Voltage quality anomaly detection and traceability method based on dynamic time warping

Abstract

The invention discloses a voltage quality anomaly detection and traceability method based on dynamic time warping. Accurate identification of voltage quality anomaly of a power distribution network and rapid traceability of event causes are realized. The method comprises the following steps: firstly, by extracting statistical characteristic quantity of equivalent voltage and combining judgment thresholds for voltage deviation and fluctuation in national standards, identifying abnormal conditions such as overvoltage, undervoltage, voltage fluctuation and the like; and secondly, performing time alignment on the voltage anomaly information and external disturbance factors such as load state, illumination and wind speed at the same moment, and constructing a voltage signal-abnormal value-event source association database, thereby realizing causal modeling of voltage anomaly and possible reasons, and providing data support and association basis for anomaly traceability. And finally, performing similarity matching on a voltage sequence input in real time and a historical sample by adopting a DTW algorithm, accurately identifying the most similar historical scene sample, and quickly positioning an abnormal cause through associated information of the most similar historical scene sample, thereby realizing efficient voltage abnormal event tracing.

Description

Technical Field

[0001] The present invention belongs to the technical field of voltage quality detection in distribution networks, and specifically provides a voltage quality anomaly detection and tracing method based on dynamic time warping. Background Art

[0002] With the integration of a large number of renewable energy sources into the power grid, the uncertainty of wind and solar power generation has increased the difficulty of grid regulation. Furthermore, the integration of power electronic converters and nonlinear loads has significantly increased the probability of voltage quality disturbances in the power grid. The cross-coupling between different disturbances has also made disturbance types increasingly complex, posing a serious threat to the safe operation of the power system. As one of the evaluation indicators of power systems, voltage quality is a key prerequisite for the targeted analysis and management of voltage quality issues, and is of great significance to the stable operation of the power system.

[0003] Most existing research on the identification and classification of voltage quality disturbances generally involves two steps: signal feature extraction and classification. Feature extraction involves extracting characteristic information from a signal through mathematical methods, primarily using time-frequency analysis techniques such as short-time Fourier transform, wavelet transform, variational mode decomposition, and S-transform. Disturbance classification methods primarily utilize neural networks, support vector machines, decision trees, and gradient boosting trees.

[0004] Current voltage quality detection and traceability technologies are focused on single voltage quality events, and the classification of disturbance factors is limited to label classification. This cannot meet the needs of problem analysis under the comprehensive operation of multiple sources in actual scenarios. After classifying voltage signal anomalies, it is difficult to further analyze the physical causes behind them, and there is a lack of effective means to locate the source of the disturbance. Summary of the Invention

[0005] In order to solve the defects and shortcomings in the above-mentioned prior art, the present invention provides a method for correlating the characteristic values ​​extracted from the signal, local historical records and surrounding environmental factors with various voltage quality disturbances on the basis of studying voltage signal feature extraction and anomaly detection, constructing a voltage signal-anomaly-event source correlation database, and tracing the anomaly after similarity matching through a dynamic time warping strategy. On the basis of realizing rapid detection of voltage anomalies, the source and cause of the disturbance are further inferred, and multi-source information can be integrated to realize the identification and cause analysis of voltage anomalies, thereby solving the problem of tracing the source of voltage quality problems caused by multi-factor coupled disturbances in the prior art.

[0006] The present invention provides the following technical solutions:

[0007] A method for detecting and tracing voltage quality anomalies based on dynamic time warping includes the following steps:

[0008] S1. Feature extraction and anomaly detection of voltage signals: After obtaining voltage data and preprocessing it, the voltage signal feature is extracted and the voltage anomaly value is determined based on it;

[0009] S2. Align abnormal voltage data with corresponding load data and environmental factors according to timestamps, establish a voltage anomaly association database, and implement causal modeling of voltage anomalies and disturbance sources;

[0010] S3. Based on the dynamic time warping algorithm, the input voltage sequence is matched with the historical samples in the database to accurately index the most similar historical disturbance source.

[0011] Preferably, step S1 specifically refers to obtaining 24-hour three-phase voltage data of the monitoring point, and after preprocessing, extracting voltage signal characteristics, including maximum value, minimum value, average value, variance and standard deviation, which are used to characterize the voltage fluctuation characteristics, and based on this, determine whether there is an overvoltage, undervoltage or abnormal voltage fluctuation event.

[0012] Preferably, the detailed steps of step S1 are as follows:

[0013] S11. Extraction of voltage signal characteristics: To effectively identify overvoltage, undervoltage, and voltage fluctuation disturbance events in voltage signals, a quantitative analysis method for voltage signals containing disturbances is proposed. For a three-phase AC voltage system, the mean three-phase voltage at each moment is calculated, and statistical characteristics are extracted based on the time series.

[0014] S12. Identification of abnormal voltage signal values: According to the national standard for power quality, the allowable deviation of the power supply voltage is ±5% of the nominal value.

[0015] Preferably, in step S11, the data is first collected and preprocessed, and the three-phase voltage instantaneous value U recorded every hour within 24 hours is recorded. a (t), U b (t), U c (t), where t represents the hour number, calculate the equivalent voltage of the node; use the arithmetic mean method to calculate the equivalent voltage mean value U per hour avg (t) is:

[0016]

[0017] The 24-hour mean result is constructed as a voltage time series {U avg (0),U avg (1),...,U avg (23)}, extract the signal feature quantity U according to the voltage sequence max 、U min 、U mean 、U std and Uvar To characterize and distinguish voltage disturbances, the definitions and calculation methods of these characteristic quantities are as follows, where N = 24;

[0018] U max :Take the voltage sequence {U avg (t)};

[0019] U max =max({U avg (t)}) (14)

[0020] U min :Take the voltage sequence {U avg (t)} minimum value;

[0021] U min =min({U avg (t)}) (15)

[0022] U mean :Voltage sequence {U avg (t)} mean;

[0023]

[0024] U var :Voltage sequence {U avg (t)} variance;

[0025]

[0026] U std :Voltage sequence {U avg (t)} standard deviation;

[0027]

[0028] Preferably, in step S12, the nominal voltage of the system is set to U n , then the overvoltage and undervoltage thresholds are:

[0029]

[0030] If the node voltage U avg (t)>U high , then transient overvoltage occurs; if the 24-hour voltage average U mean >U high , then a continuous overvoltage occurs. On the contrary, if the node voltage U avg (t)<U low , then transient undervoltage occurs; if the 24-hour voltage average U mean <U low , then a continuous undervoltage occurs.

[0031] Voltage fluctuation refers to the rapid change of voltage in a short period of time. According to GB / T 12325-2008, the threshold of voltage fluctuation should not exceed ±5%, that is:

[0032] ΔU=U max -U min <0.10U n (20)

[0033] If the voltage fluctuation range ΔU>0.10U n , it is judged that the voltage fluctuation exceeds the standard. If the voltage standard deviation ( is an approximate formula based on the standard deviation of uniform distribution), the voltage fluctuation is more serious; if the node voltage difference at adjacent time points |ΔU t |>0.05U n , then transient voltage fluctuation occurs, where |ΔU t |=U t -U t-1 (t=0,1,...,23).

[0034] Preferably, step S2 specifically involves constructing a voltage signal-anomaly-value-event source association table for rapid source tracing. First, the collected historical voltage data is preprocessed to extract features and perform outlier identification. The detected voltage anomaly events are accurately recorded with their occurrence timestamps. Voltage data, load curves, and environmental change factors are aligned by timestamp. For voltage anomaly events occurring at a specific moment, the impact of concurrent load changes and seasonal factors must be correlated to construct an association table database.

[0035] Preferably, step S3 specifically refers to performing similarity matching between the real-time voltage data and the historical samples in the database using a DTW algorithm to overcome the comparison error caused by inconsistent time series lengths and time misalignment, thereby indexing the most similar historical voltage features;

[0036] The detailed steps of the algorithm are as follows:

[0037] Given two time series X = {x1, x2, ..., x n} and Y={y1,y2,...,y m}, construct a distance matrix D according to the sequence dimension n×m :

[0038]

[0039] Where: Matrix element d(x i ,y j ) represents x i and y j The Euclidean distance,

[0040] Let the curved path be W k ={w1,w2,...,w v ,...,w f}, where w v =d(x i ,y j ), f is the path length. W k The corresponding dynamic regularization distance D k is the sum of the distances to all points.

[0041]

[0042] If the matrix D has g curved paths, then the dynamic regularization minimum distance D between sequences X and Y is min for:

[0043] D min =min{D k |k=1,2,...,g}. (23)

[0044] Preferably, the constraints of the curved path are:

[0045] (1) The regular path must satisfy w1=d(x1,y1) and w f =d(x n ,y m );

[0046] (2) In the path, w v =d(x i ,y j ) The next adjacent point w v+1 =d(x i' ,y j' ) satisfies 0≤i'-i≤1 and 0≤j'-j≤1;

[0047] DTW efficiently calculates D through dynamic programming min , avoiding exhaustive enumeration of all paths. Assume that the minimum curvature path to the current point D(i,j) has been determined to be {w1,w2,...,w v}, then the next adjacent path w v+1 Can be determined as:

[0048]

[0049] And so on, continue to determine the adjacent path w v+2 , until the path w to D(n,m) is determined f , D can be obtained by summing all points on the minimum curvature path min ;

[0050] Based on the historical samples determined by the DTW matching results, combined with the corresponding anomaly types and disturbance causes in the associated database, the source analysis of the current input voltage anomaly is quickly completed, and the possible event sources and environmental influencing factors are output.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] The present invention realizes the accurate identification of voltage quality anomalies in the distribution network and the rapid tracing of the causes of events. First, by extracting the statistical characteristic quantity of the equivalent voltage and combining it with the judgment threshold of voltage deviation and fluctuation in the national standard, the identification of abnormal situations such as overvoltage, undervoltage and voltage fluctuation is realized. Secondly, the voltage anomaly information is time-aligned with the external disturbance factors such as the load state, light, wind speed at the same moment, and a voltage signal-anomaly value-event source association database is constructed to realize causal modeling of voltage anomalies and possible causes, providing data support and association basis for anomaly tracing. Finally, the DTW algorithm is used to match the real-time input voltage sequence with the historical samples for similarity, accurately identify the most similar historical scene samples, and quickly locate the cause of the anomaly through its associated information, thereby realizing efficient voltage anomaly event tracing. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 It is a specific flow chart of the present invention. DETAILED DESCRIPTION

[0054] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0055] See also Figure 1 A detection and tracing method for distribution network voltage quality can integrate multi-source information to realize the identification and cause analysis of voltage anomalies, solving the problem of tracing the voltage quality problems caused by multi-factor coupling disturbances that cannot be effectively dealt with in existing technologies.

[0056] The present invention is achieved through the following technical steps.

[0057] Step 1: Feature extraction and anomaly detection of voltage signals. After obtaining voltage data and preprocessing it, the voltage signal features are extracted and the voltage anomaly values ​​are identified based on them.

[0058] Step 2: Align the abnormal voltage data with the corresponding load data and environmental factors according to the timestamp, establish a voltage anomaly association database, and realize causal modeling of voltage anomalies and disturbance sources.

[0059] Step 3: Based on the dynamic time warping (DTW) algorithm, the input voltage sequence is matched with the historical samples in the database to accurately index the most similar historical disturbance source.

[0060] Next, the method of the present invention is described in detail.

[0061] Step 1: Obtain 24-hour three-phase voltage data from the monitoring point. After preprocessing, extract voltage signal characteristics, including maximum, minimum, average, variance, and standard deviation, to characterize voltage fluctuation characteristics and determine whether there are abnormal events such as overvoltage, undervoltage, or voltage fluctuation.

[0062] The detailed steps of the method are as follows:

[0063] Step 1.1: Extract voltage signal features. To effectively identify disturbances such as overvoltage, undervoltage, and voltage fluctuations in voltage signals, this method proposes a quantitative analysis method for voltage signals containing disturbances. For a three-phase AC voltage system, the mean three-phase voltage at each moment is calculated, and statistical features are extracted based on the time series.

[0064] The first step is to collect and pre-process the data. The instantaneous value of the three-phase voltage U recorded every hour within 24 hours is a (t), U b (t), U c (t), where t represents the hour number, calculate the equivalent voltage of the node. Use the arithmetic mean method to calculate the equivalent voltage mean value U per hour avg (t) is:

[0065]

[0066] The 24-hour mean result is constructed as a voltage time series {U avg (0),U avg (1),...,U avg (23)}, extract the signal feature quantity U according to the voltage sequence max 、U min 、U mean 、U std and U var The definitions and calculation methods of these characteristic quantities used to characterize and distinguish voltage disturbances are as follows, where N=24.

[0067] U max :Take the voltage sequence {U avg (t)}.

[0068] U max =max({Uavg (t)}) (26)

[0069] U min :Take the voltage sequence {U avg (t)}.

[0070] U min =min({U avg (t)}) (27)

[0071] U mean :Voltage sequence {U avg (t)}.

[0072]

[0073] U var :Voltage sequence {U avg (t)}.

[0074]

[0075] U std :Voltage sequence {U avg (t)}.

[0076]

[0077] Step 1.2: Identify abnormal voltage values. According to the national standard for power quality (GB / T 12325-2008), the allowable deviation of the supply voltage is ±5% of the nominal value. Assume that the nominal voltage of the system is U n , then the overvoltage and undervoltage thresholds are:

[0078]

[0079] If the node voltage U avg (t)>U high , then transient overvoltage occurs; if the 24-hour voltage average U mean >U high , then a continuous overvoltage occurs. On the contrary, if the node voltage U avg (t)<U low , then transient undervoltage occurs; if the 24-hour voltage average U mean <U low , then a continuous undervoltage occurs.

[0080] Voltage fluctuation refers to the rapid change of voltage in a short period of time. According to GB / T 12325-2008, the threshold of voltage fluctuation should not exceed ±5%, that is:

[0081] ΔU=U max -Umin <0.10U n (32)

[0082] If the voltage fluctuation range ΔU>0.10U n , it is judged that the voltage fluctuation exceeds the standard. If the voltage standard deviation ( is an approximate formula based on the standard deviation of uniform distribution), the voltage fluctuation is more serious. If the node voltage difference at adjacent time points |ΔU t |>0.05U n , then transient voltage fluctuation occurs, where |ΔU t |=U t -U t-1 (t=0,1,...,23).

[0083] Step 2: Construct a voltage signal-outlier-event source association table for rapid source tracing. First, preprocess the collected historical voltage data, extract features, and identify outliers. Accurately record the timestamps of detected voltage anomalies. Align voltage data, load curves, and environmental factors by timestamp (e.g., hourly granularity). For voltage anomalies occurring at a specific moment, correlate the impact of concurrent load changes and seasonal factors. Construct a database of association tables, as shown in Appendix 1.

[0084] Table 1. Voltage signal-abnormal value-event source correlation table

[0085]

[0086] Step 3: Use the DTW algorithm to perform similarity matching between the real-time voltage data and the historical samples in the database to overcome the comparison errors caused by inconsistent time series lengths and time misalignment, thereby indexing the most similar historical voltage features.

[0087] The detailed steps of the algorithm are as follows:

[0088] Given two time series X = {x1, x2, ..., x n} and Y={y1,y2,...,y m}, construct a distance matrix D according to the sequence dimension n×m :

[0089]

[0090] Where: Matrix element d(x i ,y j ) represents x i and y j The Euclidean distance,

[0091] Let the curved path be W k ={w1,w2,...,w v ,...,w f}, where w v =d(x i ,y j ), f is the path length. W k The corresponding dynamic regularization distance D k is the sum of the distances to all points.

[0092]

[0093] If the matrix D has g curved paths, then the dynamic regularization minimum distance D between sequences X and Y is min for:

[0094] D min =min{D k |k=1,2,...,g} (35)

[0095] The constraints for a curved path are:

[0096] (1) The regular path must satisfy w1=d(x1,y1) and w f =d(x n ,y m ).

[0097] (2) In the path, w v =d(x i ,y j ) The next adjacent point w v+1 =d(x i' ,y j' ) satisfies 0≤i'-i≤1 and 0≤j'-j≤1.

[0098] DTW efficiently calculates D through dynamic programming min , avoiding exhaustive enumeration of all paths. Assume that the minimum curvature path to the current point D(i,j) has been determined to be {w1,w2,...,w v}, then the next adjacent path w v+1 Can be determined as:

[0099]

[0100] And so on, continue to determine the adjacent path w v+2 , until the path w to D(n,m) is determined f , D can be obtained by summing all points on the minimum curvature path min .

[0101] Based on the historical samples determined by the DTW matching results, combined with the corresponding anomaly types and disturbance causes in the associated database, the source analysis of the current input voltage anomaly is quickly completed, and the possible event sources and environmental influencing factors are output.

[0102] The main innovation of this invention focuses on improving the detection accuracy and cause identification capabilities of abnormal voltage quality events, which can be summarized as follows:

[0103] 1. Voltage signal feature extraction and anomaly identification mechanism: Extract key statistical features of the voltage signal and, in combination with national standards, determine whether anomalies such as overvoltage, undervoltage, and voltage fluctuations exist.

[0104] 2. Voltage anomaly modeling method based on multi-source data fusion. Build a relational database containing voltage characteristics, load changes, and environmental factors to achieve causal modeling and data storage of abnormal events under the coupling of multiple sources.

[0105] 3. Anomaly tracing method based on dynamic time warping. A dynamic time warping algorithm is used to perform similarity matching between real-time voltage series and historical samples in the associated database, quickly retrieving the most similar historical scenarios, thereby enabling source analysis and cause inference of the current voltage anomaly event.

[0106] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A voltage quality anomaly detection and tracing method based on dynamic time warping, characterized by: The following steps are involved: S1. Feature extraction and anomaly detection of voltage signals: After obtaining voltage data and preprocessing it, the voltage signal feature is extracted and the voltage anomaly value is determined based on it; S2. Align abnormal voltage data with corresponding load data and environmental factors according to timestamps, establish a voltage anomaly association database, and implement causal modeling of voltage anomalies and disturbance sources; S3. Based on the dynamic time warping algorithm, the input voltage sequence is matched with the historical samples in the database to accurately index the most similar historical disturbance source.

2. The method for detecting and tracing voltage quality anomalies based on dynamic time warping according to claim 1, characterized in that: The step S1 specifically refers to obtaining 24-hour three-phase voltage data of the monitoring point, performing preprocessing, and extracting voltage signal characteristics, including maximum value, minimum value, average value, variance and standard deviation, to characterize the voltage fluctuation characteristics, and based on this, determine whether there is an overvoltage, undervoltage or abnormal voltage fluctuation event.

3. The method for voltage quality anomaly detection and tracing based on dynamic time warping according to claim 2 is characterized by: The detailed steps of step S1 are as follows: S11. Extraction of voltage signal characteristics: To effectively identify overvoltage, undervoltage, and voltage fluctuation disturbance events in voltage signals, a quantitative analysis method for voltage signals containing disturbances is proposed. For a three-phase AC voltage system, the mean three-phase voltage at each moment is calculated, and statistical characteristics are extracted based on the time series. S12. Determination of abnormal voltage signal values: According to the national standard for power quality, the allowable deviation of the power supply voltage is ±5% of the nominal value.

4. The method for detecting and tracing voltage quality anomalies based on dynamic time warping according to claim 3 is characterized by: In step S11, the data is first collected and pre-processed, and the three-phase voltage instantaneous value U recorded every hour within 24 hours is calculated. a (t), U b (t), U c (t), where t represents the hour number, calculate the equivalent voltage of the node; use the arithmetic mean method to calculate the equivalent voltage mean value U per hour avg (t) is: The 24-hour mean result is constructed as a voltage time series {U avg (0),U avg (1),...,U avg (23)}, extract the signal feature quantity U according to the voltage sequence max 、U min 、U mean 、U std and U var To characterize and distinguish voltage disturbances, the definitions and calculation methods of these characteristic quantities are as follows, where N = 24; U max :Take the voltage sequence {U avg (t)}; U max =max({U avg (t)}) (2) U min :Take the voltage sequence {U avg (t)} minimum value; U min =min({U avg (t)}) (3) U mean :Voltage sequence {U avg (t)} mean; U var :Voltage sequence {U avg (t)} variance; U std :Voltage sequence {U avg (t)} standard deviation; 5. The method for voltage quality anomaly detection and tracing based on dynamic time warping according to claim 3 is characterized by: In step S12, the nominal voltage of the system is set to U n , then the overvoltage and undervoltage thresholds are: If the node voltage U avg (t)>U high , then transient overvoltage occurs; if the 24-hour voltage average U mean >U high , then a continuous overvoltage occurs; On the contrary, if the node voltage U avg (t)<U low , then transient undervoltage occurs; if the 24-hour voltage average U mean <U low , then there will be continuous undervoltage; Voltage fluctuation refers to the rapid change of voltage in a short period of time. According to GB / T 12325-2008, the threshold of voltage fluctuation should not exceed ±5%, that is: ΔU=U max -U min <0.10U n (8) If the voltage fluctuation range ΔU>0.10U n , it is judged that the voltage fluctuation exceeds the standard. If the voltage standard deviation ( is an approximate formula based on the standard deviation of uniform distribution), the voltage fluctuation is more serious; if the node voltage difference at adjacent time points |ΔU t |>0.05U n , then transient voltage fluctuation occurs, where |ΔU t |=U t -U t-1 (t=0,1,...,23).

6. The method for voltage quality anomaly detection and tracing based on dynamic time warping according to claim 1 is characterized by: The step S2 specifically refers to constructing a voltage signal-abnormal value-event source association table for rapid source tracing. First, the collected historical voltage data is preprocessed, characteristic quantities are extracted and abnormal value discrimination is performed, and the timestamp of the detected voltage abnormality event is accurately recorded; the voltage data, load curve, and environmental change factors are aligned according to the timestamp. For voltage abnormality events occurring at a certain moment, it is necessary to associate the impact of the load changes and seasonal factors during the same period to construct an association table database.

7. The method for voltage quality anomaly detection and tracing based on dynamic time warping according to claim 1 is characterized by: Step S3 specifically involves performing similarity matching between the real-time voltage data and the historical samples in the database using the DTW algorithm to overcome the comparison errors caused by inconsistent time series lengths and time misalignment, thereby indexing the most similar historical voltage features; The detailed steps of the algorithm are as follows: Given two time series X = {x1, x2, ..., x n } and Y={y1,y2,...,y m }, construct a distance matrix D according to the sequence dimension n×m : Where: Matrix element d(x i ,y j ) represents x i and y j The Euclidean distance, Let the curved path be W k ={w1,w2,...,w v ,...,w f }, where w v =d(x i ,y j ), f is the path length; W k The corresponding dynamic regularization distance D k is the sum of the distances of all points; If the matrix D has g curved paths, then the dynamic regularization minimum distance D between sequences X and Y is min for: D min =min{D k |k=1,2,...,g}。 (11) 8. The method for detecting and tracing voltage quality anomalies based on dynamic time warping according to claim 7, characterized in that: The constraints of the curved path are: (1) The regular path must satisfy w1=d(x1,y1) and w f =d(x n ,y m ); (2) In the path, w v =d(x i ,y j ) The next adjacent point w v+1 =d(x i' ,y j' ) satisfies 0≤i'-i≤1 and 0≤j'-j≤1; DTW efficiently calculates D through dynamic programming min , avoid exhausting all paths; assume that the minimum curvature path to the current point D(i,j) has been determined to be {w1,w2,...,w v }, then the next adjacent path w v+1 Can be determined as: And so on, continue to determine the adjacent path w v+2 , until the path w to D(n,m) is determined f , D can be obtained by summing all points on the minimum curvature path min ; Based on the historical samples determined by the DTW matching results, combined with the corresponding anomaly types and disturbance causes in the associated database, the source analysis of the current input voltage anomaly is quickly completed, and the possible event sources and environmental influencing factors are output.

Images