An oscillation risk early warning method and system based on correlation evaluation of multiple electrical elements

Abstract

The application discloses an oscillation risk early warning method and system based on multi-electric element correlation evaluation, the method realizes time sequence synchronous collection and rolling analysis of key electrical quantities such as power, voltage, current and power factor angle through a sliding window mechanism, extracts extreme value fluctuation characteristics and frequency characteristics of each quantity, designs a dimension threshold system, frequency correlation verification rules and a weight dynamic adaptation mechanism, establishes a coupling correlation model among multiple elements, introduces frequency consistency coefficient quantification characteristics matching degree, realizes technical upgrading from single quantity independent analysis to multi-element correlation evaluation, and realizes multi-electric quantity frequency mode cooperation. The application captures potential risks in advance through multi-quantity characteristic fusion analysis, accurately identifies risk signals in the initial stage of the oscillation fault chain, gains intervention time for power grid dispatching, effectively reduces the oscillation upgrading risk caused by multi-quantity cooperative fluctuation in the new power system, and guarantees stable operation of sensitive loads and safety of the power grid.

Description

Technical Field

[0001] This invention relates to the field of power system oscillation monitoring, and in particular to an oscillation risk early warning method and system based on the correlation evaluation of multiple electrical elements. Background Technology

[0002] Power system oscillations are a major hidden danger threatening the safe and stable operation of the power grid, with cascading and destructive consequences. Even minor oscillations can cause periodic fluctuations in voltage and current amplitudes, leading to decreased production precision in sensitive loads such as precision manufacturing and semiconductors, and increased product scrap rates. In severe cases, they can trigger generator set loss of synchronism, transmission line overload tripping, forming a fault chain of "oscillation-instability-outage," and even causing large-scale regional blackouts, resulting in huge economic losses and social impacts. Historical data shows that severe oscillation accidents can lead to power grid collapse, with power restoration taking hours to days, causing catastrophic consequences for industrial production and residential electricity use.

[0003] Oscillations have complex and diverse causes, mainly including subsynchronous oscillations caused by the high proportion of new energy sources, low-frequency oscillations caused by rapid load fluctuations, and transient oscillations caused by changes in grid topology. With the large-scale grid connection of intermittent power sources such as wind and solar power, the proportion of power electronic equipment is increasing, the system inertia is decreasing, the dynamic characteristics are deteriorating, the oscillation trigger threshold is decreasing, and the probability of occurrence is significantly increasing. In addition, the weak interconnection characteristics of long-distance transmission lines and relay protection malfunctions can also exacerbate the risk of oscillations. The traditional simple judgment mode of "single quantity threshold triggering" for oscillation alarms has technical bottlenecks such as difficulty in the coordinated identification of multiple electrical quantity frequency modes, complexity in the dynamic balancing of correlation weights, and insufficient adaptability of warning threshold scenarios. Summary of the Invention

[0004] Purpose of the invention: The purpose of this invention is to provide an oscillation risk early warning method that identifies risks in the early stage of oscillation by real-time capture of the coordinated fluctuation characteristics of multiple electrical quantities (power, voltage, current, etc.) and multi-electrical element correlation assessment; another purpose of this invention is to provide an oscillation risk early warning system based on multi-electrical element correlation assessment.

[0005] Technical solution: The oscillation risk early warning method based on multi-electrical element correlation assessment described in this invention includes:

[0006] Electrical quantity data are collected through a preset sliding window. Extreme value fluctuation features are extracted based on the extreme point identification rule of local dynamic radius. The electrical quantity data includes active power, voltage, current, and power factor angle. The extreme value fluctuation features include the maximum value sequence, the minimum value sequence, and the difference sequence between adjacent extreme points. The frequency characteristics are calculated based on the interval between the nearest adjacent maximum point, the interval between the farthest adjacent maximum point, and the average of the maximum point intervals to obtain the frequency characteristics of the electrical quantities.

[0007] Based on dimensional thresholds, frequency correlation verification rules, and dynamic weight adaptation mechanisms, a coupled correlation model among multiple elements is constructed.

[0008] Based on the preset warning threshold and reset threshold, a hierarchical warning triggering and verification mechanism is constructed. When the comprehensive evaluation result output by the coupled correlation model exceeds the warning threshold, the warning flag is set; when the comprehensive evaluation result output by the coupled correlation model falls back to below the reset threshold, the warning flag is reset.

[0009] Furthermore, for electrical quantity data collected within a preset sliding window, the time deviation of similar electrical quantity data is calculated based on the timestamp characteristics of PMU data. The time deviation is then corrected by linear interpolation based on the time deviation and the synchronization error threshold. Based on the high-frequency noise characteristics of PMU data, a moving average preprocessing is performed on each electrical quantity data within the window.

[0010] Local dynamic radius R i,t for

[0011] R i,t =R0·max(1,σ i,t / σ i,avg )

[0012]

[0013] Where R0 is the initial radius, σ i,t Let σ be the local variance at point t in the i-th window. i,avg f is the total variance within the window. s f is the sampling frequency, and f0 is the fundamental frequency of the power system. For the s-th preprocessed data of the i-th type of electrical quantity, It is a local mean.

[0014] Furthermore, based on dimensional thresholds, frequency correlation verification rules, and dynamic weight adaptation mechanisms, a coupling correlation model among multiple elements is constructed, as follows:

[0015] Based on the standard values ​​of the rated parameters of the power system, the load factor and the deviation coefficient of the span threshold of the power system are introduced to calculate the effective threshold of the differential fluctuation. Based on the effective threshold of the differential fluctuation and the effective fluctuation ratio, the noise in the difference sequence of adjacent extreme points is removed.

[0016] The correlation between different electrical quantity frequency characteristics in the frequency characteristics of electrical quantities is verified by setting a frequency proximity threshold; for electrical quantities participating in system coupling oscillation, their weights are corrected according to differentiated weight allocation rules and dynamic weight transfer strategies.

[0017] By introducing a frequency consistency coefficient, a comprehensive evaluation score is calculated based on extreme fluctuation characteristics and electrical quantity frequency characteristics to obtain a comprehensive evaluation result.

[0018] Furthermore, the threshold for differential volatility effectiveness θ i,dynamic for

[0019]

[0020] Where, θ i λ is the threshold deviation coefficient, α and β are correction coefficients, and λ is the power system load factor. Let μ be the variance of the noise within the window. i This represents the average electrical quantity within the window.

[0021] The span threshold deviation coefficient is calculated based on the differentiated fluctuation effectiveness threshold. If the span threshold deviation coefficient of at least one electrical type is greater than the threshold deviation coefficient threshold, the differentiated fluctuation effectiveness threshold of the electrical type with the largest deviation among all electrical types is corrected; otherwise, no correction is made.

[0022] If the effective fluctuation ratio is greater than or equal to the effective ratio threshold, then the current electrical quantity has an effective fluctuation signal; otherwise, the current electrical quantity is marked as noise-dominated and is removed.

[0023] Furthermore, the correlation between different electrical quantity frequency characteristics is verified by setting a frequency proximity threshold, as follows:

[0024] A frequency stability coefficient is introduced, and a weighted frequency similarity is calculated based on the frequency characteristics of electrical quantities. An association matrix is ​​constructed based on the weighted frequency similarity, the strong association frequency proximity benchmark threshold, and the weak association frequency proximity benchmark threshold. The coupling strength index is determined by the weighted frequency similarity and the association matrix.

[0025] If the coupling strength index is greater than or equal to 1, the current electrical quantity is determined to participate in the system's coupled oscillation; otherwise, the current electrical quantity is determined to be an independent fluctuation.

[0026] The differentiated weight allocation rule is to modify the preset power system basic weight vector based on the effective fluctuation ratio and the introduced frequency contribution coefficient.

[0027] The dynamic weight transfer strategy calculates a characteristic anomaly index for electrical quantities participating in system coupling oscillations based on their effective fluctuation ratio. If the characteristic anomaly index is greater than the characteristic anomaly threshold, weight transfer is performed; otherwise, weight transfer is not performed.

[0028] Furthermore, a frequency consistency coefficient is introduced, and a comprehensive evaluation score is calculated based on the extreme value fluctuation characteristics and the frequency characteristics of electrical quantities to obtain the comprehensive evaluation result, as follows:

[0029] Based on the feature anomaly index and coupling strength index, a subset of effective features is selected, the weighted average frequency of the subset of effective features is extracted, an effective frequency sequence is constructed, and the frequency range normalization factor is calculated.

[0030] Calculate the weighted frequency center based on the frequency stability coefficient and the effective frequency.

[0031] The frequency consistency coefficient is calculated based on the frequency range normalization factor, the weighted frequency center, and the sensitivity adjustment factor.

[0032] If the i-th type of electrical quantity is not in the effective feature subset, the evaluation score is 0; otherwise, the original score is calculated based on the effective fluctuation ratio, amplitude fluctuation intensity and basic weight of the electrical quantity, and frequency similarity is introduced to correct the original score to obtain the evaluation score.

[0033] Calculate the comprehensive evaluation score based on the evaluation scores of all types of electrical quantities.

[0034] The oscillation risk early warning system based on multi-electrical element correlation assessment described in this invention includes:

[0035] The feature extraction module is used to collect electrical quantity data from a preset sliding window, and extract extreme value fluctuation features based on the extreme point identification rules of the local dynamic radius. The electrical quantity data includes active power, voltage, current and power factor angle. The extreme value fluctuation features include the maximum value sequence, the minimum value sequence and the difference sequence between adjacent extreme points. The frequency characteristics are calculated based on the interval between the nearest adjacent maximum point, the interval between the farthest adjacent maximum point and the average of the maximum point intervals to obtain the frequency characteristics of the electrical quantities.

[0036] The model building module is used to construct a coupled correlation model between multiple elements based on dimensional thresholds, frequency correlation verification rules, and dynamic weight adaptation mechanisms.

[0037] The warning triggering and reset module is used to construct a hierarchical warning triggering and verification mechanism based on preset warning thresholds and reset thresholds. When the comprehensive evaluation result output by the coupled correlation model exceeds the warning threshold, the warning flag is set; when the comprehensive evaluation result output by the coupled correlation model falls back below the reset threshold, the warning flag is reset.

[0038] Furthermore, in the feature extraction module, electrical quantity data is collected for a preset sliding window. Based on the timestamp characteristics of PMU data, the time deviation of similar electrical quantity data is calculated. According to the time deviation and the synchronization error threshold, the time deviation is corrected by linear interpolation. Based on the high-frequency noise characteristics of PMU data, the moving average preprocessing is performed on each electrical quantity data in the window.

[0039] Local dynamic radius R i,t for

[0040] Ri,t =R0·max(1,σ i,t / σ i,avg )

[0041]

[0042] Where R0 is the initial radius, σ i,t Let σ be the local variance at point t in the i-th window. i,avg f is the total variance within the window. s f is the sampling frequency, and f0 is the fundamental frequency of the power system. For the s-th preprocessed data of the i-th type of electrical quantity, It is a local mean.

[0043] Furthermore, in the model building module, a coupled correlation model among multiple elements is constructed based on dimensional thresholds, frequency correlation verification rules, and dynamic weight adaptation mechanisms, as detailed below:

[0044] Based on the standard values ​​of the rated parameters of the power system, the load factor and the deviation coefficient of the span threshold of the power system are introduced to calculate the effective threshold of the differential fluctuation. Based on the effective threshold of the differential fluctuation and the effective fluctuation ratio, the noise in the difference sequence of adjacent extreme points is removed.

[0045] The correlation between different electrical quantity frequency characteristics in the frequency characteristics of electrical quantities is verified by setting a frequency proximity threshold; for electrical quantities participating in system coupling oscillation, their weights are corrected according to differentiated weight allocation rules and dynamic weight transfer strategies.

[0046] By introducing a frequency consistency coefficient, a comprehensive evaluation score is calculated based on extreme fluctuation characteristics and electrical quantity frequency characteristics to obtain a comprehensive evaluation result.

[0047] Furthermore, in the model building module, the threshold θ for differential volatility effectiveness... i,dynamic for

[0048]

[0049] Where, θ i λ is the threshold deviation coefficient, α and β are correction coefficients, and λ is the power system load factor. Let μ be the variance of the noise within the window. i This represents the average electrical quantity within the window.

[0050] The span threshold deviation coefficient is calculated based on the differentiated fluctuation effectiveness threshold. If the span threshold deviation coefficient of at least one electrical type is greater than the threshold deviation coefficient threshold, the differentiated fluctuation effectiveness threshold of the electrical type with the largest deviation among all electrical types is corrected; otherwise, no correction is made.

[0051] If the effective fluctuation ratio is greater than or equal to the effective ratio threshold, then the current electrical quantity has an effective fluctuation signal; otherwise, the current electrical quantity is marked as noise-dominated and is removed.

[0052] Furthermore, in the model building module, the correlation between different electrical quantity frequency features is verified by setting a frequency proximity threshold, as follows:

[0053] A frequency stability coefficient is introduced, and a weighted frequency similarity is calculated based on the frequency characteristics of electrical quantities. An association matrix is ​​constructed based on the weighted frequency similarity, the strong association frequency proximity benchmark threshold, and the weak association frequency proximity benchmark threshold. The coupling strength index is determined by the weighted frequency similarity and the association matrix.

[0054] If the coupling strength index is greater than or equal to 1, the current electrical quantity is determined to participate in the system's coupled oscillation; otherwise, the current electrical quantity is determined to be an independent fluctuation.

[0055] The differentiated weight allocation rule is to modify the preset power system basic weight vector based on the effective fluctuation ratio and the introduced frequency contribution coefficient.

[0056] The dynamic weight transfer strategy calculates a characteristic anomaly index for electrical quantities participating in system coupling oscillations based on their effective fluctuation ratio. If the characteristic anomaly index is greater than the characteristic anomaly threshold, weight transfer is performed; otherwise, weight transfer is not performed.

[0057] Furthermore, in the model building module, a frequency consistency coefficient is introduced to calculate a comprehensive evaluation score based on extreme value fluctuation characteristics and electrical quantity frequency characteristics, resulting in a comprehensive evaluation result, as follows:

[0058] Based on the feature anomaly index and coupling strength index, a subset of effective features is selected, the weighted average frequency of the subset of effective features is extracted, an effective frequency sequence is constructed, and the frequency range normalization factor is calculated.

[0059] Calculate the weighted frequency center based on the frequency stability coefficient and the effective frequency.

[0060] The frequency consistency coefficient is calculated based on the frequency range normalization factor, the weighted frequency center, and the sensitivity adjustment factor.

[0061] If the i-th type of electrical quantity is not in the effective feature subset, the evaluation score is 0; otherwise, the original score is calculated based on the effective fluctuation ratio, amplitude fluctuation intensity and basic weight of the electrical quantity, and frequency similarity is introduced to correct the original score to obtain the evaluation score.

[0062] Calculate the comprehensive evaluation score based on the evaluation scores of all types of electrical quantities.

[0063] Beneficial Effects: Compared with existing technologies, the significant advantages of this invention are: 1. This invention achieves synchronous time-series acquisition and rolling analysis of key electrical quantities such as power, voltage, current, and power factor angle through a sliding window mechanism, extracting extreme value fluctuation characteristics and frequency characteristics of each quantity; 2. This invention establishes a coupled correlation model among multiple elements by designing a dimensional threshold system, frequency correlation verification rules, and a dynamic weight adaptation mechanism, introducing a frequency consistency coefficient to quantify the feature matching degree between quantities, achieving a technological upgrade from independent analysis of a single quantity to multi-element correlation evaluation; 3. This invention differs from the simple judgment mode of "single quantity threshold triggering" in traditional oscillation alarms, overcoming technical bottlenecks such as the difficulty of collaborative identification of multiple electrical quantity frequency modes, the complexity of dynamic balancing of correlation weights, and insufficient adaptability of early warning threshold scenarios. By fusing and analyzing multiple quantity features, it can capture potential risks in advance, accurately identify risk signals in the early stage of oscillation fault chain formation, buy time for grid dispatching, effectively reduce the risk of oscillation escalation caused by multi-quantity collaborative fluctuations in new power systems, and ensure the stable operation of sensitive loads and grid safety. Attached Figure Description

[0064] Figure 1 This is a flowchart of the method of the present invention;

[0065] Figure 2 This is a schematic diagram showing the changes in various electrical quantities during oscillation of a certain generating unit.

[0066] Figure 3 A schematic diagram for extracting extreme points from sliding window data;

[0067] Figure 4 Line graph showing the variation of frequency stability coefficients of various electrical quantities in the oscillation case data;

[0068] Figure 5 A line graph showing the weighted frequency similarity variation of oscillation case data;

[0069] Figure 6 A line graph showing the distribution of oscillation risk scores for each electrical quantity;

[0070] Figure 7 A line graph showing the variation of the equipment threshold deviation coefficient;

[0071] Figure 8 A line graph showing the equipment oscillation risk assessment score. Detailed Implementation

[0072] The oscillation risk early warning method based on the correlation assessment of multiple electrical elements described in this invention includes the following steps:

[0073] S1: Construct a dynamic sliding window data acquisition and update mechanism to achieve time-series synchronous analysis of multiple electrical quantities. This mechanism uses a preset window length to continuously capture real-time acquired electrical quantity data, covering key elements such as power amplitude, voltage amplitude, current amplitude, and derived electrical parameters (calculated based on voltage and current phases). If the current acquired data length does not reach the preset window length, analysis is performed based on the actual acquired data length. The data window adopts a first-in, first-out (FIFO) update method, that is, the window is iterated by reading new data for a preset time period and simultaneously sliding out an equal amount of old data, providing a continuous and time-consistent data source for subsequent feature extraction and correlation evaluation.

[0074] S1.1 defines the basic parameters of the sliding window and the data integrity judgment rules. This is based on the high sampling rate characteristic of PMU data (sampling frequency f). s If the analysis window duration is set to Ts (Hz), then the window data length is as follows:

[0075] N = f s ×T (1)

[0076] Define window data matrix Where m represents the number of electrical quantity types (including power, voltage, current, and derived electrical parameters), the matrix row index corresponds to the time-series sampling point t = 1, 2, ..., N, and the column index corresponds to the electrical quantity type i = 1, 2, ..., m. When the actual collected data length N' < N, the percentage of valid data is calculated as follows:

[0077]

[0078] If η≥η th (η th To preset the data integrity threshold (with a value range of [0.8, 1]), a window matrix is ​​constructed based on the actual length N'. Otherwise, the data in the window is considered invalid.

[0079] S1.2 Design the mathematical model for FIFO sliding window updates. Define the sliding window update step size as ΔTs, corresponding to the data block length M = f s ×ΔT. Let the data matrix of the k-th window be as follows:

[0080] X k =[x k,1 ,x k,2 ,...,x k,N ] T (3)

[0081] in Let be the multi-electrical quantity vector at sampling point t. Then, the (k+1)th window is updated using the FIFO mechanism as follows:

[0082] Xk+1 =[x k,M+1 ,x k,M+2 ,...,x k,N ,x new,1 ,x new,2 ,...,x new,M ] T (4)

[0083] Where x new ,1,...,x new,M For the newly collected M sampling points, ensure that the window data is continuous in time and updated on a rolling basis.

[0084] S1.3 implements timing synchronization verification for multiple electrical quantities. Based on the timestamp characteristics of PMU data, let the timestamp of the t-th sampling point be τ. t Calculate the time deviation of different electrical quantities:

[0085] Δτ i,j (t)=|τ i,t -τ j,t | (5)

[0086] Where τ i,t (The timestamp of point t for the i-th type of electrical quantity). Define the synchronization error threshold Δτ. th (The value is no greater than 1 / f) s s), if for all (i,j) max t Δτ i,j (t)≤Δτ th If the timing of multiple electrical quantities within the window is synchronized, then the timing of the time stamps within the window is determined; otherwise, the timestamp deviation is corrected by linear interpolation calculated using equation (6).

[0087]

[0088] in The corrected timestamps ensure that multiple measurement data are aligned on the timeline.

[0089] S1.4 introduces PMU data noise suppression preprocessing. To address the high-frequency noise characteristics of PMU data, the electrical quantity data within the window are preprocessed using a moving average according to equation (7):

[0090]

[0091] Where x i,s This is the original data. For preprocessed data, (Use odd numbers to match the integer multiples of the 50Hz fundamental frequency period), suppressing white noise interference while preserving the wave characteristics.

[0092] S2: Build a multi-dimensional feature extraction module to complete the analysis of fluctuation and frequency characteristics of various electrical quantities. For each type of electrical quantity within the window, extract extreme value fluctuation characteristics and frequency characteristics respectively: the extreme value fluctuation characteristics include the maximum value, minimum value and the difference between adjacent extreme points (maximum and minimum values) in the data sequence, which are used to quantify the change in fluctuation amplitude; the frequency characteristics are calculated by the interval between the nearest adjacent maximum point, the interval between the farthest adjacent maximum point and the average of all maximum point intervals to form the frequency feature set of each electrical quantity, laying the feature foundation for subsequent correlation analysis between quantities.

[0093] S2.1 defines the extreme point identification rule based on the local dynamic radius. This is for the preprocessed data sequence of the i-th type of electrical quantity. (N is the window data length), local variance is introduced to dynamically adjust the extreme value judgment radius. Calculate the local variance at point t:

[0094]

[0095] in (f0 = 50Hz is the fundamental frequency of the power system) This is a local mean. Define the dynamic radius:

[0096] R i,t =R0·max(1,σ i,t / σ i,avg (9)

[0097] Where, σ i,avg Let V be the total variance within the window. If the following conditions are met:

[0098]

[0099] Then t is determined to be a maximum point k. peak A value ≥1.5 represents the peak significance coefficient; the rules for determining the minimum point are symmetrical, with only the inequality directions reversed. By adapting the dynamic radius to the superposition characteristics of fundamental wave fluctuations and oscillation signals in power system data, the misjudgment of the fundamental wave zero-crossing point as an extreme value is avoided.

[0100] S2.2 Constructing a quantitative model for fluctuation amplitude characteristics. This involves analyzing the identified extreme point sequence P. i =[p i,1 ,p i,2 ,...,p i,K ](p i,K (The location of the k-th extreme point, sorted by time), distinguishing maxima:

[0101]

[0102] Calculate the window extreme difference:

[0103] D i,max-min =max(M i )-min(m i (12)

[0104] The sequence of magnitude differences between adjacent extreme values ​​is:

[0105] d i =[|M i,k -m i,k |,|m i,k -M i,k+1 |,...] (13)

[0106] Where M i,k For the k-th maximum value, m i,k For the kth minimum value

[0107] Amplitude fluctuation intensity:

[0108]

[0109] in, The average difference is α, which is the weight attenuation coefficient, and α > 0 is the weight attenuation coefficient. This enhances the contribution of significant fluctuations and aligns with the non-uniform variation characteristics of power system oscillation amplitude.

[0110] S2.3 The multi-scale calculation method for design frequency characteristics is based on the sequence of maximum point locations, as shown in Equation 15:

[0111]

[0112] Where L is the number of maxima.

[0113] Calculation time interval:

[0114] Δt i,k =t i,k+1 -t i,k (16)

[0115] Define the nearest frequency that reflects the fastest fluctuation component:

[0116] f i,near =f s / min(Δt i,k (17)

[0117] Define the alienation frequency that reflects the slowest fluctuation component:

[0118] f i,far =f s / max(Δt i,k (18)

[0119] Define the weighted average frequency:

[0120]

[0121] Wherein, weight:

[0122]

[0123] in, For the average interval, β>0, highlighting the frequency contribution corresponding to the stable interval, adapting to the frequency concentration characteristics of low-frequency oscillations (0.02-25Hz) in power systems.

[0124] S2.4 implements a feature validity verification mechanism. To address potential measurement anomalies in power system data, feature validity indices are defined, including extreme point density verification.

[0125] ρ i =L / N (21)

[0126] If ρ i <ρ min , where ρ min =2f osc,min / f, f osc,min If the frequency is less than 0.02Hz, the frequency characteristic is considered invalid. Amplitude consistency check:

[0127] C i =corr(d i ,Δt i,k′ ) (twenty two)

[0128] Where, Δt i,k′ For the corresponding interval sequence, if |C i |<C th , (C th =0.3) then the amplitude feature is invalid. By verifying and eliminating false features caused by measurement noise or sudden interference, we can ensure that the extracted fluctuation and frequency features truly reflect the system oscillation state.

[0129] S3: Establish a multi-factor correlation evaluation model to achieve a technological upgrade from single-quantity analysis to multi-quantity coupled evaluation, specifically including:

[0130] S3.1 Design Dimension-Specific Judgment Threshold System sets differentiated fluctuation validity thresholds for the physical characteristics of different electrical quantities. By comparing the extreme value difference of each quantity with the corresponding threshold, fluctuation signals with practical analytical value are screened out, eliminating the interference of noise or minor disturbances on the evaluation results.

[0131] The steps for designing a dimensionless threshold system are as follows:

[0132] S3.1.1 Define the reference threshold parameters for each electrical quantity. Based on the rated parameter characteristics of the power system, set dimensionless reference values: Let the power reference value be P. base(Corresponding to the system's rated active power), the voltage reference value is U. base (corresponding to the rated voltage RMS value), current reference value is I base (Corresponding to the effective value of rated current), the power factor angle reference value is φ base (Corresponding to the midpoint of the normal operating power factor angle range). Define the basic threshold coefficient k. P ,k U ,k I ,k φ (Based on the sensitivity setting for electrical quantity fluctuations, with a value range of [0.05, 0.2]), the initial threshold is:

[0133]

[0134] It adapts to the differences in physical dimensions of different electrical quantities to ensure that the threshold has a clear physical meaning.

[0135] S3.1.2 Construct an adaptive threshold adjustment mechanism based on operating conditions. Introduce the system load factor λ (the ratio of current total active power to rated power, λ∈[0,1.2]) and the noise fluctuation coefficient γ. i Dynamically adjust the threshold. For the i-th type of electrical quantity i = P, U, I, φ, calculate the noise variance within the window. (Estimated using the residual sequence after wavelet thresholding), the dynamic threshold is:

[0136]

[0137] Where μ i The value represents the average electrical quantity within the window, with α and β being correction coefficients α∈[0.2,0.5] and β∈[0.1,0.3]. Under high load conditions, the threshold is appropriately amplified if λ>0.8 to avoid misjudgment during normal load fluctuations; when noise increases, the threshold is dynamically raised to enhance anti-interference capability.

[0138] S3.1.3 defines the rules for determining the validity of fluctuations. For the extreme value difference sequence d of the i-th type of electrical quantity... i Calculate the effective volatility percentage:

[0139]

[0140] in, This is an indicator function (it takes the value 1 if the condition is met, and 0 otherwise). If η i ≥η eff , where η eff If the effective percentage threshold is ∈[0.3,0.5], then the electrical quantity is determined to have an effective fluctuation signal; otherwise, it is marked as noise-dominated and its features are removed from subsequent correlation analysis.

[0141] S3.1.4 introduces a cross-quantity threshold consistency check. Considering the multi-quantity linkage characteristics of power systems, the threshold deviation coefficient is defined as:

[0142]

[0143] Where m = 4 represents the number of electrical quantity types. If δ > δ th , where (δ th If the value is 0.3, then a second correction is performed on the threshold with the largest deviation:

[0144]

[0145] Ensure that the threshold values ​​of different electrical quantities remain consistent in relative fluctuation intensity to avoid bias in correlation analysis caused by abnormal threshold values ​​of individual quantities.

[0146] S3.2 Constructs a frequency correlation verification mechanism, which verifies the correlation of frequency characteristics of different electrical quantities by setting a preset frequency proximity threshold. When the difference in frequency characteristics between two types of electrical quantities is within the threshold range, it is determined that there is a potential coupling relationship between them. Combining the differentiated weight allocation rule (setting the basic weight according to the importance of each electrical quantity in the oscillation risk assessment) and the dynamic weight transfer strategy, when the characteristics of a certain type of electrical quantity are abnormally missing, its weight is transferred to other electrical quantities with high correlation according to a preset ratio, ensuring the robustness of the evaluation model.

[0147] S3.2.1 Frequency Feature Similarity Measurement Model. For electrical quantities of type i and type j, extract their core frequency feature set f. i =[f i,avg ,f i,near ,f i,far ] and f j =[f j,avg ,f j,near ,f j,far ].

[0148] Introducing a frequency stability coefficient that reflects the degree of frequency concentration:

[0149]

[0150] s i ∈(0,1], calculate the weighted frequency similarity:

[0151]

[0152] Among them, (w k =[0.6,0.2,0.2]) is the frequency feature weight (emphasizing the contribution of average frequency), α=5 is the sensitivity coefficient, γ i,j The larger the value of ∈(0,1], the higher the degree of correlation.

[0153] S3.2.2 Association Level Classification and Coupling Determination. A frequency proximity reference threshold γ is set. th1 =0.7 (strong correlation) and γ th2 =0.4 (weak association), construct the association matrix

[0154]

[0155] Where m = 4.

[0156] Define the coupling strength index If C i If the value is greater than or equal to 1, the i-th type of electrical quantity is determined to participate in the system's coupled oscillation; otherwise, it is considered an independent fluctuation.

[0157] S3.2.3 Differentiated Basic Weight Allocation Model. Based on the physical role of electrical quantities in oscillation propagation, a basic weight vector w is defined. base =[w P ,w U ,w I ,w φ (Reflecting importance ranking). Introducing a frequency contribution coefficient:

[0158]

[0159] f osc,ref =1Hz (the reference oscillation frequency), then the corrected basic weight is:

[0160]

[0161] Where η i As an indicator of volatility effectiveness, the weighting of effective volatility and highly correlated quantities is strengthened.

[0162] S3.2.4 Dynamic Weight Transfer Strategy Design. Define the Feature Anomaly Index:

[0163]

[0164] a i =1 indicates complete abnormality), when a i ≥a th Weight transfer is triggered when the coefficient is 0.8. Calculate the association transfer coefficient:

[0165]

[0166] The weights after the transfer are:

[0167] w j′ =w j,base′ +w i,base′ ·t i→j ·exp(-a j(35)

[0168] Ensure that the weights of outliers are transferred to strongly correlated electrical quantities with normal characteristics, and conserve the total weight ∑w. j′ =Σw i,base′ It is adapted to some measurement anomaly scenarios in the power system.

[0169] S3.3 introduces a frequency consistency coefficient to quantify the matching degree of multiple quantity characteristics. This coefficient is calculated based on the range of the frequency characteristics of all effective electrical quantities. The smaller the range, the closer the coefficient is to 1, which represents the higher the synergy of the fluctuation characteristics of multiple quantities. The analysis scores of each electrical quantity are weighted and fused through this coefficient to form a comprehensive evaluation result, realizing the coupled analysis of multiple element characteristics.

[0170] The steps to establish a multi-factor correlation evaluation model are as follows:

[0171] S3.3.1 Construction of the effective frequency feature set. Select an effective feature subset Ω={i|a} from all electrical quantities. i <a th C i ≥1},(a i C is the characteristic anomaly index. i (where f is the coupling strength index) is used to extract the core frequency features f of each electrical quantity within the subset. i,core =f i,avg (Using average frequency as the core feature). If |Ω| < 2, then the default frequency consistency coefficient α3 = 0.5 (minimum coherence benchmark); otherwise, construct an effective frequency sequence:

[0172] f Ω =[f i,core |i∈Ω] (36)S3.3.2 Frequency Consistency Coefficient Calculation Model Definition Frequency Range Normalization Factor:

[0173] R Ω =max(f Ω )-min(f Ω (37)

[0174] Introducing frequency stability weight w Ω :

[0175]

[0176] Where s i This is the frequency stability coefficient.

[0177] Calculate the weighted frequency center:

[0178]

[0179] This yields the frequency consistency coefficient:

[0180]

[0181] Where β = 8 is the sensitivity adjustment coefficient, and |Ω| is the number of effective electrical quantities. This coefficient satisfies α3∈(0,1], and the smaller the range, the more effective quantities, and the more concentrated the frequency, the closer the coefficient is to 1, which fits the frequency consistency characteristics of the multi-quantity coordinated oscillation of the power system.

[0182] S3.3.3 Electrical Quantity Analysis Score Calculation: For the i-th type of electrical quantity, the original score s is calculated based on its fluctuation characteristics (extreme value difference, frequency characteristics). i (Calculation logic: effective fluctuation ratio η) i Multiply by amplitude fluctuation intensity S i,amp Multiplied by the base weight w i,base′ For invalid features Forced S i =0; For valid features, the corrected score is:

[0183]

[0184] Where γ = 2 is the association correction coefficient, reinforcing the measurement score consistent with group characteristics; γ i,Ω for

[0185]

[0186] Where, γ i,j For frequency similarity

[0187] S3.3.4 The comprehensive evaluation results are fused based on the frequency consistency coefficient to achieve multi-quantity fractional coupling. The comprehensive evaluation results are as follows:

[0188]

[0189] When the frequency coordination of multiple quantities is high (α3≈1), the weighted sum is mainly used to reflect the overall risk; when the coordination is low (α3≈0), the maximum single quantity fraction is mainly used to avoid misjudgment and adapt to the evolution characteristics of power system oscillations from local to global.

[0190] S4: Construct a tiered early warning triggering and verification mechanism to improve the accuracy and timeliness of early warnings. Set early warning thresholds and reset thresholds. When the comprehensive evaluation result exceeds the early warning threshold, set the early warning flag; when the result falls below the reset threshold (a preset percentage of the early warning threshold), reset the early warning flag. Record the average frequency characteristic when the early warning flag is set. If the subsequent comprehensive evaluation result exceeds the early warning threshold again, and the difference between the new average frequency characteristic and the historical record is within a preset range, the early warning flag will not be set again to avoid redundant early warnings for the same type of risk.

[0191] S5: Perform data quality verification and anomaly filtering to ensure the reliability of the analysis basis. Conduct multi-dimensional verification on the data within the window: If the average value of the core electrical quantities is lower than the preset examination threshold, or if there are more than a preset number of positive and negative crossovers, abnormal zero values, or extreme value differences of key parameters that exceed the reasonable range, then the data in that window is deemed invalid, the analysis scores of all electrical quantities are set to zero, and they are not included in the comprehensive evaluation; invalid data windows are eliminated through data verification to ensure that the early warning model is based on reliable data for analysis.

[0192] The oscillation risk early warning system based on multi-electrical element correlation assessment described in this invention includes:

[0193] The feature extraction module is used to collect electrical quantity data from a preset sliding window, and extract extreme value fluctuation features based on the extreme point identification rules of the local dynamic radius. The electrical quantity data includes active power, voltage, current and power factor angle. The extreme value fluctuation features include the maximum value sequence, the minimum value sequence and the difference sequence between adjacent extreme points. The frequency characteristics are calculated based on the interval between the nearest adjacent maximum point, the interval between the farthest adjacent maximum point and the average of the maximum point intervals to obtain the frequency characteristics of the electrical quantities.

[0194] The model building module is used to construct a coupled correlation model between multiple elements based on dimensional thresholds, frequency correlation verification rules, and dynamic weight adaptation mechanisms.

[0195] The warning triggering and reset module is used to construct a hierarchical warning triggering and verification mechanism based on preset warning thresholds and reset thresholds. When the comprehensive evaluation result output by the coupled correlation model exceeds the warning threshold, the warning flag is set; when the comprehensive evaluation result output by the coupled correlation model falls back below the reset threshold, the warning flag is reset.

[0196] Furthermore, in the feature extraction module, electrical quantity data is collected for a preset sliding window. Based on the timestamp characteristics of PMU data, the time deviation of similar electrical quantity data is calculated. According to the time deviation and the synchronization error threshold, the time deviation is corrected by linear interpolation. Based on the high-frequency noise characteristics of PMU data, the moving average preprocessing is performed on each electrical quantity data in the window.

[0197] Local dynamic radius R i,t for

[0198] R i,t =R0·max(1,σ i,t / σ i,avg )

[0199]

[0200] Where R0 is the initial radius, σ i,t Let σ be the local variance at point t in the i-th window. i,avgf is the total variance within the window. s f is the sampling frequency, and f0 is the fundamental frequency of the power system. For the s-th preprocessed data of the i-th type of electrical quantity, It is a local mean.

[0201] Furthermore, in the model building module, a coupled correlation model among multiple elements is constructed based on dimensional thresholds, frequency correlation verification rules, and dynamic weight adaptation mechanisms, as detailed below:

[0202] Based on the standard values ​​of the rated parameters of the power system, the load factor and the deviation coefficient of the span threshold of the power system are introduced to calculate the effective threshold of the differential fluctuation. Based on the effective threshold of the differential fluctuation and the effective fluctuation ratio, the noise in the difference sequence of adjacent extreme points is removed.

[0203] The correlation between different electrical quantity frequency characteristics in the frequency characteristics of electrical quantities is verified by setting a frequency proximity threshold; for electrical quantities participating in system coupling oscillation, their weights are corrected according to differentiated weight allocation rules and dynamic weight transfer strategies.

[0204] By introducing a frequency consistency coefficient, a comprehensive evaluation score is calculated based on extreme fluctuation characteristics and electrical quantity frequency characteristics to obtain a comprehensive evaluation result.

[0205] Furthermore, in the model building module, the threshold θ for differential volatility effectiveness... i,dynamic for

[0206]

[0207] Where, θ i α is the threshold deviation coefficient, used to measure the effectiveness threshold of differentiated fluctuations in the i-th type of electrical quantity (such as active power, voltage, current, power factor angle, etc.). It is a key parameter for the degree of deviation between the threshold values ​​of other electrical quantities. Its core function is to ensure the consistency of the relative fluctuation intensity of multiple electrical quantity thresholds and avoid correlation judgment deviations caused by abnormal individual quantity thresholds. α and β are correction coefficients, and λ is the power system load factor. Let μ be the variance of the noise within the window. i This represents the average electrical quantity within the window.

[0208] The span threshold deviation coefficient is calculated based on the differentiated fluctuation effectiveness threshold. If the span threshold deviation coefficient of at least one electrical type is greater than the threshold deviation coefficient threshold, the differentiated fluctuation effectiveness threshold of the electrical type with the largest deviation among all electrical types is corrected; otherwise, no correction is made.

[0209] If the effective fluctuation ratio is greater than or equal to the effective ratio threshold, then the current electrical quantity has an effective fluctuation signal; otherwise, the current electrical quantity is marked as noise-dominated and is removed.

[0210] Furthermore, in the model building module, the correlation between different electrical quantity frequency features is verified by setting a frequency proximity threshold, as follows:

[0211] A frequency stability coefficient is introduced, and a weighted frequency similarity is calculated based on the frequency characteristics of electrical quantities. An association matrix is ​​constructed based on the weighted frequency similarity, the strong association frequency proximity benchmark threshold, and the weak association frequency proximity benchmark threshold. The coupling strength index is determined by the weighted frequency similarity and the association matrix.

[0212] If the coupling strength index is greater than or equal to 1, the current electrical quantity is determined to participate in the system's coupled oscillation; otherwise, the current electrical quantity is determined to be an independent fluctuation.

[0213] The differentiated weight allocation rule is to modify the preset power system basic weight vector based on the effective fluctuation ratio and the introduced frequency contribution coefficient.

[0214] The dynamic weight transfer strategy calculates a characteristic anomaly index for electrical quantities participating in system coupling oscillations based on their effective fluctuation ratio. If the characteristic anomaly index is greater than the characteristic anomaly threshold, weight transfer is performed; otherwise, weight transfer is not performed.

[0215] Furthermore, in the model building module, a frequency consistency coefficient is introduced to calculate a comprehensive evaluation score based on extreme value fluctuation characteristics and electrical quantity frequency characteristics, resulting in a comprehensive evaluation result, as follows:

[0216] Based on the feature anomaly index and coupling strength index, a subset of effective features is selected, the weighted average frequency of the subset of effective features is extracted, an effective frequency sequence is constructed, and the frequency range normalization factor is calculated.

[0217] Calculate the weighted frequency center based on the frequency stability coefficient and the effective frequency.

[0218] The frequency consistency coefficient is calculated based on the frequency range normalization factor, the weighted frequency center, and the sensitivity adjustment factor.

[0219] If the i-th type of electrical quantity is not in the effective feature subset, the evaluation score is 0; otherwise, the original score is calculated based on the effective fluctuation ratio, amplitude fluctuation intensity and basic weight of the electrical quantity, and frequency similarity is introduced to correct the original score to obtain the evaluation score.

[0220] Calculate the comprehensive evaluation score based on the evaluation scores of all types of electrical quantities.

[0221] 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.

[0222] 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.

[0223] 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.

[0224] 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 early warning of oscillation risk based on the correlation assessment of multiple electrical elements, characterized in that, include: Electrical quantity data are collected through a preset sliding window. Extreme value fluctuation features are extracted based on the extreme point identification rule of local dynamic radius. The electrical quantity data includes active power, voltage, current and power factor angle. The extreme value fluctuation features include the maximum value sequence, the minimum value sequence and the difference sequence between adjacent extreme points. The frequency characteristics are calculated based on the interval between the nearest adjacent maximum point, the interval between the farthest adjacent maximum point and the average of the maximum point intervals to obtain the frequency characteristics of the electrical quantities. Based on dimensional thresholds, frequency correlation verification rules, and dynamic weight adaptation mechanisms, a coupled correlation model among multiple elements is constructed. Based on the preset warning threshold and reset threshold, a hierarchical warning triggering and verification mechanism is constructed. When the comprehensive evaluation result output by the coupled correlation model exceeds the warning threshold, the warning flag is set. When the comprehensive evaluation result output by the coupled correlation model falls below the reset threshold, the warning flag is reset. Based on dimensional thresholds, frequency correlation verification rules, and dynamic weight adaptation mechanisms, a coupling correlation model among multiple elements is constructed, as follows: Based on the standard values ​​of the rated parameters of the power system, the load factor and the deviation coefficient of the span threshold of the power system are introduced to calculate the effective threshold of the differential fluctuation. Based on the effective threshold of the differential fluctuation and the effective fluctuation ratio, the noise in the difference sequence of adjacent extreme points is removed. The correlation between different electrical quantity frequency characteristics in the frequency characteristics of electrical quantities is verified by setting a frequency proximity threshold; For electrical quantities involved in system coupling oscillation, their weights are modified according to differentiated weight allocation rules and dynamic weight transfer strategies; By introducing a frequency consistency coefficient, a comprehensive evaluation score is calculated based on extreme fluctuation characteristics and electrical quantity frequency characteristics to obtain a comprehensive evaluation result.

2. The oscillation risk early warning method based on multi-electrical element correlation assessment according to claim 1, characterized in that, For electrical quantity data collected within a preset sliding window, the time deviation of similar electrical quantity data is calculated based on the timestamp characteristics of PMU data. The time deviation is then corrected by linear interpolation based on the time deviation and the synchronization error threshold. Based on the high-frequency noise characteristics of PMU data, a moving average preprocessing is performed on each electrical quantity data within the window. Local dynamic radius for Where R0 is the initial radius. Let be the local variance of point t in the i-th window. The total variance within the window. Sampling frequency, The fundamental frequency of the power system. For the s-th preprocessed data of the i-th type of electrical quantity, It is a local mean.

3. The oscillation risk early warning method based on multi-electrical element correlation assessment according to claim 1, characterized in that, Differential volatility effectiveness threshold for in, This is the threshold deviation coefficient for the span. , For correction factor, For the power system load factor, The variance of noise within the window. This represents the average electrical quantity within the window. The span threshold deviation coefficient is calculated based on the differentiated fluctuation effectiveness threshold. If the span threshold deviation coefficient of at least one electrical type is greater than the threshold deviation coefficient threshold, the differentiated fluctuation effectiveness threshold of the electrical type with the largest deviation among all electrical types is corrected; otherwise, no correction is made. If the effective fluctuation ratio is greater than or equal to the effective ratio threshold, then the current electrical quantity has an effective fluctuation signal; otherwise, the current electrical quantity is marked as noise-dominated and is removed.

4. The oscillation risk early warning method based on multi-electrical element correlation assessment according to claim 3, characterized in that, The correlation between different electrical quantity frequency features in the electrical quantity frequency features is verified by setting a frequency proximity threshold. Specifically, the frequency stability coefficient is introduced, and the weighted frequency similarity is calculated based on the electrical quantity frequency features. Based on the weighted frequency similarity, the strong correlation frequency proximity benchmark threshold and the weak correlation frequency proximity benchmark threshold, the correlation matrix is ​​constructed. The coupling strength index is determined by the weighted frequency similarity and the correlation matrix. If the coupling strength index is greater than or equal to 1, the current electrical quantity is determined to participate in the system's coupled oscillation; otherwise, the current electrical quantity is determined to be an independent fluctuation. The differentiated weight allocation rule is to modify the preset power system basic weight vector based on the effective fluctuation ratio and the introduced frequency contribution coefficient. The dynamic weight transfer strategy calculates a characteristic anomaly index for electrical quantities participating in system coupling oscillations based on their effective fluctuation ratio. If the characteristic anomaly index is greater than the characteristic anomaly threshold, weight transfer is performed; otherwise, weight transfer is not performed.

5. The oscillation risk early warning method based on multi-electrical element correlation assessment according to claim 4, characterized in that, A frequency consistency coefficient is introduced, and a comprehensive evaluation score is calculated based on extreme fluctuation characteristics and electrical quantity frequency characteristics to obtain a comprehensive evaluation result, as follows: Based on the feature anomaly index and coupling strength index, a subset of effective features is selected, the weighted average frequency of the subset of effective features is extracted, an effective frequency sequence is constructed, and the frequency range normalization factor is calculated. Calculate the weighted frequency center based on the frequency stability coefficient and the effective frequency. The frequency consistency coefficient is calculated based on the frequency range normalization factor, the weighted frequency center, and the sensitivity adjustment factor. If the i-th type of electrical quantity is not in the effective feature subset, the evaluation score is 0; Conversely, based on the effective fluctuation ratio, amplitude fluctuation intensity, and basic weight of this type of electrical quantity, the original score is calculated, and frequency similarity is introduced to correct the original score to obtain the evaluation score. Calculate the comprehensive evaluation score based on the evaluation scores of all types of electrical quantities.

6. An oscillation risk early warning system based on the correlation assessment of multiple electrical elements, characterized in that, include: The feature extraction module is used to collect electrical quantity data from a preset sliding window, and extract extreme value fluctuation features based on the extreme point identification rules of the local dynamic radius. The electrical quantity data includes active power, voltage, current and power factor angle. The extreme value fluctuation features include the maximum value sequence, the minimum value sequence and the difference sequence between adjacent extreme points. The frequency characteristics are calculated based on the interval between the nearest adjacent maximum point, the interval between the farthest adjacent maximum point and the average of the maximum point intervals to obtain the frequency characteristics of the electrical quantities. The model building module is used to construct a coupled correlation model between multiple elements based on dimensional thresholds, frequency correlation verification rules, and dynamic weight adaptation mechanisms. The warning triggering and reset module is used to construct a hierarchical warning triggering and verification mechanism based on preset warning thresholds and reset thresholds. When the comprehensive evaluation result output by the coupled correlation model exceeds the warning threshold, the warning flag is set. When the comprehensive evaluation result output by the coupled correlation model falls below the reset threshold, the warning flag is reset. Based on dimensional thresholds, frequency correlation verification rules, and dynamic weight adaptation mechanisms, a coupling correlation model among multiple elements is constructed, as follows: Based on the standard values ​​of the rated parameters of the power system, the load factor and the deviation coefficient of the span threshold of the power system are introduced to calculate the effective threshold of the differential fluctuation. Based on the effective threshold of the differential fluctuation and the effective fluctuation ratio, the noise in the difference sequence of adjacent extreme points is removed. The correlation between different electrical quantity frequency characteristics in the frequency characteristics of electrical quantities is verified by setting a frequency proximity threshold; For electrical quantities involved in system coupling oscillation, their weights are modified according to differentiated weight allocation rules and dynamic weight transfer strategies; By introducing a frequency consistency coefficient, a comprehensive evaluation score is calculated based on extreme fluctuation characteristics and electrical quantity frequency characteristics to obtain a comprehensive evaluation result.

7. The oscillation risk early warning system based on multi-electrical element correlation evaluation according to claim 6, characterized in that, In the feature extraction module, electrical quantity data is collected for a preset sliding window. Based on the timestamp characteristics of PMU data, the time deviation of similar electrical quantity data is calculated. According to the time deviation and the synchronization error threshold, the time deviation is corrected by linear interpolation. Based on the high-frequency noise characteristics of PMU data, the moving average preprocessing is performed on each electrical quantity data in the window. Local dynamic radius for Where R0 is the initial radius. Let be the local variance of point t in the i-th window. The total variance within the window. Sampling frequency, The fundamental frequency of the power system. For the s-th preprocessed data of the i-th type of electrical quantity, It is a local mean.

8. The oscillation risk early warning system based on multi-electrical element correlation assessment according to claim 6, characterized in that, In the model building module, the threshold for differential volatility effectiveness. for in, This is the threshold deviation coefficient for the span. , For correction factor, For the power system load factor, The variance of noise within the window. This represents the average electrical quantity within the window. The span threshold deviation coefficient is calculated based on the differentiated fluctuation effectiveness threshold. If the span threshold deviation coefficient of at least one electrical type is greater than the threshold deviation coefficient threshold, the differentiated fluctuation effectiveness threshold of the electrical type with the largest deviation among all electrical types is corrected; otherwise, no correction is made. If the effective fluctuation ratio is greater than or equal to the effective ratio threshold, then the current electrical quantity has an effective fluctuation signal; otherwise, the current electrical quantity is marked as noise-dominated and is removed. In the model building module, the correlation between different electrical quantity frequency features in the electrical quantity frequency features is verified by setting a frequency proximity threshold. Specifically, the frequency stability coefficient is introduced, the weighted frequency similarity is calculated based on the electrical quantity frequency features, and the correlation matrix is ​​constructed based on the weighted frequency similarity, the strong correlation frequency proximity benchmark threshold and the weak correlation frequency proximity benchmark threshold. The coupling strength index is determined by the weighted frequency similarity and the correlation matrix. If the coupling strength exponent is greater than or equal to 1, the current electrical quantity is determined to participate in the system's coupled oscillation; otherwise, the current electrical quantity is determined to be an independent fluctuation. The differentiated weight allocation rule is to modify the preset power system basic weight vector based on the effective fluctuation ratio and the introduced frequency contribution coefficient. The dynamic weight transfer strategy calculates a characteristic anomaly index for electrical quantities participating in system coupling oscillations based on their effective fluctuation ratio. If the characteristic anomaly index is greater than the characteristic anomaly threshold, weight transfer is performed; otherwise, weight transfer is not performed. In the model building module, a frequency consistency coefficient is introduced to calculate a comprehensive evaluation score based on extreme value fluctuation characteristics and electrical quantity frequency characteristics, resulting in a comprehensive evaluation result, as follows: Based on the feature anomaly index and coupling strength index, a subset of effective features is selected, the weighted average frequency of the subset of effective features is extracted, an effective frequency sequence is constructed, and the frequency range normalization factor is calculated. Calculate the weighted frequency center based on the frequency stability coefficient and the effective frequency. The frequency consistency coefficient is calculated based on the frequency range normalization factor, the weighted frequency center, and the sensitivity adjustment factor. If the i-th type of electrical quantity is not in the effective feature subset, the evaluation score is 0; otherwise, the original score is calculated based on the effective fluctuation ratio, amplitude fluctuation intensity and basic weight of the type of electrical quantity, and frequency similarity is introduced to correct the original score to obtain the evaluation score. Calculate the comprehensive evaluation score based on the evaluation scores of all types of electrical quantities.

Images