Non-periodic voltage fluctuation determination method, system and device for variable scale search, and storage medium

Abstract

The invention discloses an aperiodic voltage fluctuation determination method, system and device for variable scale search, and a storage medium, and the method comprises the steps: achieving the ordered analysis of voltage data in a time domain through a dynamic sliding window, and providing a time sequence basis for fluctuation feature capture; aiming at the data in the window, a partition searching mode is adopted, the characteristics that the voltage fluctuation frequency mode is not fixed and the amplitude mode is random are adaptively adapted, and different fluctuation forms are accurately matched; and in combination with a fluctuation judgment threshold, analyzing a data change trend in a partition, determining an extreme value judgment range, and effectively avoiding the problem of local optimum or extreme value omission caused by unreasonable extreme value search. According to the invention, the capability of capturing rapid voltage fluctuation is remarkably enhanced, an accurate voltage fluctuation monitoring means is provided for power grids and other scenes with dense loads and large new energy installation capacity, power grid voltage stability control can be effectively supported, the production loss of precision manufacturing, semiconductors and other sensitive industries is reduced, and the method is suitable for popularization and application. And the power grid safety and the high-quality power utilization demand under the high-proportion access of the new energy are ensured.

Description

Technical Field

[0001] This invention relates to the field of voltage fluctuation identification, and in particular to a method, system, device, and storage medium for determining non-periodic voltage fluctuations using variable-scale searching. Background Technology

[0002] Currently, the hazards of voltage fluctuations permeate the entire power generation, transmission, and distribution chain. At the equipment level, failures caused by overheating of the iron core in asynchronous motors account for 42%, and their lifespan is generally shortened by more than 30%. In the precision manufacturing field, fluctuations of ±1% can cause wafer scrapping and precision instrument calibration failure. At the system level, the active and reactive power imbalances caused by voltage fluctuations have become a major threat to the stable operation of the power grid. With the significant increase in the installed capacity of new energy sources, the intermittent output of wind power and photovoltaics has increased the frequency of voltage fluctuations by 5-8 times compared to ten years ago. The broadband fluctuations of wind power bases, the seasonal fluctuations of hydropower, and the random fluctuations of distributed power sources have a superimposed effect, leading to a sharp increase in the difficulty of power grid regulation. Currently, the FFT analysis method suffers from spectral leakage problems near the 50Hz fundamental frequency, resulting in an identification error of over 15% for broadband fluctuations. In tests, the capture rate for fluctuations in the 2-50Hz range is less than 60%. The PRONY algorithm, due to its reliance on a preset order, has a failure probability of 40% in scenarios with random amplitude fluctuations. The "extreme value search" method has a false negative rate of up to 32% in white noise interference environments. A pilot project showed that its local optima problem caused a fluctuation warning delay of more than 0.5 seconds, missing the best control opportunity. Summary of the Invention

[0003] Purpose of the invention: The purpose of this invention is to provide a method for determining non-periodic voltage fluctuations based on dynamic sliding window, partition search, and variable scale determination; another purpose of this invention is to provide a system, device, and storage medium for determining non-periodic voltage fluctuations using scale search.

[0004] Technical solution: The method for determining non-periodic voltage fluctuations through variable-scale searching as described in this invention includes...

[0005] The voltage data read by the FIFO sliding window is divided into several sub-window datasets;

[0006] Based on the maximum, minimum, and maximum values ​​within a sub-window, count the voltage fluctuations in that sub-window.

[0007] If the voltage fluctuation count variable in any sub-window is greater than the fluctuation alarm threshold, an alarm is triggered; otherwise, no alarm is triggered.

[0008] Furthermore, the dataset can be divided into several sub-windows using methods such as equal partitioning, variance adaptive partitioning, or extreme value distance clustering.

[0009] Furthermore, based on the maximum, minimum, and maximum values ​​within the sub-window, fluctuation counts are performed on the sub-window containing those values, as detailed below:

[0010] For the i-th sub-window dataset, determine the maximum value and its position. Start searching from the maximum value point to find the first voltage fluctuation judgment point smaller than the maximum value. Calculate the position index difference between this point and the maximum value. Divide the position index difference by the radius scaling factor to obtain the first local extreme value judgment radius.

[0011] Starting from the first voltage fluctuation judgment point, and using the first local extreme value judgment radius as the search radius, the minimum value is found; when the difference between the maximum value and the minimum value is greater than or equal to the voltage fluctuation judgment threshold, the voltage fluctuation count variable of the i-th sub-window is incremented by 1; when the difference between the maximum value and the minimum value is less than the voltage fluctuation judgment threshold, the voltage fluctuation count variable of the i-th sub-window remains unchanged.

[0012] Starting from the minimum value, find the second voltage fluctuation judgment point that is larger than the minimum value, calculate the position index difference between the point and the minimum value, and divide the position index difference by the radius scaling factor to obtain the second local extreme value judgment radius.

[0013] Starting from the minimum value, the second local extremum determination radius is used as the search radius to find the maximum value; when the difference between the maximum and the minimum value is greater than or equal to the voltage fluctuation determination threshold, the voltage fluctuation count variable of the i-th sub-window is incremented by 1; when the difference between the maximum and the minimum value is less than the voltage fluctuation determination threshold, the voltage fluctuation count variable of the i-th sub-window remains unchanged.

[0014] Starting from the maximum value, begin the search to find the third voltage fluctuation judgment point smaller than the maximum value. Calculate the position index difference between this point and the maximum value, and divide the position index difference by the radius scaling factor to obtain the third local extreme value judgment radius.

[0015] Starting from the third voltage fluctuation judgment point and using the third local extreme value judgment radius as the search radius, find the next minimum value; when the difference between the maximum value and the next minimum value is greater than or equal to the voltage fluctuation judgment threshold, the voltage fluctuation count variable of the i-th sub-window is incremented by 1; when the difference between the maximum value and the next minimum value is less than the voltage fluctuation judgment threshold, the voltage fluctuation count variable of the i-th sub-window remains unchanged.

[0016] Search for the next maximum value from the next minimum value, and count voltage fluctuations, as above; until the beginning and end of the i-th sub-window are found.

[0017] Furthermore, the radius R for determining the first local extremum is:

[0018]

[0019]

[0020] Where R3 is the radius after radius robustness verification, H is the entropy of the historical extreme value distribution within the sub-sliding window, and p k Let R² be the probability of the distance between the k-th type of extrema, R² be the radius of the extremum density constraint, and L be the value of the distance between the k-th type of extrema. i Let R1 be the length of the i-th sub-window, R1 be the corrected radius, ρ be the extremum density, and E be the value of the extremum density. i Let IQR(X) be the set of potential extreme points within the i-th sub-window, R0 be the initial radius benchmark, γ be the fluctuation intensity index, and X be the value of IQR(X). i ) represents the interquartile range of the data in the i-th sub-window. Let λ be the mean of the i-th sub-window, and λ be the length ratio coefficient.

[0021] The variable-scale search non-periodic voltage fluctuation determination system of the present invention includes:

[0022] The data segmentation module is used to divide the voltage data read by the FIFO sliding window into several sub-window datasets;

[0023] The voltage fluctuation counting module is used to count the voltage fluctuations in a sub-window based on the maximum, minimum, and maximum values ​​within that sub-window.

[0024] The early warning module is used to trigger an alarm if the voltage fluctuation count variable in any sub-window is greater than the fluctuation alarm threshold; otherwise, no alarm is triggered.

[0025] Furthermore, in the data segmentation module, the data is divided into several sub-window datasets using methods such as equal division, variance adaptive segmentation, or extreme value distance clustering.

[0026] Furthermore, in the voltage fluctuation counting module, fluctuations are counted within a sub-window based on its maximum, minimum, and maximum values, as detailed below:

[0027] For the i-th sub-window dataset, determine the maximum value and its position. Start searching from the maximum value point to find the first voltage fluctuation judgment point smaller than the maximum value. Calculate the position index difference between this point and the maximum value. Divide the position index difference by the radius scaling factor to obtain the first local extreme value judgment radius.

[0028] Starting from the first voltage fluctuation judgment point, and using the first local extreme value judgment radius as the search radius, the minimum value is found; when the difference between the maximum value and the minimum value is greater than or equal to the voltage fluctuation judgment threshold, the voltage fluctuation count variable of the i-th sub-window is incremented by 1; when the difference between the maximum value and the minimum value is less than the voltage fluctuation judgment threshold, the voltage fluctuation count variable of the i-th sub-window remains unchanged.

[0029] Starting from the minimum value, find the second voltage fluctuation judgment point that is larger than the minimum value, calculate the position index difference between the point and the minimum value, and divide the position index difference by the radius scaling factor to obtain the second local extreme value judgment radius.

[0030] Starting from the minimum value, the second local extremum determination radius is used as the search radius to find the maximum value; when the difference between the maximum and the minimum value is greater than or equal to the voltage fluctuation determination threshold, the voltage fluctuation count variable of the i-th sub-window is incremented by 1; when the difference between the maximum and the minimum value is less than the voltage fluctuation determination threshold, the voltage fluctuation count variable of the i-th sub-window remains unchanged.

[0031] Starting from the maximum value, begin the search to find the third voltage fluctuation judgment point smaller than the maximum value. Calculate the position index difference between this point and the maximum value, and divide the position index difference by the radius scaling factor to obtain the third local extreme value judgment radius.

[0032] Starting from the third voltage fluctuation judgment point and using the third local extreme value judgment radius as the search radius, find the next minimum value; when the difference between the maximum value and the next minimum value is greater than or equal to the voltage fluctuation judgment threshold, the voltage fluctuation count variable of the i-th sub-window is incremented by 1; when the difference between the maximum value and the next minimum value is less than the voltage fluctuation judgment threshold, the voltage fluctuation count variable of the i-th sub-window remains unchanged.

[0033] Search for the next maximum value from the next minimum value, and count voltage fluctuations, as above; until the beginning and end of the i-th sub-window are found.

[0034] Furthermore, the radius R for determining the first local extremum is:

[0035]

[0036]

[0037]

[0038]

[0039] Where R3 is the radius after radius robustness verification, H is the entropy of the historical extreme value distribution within the sub-sliding window, and p k Let R² be the probability of the distance between the k-th type of extrema, R² be the radius of the extremum density constraint, and L be the value of the distance between the k-th type of extrema. i Let R1 be the length of the i-th sub-window, R1 be the corrected radius, ρ be the extremum density, and E be the value of the extremum density. i Let IQR(X) be the set of potential extreme points within the i-th sub-window, R0 be the initial radius benchmark, γ be the fluctuation intensity index, and X be the value of IQR(X). i ) represents the interquartile range of the data in the i-th sub-window. Let λ be the mean of the i-th sub-window, and λ be the length ratio coefficient.

[0040] The computer device of the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above method.

[0041] The computer-readable storage medium of the present invention stores a computer program thereon, which, when executed by a processor, implements the steps of the above-described method.

[0042] Beneficial Effects: Compared with the prior art, the significant advantages of this invention are: 1. This invention adopts a dynamic sliding window mechanism to achieve ordered analysis of voltage data in the time domain, providing a temporal basis for capturing fluctuation characteristics; for the data within the window, a partitioned search method is adopted to adaptively adapt to the characteristics of voltage fluctuation frequency modes being non-fixed and amplitude modes being random, accurately matching different fluctuation patterns; combined with the fluctuation judgment threshold, the trend of data change within the partition is analyzed, the extreme value judgment range is clarified, and the problem of local optima or extreme value omission caused by unreasonable extreme value search schemes is effectively avoided; finally, the partitioned search results are integrated to form a comprehensive judgment of voltage fluctuations within the entire analysis window; 2. This invention overcomes the traditional This invention addresses the shortcomings of amplitude-frequency analysis methods such as FFT and PRONY, as well as morphological analyses such as "climbing" and "extreme search," including issues like spectral leakage, insufficient expected order, local optima, and gradient search failures. It significantly improves the accuracy and reliability of monitoring, particularly for voltage fluctuation scenarios exhibiting white noise characteristics at the 50Hz data level. Furthermore, this invention significantly enhances the ability to capture rapid voltage fluctuations, providing precise voltage fluctuation monitoring methods for scenarios with dense loads and large installed capacity of new energy sources, such as power grids. This effectively supports stable voltage control of the power grid, reduces production losses in sensitive industries such as precision manufacturing and semiconductors, and ensures power grid safety and high-quality electricity demand under high-proportion new energy access. Attached Figure Description

[0043] Figure 1 This is a flowchart of the method of the present invention;

[0044] Figure 2 This is a schematic diagram for determining voltage fluctuations. Detailed Implementation

[0045] like Figure 1 As shown, the variable-scale search method for determining non-periodic voltage fluctuations according to the present invention includes...

[0046] S1: Construct a FIFO (First In First Out) sliding window function, analyze the data length (calculated based on the sampling rate, with a maximum length not exceeding the decision window length), and pass the read data to the sub-window segmentation module; where the decision window length is preset to L, and the number of sub-windows is n (n≥1).

[0047] S1.1: Read the target path data, calculate the data length based on the sampling rate, and determine the valid data sequence within the decision window as Equation 1:

[0048] X = [x1, x2, ..., x m (1)

[0049] (m is the number of data points in the decision window, m ≤ the number of points corresponding to the length of the decision window);

[0050] S1.2: Set the sub-window splitting mode selection mechanism, supporting three scientific splitting modes (equal division is enabled by default, which can be switched through configuration). The specific splitting steps are as follows:

[0051] S1.2.1 Equal Division Mode: Assuming the number of child windows is n≥1, the theoretical length of each child window is calculated as Equation 2:

[0052]

[0053] ( (Rounded down), the remaining number of data points conforms to equation 3:

[0054] r=mn×L0(3)

[0055] The length of the first nr sub-windows is L0, and the length of the last r sub-windows is L0+1. The dividing point conforms to the following relationship (4):

[0056] {k×L0+min(k,r)|k=1,2,...,n-1}(4)

[0057] The above can be used to divide X into n subsequences X1, X2, ..., Xn. n .

[0058] S1.2.2 Variance-Adaptive Segmentation Mode:

[0059] ① Calculate the global variance of X using Formula 5:

[0060]

[0061] Set the variance sensitivity threshold α, α∈(0,1), default 0.3, initial number of segments n0=2;

[0062] ② For the current number of segments n, calculate the local variance of each candidate sub-window according to Formula 6:

[0063]

[0064] l i Let i be the length of the i-th sub-window. Let η be the average of the i-th sub-window, and calculate the percentage of sub-windows that satisfy Equation 7.

[0065]

[0066] ③ If the proportion of the sub-window η < 0.8, increase n (n = n + 1) and repeat ② until η > 0.8 or n reaches the maximum limit n ≤ 10;

[0067] ④ Determine the optimal split point using dynamic programming: With the objective of minimizing the sum of variances of each sub-window, solve for the set of split points k1, k2, ..., k. n-1 The condition k1 < k2 < ... < k n-1 And X i =X[k i-1 +1:k i ], where k0=0, k n =m.

[0068] S1.2.3 Extreme Distance Clustering Segmentation Pattern:

[0069] ① Extract the set of potential extreme points of X, E = {e1, e2, ..., e} p The 3σ criterion is used to identify points that significantly deviate from the mean. Let the dataset be x1, x2, ..., x... n Calculate its mean μ and standard deviation σ: the calculation method satisfies equation (8).

[0070]

[0071] The sample standard deviation is calculated by using n-1 in the denominator to correct for bias.

[0072] The 3σ criterion states that if a certain data x i Satisfy |x i If -μ|>3σ, then x is determined to be... i This is an outlier.

[0073] According to Equation 9, calculate the distance between any two points:

[0074] d ij =|e i -e j | (i,j=1,2,...,p) (9)

[0075] ② with d ij Using the characteristic as a basis, the K-means clustering algorithm is used to divide E into n classes, where n is determined by the elbow method, and the cluster centers are c1, c2, ..., c3. n The objective function is to minimize the sum of squared intra-class distances, following Equation 10:

[0076]

[0077] Where c kLet be the set of samples of the kth class.

[0078] The core idea of ​​the elbow method is to analyze the relationship between the number of clusters (k) and the clustering error (SSE), and find the point where the error decreasing trend clearly reverses, i.e., the "elbow," to determine the most reasonable value of k. The SSE calculation follows Equation 11:

[0079]

[0080] Where: k is the number of clusters; C i Let x represent the i-th cluster; x is the cluster C. i The sample in; μ i Cluster C i The center (mean vector); ||x-μ i || 2 For sample x and cluster center μ i The square of the Euclidean distance.

[0081] ③ Calculate the position distribution of each type of extreme point in X, take the median of each position interval as the segmentation reference point, smooth the segmentation points through linear interpolation, and obtain the final set of segmentation points to ensure that each sub-window contains at least one extreme interval corresponding to a cluster center.

[0082] S2: The sub-window segmentation module divides the data in the judgment window into n preset sub-windows, extracts the array data of the first to the nth sub-windows in sequence, and passes them to the maximum value finding module of the corresponding sub-window.

[0083] S3: For each sub-window (i = 1, 2, ..., n):

[0084] S3.1: The maximum value finding module for the i-th sub-window processes the sub-window array, identifies the maximum value in the array and its position, and then finds the maximum value within each sub-window.

[0085] S3.2: After receiving the maximum value information of the sub-window, the "Right Search Module" searches for values ​​smaller than the maximum value.

[0086] For the point DEFAULT_SET_V (voltage fluctuation judgment threshold), calculate the difference in index between that point and the maximum value. Divide the index difference by DEFAULT_DIV (radius scaling factor) to obtain the radius R (local extreme value judgment range). Then, pass R and the index of that point to the "right minimum value search module".

[0087] S3.2.1 Calculation of initial radius reference value:

[0088] The initial radius reference R0 is determined based on the characteristics of the sub-window data, and the calculation formula is Equation 12:

[0089]

[0090] Among them, L i λ represents the current sub-window length, and λ is the length percentage coefficient (default 5%, which can be dynamically adjusted based on data fluctuation intensity). To round down, ensure R0≥3 to avoid local optima caused by excessively small radii.

[0091] S3.2.2 Adaptive correction of fluctuation intensity:

[0092] The fluctuation intensity index γ of the sub-window data is calculated according to Equation 13:

[0093]

[0094] Among them, IQR(X) i ) represents the interquartile range of the sub-window data (the difference between the upper quartile Q3 and the lower quartile Q1). This is the mean of the sub-window. R0 is corrected according to γ, following Equation 14:

[0095]

[0096] S3.2.3 Extreme Density Constraint Adjustment:

[0097] Extract the set of potential extreme points E within the sub-window i (Using the 3σ criterion), calculate the extreme value density ρ using Equation 15:

[0098]

[0099] Where E i This represents the number of extreme points.

[0100] When ρ > 0.2 (high extreme density), a constraint factor β is introduced to compress the radius, for example, β = 0.3, resulting in Equation 16:

[0101] R2=R1·(1-0.3·min(ρ - 0.2,0.5))(16)

[0102] To avoid duplicate judgments caused by dense extreme values; when ρ < 0.05 (low extreme value density), to prevent omissions, β = 0.5, and the expansion radius follows Equation 17:

[0103] R2=R1·(1+0.5·(0.05-ρ))(17)S3.2.4 Robustness verification of final radius:

[0104] Apply boundary constraints to the corrected radius R2 to ensure that R2 conforms to Equation 18:

[0105]

[0106] in The maximum radius is limited (not exceeding 1 / 5 of the sub-window length), and the lower bound of 3 guarantees the minimum search range. Simultaneously, secondary verification is performed using the entropy H of the historical extreme value distribution within the sliding window, following Equation 19:

[0107]

[0108] p k H represents the probability of the interval between extreme values ​​of the kth class. If H > 1.5 (distribution is chaotic), then the trigger radius is fine-tuned to R = R·1.2 to enhance robustness.

[0109] S3.2.5 Radius Transfer and Update Mechanism: The final determined radius R and target point index are transferred to the "Right Minimum Search Module," while the current radius calculation parameters are recorded as the initial prior values ​​for the next round of search (S3.5). These parameters are dynamically updated using exponential smoothing, satisfying Equation 20:

[0110] R next =0.7·R current +0.3·R prev (20)R prev This improves the efficiency and stability of continuous searching by extending the effective radius of the previous round.

[0111] The aforementioned radius calculation method, through multi-dimensional data feature fusion (length, fluctuation intensity, extreme value density, distribution entropy), effectively suppresses local optima problems while ensuring the completeness of the extreme value search, and controls the computational complexity to O(L). i (Level) , balancing mathematical rigor with engineering practicality.

[0112] S3.3: The "Right Minimum Search Module" starts searching to the right from the position (inclusive) passed in from S3.2 (limited to the i-th sub-window) to find the minimum point G that satisfies "smaller than R points to the left and R points to the right"; calculates the difference between the maximum value and G, and if the difference is ≥ DEFAULT_SET_V, then the count_i (fluctuation count variable of the i-th sub-window) of the sub-window is incremented by 1, and the position and value of G are passed back to the "Right Search Module".

[0113] S3.4: The "Right Search Module" receives the minimum value information transmitted in S3.3, searches to the right for a point that is greater than DEFAULT_SET_V, calculates the difference between the index of the point and the minimum value, divides it by DEFAULT_DIV to obtain the radius R, and transmits R and the index of the point to the "Right Maximum Search Module".

[0114] S3.5: The "Right-to-Maximum Search Module" starts searching to the right from the position (inclusive) passed in from S3.4 (limited to the i-th sub-window) to find the maximum point G that satisfies "larger than R points to the left and R points to the right"; calculate the difference between G and the minimum value, and if the difference is ≥ DEFAULT_SET_V, then count_i is incremented by 1, and the position and value of G are passed back to the "Right-to-Maximum Search Module".

[0115] S3.6: The "Right Search Module" receives the maximum value information passed from S3.5, repeats the operation of S3.2 (finds the point that is smaller than the maximum value DEFAULT_SET_V, calculates R and passes it to the "Right Minimum Search Module"), forming a loop until the end of the i-th sub-window is searched.

[0116] S3.7: The "left search module" of the i-th sub-window adopts the logic symmetrical to S3.2-S3.6, and searches to the left from the position of the maximum value of the sub-window (limited to the i-th sub-window) until the beginning of the i-th sub-window is reached, and count_i is updated in the process.

[0117] S4: After all sub-windows (i = 1, 2, ..., n) have been processed, summarize the count_i of each sub-window. If any count_i > DEFAULT_SET_N (fluctuation alarm threshold), then trigger an alarm.

[0118] S5: Update data using FIFO method: Read the new sliding length (preset sliding window update time length) data, slide out the old sliding length data, and return to S2 to repeat the subsequent steps.

[0119] Real-time acquisition of line measurement data on the 500kV side of the transformer using a wide-area measurement system. Figure 2 The figure shows the waveform of voltage data on the 500kV side of a transformer at a power plant on July 5, 2024, when line oscillation occurred. As can be seen from the figure, the voltage signal contains noise, and the waveform exhibits complex characteristics of non-stationarity and multi-mode coupling, making it difficult to determine the number of fluctuations through extreme points.

[0120] Step 2: Set the threshold parameters of the voltage fluctuation monitoring algorithm, including the fluctuation amplitude threshold set_Amp, the fluctuation number threshold set_N, the monitoring period window_Size, the extreme value search radius divisor set_Div, and the data frame number sample_Rate;

[0121] Table 1 Parameter Settings for Voltage Fluctuation Judgment Method

[0122]

[0123] Step 3: Using the 20 sub-windows obtained in Step 2, sequentially search for the maximum value A in each sub-window as the anchor point and the starting index point for the search. Search left and right for other extreme points exceeding the threshold. First, start searching for the minimum value to the left. The first step is to find point B, whose amplitude exceeds the threshold set_Amp with respect to point A. The minimum value determination range radiusMin is obtained by using the index difference between points A and B (idxA-idxB) / set_Div. Starting from point B, determine if the original voltage data Ui is less than or equal to all other voltage values ​​within the radius; if so, this point Ui is a minimum value. If the difference between the minimum and maximum values ​​exceeds the threshold, increment the fluctuation count by 1. Then, starting from the index of the minimum value, continue searching left for point C, whose amplitude difference exceeds the threshold. Use the index difference between Ui and point C (idxUi-idxC) / set_Div to determine the maximum value determination range radiusMax. Starting from point C, determine if the original voltage data Ui is greater than or equal to all other voltage values ​​within the radius; if so, this point Ui is a maximum value. If the difference between the maximum and minimum values ​​exceeds the threshold, the fluctuation count is incremented by 1. This search continues until no point exceeding the amplitude threshold is found to the left, or the last point has been reached, at which point the search ends.

[0124] Step 4: Using the 20 sub-windows obtained in Step 2, sequentially search for the maximum value A in each sub-window as the anchor point and the starting index point for the search. Search for the remaining extreme points that exceed the threshold to the left and right respectively. After the leftward search in Step 3 is completed, start searching for the minimum and maximum values ​​alternately to the right. The method is the same as the leftward search logic, which is to alternately search for the minimum and maximum values ​​in the voltage data of the monitoring period of 100s.

[0125] Step 5: Using the anchor points determined in each sub-window in Step 3, search for extreme points to the left and right sequentially. The total number of fluctuations is the sum of the fluctuations on the left and the fluctuations on the right. This case has a total of 20 sub-windows. An alarm signal is issued if the extreme point search count of any sub-window meets the counting threshold.

[0126] The extreme point statistics of the analysis case are shown in Table 2. In the search of the 16th sub-window, 6 fluctuations exceeding the threshold of 500V were found. The total length of the alarm phase data is 45 seconds, or 2250 points. Due to the FIFO principle, in the 100-second storage space, the last 2250 points, i.e., the first 2750 points, are 0 data. The statistical results in Table 2 are obtained after indexing and transformation calculation.

[0127] Table 2. Statistics of Extreme Points in Voltage Fluctuation Cases

[0128]

[0129] Within the 16th sub-window, the maximum value of 1195 is found as the starting search point. The search continues to the left, but no minimum value is found. The difference between the minimum value and 1195 exceeds 500V, so the search on the left ends. The search continues to the right, and a minimum value of 1797 is found. The difference between the minimum value and 1195 exceeds 500V, which is considered a valid fluctuation. The search continues to the right, alternating between valid maximum and minimum values, and the number of fluctuations is accumulated. When the number of fluctuations reaches the threshold of 6 at point 2210, a voltage fluctuation alarm is issued.

[0130] The variable-scale search non-periodic voltage fluctuation determination system of the present invention includes:

[0131] The data segmentation module is used to divide the voltage data read by the FIFO sliding window into several sub-window datasets;

[0132] The voltage fluctuation counting module is used to count the voltage fluctuations in a sub-window based on the maximum, minimum, and maximum values ​​within that sub-window.

[0133] The early warning module is used to trigger an alarm if the voltage fluctuation count variable in any sub-window is greater than the fluctuation alarm threshold; otherwise, no alarm is triggered.

[0134] Furthermore, in the data segmentation module, the data is divided into several sub-window datasets using methods such as equal division, variance adaptive segmentation, or extreme value distance clustering.

[0135] Furthermore, in the voltage fluctuation counting module, fluctuations are counted within a sub-window based on its maximum, minimum, and maximum values, as detailed below:

[0136] For the i-th sub-window dataset, determine the maximum value and its position. Start searching from the maximum value point to find the first voltage fluctuation judgment point smaller than the maximum value. Calculate the position index difference between this point and the maximum value. Divide the position index difference by the radius scaling factor to obtain the first local extreme value judgment radius.

[0137] Starting from the first voltage fluctuation judgment point, and using the first local extreme value judgment radius as the search radius, the minimum value is found; when the difference between the maximum value and the minimum value is greater than or equal to the voltage fluctuation judgment threshold, the voltage fluctuation count variable of the i-th sub-window is incremented by 1; when the difference between the maximum value and the minimum value is less than the voltage fluctuation judgment threshold, the voltage fluctuation count variable of the i-th sub-window remains unchanged.

[0138] Starting from the minimum value, find the second voltage fluctuation judgment point that is larger than the minimum value, calculate the position index difference between the point and the minimum value, and divide the position index difference by the radius scaling factor to obtain the second local extreme value judgment radius.

[0139] Starting from the minimum value, the second local extremum determination radius is used as the search radius to find the maximum value; when the difference between the maximum and the minimum value is greater than or equal to the voltage fluctuation determination threshold, the voltage fluctuation count variable of the i-th sub-window is incremented by 1; when the difference between the maximum and the minimum value is less than the voltage fluctuation determination threshold, the voltage fluctuation count variable of the i-th sub-window remains unchanged.

[0140] Starting from the maximum value, begin the search to find the third voltage fluctuation judgment point smaller than the maximum value. Calculate the position index difference between this point and the maximum value, and divide the position index difference by the radius scaling factor to obtain the third local extreme value judgment radius.

[0141] Starting from the third voltage fluctuation judgment point and using the third local extreme value judgment radius as the search radius, find the next minimum value; when the difference between the maximum value and the next minimum value is greater than or equal to the voltage fluctuation judgment threshold, the voltage fluctuation count variable of the i-th sub-window is incremented by 1; when the difference between the maximum value and the next minimum value is less than the voltage fluctuation judgment threshold, the voltage fluctuation count variable of the i-th sub-window remains unchanged.

[0142] Search for the next maximum value from the next minimum value, and count voltage fluctuations, as above; until the beginning and end of the i-th sub-window are found.

[0143] Furthermore, the radius R for determining the first local extremum is:

[0144]

[0145]

[0146]

[0147] Where R3 is the radius after radius robustness verification, H is the entropy of the historical extreme value distribution within the sub-sliding window, and p k Let R² be the probability of the distance between the k-th type of extrema, R² be the radius of the extremum density constraint, and L be the value of the distance between the k-th type of extrema. i Let R1 be the length of the i-th sub-window, R1 be the corrected radius, ρ be the extremum density, and E be the value of the extremum density. i Let IQR(X) be the set of potential extreme points within the i-th sub-window, R0 be the initial radius benchmark, γ be the fluctuation intensity index, and X be the value of IQR(X). i ) represents the interquartile range of the data in the i-th sub-window. Let λ be the mean of the i-th sub-window, and λ be the length ratio coefficient.

[0148] The computer device of the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above method.

[0149] The computer-readable storage medium of the present invention stores a computer program thereon, which, when executed by a processor, implements the steps of the above-described method.

[0150] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of the present invention can be implemented using various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0151] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0152] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0153] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

Claims

1. A method for determining non-periodic voltage fluctuations by variable scale search, characterized by, Comprising For the voltage data of FIFO sliding window reading, it is divided into several sub-window data sets; According to the maximum, minimum and maximum values in the sub-window, the voltage fluctuation count of the sub-window is counted; If the voltage fluctuation count variable of any sub-window is greater than the fluctuation alarm threshold, an alarm is triggered; otherwise, no alarm is triggered.

2. The method for determining non-periodic voltage fluctuations using variable-scale searching according to claim 1, characterized in that, The method for dividing into several sub-window data sets is equal division method or variance adaptive division method or extreme distance clustering division method.

3. The method of claim 1, wherein the step of determining the non-periodic voltage fluctuation is performed by a variable scale search. According to the maximum, minimum and maximum values in the sub-window, the voltage fluctuation count of the sub-window is counted, specifically as follows: For the i-th sub-window data set, the maximum value and its position are determined, starting from the maximum point, finding the first voltage fluctuation judgment point smaller than the maximum value, calculating the position sequence number difference between the point and the maximum value, dividing the position sequence number difference by the radius scaling coefficient to obtain the first local extreme value judgment radius; Taking the first voltage fluctuation judgment point as the starting point and the first local extreme value judgment radius as the search radius, the minimum value is found; When the difference between the maximum value and the minimum value is greater than or equal to the voltage fluctuation judgment threshold, the voltage fluctuation count variable of the i-th sub-window is increased by 1; when the difference between the maximum value and the minimum value is less than the voltage fluctuation judgment threshold, the voltage fluctuation count variable of the i-th sub-window remains unchanged; Starting from the minimum value, the second voltage fluctuation judgment point greater than the minimum value is found, the position sequence number difference between the point and the minimum value is calculated, and the second local extreme value judgment radius is obtained by dividing the position sequence number difference by the radius scaling coefficient; Taking the minimum value as the starting point and the second local extreme value judgment radius as the search radius, the maximum value is found; when the difference between the maximum value and the minimum value is greater than or equal to the voltage fluctuation judgment threshold, the voltage fluctuation count variable of the i-th sub-window is increased by 1; when the difference between the maximum value and the minimum value is less than the voltage fluctuation judgment threshold, the voltage fluctuation count variable of the i-th sub-window remains unchanged; Taking the maximum value as the starting point, the third voltage fluctuation judgment point smaller than the maximum value is found, the position sequence number difference between the point and the maximum value is calculated, and the third local extreme value judgment radius is obtained by dividing the position sequence number difference by the radius scaling coefficient; Taking the third voltage fluctuation judgment point as the starting point and the third local extreme value judgment radius as the search radius, the next minimum value is found; When the difference between the maximum value and the next minimum value is greater than or equal to the voltage fluctuation judgment threshold, the voltage fluctuation count variable of the i-th sub-window is increased by 1; when the difference between the maximum value and the next minimum value is less than the voltage fluctuation judgment threshold, the voltage fluctuation count variable of the i-th sub-window remains unchanged; The next maximum value is searched from the next minimum value, and the voltage fluctuation count is performed, as above; until the beginning and end of the i-th sub-window are searched.

4. The method of claim 3, wherein the step of determining the non-periodic voltage fluctuation is performed by a variable scale search. The first local extreme value judgment radius R is wherein R3 is the radius after radius robustness verification, H is the historical extreme value distribution entropy in the sub sliding window, p k is the probability of the kth extreme value interval, R2 is the radius of the extreme value density constraint, L i is the length of the ith sub-window, R1 is the corrected radius, p is the extreme value density, E i is the potential extreme value point set in the ith sub-window, R0 is the initial radius reference, g is the volatility intensity index, IQR(X i ) is the interquartile range of the data in the ith sub-window, is the mean of the ith sub-window, and l is the length ratio coefficient.

5. A variable scale search non-periodic voltage fluctuation determination system characterized by, Comprising A data segmentation module for dividing the voltage data of FIFO sliding window reading into several sub-window data sets; A voltage fluctuation counting module for counting the voltage fluctuation of the sub-window according to the maximum, minimum and maximum values in the sub-window; An early warning module for triggering an alarm if the voltage fluctuation count variable of any sub-window is greater than the fluctuation alarm threshold; otherwise, no alarm is triggered.

6. The non-periodic voltage fluctuation determination system of claim 5, wherein In the data segmentation module, the data set is divided into several sub-window data sets by equal division method, variance adaptive segmentation method or extreme distance clustering segmentation method.

7. The non-periodic voltage fluctuation determination system of claim 5, wherein In the voltage fluctuation counting module, the maximum value, the minimum value and the maximum value in the sub-window are used to count the voltage fluctuation of the sub-window, and the counting is as follows: For the i-th sub-window data set, the maximum value and its position are determined, the first voltage fluctuation determination point smaller than the maximum value is found by searching from the maximum value point, the position sequence number difference between the point and the maximum value is calculated, and the position sequence number difference is divided by the radius scaling coefficient to obtain the first local extreme value determination radius; The first voltage fluctuation determination point is taken as the starting point, and the first local extreme value determination radius is taken as the search radius to find the minimum value; When the difference between the maximum value and the minimum value is greater than or equal to the voltage fluctuation determination threshold, the voltage fluctuation count variable of the i-th sub-window is increased by 1; when the difference between the maximum value and the minimum value is less than the voltage fluctuation determination threshold, the voltage fluctuation count variable of the i-th sub-window remains unchanged; The second voltage fluctuation determination point greater than the minimum value is found by searching from the minimum value, the position sequence number difference between the point and the minimum value is calculated, and the position sequence number difference is divided by the radius scaling coefficient to obtain the second local extreme value determination radius; The minimum value is taken as the starting point, and the second local extreme value determination radius is taken as the search radius to find the maximum value; when the difference between the maximum value and the minimum value is greater than or equal to the voltage fluctuation determination threshold, the voltage fluctuation count variable of the i-th sub-window is increased by 1; when the difference between the maximum value and the minimum value is less than the voltage fluctuation determination threshold, the voltage fluctuation count variable of the i-th sub-window remains unchanged; The third voltage fluctuation determination point smaller than the maximum value is found by searching from the maximum value, the position sequence number difference between the point and the maximum value is calculated, and the position sequence number difference is divided by the radius scaling coefficient to obtain the third local extreme value determination radius; The third voltage fluctuation determination point is taken as the starting point, and the third local extreme value determination radius is taken as the search radius to find the next minimum value; When the difference between the maximum value and the next minimum value is greater than or equal to the voltage fluctuation determination threshold, the voltage fluctuation count variable of the i-th sub-window is increased by 1; when the difference between the maximum value and the next minimum value is less than the voltage fluctuation determination threshold, the voltage fluctuation count variable of the i-th sub-window remains unchanged; The next maximum value is found by searching from the next minimum value, and the voltage fluctuation counting is performed, which is the same as above; until the beginning and end of the i-th sub-window are searched.

8. The non-periodic voltage fluctuation determination system of claim 7, wherein, The first local extreme value determination radius R is wherein R3 is the radius after radius robustness verification, H is the historical extreme value distribution entropy in the sub sliding window, p k is the probability of the kth extreme value interval, R2 is the radius of the extreme value density constraint, L i is the length of the ith sub-window, R1 is the corrected radius, p is the extreme value density, E i is the potential extreme value point set in the ith sub-window, R0 is the initial radius reference, g is the volatility intensity index, IQR(X i ) is the interquartile range of the data in the ith sub-window, is the mean of the ith sub-window, and l is the length ratio coefficient.

9. A computer device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the computer program to realize the steps of the method of any one of claims 1 to 4.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to realize the steps of the method of any one of claims 1 to 4.

Images