Online self-adaptive generation method and system for power fluctuation of electric power system

Abstract

The invention discloses a power system power fluctuation online adaptive generation method and system, and the method comprises the steps: carrying out the feature extraction based on power grid historical operation data and weather data, and obtaining feature data; performing feature screening by taking the wind power fluctuation power, the photovoltaic fluctuation power and the load fluctuation power as target variables to screen out target features; determining an optimal clustering number based on the target features, carrying out clustering based on the optimal clustering number, and respectively determining a clustering group corresponding to each target variable; obtaining a real-time feature sample, calculating the Euclidean distance between the real-time feature sample and the clustering center of each clustering group, and determining a target clustering group corresponding to each target variable; calculating a power fluctuation interval of the real-time feature sample under a preset confidence coefficient based on the statistical features of the target variables in each target clustering group so as to determine a confidence interval; and determining upward power fluctuation disturbance and downward power fluctuation disturbance based on the confidence interval so as to automatically generate system fluctuation power.

Description

Technical Field

[0001] The present invention relates to the technical field of power system operation and control, and more particularly to a method and system for online adaptive generation of power fluctuations in a power system. Background Art

[0002] Power systems with a high proportion of renewable energy face significant challenges. On the one hand, load-side demand exhibits highly time-varying characteristics, with fluctuations and peak-to-valley differences significantly increasing with the large-scale integration of new power-consuming equipment. On the other hand, renewable energy generation, such as wind and photovoltaic power, is strongly coupled with meteorological conditions, resulting in random and intermittent output across multiple timescales. The combined uncertainty of both the source and the load leads to an exponential increase in the complexity of system power balance. Against this backdrop, effectively generating active power fluctuations has become a core foundation for ensuring the safe and economic operation of the power grid. Dispatching agencies urgently need to quantitatively assess power disturbance bounds at different timescales (from seconds to hours) to provide safety constraints for day-ahead and real-time multi-stage dispatching.

[0003] The existing static fault set construction method based on deterministic scenarios is difficult to characterize the continuous disturbance characteristics of the new energy-dominated system. Therefore, there is an urgent need to develop a dynamic disturbance scenario generation technology that integrates meteorological-time-power correlation characteristics to provide theoretical support for the new power system panoramic security defense system.

[0004] Therefore, an online adaptive generation method for power system power fluctuations is needed. Summary of the Invention

[0005] The present invention proposes a method and system for online adaptive generation of power system power fluctuations to solve the problem of how to online generate a comprehensive active power fluctuation range that can support the safe operation of the power grid, so as to be applicable to regional or provincial synchronous power grids with a new energy penetration rate exceeding 30%.

[0006] In order to solve the above problem, according to one aspect of the present invention, a method for online adaptive generation of power fluctuations in an electric power system is provided, characterized in that the method comprises:

[0007] Extract features based on historical operating data of wind power generation, photovoltaic power generation, and load power with time stamps of the power grid and weather data at the location of the power grid to obtain feature data;

[0008] Feature screening is performed with wind power fluctuation power, photovoltaic power fluctuation power and load power fluctuation power as target variables, and target features with the greatest impact on wind power fluctuation power, photovoltaic power fluctuation power and load power fluctuation power are screened out;

[0009] Determining the optimal number of clusters corresponding to different target variables based on the target features corresponding to different target variables, and performing clustering based on the optimal number of clusters to determine the cluster group corresponding to each target variable;

[0010] Performing feature extraction based on real-time grid operation data and weather data to obtain real-time feature samples, calculating the Euclidean distance between the real-time feature samples and the cluster center of each cluster group, and determining the target cluster group corresponding to each target variable based on the nearest Euclidean distance;

[0011] Calculating statistical features of the target variables corresponding to each target cluster group respectively, and calculating the power fluctuation interval of the real-time feature sample under a preset confidence level based on the statistical features to determine a confidence interval;

[0012] An upward power fluctuation disturbance and a downward power fluctuation disturbance are determined based on the confidence interval, and system fluctuation power is automatically generated based on the upward power fluctuation disturbance and the downward power fluctuation disturbance.

[0013] Preferably, the characteristic data includes three categories: power fluctuation characteristics, meteorological characteristics and time characteristics;

[0014] Among them, the power fluctuation characteristics include: wind power active power, photovoltaic active power, load active power, wind power fluctuation power, photovoltaic fluctuation power and load fluctuation power; the meteorological characteristics include: temperature, wind speed and irradiance; the time characteristics include: month, hour, day of the week, whether it is a weekend and whether it is a holiday.

[0015] Preferably, the method further comprises:

[0016] Before performing feature screening using wind power fluctuation power, photovoltaic power fluctuation power and load power fluctuation power as target variables, the feature data are standardized.

[0017] Preferably, the feature screening is performed using wind power fluctuation power, photovoltaic power fluctuation power and load power fluctuation power as target variables, and target features having a large impact on wind power fluctuation power, photovoltaic power fluctuation power and load power fluctuation power are screened out, including:

[0018] Based on the tree model, the features that have a great impact on wind power fluctuation are selected as target features;

[0019] Based on PCA principal component analysis, the features that have the greatest impact on photovoltaic power fluctuation are selected as target features;

[0020] Based on the tree model, recursive feature elimination is used to select features that have a great impact on load fluctuation power and use them as target features.

[0021] Preferably, the determining of the optimal number of clusters corresponding to the target variable based on the target features corresponding to different target variables includes:

[0022] Based on the target characteristics corresponding to wind power fluctuation power, photovoltaic fluctuation power and load fluctuation power as target variables, SSE curves are drawn respectively, and the elbow method is used for calculation to determine the optimal number of clusters corresponding to wind power fluctuation power, photovoltaic fluctuation power and load fluctuation power respectively.

[0023] Preferably, the method further comprises:

[0024] The real-time feature samples are standardized.

[0025] Preferably, calculating the power fluctuation interval of the real-time feature sample at a preset confidence level based on the statistical feature to determine the confidence interval includes:

[0026] Determining a preset reliability, and converting the preset reliability into a Z value according to a standard normal distribution table;

[0027] For any target variable, according to the power fluctuation mean and standard deviation of any target variable in the target cluster group corresponding to the target variable, the confidence interval range corresponding to the target variable under the preset confidence level is determined as follows: [μ i -Z·σ i ,μ i +Z·σ i ];

[0028] Among them, μ i and σ i are the power fluctuation mean and fluctuation standard deviation of the i-th target variable in the target cluster group corresponding to the i-th target variable; Z represents the Z value corresponding to the preset reliability.

[0029] Preferably, determining the upward power fluctuation disturbance and the downward power fluctuation disturbance based on the confidence interval comprises:

[0030]

[0031] Where ΔP up and ΔP down are upward power fluctuation disturbance and downward power fluctuation disturbance respectively; U load,Confidence_Level and L load,Confidence_Level are the upper and lower limits of the confidence interval corresponding to the load fluctuation power; U wind,Confidence_Level and L wind,Confidence_Level are the upper and lower limits of the confidence interval corresponding to the wind power fluctuation power; U solar,Confidence_Level and L solar,Confidence_Level are the upper and lower limits of the confidence interval corresponding to the photovoltaic fluctuation power.

[0032] According to another aspect of the present invention, there is provided an online adaptive generation system for power fluctuations in an electric power system, the system comprising:

[0033] a feature data acquisition unit, configured to extract features based on the historical operating data of wind power generation, photovoltaic power generation, and load power with time stamps of the power grid and weather data at the location of the power grid, thereby acquiring feature data;

[0034] A feature screening unit is used to perform feature screening using wind power fluctuation power, photovoltaic power fluctuation power and load power fluctuation power as target variables, and screen out target features that have a great impact on wind power fluctuation power, photovoltaic power fluctuation power and load power fluctuation power;

[0035] A clustering unit, configured to determine an optimal number of clusters corresponding to different target variables based on target features corresponding to different target variables, and to perform clustering based on the optimal number of clusters to determine a cluster group corresponding to each target variable;

[0036] a target cluster group determination unit, configured to perform feature extraction based on real-time grid operation data and weather data, obtain real-time feature samples, calculate the Euclidean distance between the real-time feature samples and the cluster center of each cluster group, and determine the target cluster group corresponding to each target variable based on the nearest Euclidean distance;

[0037] A confidence interval determination unit, configured to respectively calculate statistical features of the target variable corresponding to each target cluster group, and calculate a power fluctuation interval of the real-time feature sample under a preset confidence level based on the statistical features to determine a confidence interval;

[0038] The power fluctuation disturbance determining unit is configured to determine an upward power fluctuation disturbance and a downward power fluctuation disturbance based on the confidence interval, and automatically generate system fluctuation power based on the upward power fluctuation disturbance and the downward power fluctuation disturbance.

[0039] Preferably, the characteristic data includes three categories: power fluctuation characteristics, meteorological characteristics and time characteristics;

[0040] Among them, the power fluctuation characteristics include: wind power active power, photovoltaic active power, load active power, wind power fluctuation power, photovoltaic fluctuation power and load fluctuation power; the meteorological characteristics include: temperature, wind speed and irradiance; the time characteristics include: month, hour, day of the week, whether it is a weekend and whether it is a holiday.

[0041] Preferably, the system further comprises:

[0042] The standardization processing unit is used to perform standardization processing on the characteristic data before performing characteristic screening with wind power fluctuation power, photovoltaic fluctuation power and load fluctuation power as target variables respectively.

[0043] Preferably, the feature screening unit performs feature screening using wind power fluctuation power, photovoltaic power fluctuation power and load power fluctuation power as target variables, and screens out target features that have a great impact on wind power fluctuation power, photovoltaic power fluctuation power and load power fluctuation power, including:

[0044] Based on the tree model, the features that have a great impact on wind power fluctuation are selected as target features;

[0045] Based on PCA principal component analysis, the features that have the greatest impact on photovoltaic power fluctuation are selected as target features;

[0046] Based on the tree model, recursive feature elimination is used to select features that have a great impact on load fluctuation power and use them as target features.

[0047] Preferably, the clustering unit determines the optimal number of clusters corresponding to the target variable based on the target features corresponding to different target variables, including:

[0048] Based on the target characteristics corresponding to wind power fluctuation power, photovoltaic fluctuation power and load fluctuation power as target variables, SSE curves are drawn respectively, and the elbow method is used for calculation to determine the optimal number of clusters corresponding to wind power fluctuation power, photovoltaic fluctuation power and load fluctuation power respectively.

[0049] Preferably, the system further comprises:

[0050] The standardization processing unit is used to perform standardization processing on the real-time feature samples.

[0051] Preferably, the confidence interval determining unit calculates the power fluctuation interval of the real-time feature sample under a preset confidence level based on the statistical feature to determine the confidence interval, including:

[0052] Determining a preset reliability, and converting the preset reliability into a Z value according to a standard normal distribution table;

[0053] For any target variable, according to the power fluctuation mean and standard deviation of any target variable in the target cluster group corresponding to the target variable, the confidence interval range corresponding to the target variable under the preset confidence level is determined as follows: [μ i -Z·σ i ,μ i +Z·σ i ];

[0054] Among them, μ i and σi are the power fluctuation mean and fluctuation standard deviation of the i-th target variable in the target cluster group corresponding to the i-th target variable; Z represents the Z value corresponding to the preset reliability.

[0055] Preferably, the power fluctuation disturbance determining unit determines the upward power fluctuation disturbance and the downward power fluctuation disturbance based on the confidence interval, comprising:

[0056]

[0057] Where ΔP up and ΔP down are upward power fluctuation disturbance and downward power fluctuation disturbance respectively; U load,Confidence_Level and L load,Confidence_Level are the upper and lower limits of the confidence interval corresponding to the load fluctuation power; U wind,Confidence_Level and L wind,Confidence_Level are the upper and lower limits of the confidence interval corresponding to the wind power fluctuation power; U solar,Confidence_Level and L solar,Confidence_Level are the upper and lower limits of the confidence interval corresponding to the photovoltaic fluctuation power.

[0058] According to another aspect of the present invention, a computer-readable storage medium is provided, on which a computer program is stored. When the program is executed by a processor, any step of a method for online adaptive generation of power fluctuations in an electric power system is implemented.

[0059] According to another aspect of the present invention, the present invention provides an electronic device, including:

[0060] The computer-readable storage medium mentioned above; and one or more processors configured to execute the program in the computer-readable storage medium.

[0061] The present invention provides a method and system for online adaptive generation of power fluctuations in an electric power system, comprising: extracting features based on historical operating data of wind power generation power, photovoltaic power generation power, and load power with time stamps on a power grid and weather data at the location of the power grid to obtain feature data; performing feature screening using wind power fluctuation power, photovoltaic power fluctuation power, and load fluctuation power as target variables to screen out target features that have a significant impact on wind power fluctuation power, photovoltaic power fluctuation power, and load fluctuation power; determining the optimal number of clusters corresponding to different target variables based on the target features corresponding to different target variables, performing clustering based on the optimal number of clusters, and determining the cluster groups corresponding to each target variable. group; perform feature extraction based on the real-time operation data and weather data of the power grid, obtain real-time feature samples, calculate the Euclidean distance between the real-time feature samples and the cluster center of each cluster group, and determine the target cluster group corresponding to each target variable based on the nearest Euclidean distance; calculate the statistical characteristics of the target variables corresponding to each target cluster group respectively, and calculate the power fluctuation interval of the real-time feature samples under a preset confidence level based on the statistical characteristics to determine the confidence interval; determine the upward power fluctuation disturbance and the downward power fluctuation disturbance based on the confidence interval, and automatically generate the system fluctuation power based on the upward power fluctuation disturbance and the downward power fluctuation disturbance. The method of the present invention is designed for the power fluctuation problem of the power system, and can generate expected disturbances that meet the actual scenario, providing important support for power grid scheduling and safety and stability analysis. It can be widely used in power system scheduling, new energy power generation forecasting and load fluctuation analysis and other fields. In the context of the gradual increase in the proportion of new energy, it can effectively deal with the uncertainty problem brought about by load fluctuations, and help the efficient use of clean energy and the stable operation of the power grid. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] A more complete understanding of exemplary embodiments of the present invention may be obtained by referring to the following drawings:

[0063] Figure 1 Flowchart of an online adaptive generation method 100 for power system power fluctuation according to an embodiment of the present invention;

[0064] Figure 2 A general flow chart of an online adaptive generation process of power fluctuations in a power system according to an embodiment of the present invention;

[0065] Figure 3 (a), (b), and (c) are SSE curve diagrams of wind power fluctuation, photovoltaic fluctuation, and load fluctuation according to an embodiment of the present invention, respectively;

[0066] Figure 4 (a), (b), and (c) are respectively clustering effect diagrams of wind power fluctuation, photovoltaic fluctuation, and load fluctuation according to an embodiment of the present invention;

[0067] Figure 5 (a), (b), and (c) are respectively clustered probability distribution diagrams of wind power fluctuation, photovoltaic fluctuation, and load fluctuation according to an embodiment of the present invention;

[0068] Figure 6 A power fluctuation disturbance power map at different confidence levels throughout the day generated according to an embodiment of the present invention;

[0069] Figure 7 2 is a schematic structural diagram of an online adaptive generation system 700 for power system power fluctuations according to an embodiment of the present invention. DETAILED DESCRIPTION

[0070] Exemplary embodiments of the present invention will now be described with reference to the accompanying drawings. However, the present invention may be embodied in many different forms and is not limited to the embodiments described herein. These embodiments are provided to provide a thorough and complete disclosure of the present invention and to fully convey the scope of the present invention to those skilled in the art. The terminology used in the exemplary embodiments shown in the accompanying drawings is not intended to limit the present invention. In the accompanying drawings, identical elements are denoted by the same reference numerals.

[0071] Unless otherwise specified, the terms used herein (including technical terms) have the meanings commonly understood by those skilled in the art. In addition, it is understood that terms defined in commonly used dictionaries should be understood to have the same meanings as those in the context of the relevant fields, and should not be understood as idealized or overly formal meanings.

[0072] Figure 1 FIG. 1 is a flow chart of an online adaptive generation method 100 for power system power fluctuation according to an embodiment of the present invention. Figure 1 As shown, the online adaptive generation method for power system power fluctuations provided by the embodiment of the present invention is designed for the power fluctuation problem of the power system, can generate expected disturbances that conform to actual scenarios, and provide important support for power grid scheduling and security and stability analysis. It can be widely used in power system scheduling, new energy power generation forecasting, and load fluctuation analysis. In the context of the gradual increase in the proportion of new energy, it can effectively deal with the uncertainty problems caused by load fluctuations, and promote the efficient utilization of clean energy and the stable operation of the power grid. The online adaptive generation method 100 for power system power fluctuations provided by the embodiment of the present invention starts at step 101. In step 101, feature extraction is performed based on the historical operating data of wind power generation power, photovoltaic power generation power, and load power with time stamps of the power grid and the weather data of the location of the power grid to obtain feature data.

[0073] Preferably, the characteristic data includes three categories: power fluctuation characteristics, meteorological characteristics and time characteristics;

[0074] Among them, the power fluctuation characteristics include: wind power active power, photovoltaic active power, load active power, wind power fluctuation power, photovoltaic fluctuation power and load fluctuation power; the meteorological characteristics include: temperature, wind speed and irradiance; the time characteristics include: month, hour, day of the week, whether it is a weekend and whether it is a holiday.

[0075] In the present invention, multi-source data mainly includes historical operation data of the power system and weather data of the location of the power system. Specifically, it includes:

[0076] (1) Historical operating active power data of the power system

[0077] The present invention requires storage and analysis based on a large amount of historical data. Based on the D5000 dispatch automation platform, the real-time power values of wind power, photovoltaic power, and load of the target power grid are obtained from the .QS file in the real-time database; the historical data of wind power, photovoltaic power, and load of the target power grid with time stamps are obtained from the historical database, as shown in Table 1.

[0078] Table 1 Historical data of wind power, photovoltaic power and load with time stamp

[0079] Date_Time Value_Load Value_Wind Value_Solar 2022 / 1 / 1 14:00 62503.25 8919.25 3220.23 2022 / 1 / 1 14:05 62782.06 9079.34 3130.73 2022 / 1 / 1 14:10 62874.40 9434.02 3051.29 2022 / 1 / 1 14:15 63062.20 9719.09 2990.22 2022 / 1 / 1 14:20 63349.55 10013.81 2864.36 2022 / 1 / 1 14:25 63156.38 10402.01 2672.39 2022 / 1 / 1 14:30 63149.80 10592.15 2651.76 2022 / 1 / 1 14:35 63061.08 10794.23 2511.47 2022 / 1 / 1 14:40 63051.48 11068.21 2345.36 2022 / 1 / 1 14:45 63125.46 11148.15 2208.71 2022 / 1 / 1 14:50 63322.14 11240.25 2079.52 2022 / 1 / 1 14:55 63172.79 11366.27 1909.18 2022 / 1 / 1 15:00 63530.52 11510.89 1773.62

[0080] (2) Weather data at the power grid location

[0081] The present invention requires selecting several locations (longitude and latitude) that are representative of the weather conditions in the target grid's area based on its geographic location. Historical and real-time weather information, including temperature, wind speed, and radiation intensity, is obtained from sources such as the National Meteorological Information Center. The weather data is shown in Table 2.

[0082] Table 2 Meteorological data table with time stamp

[0083] Date_Time Temperature Windspeed Radiation 2022 / 1 / 1 6:00 -17.40 7.60 0.00 2022 / 1 / 1 7:00 -17.00 6.90 0.00 2022 / 1 / 1 8:00 -16.80 5.90 0.00 2022 / 1 / 1 9:00 -16.00 7.00 9.00 2022 / 1 / 1 10:00 -13.70 8.40 51.00 2022 / 1 / 1 11:00 -10.20 7.10 252.00 2022 / 1 / 1 12:00 -5.20 15.30 292.00 2022 / 1 / 1 13:00 -6.60 26.90 242.00 2022 / 1 / 1 14:00 -7.60 27.80 142.00 2022 / 1 / 1 15:00 -8.40 26.00 173.00

[0084] In the present invention, based on the historical wind power, photovoltaic, load power data and meteorological data of the power grid, the extracted feature data include: power fluctuation characteristics, meteorological characteristics and time characteristics, a total of three types of characteristics.

[0085] 1) Extraction of power and fluctuation characteristics

[0086] In the present invention, the extracted power and fluctuation characteristics are shown in Table 3.

[0087] Table 3 Power and fluctuation characteristics

[0088]

[0089] 2) Extraction of meteorological characteristics

[0090] In the present invention, local temperature, wind speed, and irradiance characteristics are extracted according to the area where the power grid is located (multiple locations can be sampled), as shown in Table 4.

[0091] Table 4 Meteorological characteristics

[0092] Serial number feature symbol meaning 1 temperature Temperature(t) The temperature or average temperature of the area where the power grid is located at time t 2 wind speed Windspeed(t) Wind speed or average wind speed in the area where the power grid is located at time t 3 irradiance Radiation(t) Irradiance or average irradiance in the area where the power grid is located at time t

[0093] 3) Temporal feature extraction

[0094] For time features, features such as month, hour, day of the week, whether it is a weekend, and whether it is a holiday can be extracted, as shown in Table 5.

[0095] Table 5 Time characteristics table

[0096] Serial number feature symbol meaning 1 month Month The month at time t 2 Hour Hour The hour at time t 3 Day of the week Day of Week The day of the week at time t 4 Is it the weekend? IsWeekend Is time t a weekend? 5 Is it a holiday? IsHoliday Whether time t is a holiday

[0097] In step 102, feature screening is performed using wind power fluctuation power, photovoltaic power fluctuation power and load power fluctuation power as target variables, and target features having a great impact on wind power fluctuation power, photovoltaic power fluctuation power and load power fluctuation power are screened out.

[0098] Preferably, the method further comprises:

[0099] Before performing feature screening using wind power fluctuation power, photovoltaic power fluctuation power and load power fluctuation power as target variables, the feature data are standardized.

[0100] Preferably, the feature screening is performed using wind power fluctuation power, photovoltaic power fluctuation power and load power fluctuation power as target variables, and target features having a large impact on wind power fluctuation power, photovoltaic power fluctuation power and load power fluctuation power are screened out, including:

[0101] Based on the tree model, the features that have a great impact on wind power fluctuation are selected as target features;

[0102] Based on PCA principal component analysis, the features that have the greatest impact on photovoltaic power fluctuation are selected as target features;

[0103] Based on the tree model, recursive feature elimination is used to select features that have a great impact on load fluctuation power and use them as target features.

[0104] In the present invention, the feature data is standardized and then feature selection is performed.

[0105] In the present invention, the data normalization process includes:

[0106] ① Missing value processing: linear interpolation is used to fill missing values;

[0107] ② Time feature coding: encode periodic time features into sine and cosine codes;

[0108] ③Perform Z-score standardization on feature data.

[0109] In the present invention, in order to accurately grasp the characteristics of power grid power fluctuation disturbance, wind power fluctuation power, photovoltaic power fluctuation power, and load power fluctuation power are respectively used as target variables to carry out feature correlation analysis for feature selection. Among them, wind power fluctuation is mainly closely related to wind speed, irradiance and other characteristics. In order to effectively capture these nonlinear relationships, feature analysis is carried out based on the tree model and the features that have the greatest impact on wind power fluctuation are screened out. Photovoltaic fluctuations mainly have strong collinearity with time features such as irradiance and hourly sine-cosine coding. By extracting the principal components through PCA dimensionality reduction, the comprehensive information of multiple features can be captured. Load fluctuations are mainly closely related to load power, time features (such as sine-cosine coding of the day of the week), etc. Based on recursive feature elimination (RFE), the feature subset is gradually optimized to screen out the features that have the greatest impact on load fluctuations as target features.

[0110] In step 103, the optimal number of clusters corresponding to different target variables is determined based on the target features corresponding to different target variables, and clustering is performed based on the optimal number of clusters to determine the cluster group corresponding to each target variable.

[0111] Preferably, the determining of the optimal number of clusters corresponding to the target variable based on the target features corresponding to different target variables includes:

[0112] Based on the target characteristics corresponding to wind power fluctuation power, photovoltaic fluctuation power and load fluctuation power as target variables, SSE curves are drawn respectively, and the elbow method is used for calculation to determine the optimal number of clusters corresponding to wind power fluctuation power, photovoltaic fluctuation power and load fluctuation power respectively.

[0113] Combine Figure 2 As shown, in the present invention, a fluctuating power cluster analysis is performed based on historical data to determine cluster groups with wind power fluctuating power, photovoltaic power fluctuating power, and load power fluctuating power as target variables, and statistical characteristics of the target variables in each cluster group are counted.

[0114] Specifically, it includes:

[0115] (1) Determine the optimal number of clusters

[0116] The Elbow Method is a commonly used method for determining the optimal number of clusters. Its core idea is to plot the Sum of Squared Errors (SSE) curve for different cluster numbers k and find the "elbow" point of the curve, where the rate of decrease of the SSE slows down significantly, as the optimal number of clusters.

[0117] For a given number of clusters k, the SSE is calculated as:

[0118]

[0119] Among them, C i represents all sample points in the i-th cluster; μ i represents the center point of the i-th cluster; x represents a sample point belonging to the i-th cluster.

[0120] As k increases, the SSE decreases. When k reaches a certain value, the rate of SSE decreases significantly, forming an "elbow." Plot the SSE versus k curve and observe the location of the "elbow." Select the k value corresponding to the "elbow" as the optimal number of clusters.

[0121] (2) Conducting Fluctuation Power Cluster Analysis

[0122] K-Means++ is an improved K-Means clustering initialization method that solves the problem in the traditional K-Means algorithm that randomly initializing cluster centers may lead to slow convergence or unstable results.

[0123] The clustering steps based on the K-Means++ algorithm are as follows:

[0124] 1) Initialize the first center point: randomly select a point from the data set as the first cluster center.

[0125] 2) Select subsequent center points: For each data sample, calculate its distance D(x) to the nearest selected center point, and select the subsequent center point with probability P(x)∝D(x).

[0126] 3) Assign samples: Assign each sample to the group to which the nearest center point belongs.

[0127] 4) Update the center point: Calculate the mean of each group and update it to the new center point.

[0128] 5) Iterative optimization: Repeat steps 2 and 3 until the center point no longer changes significantly or the maximum number of iterations is reached.

[0129] 6) Save clustering results: Save the normalizer, cluster centers, and cluster labels, and add the cluster labels to the original dataset.

[0130] (3) Analyze clustering results

[0131] For target variables such as wind power fluctuation power, photovoltaic power fluctuation power, and load power fluctuation power, the statistical characteristics of the target variables in the cluster group corresponding to each variable are calculated, including the mean, standard deviation, minimum value, and maximum value, and the results are stored.

[0132] In step 104, feature extraction is performed based on the real-time operation data and weather data of the power grid to obtain real-time feature samples, the Euclidean distance between the real-time feature samples and the cluster center of each cluster group is calculated, and the target cluster group corresponding to each target variable is determined based on the nearest Euclidean distance.

[0133] Preferably, the method further comprises:

[0134] The real-time feature samples are standardized.

[0135] In the present invention, in order to generate power fluctuations, it is also necessary to obtain the real-time data of the grid's weather-time-power and meteorological characteristics, date characteristics, and integrate them to form real-time samples; then load the normalizer saved in the process of fluctuating power clustering analysis, and perform preprocessing and standardization on the real-time samples similar to historical data to obtain real-time feature samples.

[0136] In this method, the real-time feature samples are identified by their clustering to determine the target clustering group. Specifically, for the standardized real-time data samples, the Euclidean distance between the real-time feature samples and the clustering results of wind power fluctuation, photovoltaic power fluctuation, and load power fluctuation is calculated. Finally, for the clustering results corresponding to any target variable, the clustering group with the smallest Euclidean distance is selected as the target clustering group for that target variable.

[0137] Among them, the Euclidean distance formula is: d i =||x-μ i ||.

[0138] In step 105, statistical features of the target variables corresponding to each target cluster group are calculated respectively, and the power fluctuation interval of the real-time feature sample under a preset confidence level is calculated based on the statistical features to determine a confidence interval.

[0139] Preferably, calculating the power fluctuation interval of the real-time feature sample at a preset confidence level based on the statistical feature to determine the confidence interval includes:

[0140] Determining a preset reliability, and converting the preset reliability into a Z value according to a standard normal distribution table;

[0141] For any target variable, according to the power fluctuation mean and standard deviation of any target variable in the target cluster group corresponding to the target variable, the confidence interval range corresponding to the target variable under the preset confidence level is determined as follows: [μ i -Z·σ i ,μ i +Z·σ i ];

[0142] Among them, μ i and σ i are the power fluctuation mean and fluctuation standard deviation of the i-th target variable in the target cluster group corresponding to the i-th target variable; Z represents the Z value corresponding to the preset reliability.

[0143] In the present invention, for each target cluster group corresponding to the target variable, confidence interval ranges at different confidence levels are generated according to the probability distribution of wind power fluctuation, photovoltaic fluctuation, and load fluctuation.

[0144] For any target variable, the process of generating a confidence interval includes:

[0145] According to the standard normal distribution table, convert the confidence level into the corresponding Z value;

[0146]

[0147] Then, the confidence interval is: [μ i -Z·σ i ,μ i +Z·σ i ],

[0148] Among them, μ i and σ i are the power fluctuation mean and fluctuation standard deviation of the i-th target variable in the target cluster group corresponding to the i-th target variable; Z represents the Z value corresponding to the preset reliability.

[0149] In step 106, an upward power fluctuation disturbance and a downward power fluctuation disturbance are determined based on the confidence interval, and system fluctuation power is automatically generated based on the upward power fluctuation disturbance and the downward power fluctuation disturbance.

[0150] Preferably, determining the upward power fluctuation disturbance and the downward power fluctuation disturbance based on the confidence interval comprises:

[0151]

[0152] Where ΔP up and ΔP down are upward power fluctuation disturbance and downward power fluctuation disturbance respectively; U load,Confidence_Level and Lload,Confidence_Level are the upper and lower limits of the confidence interval corresponding to the load fluctuation power; U wind,Confidence_Level and L wind,Confidence_Level are the upper and lower limits of the confidence interval corresponding to the wind power fluctuation power; U solar,Confidence_Level and L solar,Confidence_Level are the upper and lower limits of the confidence interval corresponding to the photovoltaic fluctuation power.

[0153] The confidence interval is a range for estimating a population parameter based on sample data. For a normal distribution, the upper and lower bounds of the confidence interval (denoted as L and U, respectively) can be calculated using the following formula:

[0154] L=μ-Z·σ

[0155] U=μ+Z·σ

[0156] According to the above analysis, we can obtain the probability distribution of the cluster to which the current data sample belongs in the wind power fluctuation clustering, photovoltaic fluctuation clustering, and load fluctuation clustering results. Taking the confidence level Confidence_Level = 99.9% as an example, we can obtain the confidence intervals [L wind,99.9％ ,U wind,99.9％ ]、[L solar,99.9％ ,U solar,99.9％ ]、[L load,99.9％ ,U load,99.9％ ].

[0157] Strictly speaking, the power fluctuation disturbance of the power system is the sum of the load, wind power and photovoltaic power fluctuations (the power increase direction is defined as positive), and the expression is:

[0158] ΔP=ΔP load +ΔP wind +ΔP solar

[0159] Combined with the actual operation of the power grid and considering the reservation of a certain margin for online safety analysis application, the present invention sets the power fluctuation expected disturbance to be

[0160] ΔP=|ΔP load |+|ΔP wind |+|ΔP solar |≥|ΔP load +ΔP wind +ΔP solar |

[0161] Specifically, the upward power fluctuation disturbance and the downward power fluctuation disturbance are respectively:

[0162]

[0163] Where ΔP up and ΔPdown are upward power fluctuation disturbance and downward power fluctuation disturbance respectively; U load,Confidence_Level and L load,Confidence_Level are the upper and lower limits of the confidence interval corresponding to the load fluctuation power; U wind,Confidence_Level and L wind,Confidence_Level are the upper and lower limits of the confidence interval corresponding to the wind power fluctuation power; U solar,Confidence_Level and L solar,Confidence_Level are the upper and lower limits of the confidence interval corresponding to the photovoltaic fluctuation power.

[0164] In summary, based on the fluctuating power of wind power, photovoltaic power, and load, combined with meteorological and temporal characteristics, a comprehensive fluctuation range of expected power system power fluctuations was generated. This comprehensive fluctuation range can support the power grid in carrying out online safety warnings, online verification of startup methods, and optimization of reserve capacity. It also provides a disturbance input basis for minute-to-hour power generation and consumption balance analysis and dispatch control.

[0165] The effects of the present invention are:

[0166] (1) Innovation in disturbance generation methods

[0167] A method for generating anticipated disturbances suitable for time scales from minutes to hours is proposed, which provides a high-quality disturbance set basis for long-time scale safety analysis.

[0168] (2) Multi-feature fusion improves accuracy

[0169] The introduction of meteorological features (such as temperature, wind speed, radiation) and time features (such as hour, week, month) significantly enhances the accuracy of online generation of disturbance morphology.

[0170] (3) Multi-dimensional decision support

[0171] A fluctuation range with different confidence levels (such as 99%, 99.9%, etc.) is generated for each group, providing a quantitative basis for risk assessment and scientific decision-making of the power system.

[0172] (4) Identification of key factors

[0173] Through in-depth analysis of group characteristics, the key driving factors affecting power system fluctuations (such as weather conditions, time periods, etc.) can be effectively identified to provide guidance for system optimization.

[0174] (5) Wide application value

[0175] It can be widely used in power system scheduling, renewable energy generation forecasting, and load fluctuation analysis. As the proportion of renewable energy gradually increases, it can effectively address the uncertainty caused by load fluctuations, helping to achieve efficient utilization of clean energy and stable grid operation.

[0176] The following examples illustrate the embodiments of the present invention.

[0177] In an embodiment of the present invention, the historical operating data of a certain actual regional power grid in 2022 is used as an example of the above-mentioned scheme to illustrate the effectiveness of the present invention.

[0178] Specifically, the online adaptive generation process of power system power fluctuations includes:

[0179] (1) Multi-source data fusion and feature engineering

[0180] The historical data of wind power, photovoltaic power and load from 2022-01-01 00:05:00 to 2022-12-30 23:55:00 are used to integrate meteorological data and time characteristics, extract data features and perform standardization.

[0181] After processing, the historical dataset integrating weather-time-power correlation characteristics includes the following characteristic variables:

[0182] 1) Wind power, photovoltaic power, load power and its fluctuating power, a total of 6 power characteristic variables;

[0183] 2) Set up one meteorological data (including temperature, wind speed, and irradiance) sampling point in each of the four provinces within the regional power grid, for a total of 12 meteorological characteristic variables;

[0184] 3) In terms of time characteristics, in addition to the features of "whether it is a weekend" and "whether it is a holiday", sine and cosine encoding is performed on the periodic features (month, hour, day of the week), and a total of 8 time feature variables are obtained.

[0185] Feature correlation analysis was conducted using wind power fluctuation, photovoltaic power fluctuation, and load power fluctuation as target variables. To reduce the dimensionality of the dataset, the meteorological data was averaged, with average temperature, average wind speed, and average irradiance as features.

[0186] The feature selection results are as follows: based on the tree model, 5 features strongly correlated with wind power fluctuation power were selected, as shown in Table 6; based on PCA principal component analysis, 6 features strongly correlated with photovoltaic power fluctuation power were selected, as shown in Table 7; based on the recursive feature elimination of the tree model, 8 features strongly correlated with load power fluctuation were selected, as shown in Table 8.

[0187] Table 6 Wind power fluctuation power selection characteristics

[0188] Serial number feature symbol Remark Wind power fluctuation ΔP_Wind Target variable 1 Regional average temperature Avg_Temperature 2 Wind power Value_Wind 3 Regional average wind speed Avg_Windspeed 4 Regional average irradiance Avg_Radiation 5 Hour sine code Hour_sin

[0189] Table 7 Photovoltaic Fluctuation Power Selection Characteristics

[0190]

[0191]

[0192] Table 8 Load Fluctuation Power Selection Characteristics

[0193] Serial number feature symbol Remark Load fluctuation power ΔP_Load Target variable 1 Load power Value_Load 2 Hour sine code Hour_sin 3 Sine encoding of the day of the week DayOfWeek_sin 4 Hour cosine encoding Hour_cos 5 Month cosine code Month_cos 6 Regional average wind speed Avg_Windspeed 7 Regional average irradiance Avg_Radiation 8 Regional average temperature Avg_Temperature

[0194] (2) Determining the optimal number of clusters and K-Means++ clustering

[0195] 1) Determine the optimal number of clusters

[0196] After selecting the main features, the wind power fluctuation power, photovoltaic power fluctuation power and load power fluctuation power are used as target variables to draw SSE curves, as shown in the following example: Figure 3 As shown in (a), (b), and (c), according to the elbow rule, the optimal clustering number of wind power fluctuation power, photovoltaic power fluctuation power, and load power fluctuation power is 4.

[0197] 2) Conducting Fluctuation Power Cluster Analysis

[0198] Clustering is performed based on the K-Means++ algorithm. After completion, the normalizer, cluster center, and cluster label are saved and the cluster label is added to the original dataset. The clustering results are as follows: Figure 4 As shown in (a), (b) and (c).

[0199] 3) Analyze clustering results

[0200] For target variables such as wind power fluctuation power, photovoltaic power fluctuation power, and load power fluctuation power, the statistical characteristics (mean, standard deviation, minimum, and maximum) of the target variables in the cluster group are calculated for each variable, and the results are stored. The statistical characteristic results are as follows: Figure 5 As shown in (a), (b) and (c).

[0201] (3) Online generation of power fluctuation prediction disturbances

[0202] 1) Real-time data generation of power grid weather, time and power

[0203] Data from 00:05:00 on December 31, 2022, to 23:55:00 on December 31, 2022, was used as real-time operational data. During model development, data from December 31, 2022, was not included in training to prevent future samples from leaking into the model data. Real-time grid data was loaded, meteorological and date features were obtained, and integrated to form real-time samples. The normalizer saved during the fluctuating power cluster analysis was also loaded, and the real-time samples were preprocessed and normalized similarly to historical data.

[0204] 2) Identification of the group to which the current real-time sample point belongs

[0205] Taking the data sample at 00:05:00 on December 31, 2022 as an example, we analyze the groups in the wind power fluctuation power clustering results, photovoltaic power fluctuation power clustering results, and load fluctuation power clustering results in turn. The results are as follows:

[0206] Processing Wind Fluctuations:

[0207] Closest Cluster Index:0

[0208] Confidence Interval(99.9%):[-1094.79,1025.92]

[0209] Processing Solar Fluctuations:

[0210] Closest Cluster Index:1

[0211] Confidence Interval(99.9%):[-819.74,722.19]

[0212] Processing Load Fluctuations:

[0213] Closest Cluster Index:2

[0214] Confidence Interval(99.9%):[-1688.11,1709.14]

[0215] 3) Calculation of power fluctuation range at the current sample point with 99.9% confidence

[0216] According to the probability distribution of wind power fluctuation, photovoltaic fluctuation, and load fluctuation of the group, the fluctuation range at confidence level Confidence_Level = 99.9% is calculated, and the distribution is [-1094.79, 1025.92], [-819.74, 722.19], and [-1688.11, 1709.14].

[0217] 4) According to the online generation and calculation method of power fluctuation expected disturbance proposed by the present invention,

[0218] ΔP up =|U load,Confidence_Level |+|L wind,Confidence_Level |+|L solar,Confidence_Level |

[0219] ΔP down =|Lload,Confidence_Level |+|U wind,Confidence_Level |+|U solar,Confidence_Level |

[0220] Substituting the fluctuation range results under Confidence_Level = 99.9%, we can get:

[0221]

[0222] Finally, the aforementioned scheme is used to verify the power fluctuation at different confidence levels for a large amount of historical actual operation data. The power fluctuation expected disturbance generated by the scheme proposed in the present invention can cover the actual power fluctuation disturbance power of the power grid. The power fluctuation disturbance value determined by the present invention does not deviate seriously from the actual power curve, which can effectively avoid the excessive configuration of standby power such as adjustable power supply, thereby improving the economy of power grid operation. Figure 6 shown.

[0223] Figure 7 FIG. 7 is a schematic diagram of the structure of an online adaptive generation system 700 for power system power fluctuations according to an embodiment of the present invention. Figure 7 As shown, the online adaptive generation system 700 of power system power fluctuations provided by the embodiment of the present invention includes: a feature data acquisition unit 701, a feature screening unit 702, a clustering unit 703, a target cluster group determination unit 704, a confidence interval determination unit 705 and a power fluctuation disturbance determination unit 706.

[0224] Preferably, the feature data acquisition unit 701 is used to extract features based on the historical operating data of wind power generation power, photovoltaic power generation power and load power with time stamps of the power grid and weather data of the location of the power grid to acquire feature data.

[0225] Preferably, the characteristic data includes three categories: power fluctuation characteristics, meteorological characteristics and time characteristics;

[0226] Among them, the power fluctuation characteristics include: wind power active power, photovoltaic active power, load active power, wind power fluctuation power, photovoltaic fluctuation power and load fluctuation power; the meteorological characteristics include: temperature, wind speed and irradiance; the time characteristics include: month, hour, day of the week, whether it is a weekend and whether it is a holiday.

[0227] Preferably, the feature screening unit 702 is used to perform feature screening using wind power fluctuation power, photovoltaic power fluctuation power and load power fluctuation power as target variables, and screen out target features that have a great impact on wind power fluctuation power, photovoltaic power fluctuation power and load power fluctuation power.

[0228] Preferably, the system further comprises:

[0229] The standardization processing unit is used to perform standardization processing on the characteristic data before performing characteristic screening with wind power fluctuation power, photovoltaic fluctuation power and load fluctuation power as target variables respectively.

[0230] Preferably, the feature screening unit 703 performs feature screening using wind power fluctuation power, photovoltaic power fluctuation power, and load power fluctuation power as target variables, and screens out target features that have a great impact on wind power fluctuation power, photovoltaic power fluctuation power, and load power fluctuation power, including:

[0231] Based on the tree model, the features that have a great impact on wind power fluctuation are selected as target features;

[0232] Based on PCA principal component analysis, the features that have the greatest impact on photovoltaic power fluctuation are selected as target features;

[0233] Based on the tree model, recursive feature elimination is used to select features that have a great impact on load fluctuation power and use them as target features.

[0234] Preferably, the clustering unit 703 is configured to determine the optimal number of clusters corresponding to different target variables based on target features corresponding to different target variables, and perform clustering based on the optimal number of clusters to determine the cluster group corresponding to each target variable.

[0235] Preferably, the clustering unit 703 determines the optimal number of clusters corresponding to the target variable based on the target features corresponding to different target variables, including:

[0236] Based on the target characteristics corresponding to wind power fluctuation power, photovoltaic fluctuation power and load fluctuation power as target variables, SSE curves are drawn respectively, and the elbow method is used for calculation to determine the optimal number of clusters corresponding to wind power fluctuation power, photovoltaic fluctuation power and load fluctuation power respectively.

[0237] Preferably, the target cluster group determination unit 704 is used to perform feature extraction based on real-time operation data and weather data of the power grid, obtain real-time feature samples, calculate the Euclidean distance between the real-time feature samples and the cluster center of each cluster group, and determine the target cluster group corresponding to each target variable based on the nearest Euclidean distance.

[0238] Preferably, the system further comprises:

[0239] The standardization processing unit is used to perform standardization processing on the real-time feature samples.

[0240] Preferably, the confidence interval determination unit 705 is used to respectively calculate the statistical characteristics of the target variable corresponding to each target cluster group, and calculate the power fluctuation interval of the real-time feature sample under a preset confidence level based on the statistical characteristics to determine the confidence interval.

[0241] Preferably, the confidence interval determining unit 705 calculates the power fluctuation interval of the real-time feature sample under a preset confidence level based on the statistical feature to determine the confidence interval, including:

[0242] Determining a preset reliability, and converting the preset reliability into a Z value according to a standard normal distribution table;

[0243] For any target variable, according to the power fluctuation mean and standard deviation of any target variable in the target cluster group corresponding to the target variable, the confidence interval range corresponding to the target variable under the preset confidence level is determined as follows: [μ i -Z·σ i ,μ i +Z·σ i ];

[0244] Among them, μ i and σ i are the power fluctuation mean and fluctuation standard deviation of the i-th target variable in the target cluster group corresponding to the i-th target variable; Z represents the Z value corresponding to the preset reliability.

[0245] Preferably, the power fluctuation disturbance determining unit 706 is configured to determine an upward power fluctuation disturbance and a downward power fluctuation disturbance based on the confidence interval, and automatically generate system fluctuation power based on the upward power fluctuation disturbance and the downward power fluctuation disturbance.

[0246] Preferably, the power fluctuation disturbance determining unit 706 determines the upward power fluctuation disturbance and the downward power fluctuation disturbance based on the confidence interval, including:

[0247]

[0248] Where ΔP up and ΔP down are upward power fluctuation disturbance and downward power fluctuation disturbance respectively; U load,Confidence_Level and L load,Confidence_Level are the upper and lower limits of the confidence interval corresponding to the load fluctuation power; U wind,Confidence_Level and L wind,Confidence_Level are the upper and lower limits of the confidence interval corresponding to the wind power fluctuation power; U solar,Confidence_Level and L solar,Confidence_Level are the upper and lower limits of the confidence interval corresponding to the photovoltaic fluctuation power.

[0249] The online adaptive generation system 700 of power system power fluctuations in the embodiment of the present invention corresponds to the online adaptive generation method 100 of power system power fluctuations in another embodiment of the present invention, and will not be described in detail here.

[0250] According to another aspect of the present invention, a computer-readable storage medium is provided, on which a computer program is stored. When the program is executed by a processor, any step of a method for online adaptive generation of power fluctuations in an electric power system is implemented.

[0251] According to another aspect of the present invention, the present invention provides an electronic device, including:

[0252] The computer-readable storage medium described above; and

[0253] One or more processors are configured to execute the program in the computer-readable storage medium.

[0254] The present invention has been described with reference to a few embodiments. However, it is apparent to those skilled in the art that other embodiments than the one disclosed above are equally within the scope of the present invention.

[0255] Generally, all terms used in this disclosure are to be interpreted according to their ordinary meaning in the art, unless explicitly defined otherwise herein. All references to "a / the / the [device, component, etc.]" are to be interpreted openly as referring to at least one instance of the device, component, etc., unless explicitly stated otherwise. The steps of any method disclosed herein do not necessarily need to be performed in the exact order disclosed, unless explicitly stated otherwise.

[0256] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0257] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the steps in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0258] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0259] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 The steps for the function specified in one or more boxes.

[0260] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered by the scope of protection of the present invention.

Claims

1. A method for online adaptive generation of power fluctuations in an electric power system, characterized in that: The method comprises: Extract features based on historical operating data of wind power generation, photovoltaic power generation, and load power with time stamps of the power grid and weather data of the location of the power grid to obtain feature data; Feature screening is performed with wind power fluctuation power, photovoltaic power fluctuation power and load power fluctuation power as target variables, and target features with the greatest impact on wind power fluctuation power, photovoltaic power fluctuation power and load power fluctuation power are screened out; Determining the optimal number of clusters corresponding to different target variables based on the target features corresponding to different target variables, and performing clustering based on the optimal number of clusters to determine the cluster group corresponding to each target variable; Performing feature extraction based on real-time grid operation data and weather data to obtain real-time feature samples, calculating the Euclidean distance between the real-time feature samples and the cluster center of each cluster group, and determining the target cluster group corresponding to each target variable based on the nearest Euclidean distance; Calculating statistical features of the target variables corresponding to each target cluster group respectively, and calculating the power fluctuation interval of the real-time feature sample under a preset confidence level based on the statistical features to determine a confidence interval; An upward power fluctuation disturbance and a downward power fluctuation disturbance are determined based on the confidence interval, and system fluctuation power is automatically generated based on the upward power fluctuation disturbance and the downward power fluctuation disturbance.

2. The method according to claim 1, characterized in that The characteristic data includes three categories: power fluctuation characteristics, meteorological characteristics and time characteristics; Among them, the power fluctuation characteristics include: wind power active power, photovoltaic active power, load active power, wind power fluctuation power, photovoltaic fluctuation power and load fluctuation power; the meteorological characteristics include: temperature, wind speed and irradiance; the time characteristics include: month, hour, day of the week, whether it is a weekend and whether it is a holiday.

3. The method according to claim 1, characterized in that The method further comprises: Before performing feature screening using wind power fluctuation power, photovoltaic power fluctuation power and load power fluctuation power as target variables, the feature data are standardized.

4. The method according to claim 1, wherein The feature screening is performed using wind power fluctuation power, photovoltaic power fluctuation power and load power fluctuation power as target variables, and target features with a large impact on wind power fluctuation power, photovoltaic power fluctuation power and load power fluctuation power are screened out, including: Based on the tree model, the features that have a great impact on wind power fluctuation are selected as target features; Based on PCA principal component analysis, the features that have the greatest impact on photovoltaic power fluctuation are selected as target features; Based on the tree model, recursive feature elimination is used to select features that have a great impact on load fluctuation power and use them as target features.

5. The method according to claim 1, characterized in that The determining of the optimal number of clusters corresponding to the target variable based on the target features corresponding to different target variables includes: Based on the target characteristics corresponding to wind power fluctuation power, photovoltaic fluctuation power and load fluctuation power as target variables, SSE curves are drawn respectively, and the elbow method is used for calculation to determine the optimal number of clusters corresponding to wind power fluctuation power, photovoltaic fluctuation power and load fluctuation power respectively.

6. The method according to claim 1, wherein The method further comprises: The real-time feature samples are standardized.

7. The method according to claim 1, characterized in that Calculating a power fluctuation interval of the real-time feature sample at a preset confidence level based on the statistical feature to determine a confidence interval includes: Determining a preset reliability, and converting the preset reliability into a Z value according to a standard normal distribution table; For any target variable, according to the power fluctuation mean and standard deviation of any target variable in the target cluster group corresponding to the target variable, the confidence interval range corresponding to the target variable under the preset confidence level is determined as follows: [μ i -Z·σ i ,μ i +Z·σ i ]; Among them, μ i and σ i are the power fluctuation mean and fluctuation standard deviation of the i-th target variable in the target cluster group corresponding to the i-th target variable; Z represents the Z value corresponding to the preset reliability.

8. The method according to claim 1, characterized in that Determining an upward power fluctuation disturbance and a downward power fluctuation disturbance based on the confidence interval includes: Where ΔP up and ΔP down are upward power fluctuation disturbance and downward power fluctuation disturbance respectively; U load,Confidence_Level and L load,Confidence_Level are the upper and lower limits of the confidence interval corresponding to the load fluctuation power; U wind,Confidence_Level and L wind,Confidence_Level are the upper and lower limits of the confidence interval corresponding to the wind power fluctuation power; U solar,Confidence_Level and L solar,Confidence_Level are the upper and lower limits of the confidence interval corresponding to the photovoltaic fluctuation power.

9. An online adaptive generation system for power fluctuations in an electric power system, characterized in that: The system comprises: a feature data acquisition unit, configured to extract features based on the historical operating data of wind power generation, photovoltaic power generation, and load power with time stamps of the power grid and weather data at the location of the power grid, thereby acquiring feature data; A feature screening unit is used to perform feature screening using wind power fluctuation power, photovoltaic power fluctuation power and load power fluctuation power as target variables, and screen out target features that have a great impact on wind power fluctuation power, photovoltaic power fluctuation power and load power fluctuation power; A clustering unit, configured to determine an optimal number of clusters corresponding to different target variables based on target features corresponding to different target variables, and to perform clustering based on the optimal number of clusters to determine a cluster group corresponding to each target variable; a target cluster group determination unit, configured to perform feature extraction based on real-time grid operation data and weather data, obtain real-time feature samples, calculate the Euclidean distance between the real-time feature samples and the cluster center of each cluster group, and determine the target cluster group corresponding to each target variable based on the nearest Euclidean distance; A confidence interval determination unit, configured to respectively calculate statistical features of the target variable corresponding to each target cluster group, and calculate a power fluctuation interval of the real-time feature sample under a preset confidence level based on the statistical features to determine a confidence interval; The power fluctuation disturbance determining unit is configured to determine an upward power fluctuation disturbance and a downward power fluctuation disturbance based on the confidence interval, and automatically generate system fluctuation power based on the upward power fluctuation disturbance and the downward power fluctuation disturbance.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

11. An electronic device, characterized in that: include: The computer-readable storage medium of claim 10; as well as One or more processors are configured to execute the program in the computer-readable storage medium.

Images