A power quality monitoring method and system of a power collection terminal

Abstract

This invention relates to the field of time-series data processing, specifically to a method and system for power quality monitoring in a power acquisition terminal. The method includes: dividing a historical voltage sequence acquired by a power meter into several sub-sequences, including stationary sub-sequences and fluctuating sub-sequences; filtering the sub-sequences based on time-series analysis to obtain similar sequences to the real-time voltage sequence, and designating the next sub-sequence adjacent to the similar sequence as a neighboring sequence, wherein one similar sequence corresponds to one neighboring sequence; predicting the characteristics of future voltage sequences based on all the neighboring sequences, further monitoring voltage anomalies, and achieving early warning. This invention can improve the reliability and timeliness of voltage anomaly monitoring.

Description

Technical Field

[0001] This invention relates to the field of time-series data processing. More specifically, this invention relates to a method and system for power quality monitoring of a power acquisition terminal. Background Technology

[0002] In industrial settings, industrial production equipment (especially precision equipment and continuous production lines) is highly sensitive to power supply stability. Abnormal voltage fluctuations (such as sudden rises, falls, and high-frequency oscillations) can easily interfere with the normal operation of equipment, leading to process interruptions, product scrapping, and consequently, significant economic losses. Therefore, it is necessary to conduct time-series analysis on the continuous voltage time-series data collected by electricity meters in industrial areas to monitor voltage anomalies.

[0003] Traditional techniques for time-series analysis of voltage anomaly monitoring typically use a fixed-increment expansion window to obtain fluctuation sequences. However, due to the differences in the length of fluctuation sequences for different fluctuation types (e.g., high-intensity fluctuations have large peak values ​​and high frequencies, while low-intensity fluctuations have small peak values ​​and low frequencies), fluctuation sequences obtained using fixed increments may be incomplete or redundant, reducing the accuracy of voltage data in time-series analysis and affecting the reliability and timeliness of voltage anomaly monitoring. Summary of the Invention

[0004] To address the problem that the traditional fixed incremental window method cannot adapt to the differences in sequence length due to voltage fluctuations of varying intensities, which can easily lead to incomplete or redundant sequences, reduce analytical accuracy, and affect the reliability and timeliness of monitoring, this invention provides solutions in the following aspects.

[0005] In a first aspect, a power quality monitoring method for a power acquisition terminal is provided. The method includes: dividing an acquired historical voltage sequence into several sub-sequences, the sub-sequences including stationary sub-sequences and fluctuating sub-sequences; filtering the sub-sequences based on a time series analysis method to obtain similar sequences to the real-time voltage sequence, and taking the next sub-sequence adjacent to the similar sequence as an adjacent sequence, wherein one similar sequence corresponds to one adjacent sequence; predicting the characteristics of future voltage sequences based on all the adjacent sequences to complete power quality analysis; wherein the process of obtaining the fluctuating sub-sequences is as follows: setting a sampling window on the historical voltage sequence, the window sequence responding to the sampling window satisfying a preset fluctuation threshold. The process involves: obtaining a first newly added sequence corresponding to a preset first window increment; expanding the sampling window by the first window increment in response to the first newly added sequence satisfying a preset fluctuation condition; calculating the fluctuation intensity of the first newly added sequence; performing a negative correlation mapping on the fluctuation intensity using an exponential function; using the product of the mapping result and the first window increment as a second window increment; obtaining a second newly added sequence corresponding to the second window increment; iterating the process of expanding the window and obtaining the second newly added sequence in response to the second newly added sequence satisfying the preset fluctuation condition until the second newly added sequence no longer satisfies the preset fluctuation condition, ending the iteration; and taking the window sequence after the last window expansion as the fluctuation subsequence.

[0006] Preferably, the preset fluctuation condition is: Calculate the voltage deviation between each sampling point in the sampling window and the standard voltage, and take the difference between the maximum and minimum deviation as the window voltage deviation range; Count the number of inflection points of the second derivative of the window sequence of the sampling window; The preset fluctuation condition is that the range of the window voltage deviation is not less than the preset range threshold or the number of inflection points of the second derivative is not zero.

[0007] Preferably, calculating the fluctuation intensity of the first newly added sequence includes: Calculate the voltage deviation between each sampling point of the first newly added sequence and the standard voltage, take the difference between the maximum and minimum deviation as the newly added voltage deviation range, and take the ratio of the newly added voltage deviation range to the preset maximum range as the degree of deviation. The ratio of the number of second derivative inflection points of the first newly added sequence to the sequence length of the first newly added sequence is used as the inflection point density. The normalized value of the sum of the deviation degree and the inflection point density is taken as the fluctuation intensity.

[0008] Preferably, the voltage anomaly monitoring method further includes: In response to the window sequence of the sampling window satisfying a preset stationarity condition, a third newly added sequence corresponding to a preset third window increment is obtained. In response to the third newly added sequence satisfying the preset stationarity condition, the sampling window is expanded with the third window increment. The process of obtaining the third newly added sequence is iterated until the third newly added sequence no longer satisfies the preset stationarity condition. The window sequence after the last window expansion is taken as a stationary subsequence.

[0009] Preferably, the preset stability condition is: Calculate the voltage deviation between each sampling point in the sampling window and the standard voltage, and take the difference between the maximum and minimum deviation as the window voltage deviation range; Count the number of inflection points of the second derivative of the window sequence of the sampling window; The preset stability condition is that the window voltage deviation range is less than the preset range threshold and the number of inflection points of the second derivative is 0.

[0010] The preferred screening process is as follows: In response to the real-time voltage sequence satisfying the preset fluctuation condition, any fluctuation subsequence is taken as the target sequence. The correlation coefficient, voltage inflection point distribution similarity, and voltage deviation range difference between the real-time voltage sequence and the target sequence are calculated. The product of the normalized correlation coefficient, voltage inflection point distribution similarity, and the difference between 1 and the voltage deviation range difference is taken as the comprehensive similarity. The comprehensive similarity between the real-time voltage sequence and each fluctuation subsequence is obtained through iteration. The fluctuation subsequence with the comprehensive similarity greater than the preset similarity threshold is taken as the similar sequence. In response to the real-time voltage sequence satisfying a preset stationarity condition, any stationary subsequence is taken as the target sequence. The correlation coefficient and voltage deviation range difference between the real-time voltage sequence and the target sequence are calculated. The product of the normalized correlation coefficient, 1, and the difference between the voltage deviation range difference is taken as the comprehensive similarity. The comprehensive similarity between the real-time voltage sequence and each stationary subsequence is obtained through iteration. The stationary subsequence with the comprehensive similarity greater than a preset similarity threshold is taken as a similar sequence.

[0011] Preferably, calculating the similarity of the voltage inflection point distribution between the real-time voltage sequence and the target sequence includes: Calculate the second derivative of the target sequence, take the longest subsequence as the standard sequence, pad the end of the target sequence with 0s so that the target sequence after padding is the same length as the standard sequence, divide the target sequence after padding into several segments, and mark each segment according to the number of inflection points of the second derivative. After marking all segments, the inflection point distribution vector of the target sequence is obtained. Construct the inflection point distribution vector of the real-time voltage sequence according to the method for constructing the inflection point distribution vector of the target sequence; Calculate the cosine similarity between the inflection point distribution vector of the real-time voltage sequence and the inflection point distribution vector of the target sequence, and use the normalized value of the cosine similarity as the voltage inflection point distribution similarity.

[0012] Preferably, calculating the voltage deviation range difference between the real-time voltage sequence and the target sequence includes: Calculate the deviation of each voltage in the real-time voltage sequence from the standard voltage, and take the difference between the maximum and minimum deviation as the real-time voltage deviation range. Calculate the target voltage deviation range of the target sequence according to the calculation method of the real-time voltage deviation range. Calculate the absolute difference between the real-time voltage deviation range and the target voltage deviation range, and use the ratio of the absolute difference to the preset maximum range as the voltage deviation range difference.

[0013] Preferably, predicting future voltage sequence features based on all said adjacent sequences includes: The similarity between the real-time voltage sequence and each similar sequence is calculated. For each adjacent sequence, the ratio of the corresponding similarity to the sum of all similarities is used as a weight. The voltage deviation range of each adjacent sequence is calculated, and the product of the voltage deviation range of each adjacent sequence and the corresponding weight is accumulated to obtain the range prediction value. The number of second derivative inflection points of each adjacent sequence is counted, and the product of the number of second derivative inflection points of each adjacent sequence and the corresponding weight is accumulated to obtain the inflection point prediction value. An early warning is triggered if the predicted range and the predicted inflection point meet a preset fluctuation condition; no early warning is triggered if the predicted range and the predicted inflection point meet a preset stability condition.

[0014] Secondly, a power quality monitoring system for a power acquisition terminal is provided. The system includes a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the power quality monitoring method for a power acquisition terminal described above is implemented.

[0015] The present invention has the following effects: 1. This invention introduces a fluctuation intensity index and uses an exponential function to map a negative correlation between the window increment and the fluctuation intensity, thus achieving a negative correlation between the window increment and the fluctuation intensity. This design can accurately match the sequence length requirements of voltage fluctuations of different intensities. For high-intensity fluctuations, the window increment is reduced to avoid information loss, while for low-intensity fluctuations, the window increment is increased to avoid redundant data mixing.

[0016] 2. This invention obtains complete and pure wave subsequences, providing high-quality historical samples for subsequent similar sequence screening and future voltage time series feature prediction, effectively improving the accuracy of voltage time series data analysis.

[0017] 3. This invention uses high-quality fluctuation subsequences for predictive analysis, which can achieve early warning of voltage anomalies, significantly improve the reliability of voltage anomaly monitoring in industrial scenarios, and ensure the stable operation of industrial production equipment. Attached Figure Description

[0018] Figure 1 This is a flowchart of steps S1-S3 in a power quality monitoring method for a power acquisition terminal according to an embodiment of the present invention.

[0019] Figure 2 This is a structural block diagram of a power quality monitoring system for a power acquisition terminal according to an embodiment of the present invention. Detailed Implementation

[0020] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0021] Reference Figure 1 A power quality monitoring method for a power acquisition terminal includes steps S1-S3, as detailed below: Step S1: Divide the acquired historical voltage sequence into several subsequences, including stationary subsequences and fluctuating subsequences.

[0022] It should be noted that the historical voltage series is obtained through electricity meters, and all sub-series are merged into a complete historical voltage series.

[0023] It should be further noted that the electricity meter mentioned in the embodiments of the present invention specifically refers to a smart electricity meter deployed in a specific area of ​​industry (such as a workshop or production line) to realize the time-series analysis and anomaly monitoring of voltage fluctuations within the industrial area.

[0024] The process of obtaining the fluctuation subsequence is as follows: a sampling window is set on the historical voltage sequence. In response to the window sequence of the sampling window satisfying the preset fluctuation condition, the window is gradually expanded based on the continuity of time series. The first newly added sequence corresponding to the preset first window increment is obtained. In response to the first newly added sequence satisfying the preset fluctuation condition, the sampling window is expanded with the first window increment. The fluctuation intensity of the first newly added sequence is calculated. An exponential function is used to perform a negative correlation mapping on the fluctuation intensity. The product of the mapping result and the first window increment is used as the second window increment. The second newly added sequence corresponding to the second window increment is obtained. In response to the second newly added sequence satisfying the preset fluctuation condition, the process of expanding the window and obtaining the second newly added sequence is iterated until the second newly added sequence no longer satisfies the preset fluctuation condition. The time series window iterative expansion is stopped, and the window sequence after the last window expansion is taken as the fluctuation subsequence.

[0025] It should be noted that the first voltage data point of the historical voltage sequence is used as the left boundary of the sampling window. The initial size of the sampling window can be set as needed. The first window increment can also be set as needed. For example, the first window increment is 5 sampling points, and the corresponding first newly added sequence is also 5 sampling points.

[0026] In some embodiments, the preset fluctuation condition is as follows: calculate the deviation between the voltage of each sampling point in the sampling window and the standard voltage, and take the difference between the maximum and minimum deviation as the window voltage deviation range; count the number of second derivative inflection points of the window sequence of the sampling window; and take the condition that the window voltage deviation range is not less than the preset range threshold or the number of second derivative inflection points is not 0 as the preset fluctuation condition.

[0027] It should be noted that the method for determining whether the newly added sequence meets the preset fluctuation conditions in this application is the same as that in the above embodiments.

[0028] It should be noted that there are three cases for satisfying the preset fluctuation conditions: the first case is that the window voltage deviation range is less than the preset range threshold and the number of inflection points of the second derivative is not 0; the second case is that the window voltage deviation range is not less than the preset range threshold and the number of inflection points of the second derivative is 0; and the third case is that the window voltage deviation range is not less than the preset range threshold and the number of inflection points of the second derivative is not 0.

[0029] The standard voltage is specifically the rated line voltage of the power grid in the area corresponding to the meter. The window voltage deviation range is the maximum deviation of the voltage from the rated value, which can reflect the magnitude of voltage fluctuation. The range threshold is set as the allowable deviation range of the standard voltage, that is, the range threshold is the difference between the upper limit of the standard allowable voltage and the lower limit of the standard allowable voltage.

[0030] It should be further explained that the first derivative (i.e., the rate of change of voltage at each moment) of the discrete voltage sequence within the acquisition window is calculated using the central difference method. The second derivative (the rate of change of the rate of change of voltage at each moment) is then obtained by central difference again on the first derivative sequence. The second derivative sequence is then iterated through, retaining points where adjacent points satisfy the "abrupt change in the sign of the second derivative," i.e., inflection points. The number of inflection points directly reflects the number of abrupt changes in the rate of change of voltage; typically, one complete fluctuation corresponds to at least two inflection points. More inflection points indicate a higher frequency of abrupt changes in the rate of change of voltage and a higher fluctuation rate.

[0031] In some embodiments, calculating the fluctuation intensity of the first newly added sequence includes: calculating the deviation between the voltage of each sampling point of the first newly added sequence and the standard voltage; taking the difference between the maximum and minimum deviation values ​​as the newly added voltage deviation range; taking the ratio of the newly added voltage deviation range to the preset maximum range as the deviation degree; taking the ratio of the number of second derivative inflection points of the first newly added sequence to the sequence length of the first newly added sequence as the inflection point density; and taking the normalized value of the sum of the deviation degree and the inflection point density as the fluctuation intensity.

[0032] For example, the formula for calculating the volatility intensity of the first newly added sequence is as follows: ; In the formula, The fluctuation intensity of the first newly added sequence, The newly added voltage deviation range for the first newly added sequence. To preset the maximum range, To indicate the degree of deviation, The number of inflection points of the second derivative of the first newly added sequence. The length of the first newly added sequence. This represents the inflection point density.

[0033] For the above formula It should be added that the fluctuation intensity is positively correlated with the voltage deviation range and the inflection point density of the newly added sequence. This is achieved through calculations... The ratio of 2 to 1 is used to normalize the sum of the deviation degree and the inflection point density to obtain the fluctuation intensity, which ranges from 0 to 1.

[0034] It should be further explained that traditional techniques for determining whether a data series is a fluctuating subsequence often use a single voltage threshold method (e.g., setting a voltage deviation of ±5% from the rated value as the fluctuation threshold; exceeding this threshold indicates a fluctuation). This method is susceptible to random events, such as instantaneous lightning strikes or millisecond-level spikes caused by switching operations. Since the voltage value of such random events may briefly exceed the fluctuation threshold, they are not continuous fluctuations. The single threshold method may misclassify the window containing such random fluctuations as a "fluctuating subsequence," leading to a large number of false samples mixed into the subsequence library, thus affecting the reliability of subsequent predictions. Compared to traditional techniques, this invention constructs two indicators—the voltage deviation range and the number of inflection points of the second derivative—through time-series analysis, which can more accurately reflect the characteristics of historical voltage time series, thereby improving the accuracy of voltage data in time-series analysis.

[0035] For example, the formula for calculating the second window increment is as follows: ; In the formula, For the second window increment, For the first window increment, It is an exponential function. For the mapping result, The fluctuation intensity of the first newly added sequence.

[0036] It should be further explained that an exponential function is used to negatively correlate the fluctuation intensity. The product of the mapping result and the first window increment is used as the second window increment. Therefore, the second window increment is negatively correlated with the fluctuation intensity; that is, the greater the fluctuation intensity, the smaller the second window increment. This invention can dynamically adjust the length of the next adjacent newly added sequence based on the fluctuation intensity of the previous newly added sequence. For example, for high-intensity fluctuations (large fluctuation peak and high fluctuation frequency), this invention reduces the window increment; for low-intensity fluctuations (small fluctuation peak and low fluctuation frequency), this invention increases the window increment. Compared with the traditional fixed window, it can avoid the mixing of irrelevant data. While ensuring the acquisition of complete fluctuation sequences, it can also ensure the purity of the fluctuation sequences. It avoids missing key stages due to excessively small windows and introducing redundant data due to excessively large windows. It provides accurate historical samples for the acquisition of subsequent similar sequences and the prediction of future voltage time series characteristics. Therefore, this invention can improve the timeliness of voltage anomaly monitoring.

[0037] In some embodiments, the voltage anomaly monitoring method further includes: in response to the window sequence of the sampling window satisfying a preset stationarity condition, obtaining a third new sequence corresponding to a preset third window increment; in response to the third new sequence satisfying the preset stationarity condition, expanding the sampling window with the third window increment; iterating the process of obtaining the third new sequence until the third new sequence no longer satisfies the preset stationarity condition; and taking the window sequence after the last window expansion as a stationary subsequence.

[0038] In some embodiments, the preset stationarity condition is as follows: calculate the deviation between the voltage of each sampling point in the sampling window and the standard voltage, and take the difference between the maximum and minimum deviation as the window voltage deviation range; count the number of second derivative inflection points of the window sequence of the sampling window; and take the window voltage deviation range being less than a preset range threshold and the number of second derivative inflection points being 0 as the preset stationarity condition.

[0039] It should be noted that if the window sequence of the sampling window meets the preset stationarity condition, the window is expanded with a fixed window increment until the fixed window increment no longer meets the preset stationarity condition, thus obtaining a stationary subsequence.

[0040] Step S2: Based on time series analysis, filter subsequences to obtain similar sequences of the real-time voltage sequence, and take the next subsequence adjacent to the similar sequence as the neighboring sequence, where one similar sequence corresponds to one neighboring sequence.

[0041] In some embodiments, the screening process is as follows: In response to the real-time voltage sequence meeting a preset fluctuation condition, any fluctuating subsequence is taken as the target sequence. Its temporal similarity to the real-time voltage sequence is mined based on time series analysis. The correlation coefficient and voltage time series characteristics between the real-time voltage sequence and the target sequence are calculated. The voltage time series characteristics include the similarity of voltage inflection point distribution and the difference in voltage deviation range. The product of the normalized correlation coefficient, the similarity of voltage inflection point distribution, and the difference between 1 and the voltage deviation range is used as the comprehensive similarity. The comprehensive similarity between the real-time voltage sequence and each fluctuating subsequence is obtained through iteration. Fluctuating subsequences with a comprehensive similarity greater than a preset similarity threshold are considered similar sequences. In response to the real-time voltage sequence meeting a preset stationarity condition, any stationary subsequence is taken as the target sequence. The correlation coefficient and voltage deviation range difference between the real-time voltage sequence and the target sequence are calculated. The product of the normalized correlation coefficient, and the difference between 1 and the voltage deviation range is used as the comprehensive similarity. The comprehensive similarity between the real-time voltage sequence and each stationary subsequence is obtained through iteration. Stationary subsequences with a comprehensive similarity greater than a preset similarity threshold are considered similar sequences.

[0042] It should be noted that after obtaining the real-time voltage sequence, it is first determined whether the real-time voltage sequence is fluctuating or stationary. If the real-time voltage sequence is fluctuating, when calculating the comprehensive similarity, it is only necessary to calculate the comprehensive similarity between the real-time voltage sequence and each fluctuating subsequence, and it is not necessary to calculate the comprehensive similarity between the real-time voltage sequence and the stationary subsequence, which can reduce the amount of calculation.

[0043] It should be noted that the correlation coefficient between the real-time voltage sequence and the target sequence is the Pearson correlation coefficient, which can reflect the similarity of the original voltage data.

[0044] In some embodiments, calculating the voltage inflection point distribution similarity between the real-time voltage sequence and the target sequence includes: calculating the second derivative of the target sequence; taking the longest subsequence as the standard sequence; padding the end of the target sequence with zeros to make the target sequence after zero padding equal in length to the standard sequence; dividing the target sequence after zero padding into several segments; marking each segment according to the number of inflection points of the second derivative; and obtaining the inflection point distribution vector of the target sequence after marking all segments; constructing the inflection point distribution vector of the real-time voltage sequence according to the method for constructing the inflection point distribution vector of the target sequence; calculating the cosine similarity between the inflection point distribution vector of the real-time voltage sequence and the inflection point distribution vector of the target sequence; and using the normalized value of the cosine similarity as the voltage inflection point distribution similarity, which can be used to quantify the consistency between the two in terms of fluctuation frequency and time series distribution.

[0045] It should be further explained that this application uses the target sequence as an example to describe the zero-padding process. For each subsequence, if the length does not reach the length of the standard sequence, zeros are padded at the end. Therefore, the zero-padding subsequences are of equal length. The zero-padding subsequences are divided into a fixed number of segments. Thus, the dimension of the inflection point distribution vectors corresponding to different subsequences is the same, and the dimension of the inflection point distribution vector of the real-time voltage sequence is also the same. For example, the longest subsequence has a sequence length of 100 sampling points, the target sequence has a sequence length of 80 sampling points, and 20 zeros are padded at the end of the target sequence (the zero value does not affect the calculation of the relevant voltage). The zero-padding target sequence is divided into 10 segments. The number of second derivative inflection points of the 10 segments are 1, 0, 3, 0, 0, 1, 2, 0, 0, 0 respectively. Then the inflection point distribution vector of the target sequence is... Similarly, the real-time voltage sequence is divided into 10 segments, and an inflection point distribution vector is constructed based on the number of inflection points of the second derivative.

[0046] It should be noted that real-time voltage data is collected by sliding the meter through a fixed-length window, and the fixed length is the same as the length of the standard sequence.

[0047] In some embodiments, calculating the voltage deviation range difference between the real-time voltage sequence and the target sequence includes: calculating the deviation of each voltage in the real-time voltage sequence from the standard voltage, taking the difference between the maximum and minimum deviation as the real-time voltage deviation range, calculating the target voltage deviation range of the target sequence according to the calculation method of the real-time voltage deviation range, calculating the absolute difference between the real-time voltage deviation range and the target voltage deviation range, and taking the ratio of the absolute difference to the preset maximum range as the voltage deviation range difference.

[0048] For example, the formula for calculating the voltage deviation range difference is as follows: ; In the formula, This represents the voltage deviation range difference between the real-time voltage sequence and the target sequence. The target voltage deviation range, For real-time voltage deviation range, The absolute difference This is the preset maximum range.

[0049] It should be noted that, in order to distinguish the differences in voltage deviation range calculated when the real-time voltage sequence meets different conditions, separate methods are used. and express Expressions under different conditions. For example, in the following calculation of comprehensive similarity in response to a real-time voltage sequence satisfying a preset fluctuation condition, the following uses... This represents the voltage deviation range difference when the real-time voltage sequence meets a preset fluctuation condition; for example, in the following calculation of comprehensive similarity in response to the real-time voltage sequence meeting a preset stationarity condition, it uses... This represents the voltage deviation range when the real-time voltage sequence meets the preset stability conditions.

[0050] For example, in response to the real-time voltage sequence satisfying a preset fluctuation condition, the formula for calculating the comprehensive similarity is as follows: ; In the formula, To assess the comprehensive similarity between the real-time voltage sequence and the target sequence, This represents the voltage deviation range difference between the real-time voltage sequence and the target sequence. Let cosine similarity be the inflection point distribution vector of the real-time voltage sequence and the inflection point distribution vector of the target sequence. This is the normalized value of the cosine similarity (similarity of voltage inflection point distribution). The Pearson correlation coefficient between the real-time voltage sequence and the target sequence. This is the normalized Pearson correlation coefficient.

[0051] For example, in response to the real-time voltage sequence satisfying a preset stationarity condition, the formula for calculating the comprehensive similarity is as follows: ; In the formula, To assess the comprehensive similarity between the real-time voltage sequence and the target sequence, This represents the voltage deviation range difference between the real-time voltage sequence and the target sequence. The Pearson correlation coefficient between the real-time voltage sequence and the target sequence. This is the normalized Pearson correlation coefficient.

[0052] It should be noted that, compared to the comprehensive similarity corresponding to the fluctuation condition, the number of inflection points is 0 when the real-time voltage sequence meets the preset stationary condition. Therefore, the inflection point distribution vector does not need to be considered when calculating the comprehensive similarity between the real-time voltage sequence and the stationary subsequence.

[0053] It should be noted that the overall similarity is negatively correlated with the difference in voltage deviation range, while the overall similarity is positively correlated with the similarity of voltage inflection point distribution and the Pearson correlation coefficient.

[0054] Step S3: Predict the characteristics of future voltage sequences based on all adjacent sequences to complete the power quality analysis.

[0055] In some embodiments, predicting future voltage time-series characteristics based on all adjacent sequences to further monitor power quality and achieve early warning includes: calculating the similarity between the real-time voltage sequence and each similar sequence; for each adjacent sequence, using the ratio of the corresponding similarity to the sum of all similarities as a weight; calculating the voltage deviation range of each adjacent sequence, accumulating the product of the voltage deviation range of each adjacent sequence and the corresponding weight to obtain the range prediction value; counting the number of second derivative inflection points of each adjacent sequence, accumulating the product of the number of second derivative inflection points of each adjacent sequence and the corresponding weight to obtain the inflection point prediction value; the future voltage time-series characteristics are the range prediction value and the inflection point prediction value; in response to the range prediction value and the inflection point prediction value satisfying a preset fluctuation condition, determining that abnormal voltage fluctuations will occur in the future and triggering an early warning; in response to the range prediction value and the inflection point prediction value satisfying a preset stability condition, determining that the future voltage will remain stable and not triggering an early warning.

[0056] It should be noted that the method for calculating the similarity between the real-time voltage sequence and each similar sequence is the same as the method for calculating the comprehensive similarity.

[0057] It should be added that a similar sequence corresponds to a comprehensive similarity, and a similar sequence corresponds to an adjacent sequence. Therefore, an adjacent sequence corresponds to a comprehensive similarity.

[0058] For example, the formula for calculating the range prediction value is as follows: ; In the formula, This is the range prediction value. The total number of similar sequences. For the first Similarity between similar sequences and real-time voltage sequences, The sum of all similarities, For the first Weights of similar sequences, For the first Voltage deviation range of similar sequences.

[0059] For example, the formula for calculating the inflection point prediction value is as follows: ; In the formula, This is the predicted value for the inflection point. The total number of similar sequences. For the first Similarity between similar sequences and real-time voltage sequences, The sum of all similarities, For the first Weights of similar sequences, For the first The number of inflection points of the second derivative of similar sequences. This is a rounding function.

[0060] It should be noted that the thresholds involved in this application can be set as needed.

[0061] The present invention also provides a power quality monitoring system for a power acquisition terminal. The system includes a processor and a memory, the memory storing computer program instructions. When the processor executes the computer program instructions, it implements a power quality monitoring method for a power acquisition terminal according to the first aspect of the present invention. The system also includes other components well-known to those skilled in the art, such as a communication bus and a communication interface; their configuration and functions are known in the art and therefore will not be described further here.

[0062] It should be noted that the preferred embodiments of this application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of this application. For those skilled in the art, various modifications and improvements can be made without departing from the concept of the invention, and these all fall within the protection scope of the invention. Therefore, the protection scope of this patent should be determined by the appended claims.

Claims

1. A method for monitoring power quality using a power acquisition terminal, characterized in that, include: The acquired historical voltage sequence is divided into several subsequences, including stationary subsequences and fluctuating subsequences; The subsequences are filtered based on time series analysis to obtain similar sequences to the real-time voltage sequence. The next subsequence adjacent to the similar sequence is taken as the adjacent sequence, wherein one similar sequence corresponds to one adjacent sequence. Based on all the adjacent sequences, the characteristics of the future voltage sequence are predicted to complete the power quality analysis; The process of obtaining the fluctuation subsequence is as follows: a sampling window is set on the historical voltage sequence. In response to the window sequence of the sampling window satisfying the preset fluctuation condition, a first new sequence corresponding to the preset first window increment is obtained. In response to the first new sequence satisfying the preset fluctuation condition, the sampling window is expanded with the first window increment. The fluctuation intensity of the first new sequence is calculated. An exponential function is used to perform a negative correlation mapping on the fluctuation intensity. The product of the mapping result and the first window increment is used as the second window increment. A second new sequence corresponding to the second window increment is obtained. In response to the second new sequence satisfying the preset fluctuation condition, the process of expanding the window and obtaining the second new sequence is iterated until the second new sequence no longer satisfies the preset fluctuation condition. The iteration ends, and the window sequence after the last window expansion is taken as the fluctuation subsequence.

2. The power quality monitoring method for a power acquisition terminal according to claim 1, characterized in that, The preset fluctuation condition is: Calculate the voltage deviation between each sampling point in the sampling window and the standard voltage, and take the difference between the maximum and minimum deviation as the window voltage deviation range; Count the number of inflection points of the second derivative of the window sequence of the sampling window; The preset fluctuation condition is that the range of the window voltage deviation is not less than the preset range threshold or the number of inflection points of the second derivative is not zero.

3. The power quality monitoring method for a power acquisition terminal according to claim 1, characterized in that, Calculating the volatility intensity of the first newly added sequence includes: Calculate the voltage deviation between each sampling point of the first newly added sequence and the standard voltage, take the difference between the maximum and minimum deviation as the newly added voltage deviation range, and take the ratio of the newly added voltage deviation range to the preset maximum range as the degree of deviation. The ratio of the number of second derivative inflection points of the first newly added sequence to the sequence length of the first newly added sequence is used as the inflection point density. The normalized value of the sum of the deviation degree and the inflection point density is taken as the fluctuation intensity.

4. The power quality monitoring method for a power acquisition terminal according to claim 1, characterized in that, Also includes: In response to the window sequence of the sampling window satisfying a preset stationarity condition, a third newly added sequence corresponding to a preset third window increment is obtained. In response to the third newly added sequence satisfying the preset stationarity condition, the sampling window is expanded with the third window increment. The process of obtaining the third newly added sequence is iterated until the third newly added sequence no longer satisfies the preset stationarity condition. The window sequence after the last window expansion is taken as a stationary subsequence.

5. The power quality monitoring method for a power acquisition terminal according to claim 4, characterized in that, The preset stability condition is: Calculate the voltage deviation between each sampling point in the sampling window and the standard voltage, and take the difference between the maximum and minimum deviation as the window voltage deviation range; Count the number of inflection points of the second derivative of the window sequence of the sampling window; The preset stability condition is that the window voltage deviation range is less than the preset range threshold and the number of inflection points of the second derivative is 0.

6. The power quality monitoring method for a power acquisition terminal according to claim 1, characterized in that, The screening process is as follows: In response to the real-time voltage sequence satisfying the preset fluctuation condition, any fluctuation subsequence is taken as the target sequence. The correlation coefficient, voltage inflection point distribution similarity, and voltage deviation range difference between the real-time voltage sequence and the target sequence are calculated. The product of the normalized correlation coefficient, voltage inflection point distribution similarity, and the difference between 1 and the voltage deviation range difference is taken as the comprehensive similarity. The comprehensive similarity between the real-time voltage sequence and each fluctuation subsequence is obtained through iteration. The fluctuation subsequence with the comprehensive similarity greater than the preset similarity threshold is taken as the similar sequence. In response to the real-time voltage sequence satisfying a preset stationarity condition, any stationary subsequence is taken as the target sequence. The correlation coefficient and voltage deviation range difference between the real-time voltage sequence and the target sequence are calculated. The product of the normalized correlation coefficient, 1, and the difference between the voltage deviation range difference is taken as the comprehensive similarity. The comprehensive similarity between the real-time voltage sequence and each stationary subsequence is obtained through iteration. The stationary subsequence with the comprehensive similarity greater than a preset similarity threshold is taken as a similar sequence.

7. The power quality monitoring method for a power acquisition terminal according to claim 6, characterized in that, Calculating the similarity of the voltage inflection point distribution between the real-time voltage sequence and the target sequence includes: Calculate the second derivative of the target sequence, take the longest subsequence as the standard sequence, pad the end of the target sequence with 0s so that the target sequence after padding is the same length as the standard sequence, divide the target sequence after padding into several segments, and mark each segment according to the number of inflection points of the second derivative. After marking all segments, the inflection point distribution vector of the target sequence is obtained. Construct the inflection point distribution vector of the real-time voltage sequence according to the method for constructing the inflection point distribution vector of the target sequence; Calculate the cosine similarity between the inflection point distribution vector of the real-time voltage sequence and the inflection point distribution vector of the target sequence, and use the normalized value of the cosine similarity as the voltage inflection point distribution similarity.

8. The power quality monitoring method for a power acquisition terminal according to claim 6, characterized in that, Calculating the voltage deviation range difference between the real-time voltage sequence and the target sequence includes: Calculate the deviation of each voltage in the real-time voltage sequence from the standard voltage, and take the difference between the maximum and minimum deviation as the real-time voltage deviation range. Calculate the target voltage deviation range of the target sequence according to the calculation method of the real-time voltage deviation range. Calculate the absolute difference between the real-time voltage deviation range and the target voltage deviation range, and use the ratio of the absolute difference to the preset maximum range as the voltage deviation range difference.

9. The power quality monitoring method for a power acquisition terminal according to claim 1, characterized in that, Predicting future voltage sequence features based on all the adjacent sequences includes: The similarity between the real-time voltage sequence and each similar sequence is calculated. For each adjacent sequence, the ratio of the corresponding similarity to the sum of all similarities is used as a weight. The voltage deviation range of each adjacent sequence is calculated, and the product of the voltage deviation range of each adjacent sequence and the corresponding weight is accumulated to obtain the range prediction value. The number of second derivative inflection points of each adjacent sequence is counted, and the product of the number of second derivative inflection points of each adjacent sequence and the corresponding weight is accumulated to obtain the inflection point prediction value. An early warning is triggered if the predicted range and the predicted inflection point meet a preset fluctuation condition; no early warning is triggered if the predicted range and the predicted inflection point meet a preset stability condition.

10. A power quality monitoring system for a power acquisition terminal, characterized in that, include: A processor and a memory, wherein the memory stores computer program instructions that, when executed by the processor, implement a power quality monitoring method for a power acquisition terminal according to any one of claims 1-9.

Images