Power grid electric energy quality analysis method based on signal processing

Abstract

The invention discloses a power grid electric energy quality analysis method based on signal processing. The method comprises the following steps: acquiring multi-source heterogeneous electric energy quality data; processing the obtained power quality data to obtain a unified characterization parameter of the power quality of the power grid; according to the unified characterization parameter, dynamically analyzing the power grid operation state to obtain a classification result of state change; key time nodes are extracted, the electric energy quality in the abnormal state is evaluated, and a comprehensive evaluation value in the abnormal state is obtained; analyzing the frequency domain distribution characteristics of the abnormal indexes, and determining dominant frequency components of the abnormal indexes; tracing analysis is carried out on the source of the dominant frequency component, the contribution degree of each factor to abnormity is judged, and quantitative distribution of a problem source is obtained; and performing priority ranking on the indexes with relatively high contribution degrees to obtain optimization basis data of power grid operation management. According to the invention, comprehensive evaluation and abnormity diagnosis of the power quality of the power grid are realized, and the stability and reliability of power grid operation are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of power system and power quality analysis, and in particular relates to a power grid power quality analysis method based on signal processing. Background Art

[0002] Power quality analysis is a crucial component of power system operation and management. Currently, the main methods used include time-domain analysis, frequency-domain analysis, and statistical analysis. These methods, through real-time monitoring and analysis of waveforms of power parameters such as voltage and current, can identify localized issues such as voltage deviation, harmonic distortion, and three-phase imbalance. Furthermore, traditional power quality analysis relies on the measurement and evaluation of single indicators, such as voltage deviation, frequency deviation, and three-phase voltage imbalance. These indicators can, to a certain extent, reflect certain aspects of power quality and provide a basis for identifying localized issues.

[0003] However, existing technologies have numerous shortcomings. On the one hand, traditional methods struggle in complex scenarios where multiple parameters are synergistically influenced, making it difficult to form a comprehensive representation with clear physical meaning. On the other hand, existing solutions often suffer from fragmented indicators and a lack of unified quantitative standards, resulting in fragmented assessment results that make it difficult to intuitively reflect the overall operational status of the power grid. Furthermore, traditional methods are highly subjective in determining the weights of various indicators and lack a scientific basis, further weakening the reliability and practicality of the assessment. These issues directly limit the accuracy and operability of comprehensive power quality assessments, thereby impacting the decision-making efficiency of power grid operation and management. Summary of the Invention

[0004] In order to solve the above technical problems, the present invention proposes a power grid power quality analysis method based on signal processing to solve the problems existing in the above prior art.

[0005] To achieve the above objectives, in a first aspect, the present invention provides a method for analyzing power quality of a power grid based on signal processing, comprising:

[0006] Obtain multi-source heterogeneous power quality data from power grid operation data;

[0007] Process the acquired power quality data to obtain unified characterization parameters of the power grid power quality;

[0008] Based on the unified characterization parameters, the grid operation status is dynamically analyzed to obtain the classification results of the state changes;

[0009] Extract key time nodes from the classification results of state changes, evaluate the power quality under abnormal conditions, and obtain a comprehensive evaluation value under abnormal conditions;

[0010] According to the comprehensive evaluation value under abnormal conditions, the frequency domain distribution characteristics of abnormal indicators are analyzed to determine the dominant frequency components of abnormal indicators;

[0011] Conduct source analysis on the dominant frequency components, determine the contribution of each factor to the anomaly, and obtain the quantitative distribution of the problem source;

[0012] According to the quantitative distribution of problem sources, the indicators with higher contribution are prioritized to obtain the optimization basis data for power grid operation and management.

[0013] Preferably, the process of acquiring multi-source heterogeneous power quality data includes:

[0014] The original data of voltage deviation, harmonic distortion rate and three-phase imbalance are obtained through the power grid operation acquisition device;

[0015] Use signal processing technology to denoise and filter the original data;

[0016] The processed data is decomposed in the frequency domain by fast Fourier transform to obtain the time-frequency distribution characteristics of multidimensional indicators.

[0017] Preferably, the process of obtaining a unified characterization parameter of the power quality of the power grid includes:

[0018] The hierarchical clustering algorithm is used to group the characteristic vectors of voltage deviation, harmonic distortion rate and three-phase imbalance to obtain the initial clustering indicator subset.

[0019] The fuzzy aggregation method is used to perform weighted fusion on the multi-dimensional indicators in each subset to obtain the comprehensive quantitative value of the subset;

[0020] The comprehensive quantized values of all subsets are aggregated twice, and the weighted average method is used to integrate the influencing factors of voltage deviation, harmonic distortion and three-phase imbalance to obtain a unified characterization parameter of the power quality of the power grid.

[0021] Preferably, the process of obtaining the classification result of the state change includes:

[0022] A hierarchical clustering algorithm is used to dynamically analyze the time series of characterization parameters;

[0023] The changing trend of the power grid operation status is judged by calculating the parameter variance within the time window, and the classification result of the status change is obtained.

[0024] Preferably, the process of obtaining the comprehensive evaluation value under abnormal conditions includes:

[0025] Key time nodes are extracted from the classification results. If the parameter variance exceeds the preset threshold, the multi-dimensional indicators of the abnormal nodes are weighted and reconstructed through the fuzzy aggregation method to obtain a comprehensive evaluation value under the abnormal state.

[0026] Preferably, the process of determining the dominant frequency component of the abnormal indicator includes:

[0027] Use signal processing technology to perform spectrum analysis on the reconstructed data;

[0028] The dominant frequency components of the anomaly indicators are determined by short-time Fourier transform.

[0029] Preferably, the process of obtaining the quantitative distribution of problem sources includes:

[0030] The hierarchical clustering algorithm is used to trace the source of the dominant frequency components;

[0031] By calculating the correlation between each frequency component and the original data, the contribution of voltage deviation, harmonic distortion or three-phase imbalance to the anomaly is determined, and the quantitative distribution of the problem source is obtained.

[0032] Preferably, the process of obtaining optimization basis data for power grid operation management includes:

[0033] The indicators with higher contribution are prioritized by fuzzy aggregation method;

[0034] If the contribution of a certain indicator exceeds the preset threshold, its corresponding frequency domain characteristics are compared and analyzed with the unified characterization parameters to obtain the optimization basis data for power grid operation management.

[0035] Preferably, when evaluating the power quality under abnormal conditions, the method further includes:

[0036] According to the comprehensive evaluation value under abnormal state, the support vector machine algorithm is used to classify the abnormal state and obtain the prediction result of the abnormal type.

[0037] Preferably, when tracing the source of the dominant frequency component, the following steps are also included:

[0038] If the contribution of the voltage deviation exceeds a preset threshold, the deviation impact is reduced by adjusting the signal input parameters to obtain an optimized distribution feature.

[0039] Compared with the prior art, the present invention has the following advantages and technical effects:

[0040] This method comprehensively assesses power quality by acquiring heterogeneous power quality data from multiple sources from grid operation data and comprehensively processing it. This approach not only considers a single indicator but also integrates multiple indicators, resolving the difficulty of traditional methods in accurately assessing the effects of multiple parameters. This approach provides more comprehensive and accurate power quality assessment results.

[0041] This invention dynamically analyzes unified parameters characterizing power quality in the power grid, enabling real-time monitoring of changing trends in the grid's operating status. It uses a hierarchical clustering algorithm to analyze the time series of these parameters and calculates parameter variance within a time window to determine changes in the grid's operating status. This allows for timely detection of abnormalities in grid operation, providing strong support for real-time monitoring and fault warning.

[0042] This method extracts key time points from the classification results of state changes and performs weighted reconstruction and comprehensive evaluation of power quality under abnormal conditions. By processing the multidimensional indicators of abnormal nodes using fuzzy aggregation methods, it can accurately identify abnormal conditions and quantitatively assess their severity, providing clear abnormal information to power grid operators, facilitating rapid problem location and resolution.

[0043] This method uses spectrum analysis to identify the dominant frequency components of abnormal indicators based on comprehensive evaluation values during abnormal conditions, and then traces the source of these components. This method can clearly identify the contribution of factors such as voltage deviation, harmonic distortion, or three-phase imbalance to abnormal conditions, helping managers accurately identify the root cause of the problem and providing a basis for subsequent optimization measures.

[0044] This invention analyzes the quantitative distribution of problem sources and prioritizes the most contributing indicators, providing optimization data for grid operation management. Based on this data, managers can develop targeted optimization measures and adjust grid operating parameters, thereby improving grid efficiency and power quality, and reducing equipment losses and user complaints caused by power quality issues.

[0045] This invention utilizes a series of signal processing and data analysis techniques to transform complex power quality data into intuitive assessment results and optimization criteria, reducing the workload and subjective errors associated with manual analysis. Furthermore, through dynamic monitoring and real-time early warning capabilities, it helps managers promptly respond to changes in grid operations, improving the overall efficiency and scientific nature of grid management. This invention enables comprehensive assessment and anomaly diagnosis of grid power quality, helping to enhance the stability and reliability of grid operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The accompanying drawings, which constitute part of this application, are intended to provide a further understanding of this application. The exemplary embodiments and descriptions of this application are intended to explain this application and do not constitute an improper limitation on this application. In the accompanying drawings:

[0047] Figure 1 Flowchart of a method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0048] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0049] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0050] Example 1

[0051] like Figure 1 As shown, this embodiment provides a power grid power quality analysis method based on signal processing, including:

[0052] S1. Obtaining multi-source heterogeneous power quality data from power grid operation data;

[0053] Furthermore, the process of acquiring multi-source heterogeneous power quality data includes:

[0054] The original data of voltage deviation, harmonic distortion rate and three-phase imbalance are obtained through the power grid operation acquisition device;

[0055] Use signal processing technology to denoise and filter the original data;

[0056] The processed data is decomposed in the frequency domain by fast Fourier transform to obtain the time-frequency distribution characteristics of multidimensional indicators.

[0057] Specifically, the voltage deviation value collected at a certain moment was ±5%, the harmonic distortion rate was 8%, and the three-phase imbalance was 2%. Next, signal processing techniques were used to preprocess the raw data. A low-pass filter was used to remove high-frequency noise, with a cutoff frequency set at 50Hz to ensure data smoothness. The time-domain signal was then converted to the frequency domain using a fast Fourier transform (FFT), and the frequency-domain characteristics of each parameter were analyzed.

[0058] FFT decomposition of the voltage deviation revealed its spectral distribution, with the primary frequency component concentrated around 50Hz and an amplitude of 5V. FFT results for harmonic distortion revealed an amplitude of 2V for the third harmonic and 1V for the fifth harmonic. Frequency domain analysis of three-phase imbalance revealed an amplitude of 3V for the negative-sequence component and 1V for the zero-sequence component. Finally, a multi-dimensional index analysis was performed on the time-frequency distribution characteristics of each parameter to construct a comprehensive power quality evaluation model, providing data support for grid operation optimization.

[0059] S2. Processing the acquired power quality data to obtain a unified characterization parameter of the power quality of the power grid;

[0060] Furthermore, the process of obtaining a unified characterization parameter of the power quality of the power grid includes:

[0061] S201, using a hierarchical clustering algorithm to group the characteristic vectors of the voltage deviation value, harmonic distortion rate, and three-phase imbalance to obtain a preliminary clustered indicator subset;

[0062] Specifically, feature vectors are obtained through time-frequency distribution analysis, and corresponding feature data is extracted from voltage deviation, harmonic distortion, and three-phase imbalance to obtain an initial feature set. A hierarchical clustering algorithm is used to group the initial feature set. The similarity of each parameter is determined by calculating the Euclidean distance between the feature vectors, resulting in a preliminary grouping result. Based on the preliminary grouping results, the average Euclidean distance of the parameter relationships within each group is calculated. If the average distance exceeds a preset threshold, the group is regrouped to obtain an optimized grouping set. Based on the optimized grouping set, a similarity judgment index is obtained within each group. The final index subset is determined by comparing the similarity judgment index between groups.

[0063] For example, consider voltage deviation, harmonic distortion, and three-phase imbalance data collected from a substation. After time-frequency analysis, a set of eigenvectors is generated. The voltage deviation eigenvector might reveal that its variations are concentrated in the low-frequency band, the harmonic distortion eigenvector reveals the distribution of specific harmonic components, and the three-phase imbalance eigenvector highlights the influence of the negative-sequence component. This initial set of features lays the foundation for subsequent analysis.

[0064] In one analysis, the Euclidean distance between the voltage deviation vector and the harmonic distortion vector was 3.5, while the distance between the voltage deviation vector and the three-phase imbalance vector was 1.2. Initial grouping grouped the latter two together. This grouping can quickly reveal potential connections between parameters.

[0065] Based on the initial grouping results, the average Euclidean distance within each group is calculated. If a group includes both voltage deviation and three-phase imbalance, and the average distance is 2.0, exceeding the preset threshold of 1.5, a second grouping is required. After the second grouping, voltage deviation may be grouped separately, while three-phase imbalance is combined with other parameters. This optimized grouping set better reflects the inherent patterns of the data.

[0066] The similarity index for Group 1 is 0.8, and for Group 2 it is 0.3, indicating that the parameters within Group 1 are more consistent. Extracting the characteristic vector distribution characteristics from the final indicator subset reveals that the voltage deviation distribution is relatively concentrated, while the harmonic distortion exhibits discrete characteristics. This clustered association pattern helps analyze the dynamic relationship between parameters.

[0067] In one embodiment, clustered correlation patterns are used to analyze the time-frequency distribution trends. Voltage deviation may exhibit periodic fluctuations with load changes, harmonic distortion may surge due to the addition of nonlinear devices, and three-phase imbalance may vary more slowly with system operating conditions. The dynamic characteristic set can thus clearly demonstrate these trends.

[0068] For example, during a certain period of time, the dynamic characteristics of the voltage deviation showed fluctuations once per hour, while the harmonic distortion suddenly increased to 10% when the equipment started up.

[0069] When recalculating inter-group similarity using Euclidean distance, if the distance between a group and other groups is unusually large, for example, exceeding 5.0, this may indicate an unusual grouping. After re-clustering, the unusual group can be split, with the voltage deviation grouped separately and the other parameters grouped separately. This approach can effectively identify unusual patterns in the data.

[0070] For example, from multiple perspectives, the characteristic vector of voltage deviation may vary depending on the length of the power supply line, harmonic distortion is affected by the type of equipment, and three-phase imbalance is related to load distribution. These characteristics are mutually verified through clustering results, forming consistent analytical conclusions. In an extended solution, by adding a time dimension to the analysis, the dynamic characteristic set can also reflect the seasonal trends of parameters. This multi-dimensional analysis provides more comprehensive support for power grid optimization.

[0071] S202, using a fuzzy aggregation method to perform weighted fusion on the multi-dimensional indicators in each subset to obtain a comprehensive quantitative value of the subset;

[0072] Multidimensional indicator data is obtained through indicator subsets, and the data is weightedly fused using a fuzzy aggregation method to obtain a preliminary fusion result. If the correlation coefficient in the preliminary fusion result is greater than the preset threshold, the contribution rate of each indicator is calculated using the membership function to determine the contribution rate set. Based on the contribution rate set, a comprehensive quantitative value is calculated to obtain a quantitative description of each indicator subset. Based on the quantitative description, the parameter relationship is analyzed, and the similarity measurement method is used to calculate the distance between parameters to determine whether there is an abnormal parameter combination. If an abnormal parameter combination exists, the abnormal part is re-divided through clustering to obtain an optimized grouping set. Based on the optimized grouping set, the intra-group similarity measurement is calculated to determine the final comprehensive quantitative result. Based on the final comprehensive quantitative result, the distribution characteristics of the multidimensional indicators are analyzed to obtain the dynamic pattern of the parameter relationship.

[0073] Specifically, within the indicator subsets obtained through preliminary clustering, assuming a subset includes three parameters: voltage fluctuation, frequency deviation, and power factor, calculations of correlation coefficients reveal that the correlation coefficient between voltage fluctuation and frequency deviation is 85, and the correlation coefficient between power factor and voltage fluctuation is 78, both exceeding the preset threshold of 8. Using a fuzzy aggregation method, we first define a membership function. For example, the membership function for voltage fluctuation is a triangular function, the frequency deviation is a Gaussian function, and the power factor is a trapezoidal function. Based on the membership function, we calculate the contribution rate of each indicator: 4 for voltage fluctuation, 35 for frequency deviation, and 25 for power factor. Using a weighted fusion formula, we multiply the contribution rate of each indicator by its actual value and sum the results to obtain a comprehensive quantitative value for the subset.

[0074] For example, if the actual value of voltage fluctuation is 12, the frequency deviation is 0.8, and the power factor is 95, the comprehensive quantitative value is 12 × 4 + 0.8 × 35 + 95 × 25 = 3085. Further analysis revealed that the comprehensive quantitative value of this subset exhibits cyclical fluctuations throughout the day, reaching a peak of 35 between 8:00 AM and 10:00 AM and dropping to a trough of 28 between 2:00 PM and 4:00 PM. Time series analysis, combined with historical data, predicted the trend of the comprehensive quantitative value for the next week and found that it fluctuated between 28 and 36. This quantitative analysis provides data support for the dynamic adjustment of power grid operating parameters. For example, during periods of high comprehensive quantitative values, voltage fluctuation and frequency deviation can be prioritized to improve system stability.

[0075] S203. Perform secondary aggregation on the comprehensive quantized values of all subsets, and use a weighted average method to fuse the influencing factors of voltage deviation, harmonic distortion, and three-phase imbalance to obtain a unified characterization parameter of the power quality of the power grid.

[0076] The smoothed quantized data is integrated using a secondary aggregation method to obtain aggregated characteristic parameters. The influencing factors of voltage deviation are integrated using a weighted average method to obtain a weighted value for the voltage deviation. If the weighted value of the voltage deviation exceeds a preset threshold, the influencing factors of harmonic distortion are adjusted to obtain a revised weighted parameter. The influencing factors of three-phase imbalance are integrated with the revised weighted parameter to obtain a comprehensive characterization parameter for power quality. By comparing the comprehensive characterization parameter with historical data, the stability trend of power quality can be determined.

[0077] Specifically, after obtaining the comprehensive quantized values of each subset, the quantized results of voltage deviation, harmonic distortion, and three-phase imbalance are first preprocessed using signal processing technology. Fast Fourier Transform (FFT) is used to perform frequency domain analysis on the harmonic data to extract the fundamental and harmonic components. For example, the fundamental is 50Hz, the second harmonic is 100Hz, and the third harmonic is 150Hz. Their amplitude proportions are calculated respectively. For voltage deviation, the standard deviation calculation method is used. Assuming that the voltage deviation data for a certain period is [220V, 225V, 218V, 222V], the calculated standard deviation is 5V. The three-phase imbalance is calculated by calculating the positive, negative, and zero-sequence components of the three-phase voltage. Assuming that the positive-sequence voltage is 220V, the negative-sequence voltage is 5V, and the zero-sequence voltage is 3V, the calculated imbalance is 6%. Next, a weighted average method is used to integrate these three influencing factors. Assuming a weight of 4 for voltage deviation, 3 for harmonic distortion, and 3 for three-phase imbalance, the weighted average formula yields a unified characterization parameter of 8. Ultimately, this parameter can be used for a comprehensive assessment of grid power quality, providing decision support for grid operation.

[0078] S3. Dynamically analyze the grid operation status based on the unified characterization parameters to obtain classification results of the state changes;

[0079] Furthermore, the process of obtaining the classification result of the state change includes:

[0080] A hierarchical clustering algorithm is used to dynamically analyze the time series of characterization parameters;

[0081] The changing trend of the power grid operation status is judged by calculating the parameter variance within the time window, and the classification result of the status change is obtained.

[0082] Specifically, through a preset time window, time series data of characterization parameters are obtained from the power grid operation data to obtain an initial sequence set. A hierarchical clustering algorithm is used to cluster the initial sequence set to obtain a clustering grouping result. Based on the clustering grouping result, the variance of the characterization parameters in each time window is calculated to obtain a variance sequence. If the variance sequence exceeds a preset threshold, it is determined that the operating state has changed, and a state change identifier is obtained. Based on the state change identifier and the direction of the change trend, the state classification of the power grid operation is determined to obtain a classification set. By comparing the classification set with historical data, the dynamic adjustment parameters of the change trend are obtained to obtain an adjusted trend sequence. Using the adjusted trend sequence, the operating state judgment within the time window is updated to obtain the final state classification result.

[0083] For example, based on unified characterization parameters, grid operation data is first standardized, with parameters such as voltage, current, and power normalized to the range [0, 1]. For example, after normalization, the original value of 220V is mapped to 8. Next, a hierarchical clustering algorithm is used to dynamically analyze the time series of the characterization parameters. Using Euclidean distance as the distance metric and a threshold of 15, data from similar time periods are clustered together using agglomerative hierarchical clustering. The trend of grid operation status is then determined by calculating the parameter variance within a time window. The time window is set to 30 minutes, and the variance of the voltage parameter is calculated within each time window. For example, if the voltage variance within a time window is 0.2, which is lower than the set threshold of 0.3, the grid operation status during that time period is stable. Finally, based on the variance analysis and clustering results, a classification result is obtained for the state change. For example, the grid operation status is classified into three categories: "stable," "fluctuating," and "abnormal," with stable status accounting for 80%, fluctuating status accounting for 15%, and abnormal status accounting for 5%. This provides decision support for grid operation management.

[0084] S4. Extract key time nodes from the classification results of state changes, evaluate the power quality under abnormal conditions, and obtain a comprehensive evaluation value under abnormal conditions;

[0085] Furthermore, the process of obtaining the comprehensive evaluation value under abnormal conditions includes:

[0086] Key time nodes are extracted from the classification results. If the parameter variance exceeds the preset threshold, the multi-dimensional indicators of the abnormal nodes are weighted and reconstructed through the fuzzy aggregation method to obtain a comprehensive evaluation value under the abnormal state.

[0087] Furthermore, the evaluation of power quality under abnormal conditions also includes:

[0088] According to the comprehensive evaluation value under abnormal state, the support vector machine algorithm is used to classify the abnormal state and obtain the prediction result of the abnormal type.

[0089] Specifically, key time nodes are extracted from the classification results to determine whether the parameter variance exceeds the preset threshold. If so, the abnormal node is identified. Multidimensional indicator data is obtained from the abnormal node, and the multidimensional indicators are processed using a fuzzy aggregation method to obtain a preliminary aggregation result. Based on the preliminary aggregation result, the weights of the multidimensional indicators are adjusted using a weighted reconstruction method to obtain a weighted indicator set. Based on the weighted indicator set, a comprehensive evaluation value under the abnormal state is calculated to determine a quantitative description of the abnormal state. If the comprehensive evaluation value exceeds the normal range, the state change trend is analyzed through the time node to determine the persistence of the abnormality. Based on the state change trend, the support vector machine algorithm is used to classify the abnormal state and obtain a prediction result of the abnormal type. By comparing the predicted result with the historical abnormal nodes, the final confirmation result of the abnormal state is determined.

[0090] For example, in the state change classification results, time series analysis was first used to extract key time nodes. For example, in the operating data of a certain device, a sliding window algorithm detected variances of 8, 6, and 7 at the 120th, 450th, and 780th seconds, respectively, all exceeding the preset threshold of 5. Fuzzy aggregation was used to weight and reconstruct multidimensional indicators for these abnormal nodes. Assuming that the weights of temperature, pressure, and vibration were 4, 3, and 3, respectively, the fuzzy C-means clustering algorithm was used to calculate the comprehensive assessment values for abnormal conditions of 72, 68, and 70, respectively. Further analysis revealed that these abnormal nodes were highly consistent with historical equipment failure data, demonstrating that this method can effectively identify abnormal conditions and provide a reliable assessment basis. By continuously optimizing weight assignment and clustering algorithm parameters, the accuracy and practicality of anomaly detection can be further improved.

[0091] S5. Analyze the frequency domain distribution characteristics of the abnormal indicators based on the comprehensive evaluation value under the abnormal state to determine the dominant frequency components of the abnormal indicators;

[0092] Furthermore, the process of determining the dominant frequency component of the abnormal indicator includes:

[0093] Use signal processing technology to perform spectrum analysis on the reconstructed data;

[0094] The dominant frequency components of the anomaly indicators are determined by short-time Fourier transform.

[0095] Specifically, under abnormal conditions, the power quality data was first comprehensively evaluated, and its comprehensive evaluation value was calculated to be 78, indicating that the system had obvious power quality problems. In order to further analyze the cause of the problem, signal processing technology was used to perform spectral analysis on the reconstructed data. Through short-time Fourier transform, the time domain signal was converted into a frequency domain signal, and the window length was selected as 256 points with an overlap rate of 50%. After the transformation, it was found that the dominant frequency component of the abnormal indicator was 150Hz, and its amplitude was significantly higher than other frequency components, reaching 45. Combined with the frequency domain distribution characteristics, it can be determined that the anomaly is mainly caused by the third harmonic, and its energy is concentrated near 150Hz, indicating that the system has serious harmonic pollution. Based on this analysis result, it can be further inferred that there may be interference from nonlinear loads or power electronic equipment in the system, thus providing a basis for subsequent power quality management.

[0096] S6. Conduct source analysis on the dominant frequency component, determine the contribution of each factor to the anomaly, and obtain a quantitative distribution of the problem source;

[0097] Furthermore, the process of obtaining a quantitative distribution of problem sources includes:

[0098] The hierarchical clustering algorithm is used to trace the source of the dominant frequency components;

[0099] By calculating the correlation between each frequency component and the original data, the contribution of voltage deviation, harmonic distortion or three-phase imbalance to the anomaly is determined, and the quantitative distribution of the problem source is obtained.

[0100] Furthermore, when tracing the source of the dominant frequency component, the following is also included:

[0101] If the contribution of the voltage deviation exceeds a preset threshold, the deviation impact is reduced by adjusting the signal input parameters to obtain an optimized distribution feature.

[0102] Specifically, in the frequency domain distribution feature analysis, the original voltage signal is first converted from the time domain to the frequency domain by fast Fourier transform (FFT) to obtain the amplitude and phase information of the frequency component.

[0103] For example, for a 50Hz voltage signal, FFT processing can extract the fundamental component (50Hz) and its harmonic components (such as 100Hz, 150Hz, etc.). Then, a hierarchical clustering algorithm is used to cluster these frequency components. Using Euclidean distance as the similarity metric, hierarchical clustering is performed using the Ward method to divide the frequency components into different clusters.

[0104] For example, suppose the clustering results divide the frequency components into three clusters, corresponding to the fundamental wave, the second harmonic, and the third harmonic. Then, the correlation between each frequency component and the original data is calculated and quantitatively analyzed using the Pearson correlation coefficient.

[0105] For example, the correlation coefficient between the fundamental component and the original data is 95, the second harmonic is 30, and the third harmonic is 25. Based on the size of the correlation coefficient, the contribution of each frequency component to the voltage anomaly can be determined.

[0106] For example, a high correlation of the fundamental component indicates that voltage deviation is the primary source of the problem, while a low correlation of the second and third harmonics indicates that harmonic distortion has a smaller impact. Finally, by calculating the proportion of each cluster's frequency component in the total spectrum, a quantitative distribution of the problem source is obtained.

[0107] For example, the fundamental cluster accounts for 80%, the second harmonic cluster accounts for 15%, and the third harmonic cluster accounts for 5%, further quantifying the contribution of voltage deviation, harmonic distortion, and three-phase imbalance to the anomaly.

[0108] S7. Based on the quantitative distribution of problem sources, prioritize the indicators with higher contribution and obtain the optimization basis data for power grid operation and management.

[0109] Furthermore, the process of obtaining optimization basis data for power grid operation management includes:

[0110] The indicators with higher contribution are prioritized by fuzzy aggregation method;

[0111] If the contribution of a certain indicator exceeds the preset threshold, its corresponding frequency domain characteristics are compared and analyzed with the unified characterization parameters to obtain the optimization basis data for power grid operation management.

[0112] Specifically, voltage fluctuation data was categorized into three categories: high, medium, and low, accounting for 30%, 50%, and 20%, respectively. A fuzzy aggregation method was then used to prioritize indicators with higher contribution. The contribution of each indicator was calculated using a fuzzy comprehensive evaluation model. For example, the contribution of voltage fluctuation was 65, the contribution of frequency deviation was 55, and the contribution of load factor was 45. For indicators with a contribution exceeding 6, such as voltage fluctuation, their corresponding frequency domain features were further compared and analyzed with unified characterization parameters. The voltage fluctuation data was converted into frequency domain features using a Fourier transform, and the main frequency components, such as 50Hz and 100Hz, were extracted. Comparison with the unified characterization parameters revealed that the amplitude deviation of the 50Hz component was 5%, and the amplitude deviation of the 100Hz component was 3%. Based on these analysis results, optimization data for grid operation management was generated, such as adjusting voltage regulator parameters to reduce the amplitude deviation of the 50Hz component to 2%, thereby improving grid stability.

[0113] The above are merely preferred embodiments of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A method for analyzing power quality of a power grid based on signal processing, characterized in that: The following steps are involved: Obtain multi-source heterogeneous power quality data from power grid operation data; Process the acquired power quality data to obtain unified characterization parameters of the power grid power quality; Based on the unified characterization parameters, the grid operation status is dynamically analyzed to obtain the classification results of the state changes; Extract key time nodes from the classification results of state changes, evaluate the power quality under abnormal conditions, and obtain a comprehensive evaluation value under abnormal conditions; According to the comprehensive evaluation value under abnormal conditions, the frequency domain distribution characteristics of abnormal indicators are analyzed to determine the dominant frequency components of abnormal indicators; Conduct source analysis on the dominant frequency components, determine the contribution of each factor to the anomaly, and obtain the quantitative distribution of the problem source; According to the quantitative distribution of problem sources, the indicators with higher contribution are prioritized to obtain the optimization basis data for power grid operation and management.

2. The method according to claim 1, characterized in that The process of acquiring multi-source heterogeneous power quality data includes: The original data of voltage deviation, harmonic distortion rate and three-phase imbalance are obtained through the power grid operation acquisition device; Use signal processing technology to denoise and filter the original data; The processed data is decomposed in the frequency domain by fast Fourier transform to obtain the time-frequency distribution characteristics of multidimensional indicators.

3. The method according to claim 1, characterized in that The process of obtaining unified characterization parameters of grid power quality includes: The hierarchical clustering algorithm is used to group the characteristic vectors of voltage deviation, harmonic distortion rate and three-phase imbalance to obtain the initial clustering indicator subset. The fuzzy aggregation method is used to perform weighted fusion on the multi-dimensional indicators in each subset to obtain the comprehensive quantitative value of the subset; The comprehensive quantized values of all subsets are aggregated twice, and the weighted average method is used to integrate the influencing factors of voltage deviation, harmonic distortion and three-phase imbalance to obtain a unified characterization parameter of the power quality of the power grid.

4. The method according to claim 1, wherein The process of obtaining the classification results of state changes includes: A hierarchical clustering algorithm is used to dynamically analyze the time series of characterization parameters; The changing trend of the power grid operation status is judged by calculating the parameter variance within the time window, and the classification result of the status change is obtained.

5. The method according to claim 1, characterized in that The process of obtaining the comprehensive evaluation value under abnormal conditions includes: Key time nodes are extracted from the classification results. If the parameter variance exceeds the preset threshold, the multi-dimensional indicators of the abnormal nodes are weighted and reconstructed through the fuzzy aggregation method to obtain a comprehensive evaluation value under the abnormal state.

6. The method according to claim 1, characterized in that The process of determining the dominant frequency components of anomaly indicators includes: Use signal processing technology to perform spectrum analysis on the reconstructed data; The dominant frequency components of the anomaly indicators are determined by short-time Fourier transform.

7. The method according to claim 1, characterized in that The process of obtaining a quantitative distribution of problem sources involves: The hierarchical clustering algorithm is used to trace the source of the dominant frequency components; By calculating the correlation between each frequency component and the original data, the contribution of voltage deviation, harmonic distortion or three-phase imbalance to the anomaly is determined, and the quantitative distribution of the problem source is obtained.

8. The method according to claim 1, characterized in that The process of obtaining optimization basis data for power grid operation management includes: The indicators with higher contribution are prioritized by fuzzy aggregation method; If the contribution of a certain indicator exceeds the preset threshold, its corresponding frequency domain characteristics are compared and analyzed with the unified characterization parameters to obtain the optimization basis data for power grid operation management.

9. The method according to claim 1, characterized in that When evaluating power quality under abnormal conditions, it also includes: According to the comprehensive evaluation value under abnormal state, the support vector machine algorithm is used to classify the abnormal state and obtain the prediction result of the abnormal type.

10. The method according to claim 1, characterized in that When tracing the source of the dominant frequency component, the following also needs to be done: If the contribution of the voltage deviation exceeds a preset threshold, the deviation impact is reduced by adjusting the signal input parameters to obtain an optimized distribution feature.

Images