A method for determining the operating section of a distribution network based on multi-layer data fusion

Abstract

This invention relates to the field of distribution network operation data processing technology, and discloses a method for determining distribution network operation sections based on multi-layer data fusion, aiming to solve the problems of poor noise resistance and real-time lag in existing technologies. The method of this invention includes: first, acquiring and preprocessing high-frequency and low-frequency measurement data of the distribution network; second, constructing a local fluctuation fingerprint using high-frequency data to characterize the physical change patterns; next, generating instantaneous virtual boundary points based on the fluctuation fingerprint to break the dependence of interpolation calculations on future data; subsequently, calculating dynamic reliability by analyzing data deviation and mapping it to a spline relaxation factor, introducing a soft constraint mechanism; finally, constructing a real-time dynamic relaxation cubic spline model and determining the section. This invention can achieve real-time, noise-resistant, and physically consistent reconstruction of distribution network operation sections, effectively adapting to different operating conditions such as stable and drastic fluctuations.

Description

Technical Field

[0001] This invention belongs to the technical field of distribution network operation data processing, specifically a method for determining the operation section of a distribution network based on multi-layer data fusion. Background Technology

[0002] The operational profile of the distribution network is a key indicator characterizing the real-time operating status of the power grid, and its construction relies on the effective fusion of multi-source measurement data in the distribution network. In practical applications, it is usually necessary to combine high-frequency sampling data from the substation side with low-frequency sampling data from the user side, and use data interpolation methods to fill the time gaps in the low-frequency data, so as to achieve continuous monitoring of the operating status of the entire network, such as voltage and power.

[0003] In existing power distribution network cross-section construction technologies, cubic spline interpolation is a commonly used mathematical fitting algorithm. This algorithm constructs a piecewise smooth polynomial function to connect discrete low-frequency sampling points, thereby calculating the values ​​within the sampling interval. However, when facing the complex actual operating environment of power distribution networks, the traditional cubic spline interpolation method has obvious drawbacks: First, when calculating the values ​​of the current interval, the algorithm often relies on the sampling points of the next future time as boundary constraints, resulting in a serious real-time lag in cross-section generation, which cannot meet the needs of real-time monitoring; Second, the traditional algorithm is a purely mathematical "hard constraint" fitting, forcing the interpolation curve to pass through every sampling point. When the data contains measurement noise or outlier data, the algorithm cannot identify and remove the interference, which will instead lead to severe distortion of the reconstructed cross-section waveform and make it difficult to reflect the physical changes in the transmission of high-frequency fluctuations to low-frequency nodes. Summary of the Invention

[0004] This invention provides a method for determining the operating section of a distribution network based on multi-layer data fusion, aiming to solve the problem that related technologies cannot achieve real-time and noise-resistant distribution network operating sections.

[0005] The present invention provides a method for determining the operating section of a distribution network based on multi-layer data fusion, comprising: Acquire multi-source measurement data containing high-frequency and low-frequency sampling data; calculate the local fluctuation fingerprint within the current calculation window based on the high-frequency sampling data; the local fluctuation fingerprint is a weighted integral of the absolute change rate of high-frequency voltage within the current calculation window, and the weight decreases exponentially as the time difference between the absolute change rate and the current time increases; When there are unreached subsequent low-frequency sampling points, a virtual boundary point is generated based on the latest low-frequency sampling data, the latest high-frequency sampling data, and the local fluctuation fingerprint at the current moment. The value of the virtual boundary point is positively correlated with the value of the latest low-frequency sampling data, the difference between the voltage value of the latest high-frequency sampling data and the high-frequency voltage value corresponding to the latest low-frequency sampling point, and the magnitude of the local fluctuation fingerprint. Calculate the dynamic confidence level of each low-frequency sampling data, wherein the dynamic confidence level is negatively correlated with the deviation between the measured and predicted values ​​of the corresponding low-frequency sampling data; generate a relaxation factor for cubic spline interpolation based on the dynamic confidence level, wherein the relaxation factor is positively correlated with the dynamic confidence level; The low-frequency sampling data and virtual boundary points are interpolated to obtain a low-frequency data curve. The interpolation is controlled by a relaxation factor to make the curve smoother at nodes with low confidence. The measured high-frequency sampling data and the interpolated low-frequency data curve are combined to obtain the distribution network operation section.

[0006] Preferably, the local fluctuation fingerprint The calculation formula is: ; in, Indicates the current moment; express The high-frequency voltage value at any given time; The set calculation window length; This is the time decay coefficient.

[0007] Preferably, the method for calculating the local fluctuation fingerprint further includes: The calculation window length The value range is 5 to 10 minutes; the time decay coefficient The value range is 0.1 to 0.5.

[0008] Preferably, the virtual boundary point The calculation method is as follows: ; in, For the current moment Virtual boundary points, This is the latest low-frequency sampling data; The voltage value is the latest high-frequency sampling data; for The corresponding high-frequency voltage value; This is the historical correlation coefficient; This represents the local fluctuation fingerprint at the current moment.

[0009] Preferably, the historical correlation coefficient The value range is 0.8 to 1.2.

[0010] Preferably, the dynamic trust level The calculation method is as follows: ; in, These are the measured values ​​corresponding to the low-frequency sampling data; These are predicted values ​​based on historical data trends. The standard deviation of the historical data for the corresponding node; This is the sensitivity constant.

[0011] Preferably, the method for generating the relaxation factor is as follows: ; in, It is a relaxation factor; , These are the lower and upper limits of the relaxation factor, respectively. This is the adjustment coefficient for the hyperbolic tangent function; For dynamic trust levels.

[0012] Preferably, the lower limit of the relaxation factor The value can be 0.01, with an upper limit. The value is 100.

[0013] Preferably, the cubic spline interpolation includes minimizing the objective function. The calculation method is as follows: ; in, For the interpolation function at the sampling point The value at; These are measured values ​​from low-frequency sampling data; Sampling points Dynamic trust level; Sampling points The relaxation factor; is the second derivative of the interpolation function.

[0014] Preferably, the method for determining the operating section of the power distribution network further includes preprocessing the multi-source measurement data, including: cleaning and aligning the timestamps of each data point, and using the time axis of the high-frequency sampling data as the reference section time axis; and marking data with format errors or missing data.

[0015] By adopting the above technical solution, the present invention has at least one of the following beneficial effects: 1. By fusing high-frequency SCADA data with low-frequency AMI data, this invention solves the core problems of poor noise resistance and real-time lag in traditional cross-section reconstruction technology. By introducing high-frequency accompanying features and a dynamic relaxation mechanism, this invention can both use high-frequency data to compensate for the loss of details in low-frequency data and avoid noise interference through physical constraints, thus achieving a noise-resistant and real-time reconstruction of the operating cross-section of the distribution network that conforms to physical laws.

[0016] 2. By constructing a "local fluctuation fingerprint" and utilizing the time-decrease-weighted integral of the voltage change rate, the directionality and timeliness of fluctuations can be taken into account. The advantage of this technique is that it uses the physical change patterns contained in high-frequency data to quantify whether the current power grid is in a "stable period" or a "period of severe fluctuations," and assigns higher weight to recent data through a decay coefficient, thereby providing accurate and physically meaningful real-time state references for subsequent interpolation calculations, rather than blind mathematical fitting.

[0017] 3. By combining the nonlinear correction of local fluctuation fingerprints with the relative rate of change of high-frequency data, the current "virtual boundary point" can be deduced in real time before the next real low-frequency data arrives. The effect of this technology is that it breaks the dependence of interpolation algorithms on future data, and achieves "zero-wait" start-up of cross-section calculation while ensuring the correctness of physical trends, significantly improving the real-time response capability of the system.

[0018] 4. A "soft constraint" mechanism based on data deviation is introduced. Data reliability is assessed by calculating the deviation between observed values ​​and trend predictions, and the spline relaxation factor is dynamically adjusted accordingly. The technical effect is that when the data is reliable, the curve closely follows the data points to preserve true characteristics; when outlier noise occurs, the algorithm automatically increases the relaxation to allow the curve to pass smoothly, thus effectively filtering out measurement noise without losing dynamic characteristics.

[0019] 5. The calculated dynamic relaxation factor is directly applied to the smoothness constraint term. The technical effect is that it transforms fluctuation fingerprints and confidence levels into weight parameters in the mathematical optimization process. This ensures that the final generated network operation cross-section not only achieves precise alignment and stitching of high- and low-frequency data on the time axis, but also combines the dynamic details of high-frequency data with the macroscopic trends of low-frequency data in terms of morphology, thus guaranteeing the accuracy of the cross-section. Attached Figure Description

[0020] Figure 1 This is a flowchart of a method for determining the operating section of a power distribution network based on multi-layer data fusion, according to the present invention. Figure 2 This is a flowchart of step S4 in the method for determining the operating section of a power distribution network according to the present invention. Detailed Implementation

[0021] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below in conjunction with the technical details of the embodiments of the present invention. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Understandably, this invention aims to address the core technical problems of poor noise resistance (susceptibility to outlier interference) and lagging real-time performance (dependence on future data points) in existing distribution network section construction technologies. This invention proposes a method for determining distribution network operating sections based on multi-layer data fusion. By introducing a fluctuation fingerprint based on high-frequency data and a dynamic relaxation mechanism, it achieves real-time, noise-resistant, and physically consistent reconstruction of operating sections.

[0023] like Figure 1 and Figure 2 As shown, the method in this embodiment mainly includes the following steps: S1. Acquire multi-level time series measurement data and perform preprocessing.

[0024] In this embodiment, multi-source measurement data within the target area of ​​the distribution network is first acquired. This data is divided into two levels on a time scale: High-frequency sampling data set ( The data primarily originates from the SCADA (Supervisory and Data Acquisition) system, including substation outlet voltage and total power. The preferred sampling interval is 1. 5 seconds.

[0025] Low-frequency sampling data set ( The data primarily originates from smart meters within the AMI (Advanced Metering Infrastructure), including user node voltage and load data. The sampling interval is typically on the order of 15 minutes.

[0026] In the data preprocessing stage, timestamp cleaning and alignment are performed on both types of data. High-frequency data is prioritized. The time axis is the reference section time axis. If obvious formatting errors or missing data are found, perform preliminary interpolation or marking.

[0027] S2. Construct local fluctuation fingerprints based on high-frequency accompanying features.

[0028] To address the loss of detail caused by the sparsity of low-frequency data, this step utilizes the physical variation patterns inherent in high-frequency data (i.e., fluctuations at the upstream power source are transmitted to the downstream nodes) to compensate for this loss. Specifically, high-frequency data... Calculate the local fluctuation fingerprint within the current calculation window.

[0029] Since variance alone cannot reflect the directionality and timeliness of fluctuations, this embodiment constructs a local fluctuation fingerprint index. The calculation formula is as follows:

[0030] in, :express The high-frequency voltage measurement value at that moment; : This represents the length of the sliding time window. In this embodiment, to balance computational efficiency and the effectiveness of historical information, The preferred setting is 5 10 minutes. If Too short, and it cannot capture the complete trend of fluctuations; if If the data is too long, it increases the interference of invalid historical data; : represents the time decay coefficient. In this embodiment, The preferred value range is 0.1. 0.5; : Represents the absolute value of the rate of change of voltage.

[0031] This formula accumulates the rate of voltage change over a past period through integration. Wherein, As a weighted function, the term increases with... With the current moment As the distance increases, the weight decreases exponentially. This indicates that the weight decreases exponentially with increasing distance from the current time. The more recent the high-frequency fluctuation, the greater its weight in determining the current system state. It is not a direct voltage value, but a scalar characterizing a physical state: when When the value approaches 0, it indicates that the power grid is in a "stable period"; when... A significant increase in the value indicates that the power grid is in a period of "severe fluctuations".

[0032] In practical implementation, the above integrals can be discretized and numerically calculated using the trapezoidal rule or Simpson's rule to adapt to computer processing.

[0033] S3. Generate instantaneous virtual boundary points based on wave fingerprints.

[0034] Traditional cubic spline interpolation requires waiting for the "next future low-frequency point" before calculating the current interval, resulting in real-time lag. This step utilizes the local fluctuation fingerprint obtained in step S2 to deduce the virtual boundary within the current low-frequency sampling interval, thereby achieving real-time calculation.

[0035] When the latest known low-frequency sampling point is However, before the next low-frequency point is reached, for the current high-frequency moment... (satisfy The virtual boundary prediction model is constructed as follows:

[0036] in, : This refers to the virtual low-frequency node data calculated for the current moment; : This is the historical correlation coefficient, obtained through regression analysis of historical high- and low-frequency data for this node, with an optimal range of 0.8. A value between 1 and 2 indicates the degree of synchronization between high- and low-frequency data changes; : indicates from From moment to the present moment The relative rate of change of high-frequency voltage; : This is a nonlinear correction term based on wave fingerprinting.

[0037] The core logic of this formula lies in: utilizing high-frequency data. The relative change is used as the basic increment, and the fluctuation fingerprint is superimposed. As a nonlinear correction.

[0038] When power grid fluctuations intensify (i.e.) When it increases, the correction term As it increases, the virtual boundary points... The predicted values ​​exhibit dynamic, non-linear growth.

[0039] In this way, the system pre-generates a "virtual anchor point" that conforms to the physical change trend before the next real low-frequency data arrives, thereby breaking the dependence on future data and realizing the real-time start of interpolation calculation.

[0040] S4. Calculate the dynamic confidence level and spline relaxation factor.

[0041] To address the issue of waveform distortion caused by traditional algorithms forcing the data to pass through every sampling point (including noise points), this step introduces a "soft constraint" mechanism by analyzing the deviation between the data and physical laws.

[0042] S4-1. Construct dynamic data reliability based on the deviation between the virtual boundary trend and the actual received low-frequency data. Suppose the low-frequency data received at a certain historical moment is... Its corresponding trend forecast value is The trust level is calculated as follows:

[0043] in, : This represents the standard deviation of the historical data for this node, reflecting the typical dispersion of the data; : This is the sensitivity constant, preferably set to 1.0. 2.0.

[0044] This formula is a normalization function. When the observed values Compared with the predicted value When the deviation is small (i.e., the molecule approaches 0), A value close to 1 indicates that the data is highly reliable; when the deviation is significantly greater than the historical standard deviation... When outliers or bad data appear, the denominator increases sharply. A value that rapidly approaches 0 indicates that the data has extremely low credibility.

[0045] S4-2. Mapped Spline Relaxation Factor. A relaxation factor that transforms confidence levels into spline interpolation. Construct the following mapping logic:

[0046] in, : Determines the compactness of the cubic spline function at the nodes; , : These represent the lower and upper limits of the relaxation factor, respectively. In this embodiment, A value of 0.01 can be used (high relaxation, curve passes smoothly). 100 can be selected (low relaxation, close to the curve). : is an adjustment coefficient used to control the slope of the hyperbolic tangent function.

[0047] when At higher levels (data conforms to physical trends). The function value approaches 1, at which point... The interpolation curve will closely follow the sampling point, preserving the true data characteristics.

[0048] when At lower levels (possibly noise). The function value decreases. Rapid decay approaching Under the influence of the relaxation factor, the interpolation curve will automatically and smoothly pass through the noisy region, rather than forcibly passing through it, thus avoiding the introduction of errors.

[0049] S5. Construct a real-time dynamic relaxation cubic spline model and determine the cross-section.

[0050] Combined with the virtual boundary in step S3 and the relaxation factor in step S4 An improved spline objective function is constructed. At the solution time... interpolation function When, minimize the following energy equation: ; Among them, the first item : Represents the data fitting term. It requires the interpolation curve to be as close as possible to the measurement points, but is subject to confidence level. Weighted control is applied. If the confidence level at a certain point is low, the weight of that item decreases, and the algorithm allows the curve to deviate from that point. Second item : Represents the smoothness constraint term. It penalizes the second derivative (i.e., curvature) of the curve. The smaller the value, the greater the weight of that item, forcing the curve to be smoother.

[0051] By solving the objective function, the interpolation results for low-frequency nodes at all high-frequency time points are obtained. Finally, the measured high-frequency data are... Interpolated data of low-frequency nodes obtained by calculation Time alignment and splicing are performed to form a complete operational profile of the entire power distribution network.

[0052] Examples of specific use cases: Scenario 1: 03:00-04:00 AM (Stable Operation Period) Operating condition description: During this period, the user load is stable, there is no photovoltaic output, and the high-frequency voltage collected by SCADA is... It fluctuated only slightly within a small range.

[0053] Parameter response: According to step S2, the rate of change of voltage Extremely small, resulting in fluctuations in the fingerprint after integration. .

[0054] Substituting into formula S3, the nonlinear correction term .

[0055] At this time, the virtual boundary It is mainly determined by the linear trend term of low-frequency data. The algorithm exhibits "strong damping" characteristics, effectively suppressing the interference of minor noise from nighttime measurements on the cross-section. The generated cross-section curve is smooth and stable, conforming to the static characteristics of nighttime load.

[0056] Scenario 2: 11:30 AM - 12:30 PM (Period of dramatic fluctuations) Operating condition description: During this period, the photovoltaic output is high and fluctuates frequently due to cloud cover, resulting in voltage... A dramatic jump occurred, occurring on the order of seconds.

[0057] Parameter response: Detected voltage change rate The local fluctuation fingerprint after integral calculation increases sharply. Rapidly rising (e.g.) ).

[0058] At this point, the nonlinear correction term The weight of the high-frequency trend term is significantly increased in the formula of step S3.

[0059] Virtual Boundary It is rapidly "pulled" to match the direction of high-frequency fluctuations. The algorithm exhibits "high sensitivity" characteristics, and the generated cross-sectional curve can accurately capture the voltage sag caused by a sudden drop in photovoltaic output, avoiding the "peak shaving and valley filling" phenomenon caused by excessive smoothing in traditional interpolation algorithms, and truly restoring the dynamic characteristics of the power grid.

[0060] This embodiment utilizes high-frequency data to construct local fluctuation fingerprints to capture physical change patterns, thereby deducing real-time virtual boundaries and solving the time lag problem of traditional interpolation relying on future data. Simultaneously, a soft constraint mechanism is constructed by introducing a dynamic relaxation factor based on data reliability, automatically filtering noise while preserving real fluctuation characteristics. This effectively overcomes the shortcomings of existing technologies in terms of poor noise resistance and real-time lag, achieving accurate and real-time reconstruction of distribution network operation sections.

[0061] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for determining the operating section of a distribution network based on multi-layer data fusion, characterized in that, include: Acquire multi-source measurement data that includes both high-frequency and low-frequency sampling data; Based on the high-frequency sampling data, a local fluctuation fingerprint is calculated within the current calculation window; the local fluctuation fingerprint is a weighted integral of the absolute rate of change of high-frequency voltage within the current calculation window, and the weight decreases exponentially as the time difference between the absolute rate of change and the current moment increases. When there are unreached subsequent low-frequency sampling points, a virtual boundary point is generated based on the latest low-frequency sampling data, the latest high-frequency sampling data, and the local fluctuation fingerprint at the current moment. The value of the virtual boundary point is positively correlated with the value of the latest low-frequency sampling data, the difference between the voltage value of the latest high-frequency sampling data and the high-frequency voltage value corresponding to the latest low-frequency sampling point, and the magnitude of the local fluctuation fingerprint. Calculate the dynamic confidence level of each low-frequency sampling data, wherein the dynamic confidence level is negatively correlated with the deviation between the measured and predicted values ​​of the corresponding low-frequency sampling data; generate a relaxation factor for cubic spline interpolation based on the dynamic confidence level, wherein the relaxation factor is positively correlated with the dynamic confidence level; The low-frequency sampling data and virtual boundary points are interpolated to obtain a low-frequency data curve. The interpolation is controlled by a relaxation factor to make the curve smoother at nodes with low confidence. The measured high-frequency sampling data and the interpolated low-frequency data curve are combined to obtain the distribution network operation section.

2. The method for determining the operating section of a power distribution network according to claim 1, characterized in that, The local fluctuation fingerprint The calculation formula is: ； in, Indicates the current time; express The high-frequency voltage value at any given time; The set calculation window length; This is the time decay coefficient.

3. The method for determining the operating section of a power distribution network according to claim 2, characterized in that, The method for calculating local fluctuation fingerprints also includes: The calculation window length The value range is 5 to 10 minutes; the time decay coefficient The value range is 0.1 to 0.

5.

4. The method for determining the operating section of a power distribution network according to claim 1, characterized in that, The virtual boundary point The calculation method is as follows: ； in, For the current moment Virtual boundary points, This is the latest low-frequency sampling data; The voltage value is the latest high-frequency sampling data; for The corresponding high-frequency voltage value; Historical correlation coefficient; This represents the local fluctuation fingerprint at the current moment.

5. The method for determining the operating section of a power distribution network according to claim 4, characterized in that, The historical correlation coefficient The value range is 0.8 to 1.

2.

6. The method for determining the operating section of a power distribution network according to claim 1, characterized in that, The dynamic trust level The calculation method is as follows: ； in, These are the measured values ​​corresponding to the low-frequency sampling data; These are predicted values ​​based on historical data trends. The standard deviation of the historical data for the corresponding node; This is the sensitivity constant.

7. The method for determining the operating section of a power distribution network according to claim 1 or 6, characterized in that, The method for generating the relaxation factor is as follows: ； in, It is a relaxation factor; , These are the lower and upper limits of the relaxation factor, respectively. This is the adjustment coefficient for the hyperbolic tangent function; For dynamic trust levels.

8. The method for determining the operating section of a power distribution network according to claim 7, characterized in that, The lower limit of the relaxation factor The value can be 0.01, with an upper limit. The value is 100.

9. The method for determining the operating section of a power distribution network according to claim 1, characterized in that, The cubic spline interpolation includes minimizing the objective function. The calculation method is as follows: ； in, For the interpolation function at the sampling point The value at; These are measured values ​​from low-frequency sampling data; Sampling points Dynamic trust level; Sampling points The relaxation factor; is the second derivative of the interpolation function.

10. The method for determining the operating section of a power distribution network according to claim 1, characterized in that, It also includes preprocessing the multi-source measurement data, including: cleaning and aligning the timestamps of each data point, and using the time axis of the high-frequency sampling data as the reference cross-sectional time axis; and marking data with format errors or missing data.

Images