder monitoring section is in the first The first data pair at time t is denoted as : Let t be the first data pair at time t. During the model training phase, it serves as a sample point to learn the distribution boundary of normal background noise; during the real-time detection phase, it serves as a test point to calculate the boundary deviation index.

[0039] Data cleaning is performed on each of the first data pairs to obtain the first running dataset.

[0040] The step of cleaning each first data pair to obtain the first running dataset includes: Obtain the zero-sequence voltage value of the bus at each moment within a preset time period; If the zero-sequence voltage value of the bus is greater than the preset safe voltage threshold, the feeder monitoring section at the current moment is in an abnormal operating state. The first data pair at the current moment is removed to obtain the first operating dataset.

[0041] As an example, in order to train a model that can accurately characterize the distribution features of normal background noise, it is necessary to filter out samples containing only normal operating conditions from historical data within a historical time period. Therefore, it is necessary to obtain data from each moment within a preset time period. In addition, before executing subsequent calculation processes, the first running dataset needs to be cleaned to ensure that the data are all under normal operating conditions.

[0042] As an example, the preset safety voltage threshold can be 15% of the system's rated phase voltage. When a single-phase ground fault occurs in the distribution network, the zero-sequence voltage will rise significantly. Therefore, the system uses the bus zero-sequence voltage data collected by the distribution automation master station as a filter at the time of the fault. Specifically, the system examines each time point in the sequence one by one. Bus zero-sequence voltage value .like Greater than the preset safe voltage threshold If a grounding fault or abnormal operation is detected at that moment, the data pair corresponding to the first running dataset at that moment will be removed.

[0043] As an example, the remaining after the above cleaning These data points constitute the first running dataset of historical steady-state operation, denoted as . : Each of them Each represents a first data pair confirmed to be under normal operating conditions. The first operating dataset not only eliminates the interference of fault data, but also enables the model to adapt to long-term trend changes brought about by the growth of distribution network load or photovoltaic capacity expansion through the sliding update mechanism of time window.

[0044] Step S30: The first running dataset is calculated using a preset ellipsoid algorithm to construct a dynamic associated boundary model. The boundary of the dynamic associated boundary model changes in real time with the photovoltaic power.

[0045] As an example, the preset ellipsoidal algorithm can be a small-volume closed ellipsoidal algorithm (such as the Khachiyan iterative algorithm). This step utilizes convex optimization mathematical tools to adaptively train a geometric model that can tightly enclose the normal background noise distribution characteristics, i.e., a dynamically associated boundary model. This model is no longer a simple fixed threshold (rectangular boundary), but an ellipsoidal boundary that dynamically adjusts with photovoltaic power, thereby achieving an intelligent protection effect of tightening at low power to improve sensitivity and loosening at high power to prevent false triggering.

[0046] As an example, in practical engineering applications, a fault in a newly commissioned line or communication system may lead to the first operational dataset being lost. Number of samples The sample size is severely insufficient, or even empty. Directly basing data-driven modeling on a small sample size could lead to overfitting (too narrow boundaries) causing erroneous actions or causing the computation process to diverge. To address this cold-start challenge, this step incorporates protection logic based on sample size. The system first counts the number of cleaned samples. and compare it with the preset minimum sample threshold. (For example, 1000 points correspond to approximately one week's worth of valid data) for comparison.

[0047] 1. When the sample size is insufficient ( ); If the sample size is insufficient, the system determines that it is in a cold start or data missing state and automatically switches to conservative model mode. In this mode, the system abandons data training and directly uses theoretical values ​​based on line design parameters to build a wide-range initial model. The model parameters of the initial model are as follows: and The setup method is as follows: An inherent unbalanced baseline value is set as the center vector of the ellipsoidal model. Set it as the zero vector This setting is based on the assumption that the three-phase parameters of the line are balanced and there is no photovoltaic injection under ideal conditions, serving as the most basic reference center.

[0048] Set the fluctuation range constraint matrix, which is also the shape matrix. Construct it as a wide-range diagonal matrix to encompass the worst background noise conditions: in, This is the sum of the rated capacities of all photovoltaic inverters within the monitoring section of the feeder (usually 1 after normalization). This is the sum of the maximum allowable zero-sequence injection current (set according to grid connection standards, such as 5% of the rated current) and the inherent capacitance current of the line at full power. The physical meaning of this matrix is ​​that it defines an elliptical boundary centered at the origin, with its major axis covering the maximum allowable current. Although the sensitivity is low, it ensures that the system does not malfunction during the initial power-on phase, providing basic safety protection. The initialization model is used to initialize the calculation process and does not execute subsequent iterative calculations.

[0049] 2. When the sample is sufficient ( ); If the number of samples meets the requirements, the system determines that the conditions for data-driven modeling are met, and proceeds to the subsequent iterative solution steps of the association boundary model based on convex optimization, using real historical data to train a more compact and more sensitive adaptive model.

[0050] Through the aforementioned cold start protection mechanism, this invention effectively avoids the risk that a pure data-driven algorithm will be unusable in the early stages, and achieves a smooth transition from conservative protection to precise protection.

[0051] Step S30 includes steps S31 to S32: Step S31: Using a preset ellipsoid algorithm, iterative calculations are performed on the first running dataset to determine the center vector and shape matrix of the ellipsoid model.

[0052] As an example, considering the significant positive correlation between the overall increase in zero-sequence background noise injected into photovoltaic inverters and the increase in active power, and the engineering characteristic that the data points exhibit a tilted band-like or approximately elliptical distribution on a two-dimensional plane, this invention employs the minimum volume closed ellipsoid algorithm for modeling. Compared to traditional rectangular boundaries (fixed maximum value protection), the tilted ellipsoidal boundary can effectively eliminate a large number of invalid non-faulty regions in the low-power range, thereby significantly improving the ability to identify weak faults.

[0053] As an example, before training the model, calculate the Mahalanobis distance of all sample points in the first run dataset, and discard the one with the largest distance. For outliers (e.g., 1%), the remaining sample points are substituted into a predefined ellipsoid algorithm. The system constructs the following optimization problem: Find a two-dimensional ellipsoid with the smallest volume. The ellipsoid model is uniquely determined by two key parameters: Inherent unbalanced benchmark value : that is, the center vector of the ellipsoid.

[0054] Fluctuation range constraint matrix : That is, the shape matrix of the ellipsoid (the inverse of the covariance matrix).

[0055] Step S31 includes: Construct the objective function for the ellipsoid model.

[0056] The objective function is minimized using a pre-defined ellipsoid algorithm to obtain candidate center vectors and candidate shape matrices.

[0057] Based on the candidate center vector and the candidate shape matrix, the Mahalanobis distance between each data point in the first running dataset and the candidate center vector is calculated. When the Mahalanobis distance meets the preset constraints, the center vector and shape matrix of the ellipsoid model are generated.

[0058] As an example, the optimization problem described above is expressed as finding a shape matrix. and center vector To minimize the logarithmic measure of the ellipsoid's volume, the objective function is solved using the Khachiyan iterative algorithm or the dual simplex method as follows: In solving the objective function, candidate center vectors and candidate shape matrices can be generated. During the solution process, the objective function can be transformed into a Lagrange dual function by introducing Lagrange multipliers. Iterative updates are performed using the Coordinate Descent Algorithm. This continues until the KKT conditions are met. Specific steps include: initializing the weight vector u; selecting two indices in each iteration for weight updates to maximize the dual function value; and when the duality gap is less than a preset tolerance (e.g., 10...). -6 Stop iterating when the final shape matrix is ​​reached, and output the final shape matrix. and center vector .

[0059] Each data point, candidate center vector, and candidate shape matrix in the first running dataset satisfies the preset constraints (Mahanobis distance calculation method): To prevent computational divergence caused by a constant photovoltaic output over a long period (resulting in a single projection of data points onto the power axis and a near-singular matrix), this step introduces regularization during the iteration process. Specifically, when calculating the covariance matrix, a very small positive number is superimposed on its diagonal elements. (For example, the value is the data variance) to (times). This treatment mathematically guarantees the positive definiteness of the matrix, and physically it is equivalent to reserving a very small basic fault tolerance space for the model, ensuring the stability of numerical calculation.

[0060] Step S32: Based on the center vector and shape matrix, a dynamic associated boundary model is constructed.

[0061] As an example, after the above algorithm converges, the center vector and shape matrix are obtained, and the updated center vector and shape matrix are encapsulated into a dynamic associated boundary model.

[0062] As an example, the center vector The physical meaning is that it quantifies the static zero-sequence bias of the feeder monitoring section in a statistical sense, which is usually caused by the asymmetry of the three-phase parameters of the line. By subtracting it as a benchmark, the interference of systematic error on fault judgment is eliminated.

[0063] Shape Matrix The physical significance lies in: quantifying the correlation and allowable fluctuation range of the zero-sequence differential current amplitude of the section with the change of the total active power of the section inverter. Its value determines the direction and length of the major and minor axes of the ellipsoid, that is, defining the dynamic boundary that allows the zero-sequence differential current to increase by a certain amount without triggering an alarm when the photovoltaic power increases by a certain amount.

[0064] Step S40: Based on the model parameters of the dynamic correlation boundary model, the first running dataset at the current moment is calculated and processed to obtain the boundary deviation index of each feeder monitoring segment at the current moment.

[0065] As an example, the model parameters of the trained dynamic correlation boundary model are used to perform millisecond-level anomaly detection on the real-time monitored power-differential current data. By introducing statistical distance and adaptive threshold, the system can accurately capture the weak differential current change caused by a single-phase grounding fault under strong photovoltaic background noise interference, thus achieving fast and reliable fault location.

[0066] As an example, the boundary deviation index is used to represent the Mahalanobis distance of real-time running data at each moment relative to the center of the dynamically associated boundary model, and is used to measure the degree of abnormality of the current running state.

[0067] Before step S40, the following steps are also included: Compare the normalized differential current data with the preset current threshold; If the normalized differential current data is less than the preset current threshold, the normalized differential current data of the feeder monitoring section is determined to be normal measurement noise, and the boundary deviation index is not calculated.

[0068] As an example, a preset current threshold is used. It can be 5A, 7A, etc., with no specific limitation.

[0069] As an example, if normalized differential stream data The system determines that the current differential current is within the dead zone, which is a normal system background fluctuation or measurement noise. Without further fault diagnosis, it directly returns to the normal state.

[0070] If the normalized differential current data is greater than or equal to the preset current threshold, it is determined that the normalized differential current data of the feeder monitoring section belongs to abnormal measurement noise caused by fault or large power fluctuation, and the boundary deviation index calculation process is executed.

[0071] As an example, if The system determines that the current differential current may be caused by a fault or a large power fluctuation, which meets the start-up conditions and triggers the subsequent boundary deviation index calculation.

[0072] Step S40 includes: Using the Mahalanobis distance formula, the center vector, shape matrix, and the corresponding data pairs of the first running dataset at the current moment are calculated and processed to obtain the boundary deviation index of each feeder monitoring segment at the current moment. The boundary deviation index is used to characterize the degree of deviation between the corresponding data pairs of the first running dataset at the current moment and the dynamic associated boundary model.

[0073] As an example, once the startup conditions are met, the system retrieves the latest power-differential flow associated boundary model parameters, i.e., the center vector, from the memory database. and shape matrix At the same time, obtain the first data pair at the current moment. .

[0074] Using the Mahalanobis distance formula, the system calculates the boundary deviation index of the feeder monitoring segment at the current moment. : In this calculation process, the roles of each parameter are as follows: Center vector It serves to debias the circuit and eliminates the zero-sequence DC component caused by static asymmetry of the line through subtraction.

[0075] Shape Matrix This matrix serves to decorrelate and dynamically weight the data. Since it's the inverse of the covariance matrix, it automatically assigns smaller weights to directions with high variance (i.e., directions with large fluctuations in photovoltaic power). This means that if the current increase in differential current is accompanied by an increase in photovoltaic power (consistent with historical statistical patterns), the calculated deviation index will be... It will not increase significantly; on the contrary, if the differential current increases but the power does not increase (which does not conform to historical patterns and is most likely a fault), the deviation index will increase sharply.

[0076] Boundary Deviation Index The magnitude of the value directly reflects the statistical probability that the current running point deviates from the center of the normal background noise distribution.

[0077] Step S50: If the boundary deviation index meets the preset conditions, it is determined that a single-phase grounding fault has occurred in the current feeder monitoring section.

[0078] Step S50 includes: The boundary deviation index is compared with the preset anomaly detection threshold.

[0079] If the boundary deviation index is less than or equal to the preset anomaly judgment threshold, then the current feeder monitoring section is within the normal background fluctuation range and no single-phase grounding fault has occurred.

[0080] If the boundary deviation index is greater than the preset anomaly judgment threshold, then the boundary deviation index meets the preset conditions. If the boundary deviation index meets the preset conditions within multiple consecutive preset sampling periods, it is determined that a single-phase grounding fault has occurred in the current feeder monitoring section.

[0081] As an example, the preset anomaly detection threshold can be 1.1, 1.2, etc. The preset ellipsoid algorithm has defined a minimum volume envelope with a radius of 1, so anything outside of 1 is an anomaly. This can maximize the effect of the preset ellipsoid algorithm in improving sensitivity.

[0082] As an example, the calculated boundary deviation index is used. Compared with the preset anomaly detection threshold Perform a comparison and execute the following logic: like It was determined that the monitoring section of the feeder was within the normal background fluctuation range.

[0083] like It was determined that there might be an abnormal zero-sequence current injection source inside the feeder monitoring section, and a single-phase grounding fault occurred in the current feeder monitoring section.

[0084] After step S50, the following is also included: If multiple feeder monitoring sections are detected to have single-phase grounding faults, the feeder monitoring section with the largest boundary deviation index is selected as the source of the single-phase grounding fault.

[0085] Based on the source of the fault, fault location information is sent to the dispatch interface of the distribution network.

[0086] As an example, to prevent false alarms caused by sudden changes in a single measurement or communication errors, the system further employs a continuous multi-frame acknowledgment mechanism. If continuous... Within a sampling period (e.g., 3-5 frames, corresponding to approximately several hundred milliseconds) A single-phase ground fault is only officially confirmed when all values ​​exceed the preset anomaly detection threshold.

[0087] Finally, in the event of multiple sections alarming simultaneously (e.g., short circuit or fault current ride-through), the system executes a competitive line selection logic based on the maximum deviation principle: in, This method selects the feeder monitoring segment with the largest boundary deviation index from the feeder monitoring segments identified as having a fault. If multiple feeder monitoring segments are found to have single-phase-to-ground faults, and the boundary deviation indices corresponding to these multiple feeder monitoring segments all have the same maximum value, then among all feeder monitoring segments whose boundary deviation indices meet preset conditions, the feeder monitoring segment at the very end of the distribution network topology is selected as the source of the single-phase-to-ground fault, and fault location information is sent to the distribution network dispatching system. Through this competitive mechanism, this method can effectively eliminate interference from upstream segments and achieve unique fault location.

[0088] This application provides a method for locating single-phase grounding faults in distribution networks with a high proportion of renewable energy access. The method involves dividing the distribution network into multiple feeder monitoring sections, acquiring active power data and zero-sequence differential current data for each feeder monitoring section within a preset time period, and constructing a first operational dataset. A preset ellipsoidal algorithm is then used to calculate the first operational dataset, constructing a dynamic correlation boundary model characterizing the correlation between power and current. Furthermore, the model parameters of the dynamic correlation boundary model are used to calculate and process the first operational dataset at the current moment, obtaining the boundary deviation index of each feeder monitoring section at the current moment. When the boundary deviation index meets a preset condition, a single-phase grounding fault is determined to have occurred in the current feeder monitoring section. In this application, active power data and line zero-sequence differential current data are acquired, and these two types of data are processed by a preset ellipsoidal algorithm to generate a dynamic correlation boundary model. Since the boundary of the dynamic correlation boundary model changes in real time with the photovoltaic power, the boundary can automatically tighten when the photovoltaic output is low, which can identify weak high-resistance grounding faults; when the photovoltaic output is high, the boundary can automatically widen to effectively accommodate background noise fluctuations and prevent false triggering. Then, the fault status of each feeder monitoring section is determined by the boundary deviation index. Thus, the fault location can be accurately located while adapting to photovoltaic power fluctuations, dynamically adjusting the protection boundary, and reducing costs.

[0089] Reference Figure 2 , Figure 2 This is a schematic diagram of the device structure of the hardware operating environment involved in the embodiments of this application.

[0090] like Figure 2 As shown, the single-phase grounding fault location device for distribution networks with a high proportion of new energy access may include: a processor 1001, a memory 1003, and a communication bus 1002. The communication bus 1002 is used to realize the connection and communication between the processor 1001 and the memory 1003.

[0091] Optionally, the single-phase grounding fault location device for distribution networks with a high proportion of renewable energy access may also include a user interface, a network interface, a camera, RF (Radio Frequency) circuitry, sensors, a WiFi module, etc. The user interface may include a display screen and an input submodule such as a keyboard; optional user interfaces may also include standard wired and wireless interfaces. The network interface may include standard wired and wireless interfaces (such as a Wi-Fi interface).

[0092] Those skilled in the art will understand that Figure 2The structure of the single-phase grounding fault location device for distribution networks with a high proportion of renewable energy access shown in the figure does not constitute a limitation on the single-phase grounding fault location device for distribution networks with a high proportion of renewable energy access. It may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0093] like Figure 2 As shown, the memory 1003, serving as a storage medium, may include an operating system, a network communication module, and a single-phase grounding fault location program for distribution networks with high-proportion renewable energy integration. The operating system is a program that manages and controls the hardware and software resources of the single-phase grounding fault location equipment for distribution networks with high-proportion renewable energy integration, supporting the operation of the single-phase grounding fault location program and other software and / or programs. The network communication module is used to enable communication between the various components within the memory 1003, as well as communication with other hardware and software in the single-phase grounding fault location system for distribution networks with high-proportion renewable energy integration.

[0094] exist Figure 2 In the single-phase grounding fault location device for distribution networks with a high proportion of renewable energy access shown, the processor 1001 is used to execute the single-phase grounding fault location program for distribution networks with a high proportion of renewable energy access stored in the memory 1003, and implement the steps of the single-phase grounding fault location method for distribution networks with a high proportion of renewable energy access described above.

[0095] The specific implementation method of the single-phase grounding fault location device for distribution networks with a high proportion of renewable energy access in this application is basically the same as the embodiments of the single-phase grounding fault location method for distribution networks with a high proportion of renewable energy access described above, and will not be repeated here.

[0096] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0097] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0098] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0099] The above are merely preferred embodiments of this application and do not limit the scope of this application. Any equivalent structural or procedural transformations made based on the description and drawings of this application, or direct or indirect applications in other related technical fields, are similarly included within the scope of protection of this application.

[0100] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0101] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A method for locating single-phase grounding faults in distribution networks with a high proportion of renewable energy integration, characterized in that, The method includes: The distribution network is divided into multiple feeder monitoring sections, and active power data and zero-sequence differential current data of each feeder monitoring section are obtained within a preset time period. Each feeder monitoring section includes multiple distributed photovoltaic inverters. Based on the active power data and the line zero-sequence differential current data, a first operating dataset is constructed; The first running dataset is calculated using a preset ellipsoid algorithm to construct a dynamic associated boundary model. The boundary of the dynamic associated boundary model changes in real time with the photovoltaic power. Based on the model parameters of the dynamic associated boundary model, the first running dataset at the current moment is calculated and processed to obtain the boundary deviation index of each feeder monitoring segment at the current moment. If the boundary deviation index meets the preset conditions, it is determined that a single-phase grounding fault has occurred in the current feeder monitoring section.

2. The method for locating single-phase grounding faults in distribution networks with a high proportion of new energy access as described in claim 1, characterized in that, Before acquiring the active power data and zero-sequence differential current data of each feeder monitoring section within a preset time period, the method further includes: The system acquires the uploaded power data of each distributed photovoltaic inverter in each feeder monitoring section within a preset time period, the zero-sequence current value at the first-end switch, and the zero-sequence current value at the end of the first-end switch. For any feeder monitoring section, the sum of the uploaded power data is calculated to obtain the active power data, and the line zero-sequence differential current data is calculated based on the difference between the zero-sequence current value at the beginning and the zero-sequence current value at the end.

3. The method for locating single-phase grounding faults in distribution networks with a high proportion of new energy access as described in claim 1, characterized in that, The first operational dataset is constructed based on the active power data and the line zero-sequence differential current data, including: The active power data is normalized to obtain normalized power data; The zero-sequence differential current data of the line is normalized to obtain normalized differential current data; The normalized power data and the normalized differential current data are combined to construct a first data pair; Data cleaning is performed on each of the first data pairs to obtain the first running dataset.

4. The method for locating single-phase grounding faults in distribution networks with a high proportion of new energy access as described in claim 3, characterized in that, The step of performing data cleaning processing on each of the first data pairs to obtain the first running dataset includes: Obtain the zero-sequence voltage value of the bus at each moment within a preset time period; If the zero-sequence voltage value of the bus is greater than the preset safety voltage threshold, the feeder monitoring section at the current moment is in an abnormal operating state. The first data pair at the current moment is removed to obtain the first operating dataset.

5. The method for locating single-phase grounding faults in distribution networks with a high proportion of new energy access as described in claim 1, characterized in that, The step of constructing a dynamic associated boundary model by calculating the first running dataset using a preset ellipsoid algorithm includes: By using a preset ellipsoid algorithm, the center vector and shape matrix of the ellipsoid model are determined through iterative calculation on the first running dataset. Based on the center vector and shape matrix, a dynamic associated boundary model is constructed.

6. The method for locating single-phase grounding faults in distribution networks with a high proportion of new energy access as described in claim 5, characterized in that, The step of iteratively calculating the center vector and shape matrix of the ellipsoid model using a preset ellipsoid algorithm on the first running dataset includes: Construct the objective function for the ellipsoid model; The objective function is minimized using a pre-defined ellipsoid algorithm to obtain candidate center vectors and candidate shape matrices. Based on the candidate center vector and the candidate shape matrix, the Mahalanobis distance between each data point in the first running dataset and the candidate center vector is calculated. When the Mahalanobis distance satisfies the preset constraints, the center vector and shape matrix of the ellipsoid model are generated.

7. The method for locating single-phase grounding faults in distribution networks with a high proportion of new energy access as described in claim 1, characterized in that, The model parameters include a center vector and a shape matrix. The model parameters based on the dynamic associated boundary model are calculated and processed on the first running dataset at the current time to obtain the boundary deviation index of each feeder monitoring segment at the current time, including: The center vector, shape matrix, and corresponding data pairs of the first running dataset at the current moment are calculated using the Mahalanobis distance formula to obtain the boundary deviation index of each feeder monitoring segment at the current moment. The boundary deviation index is used to characterize the degree of deviation between the corresponding data pairs of the first running dataset at the current moment and the dynamic associated boundary model.

8. The method for locating single-phase grounding faults in distribution networks with a high proportion of new energy access as described in claim 3, characterized in that, Before calculating and processing the first running dataset at the current moment based on the model parameters of the dynamic correlation boundary model to obtain the boundary deviation index of each feeder monitoring segment at the current moment, the method further includes: The normalized differential current data is compared with a preset current threshold. If the normalized differential current data is less than the preset current threshold, then the normalized differential current data of the feeder monitoring section is determined to be normal measurement noise, and the boundary deviation index is not calculated. If the normalized differential current data is greater than or equal to the preset current threshold, then the normalized differential current data of the feeder monitoring section is determined to be abnormal measurement noise caused by fault or large power fluctuation, and the boundary deviation index calculation process is executed.

9. The method for locating single-phase grounding faults in distribution networks with a high proportion of new energy access as described in claim 1, characterized in that, The step of determining that a single-phase ground fault has occurred in the current feeder monitoring section when the boundary deviation index meets the preset conditions includes: The boundary deviation index is compared with a preset anomaly detection threshold; If the boundary deviation index is less than or equal to the preset anomaly judgment threshold, then the current feeder monitoring section is within the normal background fluctuation range and no single-phase grounding fault has occurred. If the boundary deviation index is greater than the preset anomaly judgment threshold, then the boundary deviation index meets the preset condition. If the boundary deviation index meets the preset condition within multiple consecutive preset sampling periods, it is determined that a single-phase grounding fault has occurred in the current feeder monitoring section.

10. The method for locating single-phase grounding faults in distribution networks with a high proportion of new energy access as described in claim 1, characterized in that, After determining that a single-phase ground fault has occurred in the current feeder monitoring section when the boundary deviation index meets the preset conditions, the method further includes: If multiple feeder monitoring sections are detected to have single-phase grounding faults, the feeder monitoring section with the largest boundary deviation index is selected as the source of the single-phase grounding fault. Based on the source of the fault, fault location information is sent to the dispatch interface of the distribution network.

Images