Power quality anomaly tracing and purification strategy generation system

Abstract

The present application relates to the technical field of power quality monitoring and treatment, and discloses a power quality abnormality tracing and purification strategy generation system, which comprises power collection, characteristic analysis, abnormality tracing, pollution assessment, parameter optimization and strategy generation modules.The power collection module collects grid key node waveform data in real time and sets a monitoring interval; the characteristic analysis module decomposes frequency spectrum characteristics to generate a pollution characteristic vector; the abnormality tracing module extracts core indexes to build a tracing rule; the pollution assessment module identifies a transmission path to calculate a pollution index; the parameter optimization module deduces optimal parameters to generate a deviation sequence; and the strategy generation module integrates the sequence to form a purification strategy.The system realizes accurate abnormality tracing and dynamic treatment, improves analysis comprehensiveness and strategy intelligentization level, and is suitable for power grid power quality management and control.

Description

Technical Field

[0001] This invention relates to the field of power quality monitoring and management technology, specifically a power quality anomaly tracing and purification strategy generation system. Background Technology

[0002] With the increasing complexity of modern power systems and the widespread application of nonlinear power electronic equipment, power quality problems in power grids are becoming increasingly prominent. Power quality anomalies such as harmonic pollution, voltage sags, and frequency fluctuations not only affect the normal operation of power equipment and reduce power supply reliability, but may also lead to significant economic losses and safety hazards. Currently, addressing power quality anomalies faces the following key challenges:

[0003] Traditional power quality monitoring systems often only monitor parameters at a single node or in a localized area, lacking a systematic analysis of pollution characteristics across the entire network. Existing technologies struggle to accurately identify the propagation paths and root causes of pollution such as harmonics, making it impossible to implement precise control measures targeting core pollution sources. For example, when multiple nodes simultaneously exhibit waveform distortion, traditional methods struggle to quickly determine whether the pollution is caused by local equipment or propagation from upstream nodes, easily leading to inaccurate control plans.

[0004] Most existing pollution assessments are based on single indicators (such as harmonic distortion rate) and do not fully consider the spatiotemporal correlation and dynamic changes of power quality parameters. For example, there is a complex coupling relationship between the duration of voltage sag, the amplitude of disturbance, and the timing factor of harmonic propagation. Traditional assessment methods cannot effectively quantify the comprehensive impact of these factors on overall power quality, resulting in insufficient targeting of compensation strategies.

[0005] Current power quality control strategies rely primarily on human experience or simple threshold comparisons, lacking dynamic optimization mechanisms based on big data and intelligent algorithms. For example, when faced with different types of power quality anomalies (such as intermittent harmonics and persistent voltage fluctuations), traditional methods struggle to quickly generate optimal combinations of control parameters, resulting in a lag in the control process and an inability to adapt to the real-time changes in the power grid.

[0006] The power grid is a complex interconnected system, and the power quality issues of each node are interconnected. Current technologies lack in-depth analysis of the power grid topology and the patterns of pollution propagation between nodes, making it difficult to achieve collaborative governance across multiple nodes. For example, once pollution at a critical node is controlled, pollution at adjacent nodes may change due to power flow redistribution. Traditional methods cannot predict this cascading effect in advance, making it difficult to continuously optimize the governance effect.

[0007] With the widespread adoption of smart meters and sensors, the power grid has accumulated a vast amount of waveform data, but existing systems lack sufficient depth in mining this data. For example, the failure to fully utilize the similarity analysis between historical pollution patterns and current data makes it impossible to quickly match the optimal treatment solution, thus limiting the improvement of the intelligence level of power quality management. Summary of the Invention

[0008] The purpose of this invention is to provide a power quality anomaly tracing and purification strategy generation system to solve the problems mentioned in the background art.

[0009] To achieve the above objectives, the present invention provides the following technical solution: a power quality anomaly tracing and purification strategy generation system, the system comprising:

[0010] The power acquisition module is used to collect voltage and current waveform data of key nodes in the power grid in real time and set the power grid monitoring range corresponding to the pollution level.

[0011] The feature parsing module is used to divide the power grid monitoring range into multiple power grid nodes, perform spectral feature decomposition on the waveform data of each power grid node, and generate pollution feature vectors corresponding to the power grid nodes.

[0012] The anomaly tracing module is used to extract core pollution indicators from pollution feature vectors, establish pollution tracing rules associated with power grid nodes, and obtain the corresponding spectrum correction parameters for the rules.

[0013] The pollution assessment module is used to identify harmonic propagation paths in the spectrum correction parameters, compensate the core pollution indicators based on the propagation paths, and calculate the pollution index of each power grid node under different compensation strategies.

[0014] The parameter optimization module is used to derive the optimal purification parameters based on the pollution index and generate a pollution deviation sequence by comparing the current waveform feature values ​​with the optimal purification parameters.

[0015] The strategy generation module is used to analyze the pollution deviation sequence and integrate it into a purification control strategy based on the pollution change trend of the power grid nodes.

[0016] Preferably, the system feature parsing module is implemented by: constructing a waveform feature library corresponding to the power grid node, wherein the waveform feature library contains waveform data and pollution parameter vectors mapped by spectral features;

[0017] Similarity spectrum matching is performed on the pollution parameter vectors, and the power grid clusters of the pollution parameter vectors are divided according to the matching results. The spectral focal points of the waveform data are extracted from the power grid clusters and set as power grid nodes.

[0018] Preferably, the power grid clustering group for dividing the pollution parameter vector by the system further includes:

[0019] Based on the power grid attributes and pollution intensity in the pollution parameter vector, harmonic distribution, phase shift and waveform distortion parameters are extracted, and waveform feature labels are generated based on the above parameters.

[0020] The waveform feature labels are associated with pollution parameter vectors. By calculating the spectral similarity between feature labels, parameter vectors with similarity higher than a preset pollution threshold are selected to form power grid cluster groups.

[0021] Preferably, the system generates pollution feature vectors corresponding to power grid nodes in the following ways:

[0022] For each power grid node, based on the node's topological location within the power grid monitoring interval, waveform distortion data of the node within a preset time period is obtained, and the waveform distortion coefficient of the node is calculated.

[0023] When the waveform distortion coefficient exceeds the first pollution threshold, the node is marked as a high pollution node, and its waveform data is extracted to form a pollution feature vector; when the waveform distortion coefficient is lower than the first pollution threshold, the node is marked as a clean node, and the waveform data of the adjacent nodes of the node are superimposed on the spectrum, and the superimposed data is reconstructed into a pollution feature vector.

[0024] Preferably, the implementation of the system anomaly tracing module includes:

[0025] Harmonic distortion parameters, voltage sag parameters, and frequency fluctuation parameters are separated from the pollution feature vector, and pollution source tracing rules for power grid nodes are generated based on the harmonic distortion parameters, voltage sag parameters, and frequency fluctuation parameters.

[0026] If the number of power grid nodes covered by the current pollution source tracing rule is less than the preset pollution threshold, then the pollution feature vectors of adjacent power grid nodes are traversed, and the spectral indicators not included in the source tracing rules of the adjacent nodes are added to the current rule.

[0027] Preferably, the system pollution assessment module is implemented by: obtaining the time series factor of the transient event time constant and the fluctuation factor of the disturbance amplitude in the harmonic propagation path;

[0028] Construct an admittance weight matrix associated with time series factors and volatility factors, and determine the pollution index under different compensation strategies based on the probability distribution of each element in the matrix.

[0029] Preferably, the system further includes constructing the admittance weight matrix as follows:

[0030] Identify the perturbation periodicity of the time series factor. If the current periodicity perfectly matches the preset event period, then set the time series factor as the starting index of the admittance weight matrix.

[0031] Calculate the harmonic phase sequence matching degree between the time series factor and the fluctuation factor, and generate the intermediate index and termination index of the admittance weight matrix in descending order of matching degree.

[0032] Path backtracking is performed on the terminating index. When the matching degree of the terminating index is lower than the preset matching threshold, it is output as the final distribution of the admittance weight matrix.

[0033] Preferably, the system calculates the pollution index in the following ways:

[0034] The mean of the time series factor and the range of the volatility factor for each terminating index in the statistical admittance weight matrix are calculated, and the global variance of all index factors is calculated.

[0035] The time series offset coefficient is obtained by subtracting the mean of the time series factors of a single terminating index from the mean of the time series factors of adjacent indices and dividing by the global variance. At the same time, the ratio of the volatility factor range to the global variance is calculated, and the two are weighted and summed to obtain the contamination index of the index.

[0036] Preferably, the system derives the optimal purification parameters in the following ways:

[0037] Extract the pollution pattern from historical data that is closest to the current pollution index, calculate the cosine similarity between the two in harmonic distribution, and use it as the first dynamic reference value;

[0038] The difference in the number of disturbance events between the current pollution index and historical pollution patterns is statistically analyzed, and the number of differences is used as a second dynamic reference value.

[0039] Based on a linear combination of the first dynamic reference value and the second dynamic reference value, the optimal purification parameters in the preset purification parameter table are matched.

[0040] Preferably, the system strategy generation module is implemented by dividing the pollution intensification interval and the pollution reduction interval according to the pollution change direction of each node in the pollution deviation sequence;

[0041] The rising rate of pollution deviation in the aggravating pollution range and the falling rate of pollution deviation in the mitigating pollution range are extracted. The two are then fused according to the topological weight of the power grid nodes to generate the adjustment parameters of the purification control strategy.

[0042] Compared with the prior art, the beneficial effects of the present invention are:

[0043] In the power data acquisition phase, the system collects voltage and current waveform data from key nodes of the power grid in real time and sets power grid monitoring intervals corresponding to pollution levels, enabling dynamic monitoring of the power grid's operating status. This differentiated monitoring mechanism based on pollution levels can flexibly adjust monitoring accuracy according to the actual pollution level of the power grid, ensuring focused monitoring of high-pollution areas while avoiding resource waste in low-pollution areas, thus improving the efficiency and relevance of data acquisition.

[0044] The feature parsing module achieves in-depth mining of pollution characteristics of power grid nodes through operations such as constructing a waveform feature library, performing similar spectrum matching, and generating waveform feature labels. Specifically, by dividing the pollution parameter vector into power grid clusters and extracting spectral focal points, the system can accurately identify groups of nodes with similar pollution characteristics in the power grid, thereby determining key monitoring nodes. This clustering analysis method effectively reveals the pollution correlation between power grid nodes, providing a solid data analysis foundation for subsequent anomaly tracing and significantly improving the accuracy and comprehensiveness of pollution feature extraction.

[0045] The anomaly tracing module establishes pollution source tracing rules associated with power grid nodes by separating core pollution indicators such as harmonic distortion, voltage sag, and frequency fluctuation. Simultaneously, it dynamically expands the tracing rules by traversing the pollution feature vectors of adjacent nodes, ensuring the comprehensiveness and adaptability of the rules. This module can accurately locate the root cause nodes of pollution and clarify the propagation paths of pollution such as harmonics, solving the problem of insufficient tracing accuracy in traditional methods and providing a clear target for targeted remediation.

[0046] The pollution assessment module constructs a multi-dimensional, dynamic pollution assessment system by introducing innovative methods such as time-series factors, fluctuation factors, and admittance weight matrices. This module can not only identify transient event characteristics in harmonic propagation paths but also quantify the treatment effects under different compensation strategies by calculating a pollution index. This assessment mechanism, based on probability distribution and parameter compensation, fully considers the spatiotemporal coupling relationship of power quality parameters, making the assessment results more reflective of the actual pollution status of the power grid and providing a scientific basis for optimizing purification strategies.

[0047] The parameter optimization module analyzes the similarity between historical pollution patterns and current data, and combines dynamic reference values ​​to match the optimal purification parameters, thus achieving intelligent derivation of purification parameters. This data-driven optimization mechanism changes the traditional parameter setting method that relies on manual experience. It can quickly generate the optimal parameter combination based on real-time pollution conditions, significantly improving the efficiency and accuracy of parameter optimization and ensuring that the purification device is always in optimal working condition.

[0048] The strategy generation module analyzes the changing trends of pollution deviation sequences and performs conflict fusion by combining grid node topology weights to generate targeted pollution control strategies. This module can dynamically adjust regulation parameters based on the intensification or deterioration trends of pollution, achieving multi-node collaborative governance and effectively addressing the complexity and dynamism of pollution propagation in the power grid. This topology-based strategy generation method ensures the systematic nature and effectiveness of the governance scheme, significantly improving the overall power quality of the power grid. Attached Figure Description

[0049] Figure 1 This is a schematic diagram of the working principle of the power quality anomaly tracing and purification strategy generation system described in this invention;

[0050] Figure 2 Design diagram of the pollution feature vector generation method;

[0051] Figure 3 Design diagram of the implementation method of the exception tracing module;

[0052] Figure 4 Design diagram for the implementation of the pollution assessment module;

[0053] Figure 5 This is a design diagram for the derivation method of the optimal purification parameters. Detailed Implementation

[0054] 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.

[0055] Please see Figures 1-5 This invention relates to a power quality anomaly tracing and remediation strategy generation system. The system includes a power acquisition module, a feature analysis module, an anomaly tracing module, a pollution assessment module, a parameter optimization module, and a strategy generation module. These modules work together to achieve full-process processing of power quality anomalies. Specific implementation methods are as follows:

[0056] Power Acquisition Module: This module uses sensors deployed at key nodes of the power grid to collect voltage and current waveform data in real time, with a sampling frequency of no less than 10kHz, ensuring data real-time performance and accuracy. Simultaneously, the system pre-sets power grid monitoring intervals corresponding to pollution levels. For example, based on national standards, pollution levels are divided into Level I (light pollution), Level II (moderate pollution), and Level III (heavy pollution), each corresponding to different monitoring interval parameters such as voltage fluctuation thresholds and harmonic content thresholds, enabling tiered monitoring of the power grid's operational status.

[0057] Feature Analysis Module: Within the power grid monitoring range, the power grid topology is divided into multiple power grid nodes, each corresponding to a monitoring unit. The waveform data from each node is processed using spectral feature decomposition algorithms such as Fast Fourier Transform (FFT) to extract characteristic parameters such as the amplitude and phase of the fundamental wave and each harmonic, generating a pollution feature vector containing multi-dimensional features, such as a 100-dimensional feature vector containing the amplitudes of harmonics up to the 50th order.

[0058] The anomaly tracing module extracts core pollution indicators from the pollution feature vector, such as total harmonic distortion (THD), voltage sag, and frequency deviation, and establishes pollution source tracing rules associated with power grid nodes. These rules are constructed using a decision tree algorithm; for example, if THD > 5% and voltage sag > 10%, the node is identified as a candidate pollution source. Simultaneously, the module obtains the corresponding spectral correction parameters for each rule, such as harmonic suppression target values ​​and voltage compensation coefficients.

[0059] Pollution Assessment Module: This module identifies harmonic propagation paths in spectral correction parameters using harmonic network analysis methods, such as harmonic power flow calculations based on admittance matrices, to determine the propagation direction and attenuation characteristics of harmonics in the power grid. Based on the propagation paths, it compensates for core pollution indicators, such as adjusting the calculated THD value considering the harmonic coupling effect between adjacent nodes. Furthermore, it uses the Analytic Hierarchy Process (AHP) to calculate the pollution index of each power grid node under different compensation strategies. The pollution index comprehensively reflects the degree of pollution across multiple dimensions, including harmonics, voltage sags, and frequency fluctuations.

[0060] Parameter optimization module: Based on the pollution index, the optimal purification parameters are derived using the Particle Swarm Optimization (PSO) algorithm, such as determining the compensation current command value of the Active Power Filter (APF). By comparing the current waveform characteristic values ​​(such as real-time THD values) with the optimal purification parameters, the difference sequence between the two is calculated to generate a pollution deviation sequence, which is used to characterize the difference between the current pollution state and the ideal purification state.

[0061] Strategy Generation Module: Performs Dynamic Time Warping (DTW) analysis on the pollution deviation sequence to analyze the pollution change trend of each node, such as increasing, decreasing, or stabilizing. Based on the topology and pollution trend, the deviation sequence is integrated into a purification control strategy scheme that includes control parameters and execution timing. For example, for high-pollution nodes, the compensation action of APF is triggered first, and the compensation start time and duration are set.

[0062] The present invention will be further described below with reference to Examples 1 to 5:

[0063] Example 1:

[0064] This embodiment involves a detailed implementation of the feature parsing module. The core function of the feature parsing module is to divide the power grid into nodes within the power grid monitoring range and generate pollution feature vectors corresponding to each node. Its specific implementation is as follows:

[0065] The feature parsing module constructs a waveform feature library corresponding to power grid nodes. This library is built based on historical monitoring data and stores power grid waveform data from different time periods and operating states, as well as pollution parameter vectors that map to these waveform data using spectral analysis. The pollution parameter vectors contain multi-dimensional feature parameters, such as the fundamental amplitude, the amplitude and phase of each harmonic, voltage fluctuation coefficient, and frequency deviation. These parameters are extracted from the original waveform data using spectral processing algorithms such as Fourier transform and wavelet analysis, comprehensively reflecting the pollution characteristics of power grid nodes. The establishment of the waveform feature library provides a historical reference benchmark for subsequent pollution parameter vector analysis, enabling the system to effectively identify and classify power grid pollution characteristics by comparing current data with historical data.

[0066] After constructing the waveform feature library, the feature parsing module performs similarity spectrum matching on the pollution parameter vectors. Similarity spectrum matching calculates the spectral similarity between different pollution parameter vectors using specific algorithms, such as dynamic time warping or cosine similarity algorithms. Taking dynamic time warping as an example, this algorithm measures the differences in spectral features between different pollution parameter vectors by calculating the nonlinear alignment distance between two time series on the time axis. The system presets a distance threshold for similarity spectrum matching; when the calculated distance between two pollution parameter vectors is less than this threshold, they are determined to have similar spectral features. Based on this, the pollution parameter vectors are divided into different power grid clusters. Each cluster contains pollution parameter vectors corresponding to power grid nodes with similar spectral features, thus achieving preliminary grouping of pollution features of power grid nodes.

[0067] Extracting spectral focal points from waveform data within power grid clusters is a key step in the feature analysis module. Spectral focal points refer to spectral feature points in waveform data that exhibit high frequencies or concentrated energy within the same power grid cluster, such as specific harmonic frequencies or typical voltage fluctuation patterns. By statistically analyzing the spectral feature distribution of pollution parameter vectors within each cluster, the spectral focal points of that cluster are determined. For example, if most pollution parameter vectors within a cluster show high amplitudes for the 5th harmonic, then the 5th harmonic frequency is set as the spectral focal point for that cluster. Determining the spectral focal points helps clarify the main pollution characteristics corresponding to each cluster, thereby allowing the physical location of the spectral focal points or the corresponding power grid equipment to be designated as power grid nodes, enabling the rational selection and layout of power grid monitoring nodes.

[0068] When grouping power grids into clusters based on pollution parameter vectors, in addition to grouping based on spectral similarity, a deeper analysis of power grid attributes and pollution intensity within the pollution parameter vectors is necessary. Power grid attributes include node types, such as generator nodes, transformer nodes, and load nodes. Different types of nodes have different functions and pollution generation mechanisms within the power grid. Pollution intensity is reflected by characteristic parameter values ​​in the pollution parameter vectors, such as the magnitude of the total harmonic distortion (THD) and the amplitude of voltage sags. Based on power grid attributes and pollution intensity, the feature parsing module further extracts harmonic distribution, phase shift, and waveform distortion parameters. Harmonic distribution reflects the proportion of each harmonic in the total harmonics; phase shift characterizes the phase difference between the fundamental wave and harmonics; and waveform distortion parameters can be measured by statistics such as kurtosis and climax factor to determine the degree to which the waveform deviates from an ideal sine wave.

[0069] After extracting the above parameters, the system generates waveform feature labels associated with the pollution parameter vector. These waveform feature labels are textual descriptions of the core features of the pollution parameter vector, such as "generator node - high 5th harmonic pollution - high voltage waveform kurtosis" or "load node - frequent voltage sags - large phase shifts." These labels intuitively reflect the pollution type and severity of the power grid nodes, providing more semantically informative classification criteria for subsequent cluster analysis.

[0070] By calculating the spectral similarity between feature labels, power grid clusters are further selected. The calculation of spectral similarity is based on the parameter information contained in the feature labels. For example, for two labels containing harmonic order and amplitude information, similarity can be calculated by comparing the amplitude differences of the corresponding harmonic orders. The system presets a pollution threshold; when the spectral similarity between feature labels exceeds this threshold, the corresponding pollution parameter vectors are assigned to the same power grid cluster. This clustering method, which combines power grid attributes, pollution intensity, and feature labels, can more accurately reflect the actual pollution status of power grid nodes, avoiding the potential bias of clustering based solely on spectral data, thereby improving the scientific rigor and rationality of power grid node division.

[0071] After completing the power grid clustering and node setting, the feature parsing module performs spectral feature decomposition on the waveform data of each power grid node to generate a corresponding pollution feature vector. For highly polluted nodes, key spectral feature parameters are directly extracted from their waveform data to form the pollution feature vector. For clean nodes, considering the potential impact of pollution from neighboring nodes, the waveform data of neighboring nodes are spectrally superimposed and then reconstructed into a pollution feature vector using a dimensionality reduction algorithm. This ensures that the pollution features of clean nodes can reflect the potential pollution impact of their surrounding areas.

[0072] Example 2:

[0073] This embodiment relates to a specific method for generating pollution feature vectors corresponding to power grid nodes. Within the power grid monitoring range, the system needs to generate differentiated pollution feature vectors based on the topological location and waveform data characteristics of each power grid node. The specific implementation steps are as follows:

[0074] For each power grid node, the system preprocesses data based on its topological location within the power grid monitoring area. Topological location encompasses the node's physical connectivity within the power grid, such as its position at the beginning of a transmission line, as an intermediate node, at the end of a line, or at the connection point of critical equipment (such as transformers or generators). Nodes at different topological locations exhibit varying pollution propagation characteristics and degrees of impact; for example, nodes at the beginning of a transmission line are more susceptible to disturbances on the power supply side, while nodes at the end of a line may accumulate pollution due to load fluctuations. The system acquires waveform distortion data from the node within a preset time period. This preset time period can be set to minutes (e.g., the previous 10 minutes) or seconds, depending on the power grid's operating characteristics, ensuring that the data reflects the current and recent pollution status.

[0075] The system calculates the waveform distortion coefficient of each node. This coefficient quantifies the difference between the measured waveform and the ideal sinusoidal waveform using the root mean square error algorithm. The calculation formula is as follows:

[0076]

[0077] in, Indicates the first The measured waveform data values ​​at each sampling time, including the instantaneous values ​​of voltage or current; Indicates the first The ideal sinusoidal waveform data values ​​at each sampling time have amplitude and frequency consistent with the rated parameters of the power grid. The number of sampling points within a preset time period is specified, and the sampling frequency must satisfy the Nyquist sampling theorem to ensure the integrity of the waveform data. This formula objectively reflects the overall degree of waveform distortion by calculating the square root of the mean square of the deviations between the measured waveform and the ideal waveform.

[0078] Based on the calculated waveform distortion coefficient, the system classifies power grid nodes. When the distortion coefficient exceeds a first pollution threshold (e.g., 0.15, the specific threshold can be determined according to national standards or power grid operation experience), the node is identified as a high-pollution node. For high-pollution nodes, the system directly extracts key feature parameters from their waveform data to form a pollution feature vector. Specifically, the extracted content includes the fundamental amplitude, the amplitude and phase of each harmonic (e.g., 1-50), voltage fluctuation amplitude, etc., forming a multi-dimensional feature vector. For example, a 50-dimensional vector containing the amplitude of the 50th harmonic, where each dimension corresponds to the pollution intensity of a specific frequency component, thus clearly characterizing the pollution spectrum distribution of the node.

[0079] When the waveform distortion coefficient is below the first pollution threshold, the system marks the node as a clean node. For clean nodes, although their own waveform data does not show significant pollution, considering that pollution in the power grid may propagate through adjacent nodes, further analysis of the potential pollution impact on their surrounding areas is necessary. In practice, the system acquires the waveform data of the clean node's adjacent nodes (i.e., upstream and downstream nodes directly connected by transmission lines) and performs spectral superposition processing on these data. During spectral superposition, the phase relationship of the waveform data of adjacent nodes must be considered. For example, if the 5th harmonic phase of adjacent node A is 30° and the 5th harmonic phase of adjacent node B is 60°, then the amplitude and phase of the synthesized 5th harmonic must be calculated using vector synthesis, rather than simple arithmetic addition.

[0080] After the spectrum is superimposed, the system uses Principal Component Analysis (PCA) to reconstruct the superimposed data by dimensionality reduction, generating a pollution feature vector for clean nodes. PCA maps high-dimensional data to a low-dimensional space through linear transformation, reducing data dimensionality while preserving key feature information and improving computational efficiency. For example, if the superimposed data contains 100 harmonic features, PCA can extract the first 20 principal components, whose cumulative variance contribution rate reaches over 90%, thus compressing the high-dimensional data into a 20-dimensional pollution feature vector. This vector can reflect the combined impact of pollution from adjacent nodes, avoiding the potential risk of surrounding pollution propagation being overlooked due to the cleanliness of data from a single node.

[0081] During the generation of pollution feature vectors, the system must ensure the real-time nature and accuracy of the data. For highly polluted nodes, the update frequency of the feature vectors is consistent with the waveform data acquisition frequency to ensure timely capture of pollution changes. For clean nodes, considering the potential lag in pollution propagation from adjacent nodes, the update cycle of the feature vectors can be appropriately extended, but it must not be less than the length of a preset time period to ensure effective monitoring of surrounding pollution dynamics.

[0082] In addition, the system needs to dynamically respond to topology changes in power grid nodes. When the power grid structure is adjusted (such as line switching or equipment start-up and shutdown), the adjacency relationships of nodes are re-identified, the set of adjacent nodes of the clean nodes is updated, and the waveform data acquisition and feature vector reconstruction process is triggered to ensure the topological consistency of pollution feature analysis.

[0083] Example 3:

[0084] This embodiment details the implementation of the anomaly tracing module. The core function of the anomaly tracing module is to extract core pollution indicators from the pollution feature vector, establish pollution tracing rules associated with power grid nodes, and dynamically optimize the rule content based on the rule coverage. The specific implementation method is as follows:

[0085] The anomaly tracing module extracts key parameters from the pollution feature vector output by the feature parsing module, including harmonic distortion parameters, voltage sag parameters, and frequency fluctuation parameters. Harmonic distortion parameters encompass total harmonic distortion (THD) and the percentage of each harmonic amplitude, used to measure the degree of harmonic pollution. Voltage sag parameters include sag amplitude, duration, and frequency of occurrence, reflecting the severity and frequency of voltage sag events. Frequency fluctuation parameters include frequency deviation (the difference between the actual frequency and the rated frequency) and fluctuation period, characterizing the frequency stability of the power grid. These parameters are directly extracted through dimensional division of the pollution feature vector; for example, the first 10 dimensions of the pollution feature vector correspond to harmonic amplitude, the middle 5 dimensions correspond to voltage sag parameters, and the last 5 dimensions correspond to frequency fluctuation parameters.

[0086] Based on the three types of parameters separated, the system uses data mining algorithms to generate pollution source tracing rules for power grid nodes. Specifically, association rule mining algorithms (such as the Apriori algorithm) or decision tree algorithms can be used to establish rule logic by analyzing the correlations between parameters and threshold conditions. Taking the decision tree algorithm as an example, the system first uses the total harmonic distortion (THD) as the root node and sets a threshold (such as THD > 5%) for branching: if THD > 5%, it further determines whether the voltage sag is > 10%; if both conditions are met, the rule "THD > 5% and voltage sag > 10% → pollution source located at this node" is generated. These rules are constructed by training on historical pollution data, with each rule corresponding to a combination of pollution features from one or more power grid nodes, achieving preliminary location of the pollution source.

[0087] After pollution source tracing rules are established, the system needs to evaluate the coverage of the rules. Rule coverage is measured by counting the number of power grid nodes that the current rule can locate. A preset pollution threshold is used to determine whether the rule is comprehensive enough (e.g., a preset threshold of 3 nodes means that a rule optimization process is triggered when the number of nodes covered by the rule is less than 3). If the number of power grid nodes covered by the current rule is less than the preset pollution threshold, it indicates that the rule's applicability is limited, and the rule content needs to be further expanded to improve the accuracy of source tracing.

[0088] Rule optimization is achieved by traversing the pollution feature vectors of adjacent power grid nodes. Adjacent nodes refer to nodes directly connected to the node associated with the current rule in the power grid topology, such as upstream and downstream nodes on transmission lines, or nodes connected to the same bus. The system extracts spectral indicators from the pollution feature vectors of adjacent nodes that are not included in the current rule. For example, if the current rule only focuses on THD and voltage sag parameters, but the pollution feature vectors of adjacent nodes contain significant frequency fluctuation parameters (such as frequency deviation > ±0.3Hz), then the frequency fluctuation parameters are added as new conditions to the rule. The selection of new indicators is completed using statistical methods such as the information gain algorithm or chi-square test to ensure that the new indicators are significantly correlated with pollution source tracing.

[0089] When adding new indicators to the current rules, the system employs a decision tree pruning strategy to avoid overly complex rules. For example, when a frequency fluctuation parameter is added, if the information gain of the rule increases by less than a preset threshold (e.g., 0.1), the indicator is discarded; if the gain is significant, a new rule branch is generated through node splitting. In this way, the rules gradually integrate more dimensions of pollution characteristics, forming a hierarchical source tracing logic, such as "THD > 5% and voltage sag > 10% and frequency deviation > ±0.3Hz → pollution source is located upstream of the transformer at this node".

[0090] To ensure the real-time nature of the rules, the system updates the pollution source tracing rules periodically (e.g., hourly). During the update process, the algorithm model is retrained based on the latest pollution feature vector data to identify newly emerging pollution patterns and their correlation with parameters, and outdated rule conditions are eliminated. For example, when a new nonlinear load in the power grid causes a significant increase in a certain harmonic pollution, the system automatically detects the frequency of that harmonic frequency in the pollution feature vector and incorporates it into the rule conditions, improving the rules' adaptability to new types of pollution.

[0091] In the actual operation of the anomaly tracing module, pollution tracing rules are stored in a tree structure. Each node represents a judgment condition, branches represent whether the condition is met or not, and leaf nodes represent the tracing conclusion (such as the location of the pollution source or the type of pollution). The execution process of the rules adopts a forward reasoning approach, starting from the root node and sequentially judging whether the pollution feature vector meets the conditions of each layer, finally reaching the leaf node to obtain the tracing result. This structure makes the visualization and maintenance of the rules convenient, and technicians can optimize the tracing logic by adjusting the thresholds or adding / removing nodes in the tree structure.

[0092] Furthermore, the anomaly tracing module supports multi-rule conflict handling. When different rules reach contradictory conclusions regarding the same pollution feature vector (e.g., rule A identifies the pollution source as node A, while rule B identifies it as node B), the system arbitrates by calculating the confidence level of each rule (based on the number of times the rule was correctly executed in historical data), selecting the rule with the highest confidence level as the final conclusion. If the confidence levels of the rules are similar, a manual review process is triggered, where maintenance personnel determine the pollution source based on the real-time operating status of the power grid.

[0093] Example 4:

[0094] This embodiment involves the collaborative implementation of a pollution assessment module and a parameter optimization module. The pollution assessment module obtains the time-series factors of transient event time constants and the fluctuation factors of disturbance amplitudes in the harmonic propagation path, and constructs an admittance weight matrix associated with the time-series factors and fluctuation factors. During matrix construction, the disturbance periodic characteristics of the time-series factors are identified. If the current periodic characteristics perfectly match the preset event period, the time-series factor is set as the starting index of the matrix. The harmonic phase sequence matching degree between the time-series factors and the fluctuation factors is calculated, and intermediate and terminal indices of the matrix are generated sequentially from high to low matching degree. Path backtracking is performed on the terminal index. When the matching degree is lower than the preset matching threshold, it is output as the final distribution of the matrix, reflecting the main path and weights of harmonic propagation.

[0095] When calculating the pollution index, the mean of the time-series factors and the range of the volatility factors for each terminating index in the statistical admittance weight matrix are calculated, and the global variance of all index factors is calculated. The difference between the mean of the time-series factors of a single terminating index and the mean of the time-series factors of its adjacent indices is divided by the global variance to obtain the time-series offset coefficient; simultaneously, the ratio of the range of the volatility factors to the global variance is calculated, and the two are weighted and summed to obtain the pollution index for that index, quantifying the pollution contribution of each propagation path. Based on the pollution index, the parameter optimization module extracts the pollution pattern in historical data that is closest to the current pollution index, calculates the cosine similarity between the two in harmonic distribution as the first dynamic reference value, and statistically analyzes the difference between the current pollution index and the historical pattern in the number of disturbance events as the second dynamic reference value. The optimal purification parameters in the preset purification parameter table are matched through linear combination.

[0096] For example, in the power grid of an industrial park, five harmonic pollution levels exceeding the standard were detected. The pollution assessment module obtained the time-series factor and fluctuation factor in the harmonic propagation path. The time-series factor exhibited periodic fluctuations with a period of 0.5 seconds, perfectly matching the preset rectifier operating cycle. Therefore, this time-series factor was set as the starting index of the admittance weight matrix. The fluctuation factor showed that the harmonic amplitude fluctuated within the range of 10% to 30% of the rated value. The harmonic phase sequence matching degree between the time-series factor and the fluctuation factor was calculated, revealing a stable 30° phase offset, indicating a high matching degree. This was then used as an intermediate index. Through path backtracking, it was found that the harmonics mainly propagated through three paths: path A passed through three transformers, path B passed through two transformers and one long-distance transmission line, and path C passed through one transformer and a large number of nonlinear loads. When the matching degree of path C was found to be lower than the preset threshold, the parameters of paths A, B, and C were output as the final distribution of the matrix.

[0097] The mean of the time series factor and the range of the volatility factor for each path in the statistical admittance weight matrix are used to calculate the global variance. The mean of the time series factor for path A is 0.48 seconds. The difference between this and the mean of the time series factor for the adjacent path B (0.52 seconds), divided by the global variance, yields a time series offset coefficient of 0.15. The range of the volatility factor for path A is 25%, which is 0.8 times the global variance. The two are then weighted and summed to obtain the pollution index for path A. Similarly, the pollution indices for paths B and C are calculated. It is found that path C has the highest pollution index, indicating that this path contributes the most to the 5th harmonic pollution.

[0098] The parameter optimization module uses the pollution index to find the closest pollution pattern from historical data. Assuming the current pollution index combination is (0.65, 0.35, 0.85), and the historical data shows pollution pattern D with an index of (0.62, 0.38, 0.87), the cosine similarity between the two is 0.98, which is used as the first dynamic reference value. The difference in the number of disturbance events between the current pollution event and pattern D is statistically analyzed, revealing two more short-term disturbances than pattern D, which is used as the second dynamic reference value. Through linear combination, the optimal purification parameters from a preset purification parameter table are matched. For example, it is determined that a 500kVAR active power filter with a rated capacity should be installed at the critical node of path C to effectively suppress 5th harmonic pollution.

[0099] Throughout the collaborative implementation process, the pollution assessment module and the parameter optimization module achieved accurate assessment and parameter optimization of power grid pollution through data interaction and algorithmic collaboration. The admittance weight matrix constructed by the pollution assessment module clearly demonstrates the harmonic propagation path and pollution contribution, providing a clear objective for parameter optimization. The parameter optimization module dynamically matches the optimal purification parameters based on the pollution index, ensuring the targetedness and effectiveness of purification measures. This collaborative mechanism can quickly adjust the governance strategy according to the real-time pollution status of the power grid, improve the efficiency of power quality governance, and provide strong support for the safe and stable operation of the power grid.

[0100] Example 5

[0101] This embodiment relates to the specific implementation of the strategy generation module. The strategy generation module divides the pollution aggravation interval and the pollution reduction interval based on the pollution change direction of each node in the pollution deviation sequence. It extracts the rate of increase of pollution deviation within the aggravation interval and the rate of decrease of pollution deviation within the pollution reduction interval, performs conflict fusion based on the topology weights of the power grid nodes, and uses a weighted average algorithm to generate the adjustment parameters for the purification control strategy scheme.

[0102] For example, in a city's power distribution network, monitoring revealed harmonic pollution at multiple nodes with dynamically changing pollution levels. The strategy generation module first analyzes the pollution deviation sequence of each node, which is the difference between the current waveform characteristic value and the optimal purification parameter. Taking node A as an example, its pollution deviation sequence shows an upward trend during time period T1 to T2, indicating increased pollution; and a downward trend during time period T2 to T3, indicating decreased pollution. By calculating the slope of adjacent data points, the upward rate is determined to be 0.05 units per minute, and the downward rate is determined to be 0.03 units per minute.

[0103] Simultaneously, the system assigns topology weights to each node based on the power grid topology. Node A, as a feeder node of the main substation, directly connects to multiple important loads, and its topology weight is set to 0.8 (range 0-1); while node B, as an end-point distribution node, has a topology weight set to 0.3. When the pollution trends of node A and node B conflict (e.g., pollution at node A intensifies while pollution at node B weakens), the system will reconcile the conflict based on the topology weights.

[0104] In the specific fusion process, the strategy generation module handles the aggravated pollution interval and the de-pollution interval separately. For the aggravated pollution interval, the sum of the products of the rise rate of all nodes and the topology weights is calculated, and then divided by the sum of the topology weights to obtain the comprehensive rise rate; for the de-pollution interval, the comprehensive fall rate is calculated using the same method. Assuming that the rise rate of node A is 0.05 and the topology weight is 0.8; and the rise rate of node B is 0.02 and the topology weight is 0.3, then the comprehensive rise rate is (0.05×0.8+0.02×0.3) / (0.8+0.3)≈0.04.

[0105] Based on the combined rise and fall rates, the system generates adjustment parameters for the purification control strategy. For nodes with a faster pollution rise rate, higher compensation priority and intensity are assigned. For example, based on node A's combined rise rate of 0.04, and using a pre-defined rate-compensation intensity mapping table, it is determined that the active power filter (APF) installed at that node needs to be activated, and its compensation capacity is set to 70% of its rated capacity. For nodes with decreasing pollution, the compensation intensity is appropriately reduced or the compensation action is delayed to avoid overcompensation.

[0106] When generating specific control strategies, the strategy generation module considers the pollution change trends of each node and their topological location relationships. For example, an increase in pollution at node A may affect multiple downstream nodes (such as nodes C and D). Even though the current pollution levels at nodes C and D are low, the system will take preventative measures for these nodes in advance to prevent pollution spread, such as increasing the amount of reactive power compensation devices. This preventative control strategy is derived by analyzing the relationship between pollution propagation paths and topological structure, ensuring that the control measures are forward-looking.

[0107] The strategy generation module also integrates the pollution deviation sequence into a purification control strategy scheme that includes control parameters and execution timing. For multiple devices that need to operate simultaneously, the system determines the action sequence based on topology weights and pollution change rates. For example, when mitigating a harmonic pollution event, the purification devices at nodes with high topology weights and fast pollution rise rates are triggered first, ensuring that the power quality of critical nodes is improved first. For different purification devices at the same node, the execution timing is also assigned according to their effectiveness in controlling different types of pollution; for example, the APF is activated first to suppress harmonics, and then the Static Var Compensator (SVC) is adjusted to improve voltage stability.

[0108] In a practical application scenario, the power grid detected simultaneous exceedances of 3rd and 5th harmonic pollution, involving multiple nodes. The strategy generation module analyzed the pollution deviation sequence of each node and found that node E had the fastest rate of increase in 3rd harmonic pollution, while node F had the fastest rate of increase in 5th harmonic pollution. Based on topology weights, node E is an industrial load concentration area with a weight of 0.7, while node F is a commercial area with a weight of 0.6. The system prioritizes the control of the 3rd harmonic at node E, activating its Active Harmonic Filter (APF) and setting specific harmonic suppression parameters; simultaneously, it controls the 5th harmonic at node F, but at a slightly later time than at node E. Through this differentiated control strategy, precise control of different types of pollution is achieved.

[0109] Furthermore, the strategy generation module periodically updates the pollution control strategy to adapt to dynamic changes in the power grid's operating status. When a new pollution event occurs or the power grid topology is adjusted, the system reanalyzes the pollution deviation sequence, calculates the rise / fall rates, and updates the adjustment parameters and execution timing. For example, when a large nonlinear load is added to the power grid, the pollution characteristics of the relevant nodes will change, and the strategy generation module will adjust the control strategy in a timely manner to ensure that the pollution control measures always match the actual pollution situation.

[0110] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0111] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A system for tracing the source of power quality anomalies and generating purification strategies, characterized in that, include: The power acquisition module is used to collect voltage and current waveform data of key nodes in the power grid in real time and set the power grid monitoring range corresponding to the pollution level. The feature parsing module is used to divide the power grid monitoring range into multiple power grid nodes, perform spectral feature decomposition on the waveform data of each power grid node, and generate pollution feature vectors corresponding to the power grid nodes. The anomaly tracing module is used to extract core pollution indicators from pollution feature vectors, establish pollution tracing rules associated with power grid nodes, and obtain the corresponding spectrum correction parameters for the rules. The pollution assessment module is used to identify harmonic propagation paths in the spectrum correction parameters, compensate the core pollution indicators based on the propagation paths, and calculate the pollution index of each power grid node under different compensation strategies. The parameter optimization module is used to derive the optimal purification parameters based on the pollution index and generate a pollution deviation sequence by comparing the current waveform feature values ​​with the optimal purification parameters. The strategy generation module is used to analyze the pollution deviation sequence and integrate the pollution deviation sequence into a purification control strategy scheme based on the pollution change trend of the power grid nodes. The implementation methods of the anomaly tracing module include: Harmonic distortion parameters, voltage sag parameters, and frequency fluctuation parameters are separated from the pollution feature vector, and pollution source tracing rules for power grid nodes are generated based on the harmonic distortion parameters, voltage sag parameters, and frequency fluctuation parameters. If the number of power grid nodes covered by the current pollution source tracing rule is less than the preset pollution threshold, then the pollution feature vectors of adjacent power grid nodes are traversed, and the spectral indicators not included in the source tracing rules of the adjacent nodes are added to the current rule.

2. The power quality anomaly tracing and purification strategy generation system according to claim 1, characterized in that, The implementation of the feature parsing module includes: constructing a waveform feature library corresponding to the power grid node, which contains waveform data and pollution parameter vectors mapped by spectral features; Similarity spectrum matching is performed on the pollution parameter vectors, and the power grid clusters of the pollution parameter vectors are divided according to the matching results. The spectral focal points of the waveform data are extracted from the power grid clusters and set as power grid nodes.

3. The power quality anomaly tracing and purification strategy generation system according to claim 2, characterized in that, The power grid clustering groups that partition pollution parameter vectors also include: Based on the power grid attributes and pollution intensity in the pollution parameter vector, harmonic distribution, phase shift and waveform distortion parameters are extracted, and waveform feature labels are generated based on the above parameters. The waveform feature labels are associated with pollution parameter vectors. By calculating the spectral similarity between feature labels, parameter vectors with similarity higher than a preset pollution threshold are selected to form power grid cluster groups.

4. The power quality anomaly tracing and purification strategy generation system according to claim 1, characterized in that, The methods for generating pollution feature vectors corresponding to power grid nodes include: For each power grid node, based on the node's topological location within the power grid monitoring interval, waveform distortion data of the node within a preset time period is obtained, and the waveform distortion coefficient of the node is calculated. When the waveform distortion coefficient exceeds the first pollution threshold, the node is marked as a high pollution node, and its waveform data is extracted to form a pollution feature vector; when the waveform distortion coefficient is lower than the first pollution threshold, the node is marked as a clean node, and the waveform data of the adjacent nodes of the node are superimposed on the spectrum, and the superimposed data is reconstructed into a pollution feature vector.

5. The power quality anomaly tracing and purification strategy generation system according to claim 1, characterized in that, The implementation of the pollution assessment module includes: obtaining the time series factor of the time constant of transient events and the fluctuation factor of the disturbance amplitude in the harmonic propagation path; Construct an admittance weight matrix associated with time series factors and volatility factors, and determine the pollution index under different compensation strategies based on the probability distribution of each element in the matrix.

6. The power quality anomaly tracing and purification strategy generation system according to claim 5, characterized in that, Constructing the admittance weight matrix also includes: Identify the perturbation periodicity of the time series factor. If the current periodicity perfectly matches the preset event period, then set the time series factor as the starting index of the admittance weight matrix. Calculate the harmonic phase sequence matching degree between the time series factor and the fluctuation factor, and generate the intermediate index and termination index of the admittance weight matrix in descending order of matching degree. Path backtracking is performed on the terminating index. When the matching degree of the terminating index is lower than the preset matching threshold, it is output as the final distribution of the admittance weight matrix.

7. The power quality anomaly tracing and purification strategy generation system according to claim 6, characterized in that, Methods for calculating the pollution index include: The mean of the time series factor and the range of the volatility factor for each terminating index in the statistical admittance weight matrix are calculated, and the global variance of all index factors is calculated. The time series offset coefficient is obtained by subtracting the mean of the time series factors of a single terminating index from the mean of the time series factors of adjacent indices and dividing by the global variance. At the same time, the ratio of the volatility factor range to the global variance is calculated, and the two are weighted and summed to obtain the contamination index of the index.

8. The power quality anomaly tracing and purification strategy generation system according to claim 1, characterized in that, The methods for deriving the optimal purification parameters include: Extract the pollution pattern from historical data that is closest to the current pollution index, calculate the cosine similarity between the two in harmonic distribution, and use it as the first dynamic reference value; The difference in the number of disturbance events between the current pollution index and historical pollution patterns is statistically analyzed, and the number of differences is used as a second dynamic reference value. Based on a linear combination of the first dynamic reference value and the second dynamic reference value, the optimal purification parameters in the preset purification parameter table are matched.

9. The power quality anomaly tracing and purification strategy generation system according to claim 1, characterized in that, The strategy generation module is implemented by dividing the pollution intensification interval and the pollution reduction interval according to the pollution change direction of each node in the pollution deviation sequence; The rising rate of pollution deviation in the aggravating pollution range and the falling rate of pollution deviation in the mitigating pollution range are extracted. The two are then fused according to the topological weight of the power grid nodes to generate the adjustment parameters of the purification control strategy.

Images