Power distribution network multi-harmonic source positioning method and device based on harmonic impedance online identification

Abstract

The embodiment provides a power distribution network multi-harmonic source positioning method based on harmonic impedance online identification, establishes a harmonic network model containing parameter uncertainty interval and selects minimum measurement point configuration satisfying observability, synchronously collects three-phase harmonic voltage and current, generates harmonic voltage phasor and harmonic current phasor through windowed interpolation analysis, extracts phasor fluctuation and identifies network impedance parameters of each harmonic through a recursive algorithm with a forgetting factor, estimates harmonic source injection current of each node under sparse constraint according to the network impedance parameters, generates harmonic source positioning results through spatial clustering and contribution separation, calculates the contribution of each harmonic source to the total harmonic distortion rate of the voltage of the point of common coupling, generates harmonic responsibility division results through Monte Carlo evaluation and normalization, continuously monitors responsibility trend, and shortens the update cycle to trigger full recalculation when detecting a power distribution network operation event.

Description

Technical Field

[0001] This application relates to the field of power quality analysis and distribution network monitoring technology, specifically to a method and device for locating multiple harmonic sources in a distribution network based on online harmonic impedance identification. Background Technology

[0002] Existing harmonic source location methods are mostly based on steady-state harmonic power flow analysis, assuming the system's harmonic impedance to be a known, fixed value. However, the harmonic impedance of actual distribution networks changes dynamically with load switching, capacitor bank switching, and distributed generation output. The fixed impedance assumption leads to significant deviations between the location results and reality, and the accuracy decreases during periods of drastic harmonic impedance changes. Existing methods often employ power direction or impedance angle criteria for single harmonic sources. When multiple harmonic sources exist simultaneously in a distribution network, their harmonic currents are superimposed and coupled within the network, making it difficult for single-source location methods to distinguish the independent contributions of each source, resulting in a high error rate in multi-source scenarios.

[0003] Existing harmonic responsibility allocation methods rely on accurate acquisition of reference impedance. Commonly used background harmonic methods and fluctuation methods, when determining the harmonic contributions on the system side and the user side, have strict requirements on measurement conditions and background harmonic stability. In distribution networks with large background harmonic fluctuations, the uncertainty of responsibility allocation results is relatively high, making it difficult to support fair power quality management. Existing methods often rely on experience to set measurement points or adopt a full-coverage strategy, lacking systematic analysis of the observability of measurement points. Insufficient measurement points lead to location blind spots, while redundant measurement points increase investment and maintenance costs. It is difficult to balance the economy and effectiveness of measurement resource allocation.

[0004] Existing methods are mostly offline analyses, requiring manual data collection and import into analysis software for calculations. The time required from identifying harmonic problems to completing location analysis is long, making it difficult to meet the real-time requirements of distribution network operation and to handle scenarios with dynamically changing harmonic source locations. Existing harmonic power direction location schemes can determine the harmonic direction at a single point, but they do not address online identification of system harmonic impedance, decoupling and locating multiple harmonic sources, quantifying and separating harmonic contributions, or optimizing the configuration of measurement points. Therefore, a method for locating multiple harmonic sources in distribution networks that combines online harmonic impedance identification, multi-source decoupling and location, harmonic responsibility quantification, and measurement configuration optimization is needed. Summary of the Invention

[0005] To address the problems of existing harmonic source location methods that assume fixed impedance, struggle to distinguish independent contributions from multiple sources, and impose strict requirements on background harmonic stability for responsibility allocation, and are mostly performed offline, this application provides a method and device for locating multiple harmonic sources in distribution networks based on online harmonic impedance identification. This method can identify time-varying harmonic impedances online through fluctuation analysis, separate and decouple multiple source contributions through sparse constraint state estimation and spatial clustering, quantify harmonic responsibility through contribution phasor synthesis and Monte Carlo evaluation, and continuously monitor responsibility trends and event triggers to re-identify and track the dynamic changes in harmonic source location and responsibility, thereby improving the accuracy of harmonic source location, the certainty of responsibility allocation, and the real-time performance of online analysis.

[0006] To solve at least one of the above problems, this application provides the following technical solution:

[0007] Firstly, this application provides a method for locating multiple harmonic sources in a distribution network based on online harmonic impedance identification, comprising: establishing a harmonic network model containing the admittance matrix of each harmonic node and the uncertainty interval of parameters according to the distribution network topology and equipment fundamental parameters, selecting a configuration of measurement points that meet observability and have the smallest number, synchronously collecting three-phase harmonic voltage and current at the configuration of the measurement points, and generating harmonic voltage phasors and harmonic current phasors through windowed interpolation harmonic analysis, sequence component decomposition, and data quality inspection; extracting the fluctuation of the harmonic voltage phasors and the harmonic current phasors, screening effective fluctuation events, and identifying each harmonic source using a recursive algorithm with a forgetting factor in conjunction with the harmonic network model. The network impedance parameters of the subharmonics are used to estimate the harmonic source injection current of each node under sparse constraints, and the harmonic source location results are generated through spatial clustering, background harmonic separation, and contribution separation. Based on the harmonic source location results, the contribution of each harmonic source to the total harmonic distortion rate of the voltage at the point of common coupling is calculated, and the harmonic responsibility allocation results are generated through Monte Carlo uncertainty assessment and normalization. The responsibility trend of the harmonic responsibility allocation results is continuously monitored and updated according to a preset update cycle. When a distribution network operation event is detected, the update cycle is shortened to trigger a full recalculation of re-identification and location. When the responsibility ratio exceeds the preset alarm threshold, an alarm is pushed and linked with billing.

[0008] Furthermore, it also includes: converting the fundamental impedance of each branch into the harmonic branch impedance based on the distribution network topology and equipment fundamental parameters, then accumulating the connection branch admittance node by node to construct a node harmonic admittance matrix, and adding parameter uncertainty intervals representing calibration error and aging offset to each element of the node harmonic admittance matrix; establishing the observability relationship between the measured quantity and the harmonic source injection current of each node based on the node harmonic admittance matrix and verifying the rank of the measurement matrix, and using a greedy strategy to successively select the node with the largest observability gain under the constraint of minimizing the number of measurement points until the measurement point configuration is fully observable.

[0009] Furthermore, it also includes: at the measurement point configuration location, sampling synchronization is achieved through unified timing and the sampling rate meets the requirements of characteristic harmonic analysis to acquire three-phase harmonic voltage and current waveforms; the amplitude and phase of each harmonic are extracted from the three-phase harmonic voltage and current waveforms using windowed interpolation harmonic analysis; the amplitude and phase of each harmonic are decomposed into positive sequence, negative sequence, and zero sequence components, and data quality checks are performed on amplitude rationality, phase continuity, and three-phase balance; data that fails the check are registered with an anomaly flag, then weighted and processed, and harmonic voltage phasors and harmonic current phasors are generated through sliding window mid-value filtering.

[0010] Furthermore, it also includes: calculating the voltage fluctuation and current fluctuation at adjacent moments for the continuously acquired harmonic voltage phasor and harmonic current phasor; screening effective fluctuation events based on the lower and upper threshold conditions of the fluctuation amplitude; estimating the equivalent harmonic impedance by weighted least squares synthesis of multiple effective fluctuation events with the current fluctuation amplitude as the weight; establishing a collaborative identification model in conjunction with the harmonic network model and tracking the time-varying harmonic impedance using a recursive least squares algorithm with a forgetting factor; and generating the network impedance parameters for each harmonic after correcting the identification results that do not meet the physical constraints to the constraint boundary.

[0011] Furthermore, it also includes: establishing a linear measurement equation between the measurement vector and the harmonic source injection current of each node based on the network impedance parameters; determining the weights using the measurement variance and solving for the harmonic source injection current of each node using weighted least squares with added sparsity constraints; performing density-based spatial clustering to identify the harmonic source region for the harmonic source injection current of each node according to the electrical distance and injection characteristics; separating the independent contribution of each source to the measurement point based on the superposition principle and modeling the background harmonics on the system side as equivalent harmonic voltage sources for separation; and generating the harmonic source localization result through multi-time Bayesian fusion.

[0012] Furthermore, it also includes: calculating the contribution phasor of each harmonic voltage at the common junction point when each harmonic source acts independently based on the harmonic source location results; synthesizing the equivalent voltage total harmonic distortion rate of each source according to the harmonic order of each contribution phasor; determining the original responsibility ratio of each source by the ratio of the contribution phasor to the total voltage total harmonic distortion rate; randomly sampling the network impedance parameters and the harmonic source injection current of each node multiple times within their confidence intervals and repeatedly calculating the responsibility ratio; evaluating the uncertainty by the mean and standard deviation of the responsibility ratio; and generating the harmonic responsibility allocation result by normalizing the original responsibility ratio of each source and adding an uncertainty interval.

[0013] Furthermore, it also includes: completing the process from data acquisition to responsibility assessment in each cycle according to a preset update cycle and performing incremental reuse of intermediate results with changes within the threshold condition; shortening the update cycle and triggering a full recalculation when distribution network switch action and load switching events are detected; recording the harmonic responsibility division results as a time series and analyzing the upward trend and abrupt events of each harmonic source responsibility; pushing an alarm to the power quality management platform and linking with power billing to complete responsibility tracing when the responsibility ratio exceeds the preset alarm threshold condition; and storing the location results and responsibility division results of each update cycle into a historical database for subsequent analysis.

[0014] Secondly, this application provides a multi-harmonic source location device for distribution networks based on online harmonic impedance identification, comprising: a network modeling module, used to convert the fundamental impedance of each branch into the impedance of each harmonic branch according to the distribution network topology and equipment fundamental parameters, construct a node harmonic admittance matrix containing parameter uncertainty intervals, and select the minimum number of measurement points that meet observability requirements after verifying observability; a data acquisition module, used to synchronously acquire three-phase harmonic voltage and current waveforms at the measurement point configuration locations through unified time synchronization, and generate harmonic voltage phasors and harmonic current phasors from the three-phase harmonic voltage and current waveforms through windowed interpolation harmonic analysis, sequence component decomposition, and data quality inspection; and an impedance identification module, used to extract the fluctuation of the harmonic voltage phasors and the harmonic current phasors and filter valid ones. For fluctuation events, the network impedance parameters of each harmonic are identified and tracked using a recursive least squares algorithm with a forgetting factor, based on the node harmonic admittance matrix. A location estimation module estimates the harmonic source injection current of each node using weighted least squares under sparse constraints based on the network impedance parameters, and generates harmonic source location results by spatial clustering, background harmonic separation, and contribution separation of the harmonic source injection current of each node. A responsibility monitoring module generates harmonic responsibility allocation results based on the harmonic source location results through contribution phasor synthesis and Monte Carlo uncertainty assessment, and continuously monitors the responsibility trend and abrupt events of the harmonic responsibility allocation results. When a distribution network operation event is detected, the update cycle is shortened to trigger a full recalculation, and an alarm is pushed and linked with billing when the responsibility ratio exceeds a preset alarm threshold.

[0015] Thirdly, this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the method for locating multiple harmonic sources in a distribution network based on online harmonic impedance identification.

[0016] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method for locating multiple harmonic sources in a distribution network based on online harmonic impedance identification.

[0017] Fifthly, this application provides a computer program product, including a computer program / instructions, which, when executed by a processor, implement the steps of the method for locating multiple harmonic sources in a distribution network based on online harmonic impedance identification.

[0018] As can be seen from the above technical solution, this application provides a method and device for locating multiple harmonic sources in a distribution network based on online harmonic impedance identification. It parametrically models the harmonic impedance of each harmonic in the distribution network using a harmonic network model and identifies and tracks time-varying impedances online. It separates the independent contributions of each source under multi-source coupling conditions and decouples and locates them by using sparse constraint state estimation and spatial clustering. It quantifies the responsibility ratio of each harmonic source to the common connection point by using contribution phasor synthesis and Monte Carlo uncertainty assessment. It forms a closed loop by continuously monitoring the responsibility trend and re-identifying and locating the harmonic source location and responsibility through event triggering. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below.

[0020] Figure 1 This is a flowchart illustrating the method for locating multiple harmonic sources in a distribution network based on online harmonic impedance identification in an embodiment of this application. Figure 2 This is a schematic diagram of the hierarchical architecture and device structure of the power quality monitoring system for the power distribution network in this application embodiment; Figure 3 This is a schematic diagram of the process for modeling, optimizing the configuration of measurement points, and performing observability analysis of the distribution network harmonic network in this application embodiment; Figure 4 This is a schematic diagram of the harmonic data synchronous acquisition and preprocessing and harmonic impedance online identification process in the embodiments of this application; Figure 5 This is a schematic diagram of the sparse constraint state estimation and multi-harmonic source decoupling and localization process in an embodiment of this application. Figure 6 This is a schematic diagram of the closed-loop process of harmonic liability quantification assessment, online continuous monitoring, and event-triggered full recalculation in the embodiments of this application. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of this application, but not all embodiments.

[0022] Considering that existing harmonic source location methods often assume that the system harmonic impedance is a fixed value, have difficulty distinguishing the independent contributions of multiple harmonic sources, and require strict background harmonic stability for responsibility allocation, and are mostly offline analyses, this application provides a distribution network multi-harmonic source location method based on online harmonic impedance identification. Through online harmonic impedance identification, sparse constraint state estimation, multi-source decoupling location, and responsibility quantification assessment, multiple harmonic pollution sources in the distribution network can be accurately located and their responsibility quantified.

[0023] This application provides an embodiment of a method for locating multiple harmonic sources in a distribution network based on online harmonic impedance identification. See [link to relevant documentation]. Figure 1 The method specifically includes the following:

[0024] Step S101: Based on the distribution network topology and equipment fundamental parameters, establish a harmonic network model containing the admittance matrix of each harmonic node and the parameter uncertainty interval, and select the measurement point configuration that satisfies observability and has the smallest number. Simultaneously collect the three-phase harmonic voltage and current at the measurement point configuration, and generate harmonic voltage phasors and harmonic current phasors through windowed interpolation harmonic analysis, sequence component decomposition and data quality inspection.

[0025] In this embodiment, the method operates on a power quality monitoring system for a power distribution network, such as... Figure 1 As shown, the system is organized into functional components including harmonic network modeling, data acquisition, impedance identification, location estimation, and responsibility monitoring. The harmonic network model is organized using fields such as node number, branch connection relationship, impedance of each harmonic branch, and uncertainty interval of parameters. The node number is a globally unique code.

[0026] In one feasible implementation, the harmonic network model is constructed based on the topology and equipment ledger parameters of the power distribution management system, such as... Figure 3 As shown, a tree topology description is established for the radial network. For the network containing tie switches, the actual operating topology is determined based on the current switch status. The fundamental resistance and inductance of each branch, the transformer short-circuit impedance and the capacitor capacitive reactance are obtained from the equipment ledger.

[0027] Based on this, the fundamental impedance parameter is converted into the impedance of each harmonic branch. The line harmonic resistance is included with the skin effect correction factor, the reactance component is proportional to the harmonic order, the transformer harmonic impedance is taken as the leakage reactance is approximately proportional to the order, the capacitor harmonic reactance is inversely proportional to the order, and the node harmonic admittance matrix is ​​constructed by accumulating the admittance of each connected branch node by node.

[0028] Optionally, an uncertainty range of parameters is added to each element of the node harmonic admittance matrix to characterize the calibration error and aging offset. The parameters of the line, transformer and capacitor are set according to their respective preset uncertainty ratios, and the ratio values ​​are based on the calibration records and operating years of the equipment ledger.

[0029] In the parallel processing path, the observable relationship between the measured quantities and the harmonic source injection currents at each node is established, and the rank of the measurement matrix is ​​verified, such as... Figure 3 As shown, a greedy strategy is used to successively select nodes that maximize the rank gain of the measurement matrix, starting from an empty configuration, until the rank of the measurement matrix equals the dimension of the state to be estimated. Key measurement points whose missing data will cause some states to be unobservable are identified, and appropriate redundant measurements are configured.

[0030] Perform synchronous acquisition and preprocessing at the selected measurement point configuration, such as... Figure 4 As shown, each measurement point achieves sampling synchronization through unified time synchronization and acquires three-phase waveforms at a sampling rate that meets the requirements of characteristic harmonic analysis. The amplitude and phase of each harmonic are extracted through windowed interpolation harmonic analysis. Data quality inspection registers and de-weights abnormal data such as amplitude exceeding limits, phase abrupt changes, and three-phase imbalance.

[0031] The amplitude and phase of each harmonic are decomposed into positive-sequence, negative-sequence, and zero-sequence components, and then filtered through a sliding window to generate harmonic voltage phasors and harmonic current phasors. The harmonic voltage phasors and harmonic current phasors are written into the measurement cache with the node number and timestamp as the primary key, so that step S401 can read them in the fluctuation quantity extraction entry according to the timestamp order. The median filter window length is set according to the time scale of interest.

[0032] Step S102: Extract the fluctuation of the harmonic voltage phasor and the harmonic current phasor and screen the effective fluctuation events. Combine the harmonic network model with a recursive algorithm with a forgetting factor to identify the network impedance parameters of each harmonic. Based on the network impedance parameters, estimate the harmonic source injection current of each node under sparse constraints and generate the harmonic source location result through spatial clustering, background harmonic separation and contribution separation.

[0033] The harmonic voltage phasor and harmonic current phasor output in step S101 are read into the impedance identification stage, such as... Figure 4 As shown, adjacent time-time data are associated with node numbers and timestamps. The difference between harmonic voltage phasors at adjacent time-times is set as voltage fluctuation, and the difference between harmonic current phasors is set as current fluctuation. Nodes with misaligned voltages are registered with an alignment missing flag and transferred to the supplementary data collection queue.

[0034] In one specific implementation, valid fluctuation events are screened based on a lower threshold condition and an upper threshold condition for the fluctuation amplitude. The lower threshold condition is set to a preset multiple of the measured noise level to ensure the signal-to-noise ratio, and the upper threshold condition is set to the statistical upper limit of the fluctuation amplitude during normal operation to exclude topological changes and fault disturbances. Fluctuations falling outside the interval are not included in the identification and are not registered as invalid fluctuations.

[0035] Based on this, the equivalent harmonic impedance is estimated by weighted least squares summation of multiple effective fluctuation events using the current fluctuation amplitude as the weight. A collaborative identification model is established for the effective fluctuation events at multiple measurement points using the harmonic network model. A recursive least squares algorithm with a forgetting factor is used to track the time-varying harmonic impedance. The forgetting factor is calibrated based on the impedance change rate. When capacitor switching is detected, the covariance matrix is ​​reset to accelerate the response to sudden changes.

[0036] The identification results must meet the physical constraints that the resistance is positive and the reactance sign is consistent with the device type. Results that do not meet the constraints are corrected to the constraint boundary. When identification fails, the previous valid value is retained and an identification anomaly is registered. The network impedance parameters of each harmonic are written into the impedance model library according to the branch number and harmonic order, for step S501 to read at the state estimation entry.

[0037] After the network impedance parameters in step S402 are ready, perform harmonic source injection current state estimation, such as... Figure 5 As shown. A linear measurement equation is established between the measurement vector and the harmonic source injection current of each node based on the network impedance parameters. The weights are determined by the measurement variance and solved by weighted least squares with additional sparsity constraints. The sparsity constraints cause the injection estimate of nodes without harmonic sources to approach zero.

[0038] Optionally, standardized residuals are calculated for each measurement. Measurements with standardized residuals exceeding a preset threshold are judged as bad data and re-estimated after weight reduction, distinguishing between excessive measurement errors, channel faults, and topology errors. The harmonic source injection current of each node is given an estimated confidence interval to provide uncertainty information for positioning decisions. The sparse penalty coefficient is cross-validated with historical data to balance fitting accuracy and sparsity.

[0039] Accordingly, the harmonic source injection current of each node is used to perform density-based spatial clustering to identify the harmonic source region based on electrical distance and injection characteristics. The independent contribution of each source to the measurement point is separated according to the superposition principle. The background harmonics on the system side are modeled as equivalent harmonic voltage sources and separated. The harmonic source location result is generated by multi-time Bayesian fusion. The harmonic source location result is written into the location ledger with the source region number as the primary key for step S601 to read.

[0040] Step S103: Calculate the contribution of each harmonic source to the total harmonic distortion rate of the voltage at the point of common coupling based on the harmonic source location results, and generate harmonic responsibility allocation results through Monte Carlo uncertainty assessment and normalization. Continuously monitor the responsibility trend of the harmonic responsibility allocation results and update them continuously according to a preset update cycle. When a distribution network operation event is detected, shorten the update cycle to trigger a full recalculation of re-identification and location. When the responsibility ratio exceeds the preset alarm threshold, push an alarm and link it with billing.

[0041] The harmonic source location result output in step S502 is read into the responsibility quantification stage, such as... Figure 6As shown. The contribution phasor of each harmonic voltage at the point of common coupling is calculated when each harmonic source acts independently. The contribution phasor is organized by source region number, harmonic order, amplitude and phase fields. Source regions with missing location results are registered with a missing location identifier and are not included in the responsibility synthesis.

[0042] In one specific implementation, the total harmonic distortion (THD) of the equivalent voltage for each source is synthesized according to the harmonic order of each contributing phasor. The original responsibility ratio of each source is determined by the ratio of the THD of the equivalent voltage of each source to the total harmonic distortion of the total voltage. Harmonic contributions can involve phase enhancement and cancellation, and the sum of the original responsibility ratios of each source may deviate from the overall value, requiring normalization processing.

[0043] The uncertainty of the responsibility calculation results is assessed accordingly. The network impedance parameters and the harmonic source injection currents at each node are randomly sampled multiple times within their confidence intervals, and the responsibility ratio is repeatedly calculated. The number of samplings is set according to the convergence criterion of the responsibility ratio. Out-of-bounds sampling values ​​are clipped to the interval boundaries. The uncertainty is assessed using the mean and standard deviation of the responsibility ratio.

[0044] After normalizing the original responsibility ratio of each source, an uncertainty interval is added to generate the harmonic responsibility division result. The harmonic responsibility division result is organized with fields such as harmonic source location and type, contribution of each harmonic, comprehensive responsibility ratio and uncertainty interval. Failed archiving of responsibility records is registered with an archiving anomaly mark and retried. The responsibility report supports docking with the power quality management system.

[0045] In the parallel processing path, the process from data acquisition to responsibility assessment is completed in each cycle according to a preset update cycle. Incremental reuse is performed on intermediate results with changes within the threshold condition. When impedance changes are within the threshold condition, the current impedance model is maintained. When the network topology remains unchanged, the observability analysis results and measurement matrix are reused. Full recalculation is triggered only when structural changes occur to reduce the computational burden.

[0046] The system monitors the switching actions, fault trips, and high-load switching events in the distribution network. When an event that may affect harmonic source location is detected, the update cycle is shortened, the impedance identification algorithm is reset, and a full recalculation is forced. Regular updates resume after the event subsides. The location results of each update cycle are stored in a historical database, and the data retention period is determined according to power quality management requirements.

[0047] The harmonic responsibility allocation results are recorded as a time series, and the upward trend and abrupt events of the responsibility of each harmonic source are analyzed. When the responsibility ratio exceeds the preset alarm threshold, an alarm is pushed to the power quality management platform and linked with power billing to complete responsibility tracing. When a distribution network operation event is detected, the update cycle is shortened to trigger a full recalculation of re-identification and relocation. The recalculation result is used to update the harmonic network model in step S101 for subsequent applications. The responsibility allocation results of each update cycle are stored in the historical database, thus forming a closed loop of continuous monitoring, event triggering, and re-identification.

[0048] As can be seen from the above description, the distribution network multi-harmonic source localization method based on online harmonic impedance identification provided in this application embodiment can identify time-varying harmonic impedance online through fluctuation analysis, separate multi-source contributions through sparse constraint state estimation and spatial clustering, quantify harmonic responsibility through contribution phasor synthesis and Monte Carlo evaluation, and continuously monitor responsibility trends and trigger re-identification and localization to track the dynamic changes in the location and responsibility of harmonic sources when distribution network operation events are detected.

[0049] In one embodiment of the distribution network multi-harmonic source location method based on online harmonic impedance identification in this application, see [link to relevant documentation]. Figure 3 It can also specifically include the following:

[0050] Step S201: Based on the distribution network topology and equipment fundamental parameters, the fundamental impedance of each branch is converted into the impedance of each harmonic branch, and then the admittance of the connecting branch is accumulated node by node to construct the node harmonic admittance matrix. The uncertainty range of parameters representing calibration error and aging offset is added to each element of the node harmonic admittance matrix.

[0051] Step S202: Based on the node harmonic admittance matrix, establish the observability relationship between the measured quantity and the injected current of each node harmonic source and verify the rank of the measurement matrix. Using a greedy strategy, select the node with the largest observability gain one by one under the constraint of minimizing the number of measurement points until the measurement point configuration is fully observable.

[0052] Step S101, the network modeling step, constructs a harmonic network model based on the distribution network topology and equipment fundamental parameters, such as... Figure 3 As shown, a tree-like topology description is established for the radial network. For networks containing tie switches, the actual operating topology is determined based on the current switch status. The fundamental resistance and inductance of each branch, the transformer short-circuit impedance and the capacitor capacitive reactance are obtained from the equipment ledger. The fundamental impedance of each branch is converted into the harmonic branch impedance. The line harmonic resistance is included with the skin effect correction factor and the reactance is proportional to the harmonic order. The transformer harmonic impedance is taken as the leakage reactance is approximately proportional to the order. The capacitor harmonic reactance is inversely proportional to the order. For nodes containing distributed power sources, the equivalent output impedance of the inverter is incorporated into the self-admittance of the node, and the admittance of the connected branches is accumulated node by node to construct the node harmonic admittance matrix. The uncertainty range of parameters representing the calibration error and aging offset is added to each element of the node harmonic admittance matrix.

[0053] Once the node harmonic admittance matrix is ​​ready, the measurement point optimization configuration is performed, such as... Figure 3As shown. Based on the node harmonic admittance matrix, an observable relationship is established between the measured quantity and the harmonic source injection current of each node. The measured quantity is the harmonic voltage and harmonic current of each measurement point, and the estimated quantity is the harmonic source injection current of each node. The measurement equation is rearranged into the form of measurement matrix multiplied by state vector, and the rank of the measurement matrix is ​​verified. Using a greedy strategy, nodes that maximize the rank gain of the measurement matrix are selected successively from the empty configuration until the rank of the measurement matrix equals the dimension of the state to be estimated. Key measurement points whose data loss will lead to the unobservability of some states are identified, and appropriate redundant measurements are configured to generate a measurement point configuration. The measurement point configuration uses the node number as the primary key for step S301 to call at the acquisition entry. The redundancy is set according to the estimation accuracy and robustness requirements.

[0054] In one embodiment of the distribution network multi-harmonic source location method based on online harmonic impedance identification in this application, see [link to relevant documentation]. Figure 4 It can also specifically include the following:

[0055] Step S301: At the measurement point configuration location, sampling synchronization is achieved through unified time synchronization, and the sampling rate that meets the characteristic harmonic analysis is used to acquire the three-phase harmonic voltage and current waveforms. The amplitude and phase of each harmonic are extracted from the three-phase harmonic voltage and current waveforms by windowed interpolation harmonic analysis.

[0056] Step S302: Decompose the amplitude and phase of each harmonic into positive sequence, negative sequence and zero sequence components, and perform data quality checks on amplitude rationality, phase continuity and three-phase balance. Data that fails the check is registered as abnormal and then processed with reduced weight. Harmonic voltage phasors and harmonic current phasors are generated by sliding window mid-value filtering.

[0057] After the measurement points are configured as described in step S202, synchronous data acquisition is performed, such as... Figure 4 As shown. At the designated measurement point location, each measurement point achieves sampling synchronization through unified time synchronization and acquires three-phase harmonic voltage and current waveforms at a sampling rate that meets the requirements of characteristic harmonic analysis. The amplitude and phase of each harmonic are extracted from the three-phase harmonic voltage and current waveforms using windowed interpolation harmonic analysis. A window type that reduces spectral leakage is selected as the window function, and interpolation correction eliminates spectral offset caused by asynchronous sampling. The phase is based on the system fundamental voltage reference phase. Measurement points that are not synchronized are registered with a synchronization anomaly flag and trigger re-synchronization.

[0058] The amplitude and phase of each harmonic are decomposed into positive-sequence, negative-sequence, and zero-sequence components, such as Figure 4As shown. Data quality checks are performed on the harmonic components obtained from the decomposition for amplitude rationality, phase continuity, and three-phase balance. Data with amplitude deviations exceeding thresholds from historical data or phase abrupt changes exceeding thresholds at adjacent times are flagged as abnormal and then weighted down. Outliers are suppressed by a sliding window median filter to generate harmonic voltage and harmonic current phasors. The harmonic voltage and harmonic current phasors are written to the measurement cache using node number and timestamp as primary keys, for sequential reading by timestamp at the fluctuation extraction entry point in step S401. The median filter window length is set according to the time scale of interest.

[0059] In one embodiment of the distribution network multi-harmonic source location method based on online harmonic impedance identification in this application, see [link to relevant documentation]. Figure 5 It can also specifically include the following:

[0060] Step S401: Calculate the voltage fluctuation and current fluctuation at adjacent moments for the continuously acquired harmonic voltage phasor and harmonic current phasor, and filter valid fluctuation events based on the lower and upper threshold conditions of the fluctuation amplitude.

[0061] Step S402: Using the current fluctuation amplitude as the weight, the equivalent harmonic impedance is estimated by weighted least squares synthesis of multiple effective fluctuation events. A collaborative identification model is established in conjunction with the harmonic network model, and the time-varying harmonic impedance is tracked by a recursive least squares algorithm with a forgetting factor. The identification results that do not meet the physical constraints are corrected to the constraint boundary, and then the network impedance parameters of each harmonic are generated.

[0062] After the harmonic voltage phasor and harmonic current phasor are ready in step S302, fluctuation quantity extraction is performed, such as... Figure 5 As shown. The continuously collected harmonic voltage phasors and harmonic current phasors are associated with adjacent time-stamped data. The difference between the harmonic voltage phasors at adjacent time-stamped moments is defined as the voltage fluctuation, and the difference between the harmonic current phasors is defined as the current fluctuation. Valid fluctuation events are screened based on the lower and upper threshold conditions of the fluctuation amplitude. The lower threshold condition is a preset multiple of the measurement noise level to ensure the signal-to-noise ratio, and the upper threshold condition is the statistical upper limit of the fluctuation amplitude during normal operation to exclude topology changes and fault disturbances. Fluctuations outside the interval are registered as invalid fluctuations and are not included in the identification.

[0063] Once the effective fluctuation event is ready, harmonic impedance identification is performed, such as... Figure 5As shown, the equivalent harmonic impedance is estimated by weighted least squares summation of multiple effective fluctuation events using the current fluctuation amplitude as the weight. A collaborative identification model is established for the effective fluctuation events at multiple measurement points using the harmonic network model. A recursive least squares algorithm with a forgetting factor is used to track the time-varying harmonic impedance. The forgetting factor is calibrated based on the impedance change rate. The covariance matrix is ​​reset when discrete events such as capacitor switching are detected. Identification results that do not meet the constraints of positive resistance and consistent reactance sign with device type are corrected to the constraint boundary. When identification fails, the previous valid value is retained and an identification anomaly is registered. Network impedance parameters for each harmonic are generated and written into the impedance model library according to branch number and harmonic order for reading in step S501 at the state estimation entry.

[0064] In one embodiment of the distribution network multi-harmonic source location method based on online harmonic impedance identification in this application, see [link to relevant documentation]. Figure 5 It can also specifically include the following:

[0065] Step S501: Establish a linear measurement equation between the measurement vector and the harmonic source injection current of each node based on the network impedance parameters, and solve for the harmonic source injection current of each node by weighting least squares and adding sparse constraints after determining the weights based on the measurement variance.

[0066] Step S502: Perform density-based spatial clustering to identify harmonic source regions for the injected current of each node harmonic source according to electrical distance and injection characteristics. Separate the independent contribution of each source to the measurement point according to the superposition principle and model the background harmonics on the system side as equivalent harmonic voltage sources for separation. Generate the harmonic source localization result through multi-time Bayesian fusion.

[0067] After the network impedance parameters are ready in step S402, state estimation is performed, such as... Figure 5 As shown, a linear measurement equation is established between the measurement vector and the harmonic source injection current of each node based on the network impedance parameters. The measurement vector includes the harmonic voltage and branch harmonic current of each measurement point. The weights are determined by the variance of each measurement, and the solution is obtained by weighted least squares with additional sparsity constraints. The sparsity constraints add a sparsity penalty term to the injection current vector in the objective function to make the injection estimation of nodes without harmonic sources approach zero. Measurements with standardized residuals exceeding the preset threshold are judged as bad data and re-estimated after weight reduction to generate the harmonic source injection current of each node. The sparsity penalty coefficient is calibrated by cross-validation with historical data.

[0068] After the harmonic source injection current at each node is ready, multi-source decoupling and localization are performed, such as... Figure 5As shown. Density-based spatial clustering is performed on the injected current of each node's harmonic source based on electrical distance and injection characteristics. Nodes with similar electrical distances and injection characteristics are clustered into a single harmonic source region. The independent contribution of each source to the measurement point is separated based on the superposition principle. The system-side background harmonics are modeled as equivalent harmonic voltage sources at common connection points. Their amplitude and phase are identified and separated. The location results from multiple sampling times are accumulated using Bayesian fusion to generate harmonic source location results based on the confidence scores of the continuously determined nodes. The source region number is used as the primary key for step S601 to read at the responsibility calculation entry.

[0069] In one embodiment of the distribution network multi-harmonic source location method based on online harmonic impedance identification in this application, see [link to relevant documentation]. Figure 6 It can also specifically include the following:

[0070] Step S601: Calculate the contribution phasor of each harmonic voltage at the common junction point when each harmonic source acts independently based on the harmonic source location results. Combine the contribution phasors according to the harmonic order to form the equivalent voltage total harmonic distortion rate of each source and determine the original responsibility ratio of each source by the ratio of the contribution phasor to the total voltage total harmonic distortion rate.

[0071] Step S602: Randomly sample the network impedance parameters and the injected current of each node harmonic source multiple times within their confidence intervals and repeatedly calculate the responsibility ratio. Evaluate the uncertainty using the mean and standard deviation of the responsibility ratio. Normalize the original responsibility ratio of each source and add an uncertainty interval to generate the harmonic responsibility division result.

[0072] After the harmonic source location result in step S502 is ready, responsibility quantization is performed, such as... Figure 6 As shown. Based on the harmonic source location results, the contribution phasors of each harmonic voltage at the common junction point are calculated when each harmonic source acts independently. The contribution phasors are then combined according to the harmonic order to form the total harmonic distortion rate of the equivalent voltage of each source. The original responsibility ratio of each source is determined by the ratio of the total harmonic distortion rate of the equivalent voltage of each source to the total harmonic distortion rate of the total voltage. The phase enhancement and cancellation of harmonic contributions may cause the sum of the original responsibility ratios of each source to deviate from the overall value. Source areas with missing location results are registered with a missing location marker and are not included in the responsibility synthesis.

[0073] Once the original liability ratio is ready, uncertainty assessment and normalization are performed, such as... Figure 6As shown. The network impedance parameters and the harmonic source injection currents at each node are randomly sampled multiple times within their confidence intervals, and the responsibility ratio is repeatedly calculated. The number of sampling times is set according to the convergence criterion of the responsibility ratio. Out-of-bounds sampling values ​​are clipped to the interval boundaries. The uncertainty is assessed using the mean and standard deviation of the responsibility ratio. After normalizing the original responsibility ratios of each source, an uncertainty interval is added to generate the harmonic responsibility allocation result. The harmonic responsibility allocation result is organized with fields for harmonic source location and type, contribution of each harmonic, comprehensive responsibility ratio, and uncertainty interval, for reading at the monitoring inlet in step S701.

[0074] In one embodiment of the distribution network multi-harmonic source location method based on online harmonic impedance identification in this application, see [link to relevant documentation]. Figure 6 It can also specifically include the following:

[0075] Step S701: Complete the process from data acquisition to responsibility assessment in each cycle according to the preset update cycle, and perform incremental reuse of intermediate results with changes within the threshold condition. When the distribution network switch action and load switching event are detected, shorten the update cycle and trigger full recalculation.

[0076] Step S702: Record the harmonic responsibility division results as a time series and analyze the upward trend and abrupt events of the responsibility of each harmonic source. When the responsibility ratio exceeds the preset alarm threshold, push an alarm to the power quality management platform and link with the power billing to complete the responsibility tracing. Store the location results and responsibility division results of each update cycle into the historical database for subsequent analysis.

[0077] After the harmonic responsibility allocation results in step S602 are ready, online continuous monitoring is performed, such as... Figure 6 As shown. The process from data acquisition to responsibility assessment is completed within each cycle according to a preset update cycle. Intermediate results with changes within a threshold condition are incrementally reused. When impedance changes are within the threshold condition, the current impedance model is maintained. When the network topology remains unchanged, observability analysis results and measurement matrices are reused. A full recalculation is triggered only when structural changes occur. The system monitors distribution network switch actions, fault trips, and high-load switching events. When an event affecting harmonic source location is detected, the update cycle is shortened, the impedance identification algorithm is reset, and a full recalculation is forced. Regular updates resume after the event subsides, and the location results for each cycle are stored in the historical database.

[0078] When the responsibility ratio exceeds the alarm threshold, perform responsibility trend analysis and alarm linkage, such as... Figure 6As shown. The harmonic responsibility allocation results are recorded as a time series, and the upward trend and abrupt events of the responsibility of each harmonic source are analyzed. Harmonic sources with continuously increasing responsibility and events with abrupt responsibility changes are identified. When the responsibility ratio exceeds the preset alarm threshold, an alarm is pushed to the power quality management platform and linked with power billing to complete responsibility tracing. When operational events such as distribution network switch operation and large load switching are detected, the update cycle is shortened to trigger a full recalculation of re-identification and positioning. The recalculation results are used to update the harmonic network model in step S101 for subsequent applications. The responsibility allocation results of each update cycle are stored in the historical database, thus forming a closed loop of continuous monitoring, event triggering, and re-identification.

[0079] To implement all or part of the aforementioned method, this application provides an embodiment of an apparatus for implementing the method for locating multiple harmonic sources in a distribution network based on online harmonic impedance identification. See [link to embodiment]. Figure 2 The device specifically includes the following components:

[0080] The network modeling module is used to convert the fundamental impedance of each branch into the harmonic branch impedance based on the distribution network topology and equipment fundamental parameters, construct the node harmonic admittance matrix containing the parameter uncertainty interval, and select the measurement points with the minimum number that meet the observability requirements after verifying the observability.

[0081] The data acquisition module is used to synchronously acquire three-phase harmonic voltage and current waveforms at the measurement point configuration location via unified time synchronization, and to generate harmonic voltage phasors and harmonic current phasors from the three-phase harmonic voltage and current waveforms through windowed interpolation harmonic analysis, sequence component decomposition, and data quality inspection.

[0082] Impedance identification module is used to extract the fluctuation amount of the harmonic voltage phasor and the harmonic current phasor and screen the valid fluctuation events. Combined with the node harmonic admittance matrix, it uses a recursive least squares algorithm with a forgetting factor to identify and track the network impedance parameters of each harmonic.

[0083] The localization estimation module is used to estimate the harmonic source injection current of each node under sparse constraints using weighted least squares based on the network impedance parameters, and to generate the harmonic source localization result by spatial clustering, background harmonic separation and contribution separation of the harmonic source injection current of each node.

[0084] The responsibility monitoring module is used to generate harmonic responsibility allocation results based on the harmonic source location results through contribution phasor synthesis and Monte Carlo uncertainty assessment, and continuously monitor the responsibility trend and abrupt events of the harmonic responsibility allocation results. When a distribution network operation event is detected, the update cycle is shortened to trigger a full recalculation, and an alarm is pushed and linked with billing when the responsibility ratio exceeds the preset alarm threshold.

[0085] The network modeling module converts the fundamental impedance of each branch into the harmonic branch impedance based on the distribution network topology and equipment fundamental parameters. Then, it accumulates the connection branch admittance node by node to construct a node harmonic admittance matrix. For each element of the node harmonic admittance matrix, it adds a parameter uncertainty range representing calibration error and aging offset. Based on the node harmonic admittance matrix, it establishes observability relationships and selects observable measurement point configurations with a greedy strategy, minimizing the number of measurement points. The data acquisition module synchronously acquires three-phase harmonic voltage and current waveforms at the configured measurement points via unified time synchronization. It extracts the amplitude and phase of each harmonic using windowed interpolation harmonic analysis and decomposes them into positive, negative, and zero-sequence components. After data quality inspection and sliding window mid-range filtering, it generates harmonic voltage and harmonic current phasors.

[0086] The impedance identification module calculates the voltage and current fluctuations at adjacent moments based on the continuously acquired harmonic voltage and current phasors, and filters valid fluctuation events according to the fluctuation amplitude threshold. Using the current fluctuation amplitude as weight, it integrates multiple valid fluctuation events through weighted least squares and combines them with the node harmonic admittance matrix using a recursive least squares algorithm with a forgetting factor to identify and track the network impedance parameters of each harmonic. The location estimation module establishes a linear measurement equation based on the network impedance parameters and determines the weights using measurement variance. It then estimates the harmonic source injection current at each node using weighted least squares with added sparsity constraints. The module performs density-based spatial clustering on the harmonic source injection current at each node based on electrical distance and injection characteristics, and separates the independent contribution of each source from the system-side background harmonics based on the superposition principle to generate the harmonic source location results.

[0087] The responsibility monitoring module calculates the contribution phasor of each harmonic source at the point of common coupling (PCC) based on the harmonic source location results, and synthesizes the total harmonic distortion rate (THD) of the equivalent voltage of each source according to the harmonic order. The original responsibility ratio is determined by the ratio of the contribution phasor to the total harmonic distortion rate of the total voltage. The uncertainty of the network impedance parameters and the injected current of each harmonic source within the confidence interval is evaluated using Monte Carlo sampling and normalized to generate the harmonic responsibility allocation result. The responsibility monitoring module further records the harmonic responsibility allocation result as a time series and analyzes the upward trend and abrupt events of each harmonic source's responsibility. When the responsibility ratio exceeds the preset alarm threshold, an alarm is pushed to the power quality management platform and linked with power billing. When a distribution network operation event is detected, the update cycle is shortened to trigger a full recalculation of re-identification and relocation. The responsibility allocation results of each update cycle are stored in the historical database.

[0088] As can be seen from the above description, the distribution network multi-harmonic source location device based on online harmonic impedance identification provided in this application embodiment can perform online location and quantitative assessment of responsibility for multiple harmonic pollution sources in the distribution network through the coordinated operation of five modules: network modeling, data acquisition, impedance identification, location estimation, and responsibility monitoring.

[0089] This invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the above-described method.

[0090] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method.

[0091] This invention also provides a computer program product, which includes a computer program that, when executed by a processor, implements the above-described method.

[0092] In this embodiment, the harmonic impedance of each harmonic in the distribution network is parameterized and modeled using a harmonic network model, and the time-varying impedance is identified and tracked online. The independent contribution of each source is separated and decoupled and located under multi-source coupling conditions by sparse constraint state estimation and spatial clustering. The responsibility ratio of each harmonic source to the common connection point is quantified by contribution phasor synthesis and Monte Carlo uncertainty assessment. By continuously monitoring the responsibility trend and shortening the update cycle when a distribution network operation event is detected, a full recalculation of detection, identification, location and recalculation is triggered.

[0093] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for locating multiple harmonic sources in a distribution network based on online harmonic impedance identification, characterized in that, The method includes: Based on the distribution network topology and equipment fundamental parameters, a harmonic network model containing the admittance matrix of each harmonic node and the parameter uncertainty interval is established. The configuration of measurement points that meet the requirements of observability and have the smallest number is selected. The three-phase harmonic voltage and current at the configuration of the measurement points are collected synchronously, and harmonic voltage phasors and harmonic current phasors are generated through windowed interpolation harmonic analysis, sequence component decomposition and data quality inspection. The fluctuations of the harmonic voltage phasor and the harmonic current phasor are extracted and effective fluctuation events are screened. Combined with the harmonic network model, a recursive algorithm with a forgetting factor is used to identify the network impedance parameters of each harmonic. Based on the network impedance parameters, the harmonic source injection current of each node is estimated under sparse constraints. The harmonic source region is identified by density-based spatial clustering of the harmonic source injection current of each node according to electrical distance and injection characteristics. Based on the superposition principle, the independent contribution of each source to the measurement point is separated, and the background harmonics on the system side are modeled as equivalent harmonic voltage sources and separated. The harmonic source localization result is generated by multi-time Bayesian fusion. Based on the harmonic source location results, the contribution of each harmonic source to the total harmonic distortion rate of the voltage at the point of common coupling is calculated, and the harmonic responsibility allocation results are generated through Monte Carlo uncertainty assessment and normalization. The responsibility trend of the harmonic responsibility allocation results is continuously monitored and updated according to a preset update cycle. Intermediate results with changes within the threshold condition are incrementally reused. When a distribution network operation event is detected, the update cycle is shortened to trigger a full recalculation of re-identification and location, and the recalculation results are used to refresh the harmonic network model. When the responsibility ratio exceeds the preset alarm threshold condition, an alarm is pushed and linked with billing.

2. The method for locating multiple harmonic sources in a distribution network based on online harmonic impedance identification according to claim 1, characterized in that, The process of establishing a harmonic network model containing the admittance matrices of each harmonic node and the uncertainty interval of parameters, and selecting a configuration of measurement points that satisfies observability and has the minimum number of points, includes: Based on the distribution network topology and equipment fundamental parameters, the fundamental impedance of each branch is converted into the impedance of each harmonic branch, and then the admittance of the connected branches is accumulated node by node to construct the node harmonic admittance matrix. The uncertainty range of parameters representing calibration error and aging offset is added to each element of the node harmonic admittance matrix. Based on the node harmonic admittance matrix, the observability relationship between the measured quantity and the injected current of each node harmonic source is established and the rank of the measurement matrix is ​​verified. Using a greedy strategy, the node with the largest observability gain is selected successively under the constraint of minimizing the number of measurement points until the measurement point configuration is fully observable.

3. The method for locating multiple harmonic sources in a distribution network based on online harmonic impedance identification according to claim 1, characterized in that, The process of synchronously acquiring the three-phase harmonic voltage and current at the measurement point configuration and generating harmonic voltage and current phasors through windowed interpolation, harmonic analysis, sequence component decomposition, and data quality verification includes: At the measurement point configuration location, sampling synchronization is achieved through unified time synchronization, and the sampling rate is collected to meet the requirements of characteristic harmonic analysis. The amplitude and phase of each harmonic are extracted from the three-phase harmonic voltage and current waveforms by windowed interpolation harmonic analysis. The amplitude and phase of each harmonic are decomposed into positive-sequence, negative-sequence, and zero-sequence components, and data quality checks are performed on amplitude rationality, phase continuity, and three-phase balance. Data that fails the check is marked as abnormal, then weighted and processed by sliding window mid-value filtering to generate harmonic voltage phasors and harmonic current phasors.

4. The method for locating multiple harmonic sources in a distribution network based on online harmonic impedance identification according to claim 1, characterized in that, The process of extracting the fluctuations of the harmonic voltage phasors and harmonic current phasors and screening effective fluctuation events, combined with the harmonic network model, uses a recursive algorithm with a forgetting factor to identify the network impedance parameters of each harmonic, including: The voltage fluctuation and current fluctuation at adjacent moments are calculated for the continuously acquired harmonic voltage phasor and harmonic current phasor, and valid fluctuation events are screened based on the lower and upper threshold conditions of the fluctuation amplitude. The equivalent harmonic impedance is estimated by weighted least squares synthesis of multiple effective fluctuation events using the current fluctuation amplitude as the weight. A collaborative identification model is established in conjunction with the harmonic network model, and the time-varying harmonic impedance is tracked by a recursive least squares algorithm with a forgetting factor. The network impedance parameters of each harmonic are generated after the identification results that do not meet the physical constraints are corrected to the constraint boundary.

5. The method for locating multiple harmonic sources in a distribution network based on online harmonic impedance identification according to claim 1, characterized in that, The estimation of harmonic source injection current at each node under sparse constraints based on the network impedance parameters includes: Based on the network impedance parameters, a linear measurement equation is established between the measurement vector and the harmonic source injection current of each node. The weights are determined by the measurement variance, and the harmonic source injection current of each node is solved by weighted least squares and sparse constraints.

6. The method for locating multiple harmonic sources in a distribution network based on online harmonic impedance identification according to claim 1, characterized in that, The calculation of the contribution of each harmonic source to the total harmonic distortion rate of the voltage at the point of common coupling, followed by Monte Carlo uncertainty assessment and normalization to generate the harmonic responsibility allocation results, includes: Based on the harmonic source location results, calculate the contribution phasor of each harmonic voltage at the common junction point when each harmonic source acts independently. Combine the contribution phasors according to the harmonic order to form the equivalent voltage total harmonic distortion rate of each source, and determine the original responsibility ratio of each source by the ratio of the contribution phasor to the total voltage total harmonic distortion rate. The network impedance parameters and the injected current of each node harmonic source are randomly sampled multiple times within their confidence intervals, and the responsibility ratio is repeatedly calculated. The uncertainty is evaluated by the mean and standard deviation of the responsibility ratio. The original responsibility ratio of each source is normalized and an additional uncertainty interval is added to generate the harmonic responsibility division result.

7. The method for locating multiple harmonic sources in a distribution network based on online harmonic impedance identification according to claim 1, characterized in that, The continuous monitoring of the harmonic responsibility allocation results and the continuous updating according to a preset update cycle, and the shortening of the update cycle to trigger a full recalculation of re-identification and relocation when a distribution network operation event is detected, includes: The process from data collection to responsibility assessment is completed within each cycle according to the preset update cycle. When the operation of the distribution network switch and load switching event are detected, the update cycle is shortened and a full recalculation is triggered. The harmonic responsibility allocation results are recorded as a time series, and the upward trend and abrupt events of the responsibility of each harmonic source are analyzed. When the responsibility ratio exceeds the preset alarm threshold, an alarm is pushed to the power quality management platform and the responsibility is traced in conjunction with the power billing. The location results and responsibility allocation results of each update cycle are stored in the historical database for subsequent analysis.

8. A multi-harmonic source location device for distribution networks based on online harmonic impedance identification, characterized in that, The device includes: The network modeling module is used to convert the fundamental impedance of each branch into the harmonic branch impedance based on the distribution network topology and equipment fundamental parameters, construct the node harmonic admittance matrix containing the parameter uncertainty interval, and select the measurement points with the minimum number that meet the observability requirements after verifying the observability. The data acquisition module is used to synchronously acquire three-phase harmonic voltage and current waveforms at the measurement point configuration location via unified time synchronization, and to generate harmonic voltage phasors and harmonic current phasors from the three-phase harmonic voltage and current waveforms through windowed interpolation harmonic analysis, sequence component decomposition, and data quality inspection. Impedance identification module is used to extract the fluctuation amount of the harmonic voltage phasor and the harmonic current phasor and screen the valid fluctuation events. Combined with the node harmonic admittance matrix, it uses a recursive least squares algorithm with a forgetting factor to identify and track the network impedance parameters of each harmonic. The localization estimation module is used to estimate the harmonic source injection current of each node under sparse constraints using weighted least squares based on the network impedance parameters, and to perform density-based spatial clustering to identify the harmonic source region based on the electrical distance and injection characteristics of the harmonic source injection current of each node. Based on the superposition principle, the independent contribution of each source to the measurement point is separated, and the background harmonics on the system side are modeled as equivalent harmonic voltage sources for separation. The harmonic source localization result is generated by multi-time Bayesian fusion. The responsibility monitoring module is used to generate harmonic responsibility allocation results based on the harmonic source location results through contribution phasor synthesis and Monte Carlo uncertainty assessment, continuously monitor the responsibility trend of the harmonic responsibility allocation results and update them continuously according to a preset update cycle, and perform incremental reuse of intermediate results with changes within the threshold condition. When a distribution network operation event is detected, the update cycle is shortened to trigger a full recalculation of re-identification and location, and the recalculation results are used to refresh the harmonic network model. When the responsibility ratio exceeds the preset alarm threshold condition, an alarm is pushed and linked with billing.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the distribution network multi-harmonic source location method based on online harmonic impedance identification as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the method for locating multiple harmonic sources in a distribution network based on online harmonic impedance identification as described in any one of claims 1 to 7.

Images