A method for identifying the topology of a distribution radio station
By introducing a modified weighted coefficient feedback mechanism based on Kirchhoff's laws of current and power conservation into the distribution area topology identification method, the problem of dynamic threshold calculation relying on the deviation of preceding links is solved, thereby improving the accuracy and stability of the topology identification results and ensuring that the threshold value adapts to physical laws.
Patent Information
- Application Number
- CN202611234252.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-14
- Publication Date
- 2026-11-10
AI Technical Summary
In existing methods for identifying the topology of distribution substations, the dynamic threshold is calculated based on the statistical distribution of features from preceding steps. This is easily affected by factors such as data packet loss, communication delays, or load fluctuations, leading to uncertainty and accumulated biases in the identification results, and lacking a reverse correction mechanism.
A modified weighted coefficient feedback mechanism based on Kirchhoff's current conservation law and power conservation law is adopted. The modified weighted coefficient is generated by calculating the current residual value and the power residual value. It is then fed back to the calculation link of the dynamic decision threshold as a feedback signal to perform iterative adjustment to correct the threshold value, construct a reverse adjustment path, and suppress the accumulation of deviation.
It effectively suppresses deviations introduced by load fluctuations, data packet loss, or communication delays, improves the accuracy and stability of topology identification results, ensures that the threshold value is adaptively corrected based on physical verification, satisfies Kirchhoff conservation constraints, and achieves a balance between the accuracy of topology identification and system stability.
Smart Images

Figure CN123073715A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of power system automation technology, and more specifically, to a method for identifying the topology of a distribution substation. Background Technology
[0002] As a crucial link between the power system and end users, the accuracy of the low-voltage distribution network's topology directly impacts the quality of operation and maintenance tasks such as line loss calculation, fault location, and load management. Distribution area topology identification typically follows a sequential process: data acquisition, feature extraction, similarity determination, relation table generation, and physical verification. Within this process, the threshold used in the similarity determination stage has a decisive influence on the accuracy of the final identification result.
[0003] Regarding the strategy for determining the threshold, existing methods have attempted to introduce dynamic adjustment mechanisms. For example, some methods calculate an adaptive threshold based on the amplitude distribution characteristics of the current data segment, while others use a sliding window to set judgment criteria for data from different time periods, allowing the threshold value to adjust according to changes in the characteristics of the data segment itself. These methods demonstrate that those skilled in the art recognize that the threshold value should change with the operating state, and have proposed various implementation schemes in this direction.
[0004] In the aforementioned technical approach, there is a problem that has not been adequately addressed: the calculation of the dynamic threshold is based on the analysis of the statistical distribution of features within the current data segment, with its input coming from the outputs of preceding steps such as feature extraction and similarity calculation. When these preceding steps introduce deviations due to factors such as data packet loss, communication delays, or load fluctuations, these deviations are incorporated into subsequent calculations as input to the threshold calculation, causing the calculated threshold value to deviate from its intended discrimination boundary. Since the topology identification process is unidirectional and serial, after the deviations from the preceding steps are propagated along the process direction to subsequent steps, there is a lack of a path to send the subsequent verification results back to the threshold calculation stage to correct the threshold value. Therefore, this invention proposes a distribution transformer area topology identification method to solve the above-mentioned problem. Summary of the Invention
[0005] To achieve the above objectives, the present invention provides the following technical solution: A method for identifying the topology of a distribution radio station includes the following steps: The first step is to collect dynamic electrical operation data sets of each monitoring node in the distribution substation within a continuous time window. The dynamic electrical operation data sets include at least voltage instantaneous sequences, current instantaneous sequences, and power instantaneous sequences. Based on the dynamic electrical operation data sets, the dynamic jump feature vector corresponding to the load switching event of each monitoring node is extracted. The second step is to receive the dynamic jump feature vector and automatically calculate the dynamic decision threshold based on the degree of dispersion of the numerical distribution of the dynamic jump feature vector within the current time window. The dynamic decision threshold replaces the fixed empirical threshold and is used to dynamically divide the boundaries for subsequent node similarity determination. The third step is to obtain the dynamic decision threshold, and based on the dynamic decision threshold, perform time-series flexible alignment processing on the dynamic electrical operation data groups of each monitoring node. Then, calculate the correlation strength value between each monitoring node after the time-series flexible alignment processing, and then generate a preliminary topology association relationship table based on the correlation strength value and the dynamic decision threshold. The fourth step involves receiving the preliminary topology association attribution table and the dynamic electrical operation data set. Based on Kirchhoff's current conservation law, the current residual value of each monitoring node under the preliminary topology association attribution table is calculated. Simultaneously, based on Kirchhoff's power conservation law, the power residual value of each monitoring node under the preliminary topology association attribution table is calculated. Then, based on the current residual value and the power residual value, a modified weighting coefficient is generated. The modified weighting coefficient is used to characterize the physical reliability of each candidate association path in the preliminary topology association attribution table. The fifth step involves sending the corrected weighting coefficients back to the second step as feedback signals. Based on these corrected weighting coefficients, the automatic calculation basis for the dynamic decision threshold in the second step is adjusted to update the value of the dynamic decision threshold. This process then drives the iterative execution of the third and fourth steps. During each iteration, when both the current residual value and the power residual value are less than the preset convergence threshold, the final topology identification result is output and the iteration is terminated. When the number of iterations reaches the preset maximum number of iterations and the current residual value or the power residual value is still not less than the preset convergence threshold, the iteration is forcibly terminated, and the current preliminary topology association table is output as the final topology identification result. Simultaneously, a timeout alarm signal is generated.
[0006] In a preferred embodiment, in the first step, the extraction process of the dynamic transition feature vector is triggered by the absolute value of the difference between adjacent sampling points of the instantaneous voltage sequence or instantaneous current sequence of each monitoring node exceeding a preset start-capture threshold value. When the trigger condition is met, the waveform segments before and after the recording time are recorded as load switching events, and the amplitude change rate, duration, and zero-crossing offset of the load switching events are combined to form the dynamic transition feature vector.
[0007] In a preferred embodiment, the specific implementation method for automatically calculating the dynamic decision threshold based on the dispersion of the numerical distribution of the dynamic jump feature vector within the current time window in the second step is as follows: The mean amplitude and standard deviation of all dynamic jump feature vectors within the current time window are statistically analyzed. The mean amplitude and standard deviation are added together and then divided by two. The resulting value is used as the initial value of the dynamic decision threshold. When the change amplitude of the dynamic decision threshold calculated in two adjacent time windows exceeds the preset stability judgment limit, the difference between the change amplitude and the stability judgment limit is added to the initial value of the dynamic decision threshold to form the dynamic decision threshold at the current moment.
[0008] In a preferred embodiment, the third step, the timing elastic alignment process is specifically executed as follows: Using the voltage instantaneous sequence in the dynamic electrical operation data set as the alignment reference, the current instantaneous sequence and power instantaneous sequence of each monitoring node are scaled proportionally according to the time offset of their respective peak times relative to the peak times of the voltage instantaneous sequence. This ensures that the current instantaneous sequence and power instantaneous sequence of all monitoring nodes achieve maximum overlap and matching with the voltage instantaneous sequence on the time axis. For data missing locations due to communication packet loss, the average value of the two sampling points before and after the location is used for interpolation to fill in the missing data, thus obtaining a complete and equal-length dynamic electrical operation data set.
[0009] In a preferred embodiment, in the third step, when calculating the correlation strength values between each monitoring node, the waveform dynamic time warping path cost calculation method is used: For any two monitoring nodes' dynamic electrical operation data sets, their instantaneous voltage sequences are used as the matching objects. The sum of the absolute differences between the two nodes at all corresponding sampling points on the time axis is calculated, and then divided by the sum of their respective amplitude standard deviations to obtain the initial path cost value. When the sum of their amplitude standard deviations is greater than zero, a division operation is performed. When the sum of their amplitude standard deviations is equal to zero, the initial path cost value is directly assigned a preset maximum constant. Then, the process is recursively traced back to find the minimum interval segmented path corresponding to the initial path cost value on the time axis. The difference between the endpoint cost and the starting cost of the minimum interval segmented path is used as the corrected normalized cost value. When the corrected normalized cost value is greater than zero, the reciprocal of the corrected normalized cost value is used as the correlation strength value. When the corrected normalized cost value is equal to zero, the correlation strength value is directly assigned a preset maximum constant. The smaller the corrected normalized cost value, the larger the correlation strength value.
[0010] In a preferred embodiment, in the fourth step, before calculating the current residual value and the power residual value, the instantaneous current sequences of all lower-level monitoring nodes belonging to the same upper-level node in the preliminary topology association table are summed at the same time to obtain the current aggregate value at the upper-level node. At the same time, the instantaneous power sequences of all lower-level monitoring nodes belonging to the same upper-level node are summed at the same time to obtain the power aggregate value at the upper-level node. Then, the difference between the current aggregate value and the actual total value of the measured current at the upper-level node is used as the current residual value, and the difference between the power aggregate value and the actual total value of the measured power at the upper-level node is used as the power residual value.
[0011] In a preferred embodiment, in the fourth step, the generation method of the modified weighting coefficients is as follows: First, each candidate association path in the preliminary topology association table is assigned an initial weight value. Then, the deviation percentages of the current residual value and power residual value relative to the measured current value and measured power value of the monitoring node are calculated. When the deviation percentage is greater than the preset physical reliability threshold, the initial weight value of the candidate association path is multiplied by an attenuation coefficient less than one. When the deviation percentage is less than or equal to the physical reliability threshold, the initial weight value of the candidate association path is multiplied by a gain coefficient greater than one. All weight values after the coefficient multiplication together constitute the corrected weighting coefficient.
[0012] In a preferred embodiment, in the fifth step, when the iteration count reaches the preset maximum iteration count and the iteration is forcibly terminated, and the current preliminary topology association table is output as the final topology identification result, a timeout alarm signal is generated and recorded, which includes the following information: the specific monitoring node number that triggered the forced exit, the current residual value and power residual value corresponding to the monitoring node, and the final iteration round number when the preset maximum iteration count is reached. The timeout alarm signal is reported to the main station system through the uplink communication channel of the distribution transformer area to prompt the operation and maintenance personnel to conduct on-site verification.
[0013] In a preferred embodiment, in the fifth step, when adjusting the automatic calculation basis of the dynamic decision threshold in the second step according to the modified weighting coefficient, an additional waveform change gradient direction consistency judgment process is introduced: The direction of numerical change between two adjacent sampling points in the dynamic electrical operation data set of each monitoring node is defined as an upward trend, a downward trend, or a flat trend. The frequency of each trend direction occurring at each monitoring node within the entire continuous time window is counted to form the trend direction frequency vector of the monitoring node. Then, the directional overlap ratio between the trend direction frequency vectors of every two monitoring nodes is calculated. When the directional overlap ratio is greater than the preset directional consistency threshold, the corresponding weight value in the corrected weighting coefficient between the two monitoring nodes is multiplied by three. This weight value multiplied by three is used to re-drive the iterative update of the dynamic decision threshold, thereby accelerating the iterative process of satisfying the preset convergence threshold.
[0014] The technical effects and advantages of this invention are as follows: This invention transforms the current and power residuals calculated based on Kirchhoff's current and power conservation laws into corrected weighting coefficients. These corrected weighting coefficients are then used as feedback signals to the dynamic decision threshold calculation stage, enabling the verification results of subsequent stages to influence the adjustment of the preceding decision criteria. Compared to existing technologies where deviations can only be transmitted unidirectionally along the serial chain of data acquisition, feature extraction, similarity determination, and relation table generation without reverse correction, this invention constructs a reverse adjustment path from subsequent stages to preceding stages in the serial process. This allows deviations introduced by load fluctuations, data loss, or communication delays in preceding stages to be gradually reduced in subsequent iterations, thereby suppressing the uncertainty in the identification results caused by the accumulation of deviations along the unidirectional chain.
[0015] This invention automatically adjusts the calculation basis of the dynamic decision threshold based on the returned modified weighting coefficients. This means the threshold update no longer relies solely on the statistical characteristics of the current data segment itself, but instead incorporates physical reliability evaluation information based on Kirchhoff's conservation laws. Since the modified weighting coefficients directly reflect the degree to which each candidate association path in the current topological association attribution table conforms to Kirchhoff's conservation laws, using these coefficients as feedback signals in the threshold calculation process enables the dynamic decision threshold to adaptively adjust based on physical verification deviations. Compared to existing technologies where the threshold value adjusts only with changes in data segment amplitude distribution, this invention simultaneously drives the threshold value with physical conservation constraints, aligning the boundary delineation of similarity determination with the inherent electrical laws of the distribution network.
[0016] This invention simultaneously sets convergence exit conditions and timeout forced exit conditions during the iteration process. When both the current residual value and the power residual value are less than a preset convergence threshold, the iteration terminates normally and the result is output. When the number of iterations reaches a preset maximum number of iterations and the residual value still has not fallen below the convergence threshold, the iteration is forcibly terminated, and the current result and a timeout alarm signal are output. This dual exit mechanism ensures that the topology identification result can achieve the physical accuracy required by Kirchhoff's conservation law under normal operating conditions, while avoiding the system falling into infinite iteration when the residual value fails to converge due to abnormal operating conditions. Thus, a balance is established between identification accuracy and system operational stability. Attached Figure Description
[0017] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings; Figure 1 This is a schematic diagram of a distribution radio station topology identification method according to the present invention. Detailed Implementation
[0018] 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, and 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.
[0019] Reference Figure 1 The following examples were obtained: Example 1: A method for identifying the topology of a distribution radio station, comprising the following steps: The first step is to collect dynamic electrical operation data sets of each monitoring node in the distribution substation within a continuous time window. The dynamic electrical operation data sets include at least voltage instantaneous sequences, current instantaneous sequences, and power instantaneous sequences. Based on the dynamic electrical operation data sets, the dynamic jump feature vector corresponding to the load switching event of each monitoring node is extracted. The second step is to receive the dynamic jump feature vector and automatically calculate the dynamic decision threshold based on the degree of dispersion of the numerical distribution of the dynamic jump feature vector within the current time window. The dynamic decision threshold replaces the fixed empirical threshold and is used to dynamically divide the boundaries for subsequent node similarity determination. The third step is to obtain the dynamic decision threshold, and based on the dynamic decision threshold, perform time-series flexible alignment processing on the dynamic electrical operation data groups of each monitoring node. Then, calculate the correlation strength value between each monitoring node after the time-series flexible alignment processing, and then generate a preliminary topology association relationship table based on the correlation strength value and the dynamic decision threshold. The fourth step involves receiving the preliminary topology association attribution table and the dynamic electrical operation data set. Based on Kirchhoff's current conservation law, the current residual value of each monitoring node under the preliminary topology association attribution table is calculated. Simultaneously, based on Kirchhoff's power conservation law, the power residual value of each monitoring node under the preliminary topology association attribution table is calculated. Then, based on the current residual value and the power residual value, a modified weighting coefficient is generated. The modified weighting coefficient is used to characterize the physical reliability of each candidate association path in the preliminary topology association attribution table. The fifth step involves sending the corrected weighting coefficients back to the second step as feedback signals. Based on these corrected weighting coefficients, the automatic calculation basis for the dynamic decision threshold in the second step is adjusted to update the value of the dynamic decision threshold. This process then drives the iterative execution of the third and fourth steps. During each iteration, when both the current residual value and the power residual value are less than the preset convergence threshold, the final topology identification result is output and the iteration is terminated. When the number of iterations reaches the preset maximum number of iterations and the current residual value or the power residual value is still not less than the preset convergence threshold, the iteration is forcibly terminated, and the current preliminary topology association table is output as the final topology identification result. Simultaneously, a timeout alarm signal is generated.
[0020] In the first step, dynamic electrical operation data sets are collected from each monitoring node in the distribution substation within a continuous time window. These dynamic electrical operation data sets include at least instantaneous voltage, current, and power sequences. Based on these data sets, dynamic transition feature vectors corresponding to load switching events at each monitoring node are extracted. The length of this continuous time window is set to 2 to 5 seconds based on the statistical results of the time distribution of typical load switching behaviors in the distribution substation. The sampling frequency is set to 128 sampling points per power frequency cycle, based on the highest effective sampling rate of the analog-to-digital conversion module specified in the monitoring terminal equipment manual.
[0021] The extraction process of dynamic jump feature vectors is triggered by the absolute value of the difference between adjacent sampling points in the instantaneous voltage or current sequence of each monitoring node exceeding a preset start-up threshold. The start-up threshold is set based on experimental data of the minimum identifiable current change caused by load switching at the monitoring node, for example, 0.1 amperes to 0.5 amperes, to avoid false triggering caused by normal load fluctuations. When the trigger condition is met, the moment when the absolute value of the difference between adjacent sampling points in the instantaneous voltage or current sequence exceeds the start-up threshold is marked as the load switching moment. Waveform segments within the range of 0.5 power frequency cycles before to 6 power frequency cycles after the load switching moment are recorded as the load switching event. The length of the waveform segment covers the transient process generated by the load switching and its steady-state process after stabilization, ensuring sufficient waveform information for subsequent feature extraction.
[0022] The dynamic transition feature vector is formed by combining the amplitude change rate, duration, and zero-crossing offset of the load switching event. The amplitude change rate is the maximum slope value of the instantaneous current sequence in the waveform segment from the steady-state amplitude before switching to the steady-state amplitude after switching. The duration is the length of time that the instantaneous current sequence in the waveform segment takes to transition from the steady-state amplitude before switching to the steady-state amplitude after switching, and the typical value range is 2 to 10 power frequency cycles. The zero-crossing offset is the time deviation value of the voltage zero-crossing time in the power frequency cycle where the load switching time is located relative to the load switching time. When neither the instantaneous voltage sequence nor the instantaneous current sequence exceeds the start-up capture threshold value at the same time, the dynamic transition feature vector at that time is set to zero vector to indicate that no load switching event has occurred at that time.
[0023] In the second step, after receiving the dynamic jump feature vector, this vector is stored in the buffer and its corresponding time window number is marked. Based on the dispersion of the numerical distribution of the dynamic jump feature vector within the current time window, a dynamic decision threshold is automatically calculated. This dynamic decision threshold replaces the fixed empirical threshold for dynamic boundary division of the similarity judgment between subsequent nodes. The fixed empirical threshold is usually a constant between 0.5 and 0.7. Its drawback is that it cannot adapt to the drift of feature value distribution caused by changes in load type or communication noise interference. The dynamic decision threshold is adjusted in real time according to the actual distribution state of the feature vector within the current window, thereby avoiding misjudgment caused by the fixed threshold being too strict or too wide.
[0024] When calculating the mean amplitude and standard deviation of all dynamic jump feature vectors within the current time window, the mean amplitude is the arithmetic mean of the amplitude change rate of each feature vector, and the standard deviation is a measure of the dispersion of the amplitude change rate of each feature vector relative to the mean. The mean amplitude and standard deviation are added together and then divided by two. The resulting value is used as the initial value of the dynamic decision threshold. This calculation method places the initial threshold in the middle position between the mean and the mean plus the standard deviation, which can cover most of the normal fluctuation range and exclude extreme outliers from the similarity judgment.
[0025] When the change in the dynamic decision threshold calculated between two adjacent time windows exceeds a preset stability threshold, the difference between the change and the stability threshold is added to the initial value of the dynamic decision threshold to form the current dynamic decision threshold. The preset stability threshold is set based on the statistical results of the natural threshold drift caused by load fluctuations under normal operating conditions of the distribution substation, for example, 0.05 to 0.1. For example, if the dynamic decision threshold in the previous window is 0.42, and the initial value in the current window is 0.51, the change is 0.09. If the stability threshold is 0.06, the difference 0.03 is added to the initial value of 0.51, resulting in a current dynamic decision threshold of 0.54. This addition operation avoids frequent jumps in the threshold value due to accidental fluctuations between windows, while ensuring that the threshold can respond in the direction of the shift when the feature distribution undergoes a substantial shift. If the change does not exceed the stability threshold, the initial value is directly used as the current dynamic decision threshold without addition correction, in order to maintain the continuity of the threshold value and reduce invalid update operations.
[0026] In the third step, after obtaining the dynamic decision threshold, the dynamic electrical operation data sets of each monitoring node are subjected to time-series flexible alignment processing based on this threshold. The purpose of time-series flexible alignment processing is to eliminate time axis misalignment caused by inconsistent data acquisition start times, crystal oscillator frequency deviations, or communication transmission delays between different monitoring nodes. The voltage instantaneous sequence in the dynamic electrical operation data set is used as the alignment reference because the voltage instantaneous sequence has highly consistent phase characteristics across all monitoring nodes within the same transformer area, and its zero-crossing and peak times are unaffected by load changes, making it a stable and reliable time reference. The current instantaneous sequence and power instantaneous sequence of each monitoring node are scaled proportionally according to the time offset of their respective peak times relative to the peak times of the voltage instantaneous sequence.
[0027] The specific operation of proportional scaling is as follows: The time difference between the peak moments of the instantaneous current or power sequences and the peak moments of the instantaneous voltage sequences is calculated. This time difference is divided by the duration of one power frequency cycle of the instantaneous voltage sequence to obtain a scaling factor. The entire time axis range of the instantaneous current and power sequences is then linearly scaled according to this scaling factor, ensuring that the power frequency cycle duration of each sequence matches that of the instantaneous voltage sequence. This maximizes the overlap and matching between the instantaneous current and power sequences and the instantaneous voltage sequence on the time axis. For example, if the peak moment of the instantaneous current sequence at a certain monitoring node lags behind the peak moment of the instantaneous voltage sequence by 3.2 milliseconds, and the power frequency cycle duration of the instantaneous voltage sequence is 20 milliseconds, then the scaling factor is 3.2 divided by 20, which equals 0.16. After linearly scaling the entire time axis length of the instantaneous current sequence at this monitoring node according to this scaling factor, the peak moments of the current and voltage sequences coincide on the time axis. Other sampling points are scaled synchronously in the same proportion, maintaining the relative shape of the waveforms.
[0028] For data loss due to communication packet loss, the average value of the two sampling points before and after this location is used for interpolation to fill in the missing data, thus obtaining a complete and equal-length dynamic electrical operation data set. For example, if the instantaneous current sequence of a monitoring node is missing at the 25th sampling point, and the 23rd, 24th, 26th, and 27th sampling points are all complete and valid data points, then the arithmetic mean of the current values of the four sampling points (23rd, 24th, 26th, and 27th) is calculated, and this average value is used as the interpolation filler value for the 25th sampling point. If there are other missing points among the two sampling points before and after the missing location, then the two most recent complete and valid sampling points are selected forward or backward to participate in the average calculation. This interpolation method is based on the premise that the power signal changes continuously and smoothly in a short time without any step changes, and can recover missing data without introducing additional phase distortion.
[0029] When calculating the correlation strength between monitoring nodes after time-series elastic alignment, the waveform dynamic time warping path cost calculation method is adopted. This method can handle local stretching and compression deformations between two waveform sequences, and can accurately measure waveform similarity even when load fluctuations cause local waveform distortion. For any two sets of dynamic electrical operation data from monitoring nodes, their instantaneous voltage sequences are used as the matching objects because the instantaneous voltage sequence is most directly affected by the transformer topology and has a higher signal-to-noise ratio than the current and power sequences.
[0030] Calculate the sum of the absolute differences between the two at all corresponding sampling points on the time axis. The specific calculation method for the sum of absolute differences is as follows: Subtract the value of each sample point in the first instantaneous voltage sequence from the value of the sample point at the same time position in the second instantaneous voltage sequence, and take the absolute value of the difference. Then, sum the absolute values of all sample points to obtain the total absolute difference. Divide this total absolute difference by the sum of the standard deviations of the amplitudes of the two sequences. The standard deviation of amplitude measures the dispersion of all sample point values in each instantaneous voltage sequence relative to its arithmetic mean. The sum of the standard deviations is the value obtained by adding the standard deviations of the amplitudes of the two sequences. The result of the division operation is the initial path cost value.
[0031] When the sum of the standard deviations of the amplitudes of the two values is greater than zero, a division operation is performed. When the sum of the standard deviations of the amplitudes of the two values is equal to zero, the initial path cost value is directly assigned a preset maximum constant. This preset maximum constant is determined to be 1000 based on the theoretical upper limit of the sum of the absolute differences when there is a complete mismatch, with the normalized amplitude of the voltage instantaneous sequence ranging from 0 to 1 and the number of sampling points per cycle being 128. That is, the theoretical maximum sum of the absolute differences of the 128 sampling points in a single cycle is 128 multiplied by 1, which equals 128. Considering the cumulative effect of multiple consecutive power frequency cycles, a safety margin factor of approximately 8 is taken, thus determining this upper limit value to be 1000. This boundary processing avoids calculation abnormalities caused by a denominator of zero, while ensuring that a completely flat waveform sequence is not incorrectly judged as highly similar.
[0032] After obtaining the initial path cost value through division, the minimum interval segmented path corresponding to this initial path cost value on the time axis is searched recursively. The minimum interval segmented path is the path with the smallest cumulative cost from the start to the end of the cost matrix. Each node on this path represents a matching relationship between a pair of sampling points. The search method is as follows: starting from the end position of the cost matrix, the adjacent position with the smallest cumulative cost is selected point by point forward as the previous node in the path. At each step, the position with the smallest cumulative cost is selected from the three adjacent positions to the left, top, and left of the current position as the direction of path continuation. This selection process is repeated until the starting position of the cost matrix is returned. The path obtained in this way is the minimum interval segmented path. The difference between the endpoint cost and the starting cost of the minimum interval segmented path is used as the corrected normalized cost value. The endpoint cost is the cumulative cost value stored at the lower right corner of the cost matrix, and the starting cost is the cumulative cost value stored at the upper left corner of the cost matrix. The difference between the two is the corrected normalized cost value. This correction operation eliminates the interference of the DC bias components of the two sequences on the cost calculation, so that the cost value only reflects the difference between the waveform shapes.
[0033] When the corrected normalization cost is greater than zero, its reciprocal is used as the correlation strength value. Taking the reciprocal ensures that a smaller normalization cost corresponds to a stronger correlation strength, consistent with intuitive physical understanding. When the corrected normalization cost is zero, the correlation strength value is directly assigned a preset maximum constant of 1000. This indicates that the two instantaneous voltage sequences completely overlap on the time axis and have identical amplitudes, corresponding to the two monitoring nodes being located at the same electrical position, i.e., the topological distance is zero. A smaller corrected normalization cost value results in a larger correlation strength value, indicating a tighter electrical connection between the two monitoring nodes.
[0034] A preliminary topological association attribution table is generated based on the association strength value and the dynamic decision threshold. The specific method is as follows: Monitoring node pairs with a correlation strength value greater than the dynamic decision threshold are determined to have a direct hierarchical connection relationship, while monitoring node pairs with a correlation strength value less than or equal to the dynamic decision threshold are determined to have no direct connection relationship. All determination results are compiled into a table, with rows and columns arranged according to the monitoring node number. Each cell in the table records the connection status between the corresponding two monitoring nodes.
[0035] In the fourth step, after receiving the preliminary topology association table and the dynamic electrical operation data set, the current residual value of each monitoring node under this preliminary topology association table is calculated based on Kirchhoff's current conservation law. Simultaneously, the power residual value of each monitoring node under this preliminary topology association table is calculated based on Kirchhoff's power conservation law. Kirchhoff's current conservation law stipulates that the algebraic sum of the current flowing into any node is zero, and Kirchhoff's power conservation law stipulates that the algebraic sum of the active power flowing into any node is zero. These two conservation laws are the fundamental criteria for the correctness of the power system's physical topology; any topology association that violates these conservation laws is physically impossible. The modified weighting coefficient is used to characterize the physical reliability of each candidate association path in the preliminary topology association table. The larger the modified weighting coefficient, the more the candidate association path conforms to Kirchhoff's conservation constraints; conversely, a smaller value indicates that the candidate association path has a large physical deviation and needs to be re-evaluated.
[0036] Before calculating the current residual and power residual, the instantaneous current sequences of all lower-level monitoring nodes belonging to the same upper-level node in the preliminary topology association table are summed at the same time to obtain the current aggregate value at this upper-level node. At the same time, the instantaneous power sequences of all lower-level monitoring nodes belonging to the same upper-level node are summed at the same time to obtain the power aggregate value at this upper-level node. Then, the difference between the current aggregate value and the actual total measured current at this upper-level node is taken as the current residual value, and the difference between the power aggregate value and the actual total measured power at this upper-level node is taken as the power residual value.
[0037] For example, a certain superior node is initially identified as having three subordinate monitoring nodes. At a certain moment, the instantaneous current sequence values of the three subordinate monitoring nodes are 2.3 amps, 1.7 amps, and 0.5 amps, respectively. After summing, the aggregated current value is 4.5 amps, while the actual measured total current at this superior node is 4.8 amps. The current residual value is 4.5 minus 4.8, which equals -0.3 amps. Its absolute value of 0.3 amps indicates a deviation between the aggregated value and the measured value. At the same time, a negative current residual value indicates that the aggregated value is less than the measured value, which may correspond to the situation where some current was not measured or the measurement was lowered due to line loss. A positive current residual value indicates that the aggregated value is greater than the measured value, which may correspond to the situation where incorrect attribution leads to repeated measurement. The residuals in both directions indicate that there is a deviation in the topology relationship and both need to be corrected in subsequent processing. Similarly, the same operation is performed on the instantaneous power sequence to obtain the power residual value. Perform the above aggregation and residual calculation operations on each candidate association path in the preliminary topology association attribution table to obtain the current residual value and power residual value corresponding to each candidate association path.
[0038] After obtaining the current and power residual values of all candidate associated paths, an initial weight value is assigned to each candidate associated path in the preliminary topology association table. Before obtaining any verification information, the initial weight value assigns the same confidence to each candidate associated path. According to the system default configuration, the initial weight value is set to 1. This initial weight value setting ensures that each candidate associated path is on an equal footing when entering the physical verification stage. Subsequent correction operations are performed independently based on the residual size of each path, avoiding the introduction of human bias due to differences in initial assignment.
[0039] Calculate the percentage deviation of the current residual value and the power residual value from the measured current value and the measured power value of the upstream node, respectively. The specific calculation method for the percentage deviation is to divide the absolute value of the current residual value by the measured current value and then multiply by 100 to obtain the current deviation percentage value, and divide the absolute value of the power residual value by the measured power value and then multiply by 100 to obtain the power deviation percentage value.
[0040] The preset physical reliability thresholds are set based on the statistical results of inherent deviations caused by metering errors and line losses during normal operation of the distribution substation. The current deviation percentage threshold is set at 5%, and the power deviation percentage threshold at 8%. These two thresholds are set independently because the current and power measurements involve different accuracy levels of the instrument transformers. The typical accuracy of a current transformer is 0.5% to 1%, and power calculation is further affected by power factor measurement errors, thus its inherent deviation is greater than that of current measurement. When the current deviation percentage exceeds the current deviation percentage threshold or the power deviation percentage exceeds the power deviation percentage threshold, i.e., either physical quantity deviates beyond its respective allowable range, the candidate correlation path is determined to be physically unreliable. The initial weight value of this candidate correlation path is multiplied by an attenuation coefficient less than one. This attenuation coefficient is set to 0.3 to 0.8 based on the statistical results of metering deviation distribution during actual operation of the distribution substation; a larger deviation percentage results in a smaller attenuation coefficient.
[0041] For example, a value of 0.4 is selected when the current deviation percentage is 20% or the power deviation percentage is 25%, and a value of 0.7 is selected when the current deviation percentage is 8% or the power deviation percentage is 10%. When both the current deviation percentage and the power deviation percentage are less than or equal to their respective physical reliability thresholds, the initial weight value of this candidate association path is multiplied by a gain coefficient greater than one. This gain coefficient is set to 1.1 to 1.5 based on the statistical results of the residual distribution of the correct path in the topology association. The smaller the deviation percentage, the larger the gain coefficient is selected. For example, a value of 1.5 is selected when the current deviation percentage is 2% and the power deviation percentage is 3%, and a value of 1.2 is selected when the current deviation percentage is 4% and the power deviation percentage is 5%.
[0042] All the weight values after the coefficients have been multiplied together constitute the modified weighted coefficient. Each element in the modified weighted coefficient corresponds to the final weight value of a candidate association path. This weight value reflects the original position information of the path in the preliminary topological association table and incorporates the physical credibility evaluation given by Kirchhoff's conservation law, thus providing a physically verified weight basis for the adaptive adjustment of the subsequent dynamic decision threshold.
[0043] In the fifth step, the corrected weighting coefficient is sent back to the second step as a feedback signal. Based on this corrected weighting coefficient, the automatic calculation basis of the dynamic decision threshold in the second step is adjusted to update the value of the dynamic decision threshold. The specific operation of this adjustment is as follows: The weight value corresponding to each element in the modified weighted coefficient is used as the weight coefficient when the amplitude mean and standard deviation are weighted and summed in the second step. The larger the weight value of the candidate association path, the higher the contribution ratio of the monitoring node corresponding to the amplitude mean and standard deviation in the statistical window. The updated dynamic decision threshold value drives the iterative execution of the third and fourth steps in turn.
[0044] When adjusting the automatic calculation basis of the dynamic decision threshold in the second step based on the modified weighting coefficient, an additional waveform change gradient direction consistency discrimination process is introduced. This discrimination process, independent of Kirchhoff's conservation check, provides an auxiliary verification dimension based on waveform morphology similarity. The two verification mechanisms are independent and complementary, and either dimension can trigger weight correction independently. The direction of numerical change of two adjacent sampling points in the dynamic electrical operation data group of each monitoring node is defined as an upward trend direction, a downward trend direction, or a flat trend direction. An upward trend direction indicates that the value of the later sampling point is greater than that of the previous sampling point, a downward trend direction indicates that the value of the later sampling point is less than that of the previous sampling point, and a flat trend direction indicates that the value of the later sampling point is equal to that of the previous sampling point. The frequency of each trend direction occurring for each monitoring node within the entire continuous time window is counted to form the trend direction frequency vector of this monitoring node. This trend direction frequency vector has three dimensions, with the three components being the number of occurrences of the upward trend direction, the number of occurrences of the downward trend direction, and the number of occurrences of the flat trend direction, respectively.
[0045] Calculate the directional overlap ratio between the trend direction frequency vectors of every two monitoring nodes. The specific calculation method for this directional overlap ratio is as follows: The upward trend direction component of the trend direction frequency vector of the first monitoring node is compared with the upward trend direction component of the trend direction frequency vector of the second monitoring node. The smaller value is divided by the larger value to obtain the upward direction overlap ratio. The same operation is performed on the downward trend direction and the flat trend direction to obtain the downward direction overlap ratio and the flat direction overlap ratio. When the smaller and larger values of a certain direction component are both zero, the overlap ratio of that direction is directly assigned to 1 to avoid the abnormal zero denominator in the division operation. The overlap ratios of the three directions are added together and divided by three to obtain the direction overlap ratio.
[0046] When the directional overlap ratio exceeds the preset directional consistency threshold, this preset directional consistency threshold is set to 0.6 to 0.8 based on empirical data of waveform morphology similarity discrimination. The higher the threshold value, the stricter the judgment, meaning that weight enhancement is only triggered when the waveform morphology is highly consistent. The lower the threshold value, the more lenient the judgment, meaning that more monitoring node pairs will be judged as directionally consistent. The corresponding weight value in the correction weighting coefficient between these two monitoring nodes is multiplied by three. This multiplier is determined to be 3 times based on experimental data of typical enhancement multipliers for directional consistency correction. This multiplier achieves the best balance between convergence speed improvement and stability within the selectable range of 1.5 to 5 times. The weight value multiplied by three is used to re-drive the iterative update of the dynamic decision threshold, thereby accelerating the iterative process of satisfying the preset convergence threshold.
[0047] For example, if the trend direction frequency vector of one monitoring node is 45 times of rising, 38 times of falling, and 17 times of remaining unchanged, and the trend direction frequency vector of another monitoring node is 42 times of rising, 40 times of falling, and 18 times of remaining unchanged, then the overlapping sub-proportion of the rising direction is 42 divided by 45, which equals 0.933; the overlapping sub-proportion of the falling direction is 38 divided by 40, which equals 0.950; and the overlapping sub-proportion of the remaining direction is 17 divided by 18, which equals 0.944. After adding the three sub-proportions and dividing by three, we get the direction overlap ratio of 0.942. This value is greater than the preset direction consistency threshold of 0.7. Therefore, the corresponding weight value in the corrected weighting coefficient of the candidate association path between these two monitoring nodes is multiplied by three to strengthen the influence of this path in subsequent iterations.
[0048] During each iteration, when both the current residual value and the power residual value are less than the preset convergence threshold, the final topology identification result is output and the iteration is terminated. The preset convergence threshold is set to 0.05 amperes for the current residual and 2 watts for the power residual, based on the inherent error of the distribution substation metering device and the statistical results of the residual caused by normal line losses. This value is calculated based on a substation rated capacity of 100 kVA and a rated voltage of 220 V. For substations with different capacities, the value can be linearly scaled according to the rated current ratio. When both the current residual value and the power residual value are less than their respective corresponding convergence thresholds, it indicates that the current topology association table has reached an acceptable accuracy in the sense of physical conservation.
[0049] When the number of iterations reaches the preset maximum number of iterations and the current residual value or power residual value is still not less than the preset convergence threshold, the iteration is forcibly terminated, and the current preliminary topology association and attribution table is output as the final topology identification result. At the same time, a timeout alarm signal is generated. The preset maximum number of iterations is set to 50 to 100 times based on the algorithm convergence speed statistics. The lower limit of 50 times ensures that there are enough iteration rounds to reduce the residual to below the convergence threshold under most normal operating conditions, while the upper limit of 100 times avoids excessive occupation of system computing resources in extreme cases.
[0050] When the iteration count reaches the preset maximum number of iterations and the iteration is forcibly terminated, and the current preliminary topology association table is output as the final topology identification result, a timeout alarm signal is generated and recorded, which includes the following information: the specific monitoring node number that triggered the forced exit, the current residual value and power residual value corresponding to this monitoring node, and the final iteration round number when the preset maximum number of iterations is reached. This timeout alarm signal is reported to the main station system through the uplink communication channel of the distribution transformer area to prompt the operation and maintenance personnel to conduct on-site verification. For example, when the current residual value is 0.12 amps and the power residual value is 3.5 watts in the 75th iteration, both exceeding the convergence threshold, and the preset maximum number of iterations is 80, the system continues to iterate until the 80th iteration. If the current residual value is still 0.10 amps and the power residual value is still 3.2 watts after the 80th iteration, the iteration is forcibly terminated and the monitoring node number that triggered the exit is recorded as node 07, the current residual value is 0.10 amps, the power residual value is 3.2 watts, and the final iteration round number is 80. This information is encapsulated as a timeout alarm signal and reported to the main station system. The main station system is a background management platform deployed in the distribution network dispatch or operation and maintenance center, used to receive data and alarm signals reported by terminal equipment in each distribution substation, and to remotely monitor, configure parameters, and manage the operating status of all terminals in the area under its jurisdiction.
[0051] The above algorithms or formulas are all dimensionless and numerical calculations, and the results are obtained by software simulation based on a large amount of collected data to obtain the most recent real-world results. The preset parameters are set by those skilled in the art according to the actual situation.
[0052] It should be understood that in the various embodiments of this application, the sequence number of each process does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.
[0053] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.
[0054] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.
[0055] The above are merely specific embodiments 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 method for identifying the topology of a distribution radio station area, characterized in that, Includes the following steps: The first step is to collect dynamic electrical operation data sets of each monitoring node in the distribution substation within a continuous time window. The dynamic electrical operation data sets include at least voltage instantaneous sequences, current instantaneous sequences, and power instantaneous sequences. Based on the dynamic electrical operation data sets, the dynamic jump feature vector corresponding to the load switching event of each monitoring node is extracted. The second step is to receive the dynamic jump feature vector and automatically calculate the dynamic decision threshold based on the degree of dispersion of the numerical distribution of the dynamic jump feature vector within the current time window. The dynamic decision threshold replaces the fixed empirical threshold and is used to dynamically divide the boundaries for subsequent node similarity determination. The third step is to obtain the dynamic decision threshold, and based on the dynamic decision threshold, perform time-series elastic alignment processing on the dynamic electrical operation data groups of each monitoring node. Then, calculate the correlation strength value between each monitoring node after the time-series elastic alignment processing, and then generate a preliminary topology association relationship table based on the correlation strength value and the dynamic decision threshold. The fourth step involves receiving the preliminary topology association attribution table and the dynamic electrical operation data set. Based on Kirchhoff's current conservation law, the current residual value of each monitoring node under the preliminary topology association attribution table is calculated. Simultaneously, based on Kirchhoff's power conservation law, the power residual value of each monitoring node under the preliminary topology association attribution table is calculated. Then, based on the current residual value and the power residual value, a modified weighting coefficient is generated. The modified weighting coefficient is used to characterize the physical reliability of each candidate association path in the preliminary topology association attribution table. The fifth step is to send the corrected weighting coefficient back to the second step as a feedback signal, and adjust the automatic calculation basis of the dynamic decision threshold in the second step according to the corrected weighting coefficient, so as to update the value of the dynamic decision threshold, and drive the iterative execution of the third and fourth steps in sequence. During each iteration, when both the current residual value and the power residual value are less than the preset convergence threshold, the final topology identification result is output and the iteration is terminated; when the number of iterations reaches the preset maximum number of iterations and the current residual value or the power residual value is still not less than the preset convergence threshold, the iteration is forcibly terminated, the current preliminary topology association table is output as the final topology identification result, and a timeout alarm signal is generated.
2. The method for identifying the topology of a distribution radio station area according to claim 1, characterized in that, In the first step, the extraction process of the dynamic jump feature vector is triggered by the absolute value of the difference between adjacent sampling points of the instantaneous voltage sequence or instantaneous current sequence of each monitoring node exceeding the preset start capture threshold value. When the trigger condition is met, the waveform segments before and after the recording time are recorded as load switching events, and the amplitude change rate, duration and zero-crossing offset of the load switching events are combined to form the dynamic jump feature vector.
3. The method for identifying the topology of a distribution radio station according to claim 2, characterized in that, In the second step, the specific implementation method for automatically calculating the dynamic decision threshold based on the dispersion of the numerical distribution of the dynamic jump feature vector within the current time window is as follows: The mean amplitude and standard deviation of all dynamic jump feature vectors within the current time window are statistically analyzed. The mean amplitude and standard deviation are added together and then divided by two. The resulting value is used as the initial value of the dynamic decision threshold. When the change amplitude of the dynamic decision threshold calculated in two adjacent time windows exceeds the preset stability judgment limit, the difference between the change amplitude and the stability judgment limit is added to the initial value of the dynamic decision threshold to form the dynamic decision threshold at the current moment.
4. The method for identifying the topology of a distribution radio station according to claim 3, characterized in that, In the third step, the specific execution method of timing elastic alignment is as follows: Using the voltage instantaneous sequence in the dynamic electrical operation data set as the alignment reference, the current instantaneous sequence and power instantaneous sequence of each monitoring node are scaled proportionally according to the time offset of their respective peak times relative to the peak times of the voltage instantaneous sequence. This ensures that the current instantaneous sequence and power instantaneous sequence of all monitoring nodes achieve maximum overlap and matching with the voltage instantaneous sequence on the time axis. For data missing locations due to communication packet loss, the average value of the two sampling points before and after the location is used for interpolation to fill in the missing data, thus obtaining a complete and equal-length dynamic electrical operation data set.
5. The method for identifying the topology of a distribution radio station according to claim 4, characterized in that, In the third step, when calculating the correlation strength between each monitoring node, the waveform dynamic time warping path cost calculation method is adopted: For any two monitoring nodes' dynamic electrical operation data sets, their instantaneous voltage sequences are used as the matching objects. The sum of the absolute differences between the two nodes at all corresponding sampling points on the time axis is calculated, and then divided by the sum of their respective amplitude standard deviations to obtain the initial path cost value. When the sum of their amplitude standard deviations is greater than zero, a division operation is performed. When the sum of their amplitude standard deviations is equal to zero, the initial path cost value is directly assigned a preset maximum constant. Then, the process is recursively traced back to find the minimum interval segmented path corresponding to the initial path cost value on the time axis. The difference between the endpoint cost and the starting cost of the minimum interval segmented path is used as the corrected normalized cost value. When the corrected normalized cost value is greater than zero, the reciprocal of the corrected normalized cost value is used as the correlation strength value. When the corrected normalized cost value is equal to zero, the correlation strength value is directly assigned a preset maximum constant. The smaller the corrected normalized cost value, the larger the correlation strength value.
6. The method for identifying the topology of a distribution radio station according to claim 5, characterized in that, In the fourth step, before calculating the current residual value and the power residual value, the instantaneous current sequences of all lower-level monitoring nodes belonging to the same upper-level node in the preliminary topology association table are summed at the same time to obtain the current aggregate value at the upper-level node. At the same time, the instantaneous power sequences of all lower-level monitoring nodes belonging to the same upper-level node are summed at the same time to obtain the power aggregate value at the upper-level node. Then, the difference between the current aggregate value and the actual total value of the measured current at the upper-level node is taken as the current residual value, and the difference between the power aggregate value and the actual total value of the measured power at the upper-level node is taken as the power residual value.
7. The method for identifying the topology of a distribution radio station according to claim 6, characterized in that, In the fourth step, the method for generating the weighting coefficients is corrected as follows: First, each candidate association path in the preliminary topology association table is assigned an initial weight value. Then, the deviation percentages of the current residual value and power residual value relative to the measured current value and measured power value of the monitoring node are calculated. When the deviation percentage is greater than the preset physical reliability threshold, the initial weight value of the candidate association path is multiplied by an attenuation coefficient less than one. When the deviation percentage is less than or equal to the physical reliability threshold, the initial weight value of the candidate association path is multiplied by a gain coefficient greater than one. All weight values after the coefficient multiplication together constitute the corrected weighting coefficient.
8. The method for identifying the topology of a distribution radio station according to claim 7, characterized in that, In the fifth step, when the number of iterations reaches the preset maximum number of iterations and the iteration is forcibly terminated, and the current preliminary topology association table is output as the final topology identification result, the timeout alarm signal generated and recorded at the same time includes the following information: the specific monitoring node number that triggered the forced exit, the current residual value and power residual value corresponding to the monitoring node, and the final iteration round number when the preset maximum number of iterations is reached. The timeout alarm signal is reported to the master station system through the uplink communication channel of the distribution radio area.
9. The method for identifying the topology of a distribution radio station according to claim 8, characterized in that, In the fifth step, when adjusting the automatic calculation basis of the dynamic decision threshold in the second step according to the modified weighting coefficient, an additional waveform change gradient direction consistency judgment process is introduced: The direction of numerical change between two adjacent sampling points in the dynamic electrical operation data set of each monitoring node is defined as an upward trend, a downward trend, or a flat trend. The frequency of each trend direction occurring in the entire continuous time window of each monitoring node is counted to form the trend direction frequency vector of the monitoring node. Then, the directional overlap ratio between the trend direction frequency vectors of every two monitoring nodes is calculated. When the directional overlap ratio is greater than the preset directional consistency threshold, the corresponding weight value in the corrected weighting coefficient between the two monitoring nodes is multiplied by three, and the weight value after multiplying by three is used to drive the iterative update of the dynamic decision threshold.
