Multi-time-scale wind-light-load extreme scenario extraction method and system and storage medium

Abstract

The application discloses a kind of multi-time scale wind-light-load extreme scene extraction method, system and storage medium;The method comprises: obtaining the time series data of each node in power grid in horizontal year wind-light-load as scene data, to constitute time series scene;The time series scene is clustered in time dimension, and a plurality of typical wind-light-load operating scenarios based on time correlation and geographical distribution characteristics combination are obtained;Based on each typical wind-light-load operating scenario, the time series scene farthest from the distance clustering center corresponding to the typical wind-light-load operating scenario is used as the extreme scene corresponding to the typical wind-light-load operating scenario.The application considers the time correlation and geographical distribution characteristics between wind-light-load by weighted clustering method, to identify the extreme operating scenario of system, to evaluate the stability and safety of system planning, to assist personnel to evaluate the operating safety of system planning scheme and carry out power grid stability scenario analysis.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power system scenario extraction, in particular to a multi-time-scale wind-solar-load extreme scenario extraction method, system and storage medium. BACKGROUND

[0002] Under the background of deepening consensus on global climate change and promoting carbon emission reduction, the utilization rate of renewable energy such as wind energy and solar energy has been significantly improved. Although these energy forms have less impact on the environment, their power generation has great volatility, which increases the complexity of power system operation. In addition, the volatility of power demand, influenced by multiple factors such as seasonality, climate conditions and socio-economic activities, together with the uncertainty of renewable energy supply, challenges the stable operation of the power grid. With the increasing frequency and intensity of extreme climate events caused by climate change, these events may have a significant impact on power supply and demand balance. Therefore, to maintain the safety and reliability of the power system, identifying and preparing for extreme scenarios in wind and solar power generation and power demand is crucial to ensure the stability and response capability of the power system when facing uncertainty challenges.

[0003] Existing research on wind-solar-load extreme scenario extraction mainly focuses on using statistical methods, machine learning techniques and system simulation to predict and analyze extreme changes in renewable energy generation (especially wind and solar energy) and power load. These studies aim to improve the adaptability and resilience of power systems to renewable energy fluctuations and uncertainties, and are an important part of the power system reliability and stability research field. However, existing models have limited ability to handle the uncertainty and variability of extreme events, making it difficult to fully cover all potential extreme situations. The generalization ability of the model is also a problem, as different regional climates and specific conditions of the power system will result in the model being unable to be directly applied. SUMMARY

[0004] In view of the above deficiencies of the prior art, the purpose of the present application is to provide a multi-time-scale wind-solar-load extreme scenario extraction method, which studies the determination method of wind-solar-load clustering feature vectors and extreme scenario judgment basis for extreme scenarios such as new energy long-term low output, long-term high output and load surge or sudden drop, and studies the wind-solar-load extreme scenario extraction method based on extreme value extrapolation for power supply and consumption scenarios.

[0005] To solve the above technical problems, the present application adopts the following technical solutions:

[0006] A multi-time-scale wind-solar-load extreme scenario extraction method, comprising the following steps:

[0007] Obtain the hourly time series data of each node in the power grid in the horizontal year wind-solar-load as scenario data to form a time series scenario;

[0008] clustering the time sequence scenarios in the time dimension to obtain a plurality of typical wind-light-load operation scenarios based on a combination of time correlation and geographical distribution characteristics;

[0009] based on each typical wind-light-load operation scenario, taking a time sequence scenario farthest from a distance clustering center corresponding to the typical wind-light-load operation scenario as an extreme scenario corresponding to the typical wind-light-load operation scenario.

[0010] As a preferred solution, the clustering of the time sequence scenarios in the time dimension to obtain a plurality of typical wind-light-load operation scenarios based on a combination of time correlation and geographical distribution characteristics specifically includes:

[0011] According to the time sequence data of each wind-light-load variable in the time sequence scenario, the wind-light-load variable is weighted, and the wind-light-load variable includes wind power, photovoltaic and load of each node in the power grid;

[0012] According to the weight of the wind-light-load variable, the time sequence scenarios are clustered based on weighted Euclidean distance to obtain a plurality of typical wind-light-load operation scenarios based on a combination of time correlation and geographical distribution characteristics.

[0013] As a preferred solution, the weighting of the wind-light-load variable according to the time sequence data of each wind-light-load variable in the time sequence scenario specifically includes:

[0014] According to the time sequence data of each wind-light-load variable in the time sequence scenario, the wind-light-load variable is weighted using an entropy weight method to obtain an objective weight of each variable;

[0015] According to the capacity of each wind-light-load variable, a subjective weight of each variable is calculated;

[0016] The objective weight and the subjective weight of each variable are weighted and averaged to obtain a combined weight of each variable, and the combined weight is taken as the weight of each wind-light-load variable.

[0017] As a preferred solution, the wind-light-load variable weight is determined using an entropy weight method according to the time sequence data of each wind-light-load variable in the time sequence scenario to obtain an objective weight of each variable, and the specific way is:

[0018] Suppose there are T time sequence scenario objects in the time dimension for clustering, each of the time sequence scenarios contains N WD +N Load +N PV dimensional variables, and each variable is defined as a characteristic index. The entropy value of each characteristic index is calculated, and the objective weight of each variable is determined according to the entropy value of each characteristic index. The method for obtaining the objective weight of each variable is as follows:

[0019]

[0020] In the formula, H i Let f be the entropy value of the i-th feature index, n be the total dimension of the temporal variables for each scene, T be the total number of objects in the temporal scene, and f be the entropy value of the i-th feature index. ij Let s be the proportion of the j-th temporal scene object at the i-th feature index. ij For the data of the variable, p i N is the entropy weight of the i-th feature index. WD N Load N PV These are the dimensional variables for each time series scenario.

[0021] As a preferred embodiment, the specific method for obtaining the combined weight of each variable by weighting the objective and subjective weights is as follows:

[0022]

[0023] In the formula, w i p represents the combined weights of the variables. i q is the objective weight of the i-th feature index determined by the entropy weight method. i The subjective weight of the determined i-th feature index.

[0024] As a preferred embodiment, the formula for calculating the weighted Euclidean distance is as follows:

[0025]

[0026] In the formula, x i Let x be the i-th element in the time-series scene matrix dataset. j Let d(x) be the j-th element in the time-series scene dataset. i ,x j ) is x i and x j The weighted Euclidean distance between them, w v Let x be the weight of the v-th dimension variable in the time-series scene data, h be the dimension of the time-series scene data, and x be the weight of the v-th dimension variable. iv Let x be the value of the i-th data point on the v-th variable. jv Let be the value of the j-th data point on the v-th variable.

[0027] As a preferred embodiment, the step of taking the time-series scenario farthest from the cluster center corresponding to each typical wind-solar-load operation scenario as the extreme scenario corresponding to the typical wind-solar-load operation scenario specifically includes:

[0028] For each typical wind-solar-load operation scenario, based on a preset number of required scenarios, from the classes corresponding to the typical wind-solar-load operation scenarios, the time series scenarios with the longest weighted Euclidean distance are selected as the extreme scenarios corresponding to the typical wind-solar-load operation scenarios.

[0029] Accordingly, the present invention also provides a multi-timescale wind-solar-load extreme scene extraction system, including: a data acquisition module, a typical scene module and an extreme scene module;

[0030] The data acquisition module is used to acquire hourly time-series data of wind-solar-load at each node in the power grid in a horizontal year as scene data to form a time-series scene;

[0031] The typical scenario module is used to cluster the time-series scenarios in the time dimension to obtain multiple typical wind-solar-load operation scenarios based on a combination of time correlation and geographical distribution characteristics.

[0032] The extreme scenario module is used to select the time series scenario that is farthest from the cluster center corresponding to each typical wind-solar-load operation scenario as the extreme scenario corresponding to the typical wind-solar-load operation scenario.

[0033] In the above-mentioned multi-timescale wind-solar-load extreme scene extraction system, as a preferred solution, the typical scene module includes: a weighting unit and a typical scene unit;

[0034] The weighting unit is used to assign weights to the wind-solar-load variables based on the time-series data of each wind-solar-load variable in the time-series scenario. The wind-solar-load variables include wind power, photovoltaic power and load at each node in the power grid.

[0035] The typical scenario unit is used to cluster the time-series scenarios based on the weights of the wind-solar-load variables and weighted Euclidean distance to obtain multiple typical wind-solar-load operation scenarios based on a combination of time correlation and geographical distribution characteristics.

[0036] In the above-mentioned multi-timescale wind-solar-load extreme scene extraction system, as a preferred solution, the weighting unit includes: an objective weighting subunit, a subjective weighting subunit, and a combined weighting subunit;

[0037] The objective weighting subunit is used to assign weights to the wind-solar-load variables based on the time series data of each wind-solar-load variable in the time series scenario using the entropy weight method, so as to obtain the objective weights of each variable. The wind-solar-load variables include wind power, photovoltaic power and load at each node in the power grid.

[0038] The subjective weighting subunit is used to calculate the subjective weight of each variable based on the capacity of each wind-solar-load variable;

[0039] The combined weighting subunit is used to perform a weighted average of the objective weighting subunit and the subjective weighting subunit to obtain the combined weight of each variable.

[0040] Thirdly, the present invention also provides a storage medium containing a computer-executable program, which, when executed by a computer processor, is used to perform the multi-timescale wind-light-load extreme scene extraction method described in any of the preceding claims.

[0041] Compared with the prior art, the present invention has the following technical effects:

[0042] This invention proposes a weighted clustering method to consider the temporal correlation and geographical distribution characteristics of wind, solar, and load, thereby rapidly identifying extreme operating scenarios of the system to assess the stability and security of system planning. Specifically, using hourly wind-solar-load time-series data of each node in the power grid as time-series scenarios, a combined weighted clustering method is employed to select typical wind-solar-load operating scenarios of the power grid. Edge operating points far from the cluster centers corresponding to typical wind-solar-load operating scenarios are extracted as extreme scenarios. The clustering process considers the fluctuations of different wind farms and loads, as well as the impact of capacity on extreme scenarios, further improving the efficiency and rationality of extreme scenario extraction. This method can identify large-scale wind and solar power during the planning stage and conduct stability assessments and power flow analyses on weak links in power grid operation, enabling planners to quickly assess the operational safety of system planning schemes and improve the rationality and scientific rigor of large-scale power grid stability scenario analysis.

[0043] Furthermore, based on the volatility of different variables over time and their impact on extreme scenarios, this invention extracts and adopts different weight values ​​in clustering. By using the entropy weight method and the capacity method, the objective and subjective weight ratios of variables are determined, and the combined weights of the variables are determined by weighted averaging. Scene clustering is performed by weighted Euclidean distance, and the farthest scene point is taken as the extreme scenario, which improves the rationality of scene extraction. Attached Figure Description

[0044] To make the objectives, technical solutions, and advantages of the invention clearer, the invention will now be described in further detail with reference to the accompanying drawings, wherein:

[0045] Figure 1 A flowchart of a method for extracting extreme wind-light-load scenes at multiple time scales provided by the present invention;

[0046] Figure 2 This is a flowchart illustrating the weighting process of the variables in this invention.

[0047] Figure 3 An illustration of an embodiment of an extreme scene extraction method provided by the present invention;

[0048] Figure 4 This is a normalized data and extreme scenario diagram of wind-solar-load in an embodiment of the present invention;

[0049] Figure 5 This is a distribution map of wind, solar load, and load using distributed weighted clustering and centralized weighted clustering, as described in an embodiment of the present invention.

[0050] Figure 6 This is a graph showing the relationship between the number of clusters and the squared error of the poles in an embodiment of the present invention. Detailed Implementation

[0051] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but only to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0052] The present invention will now be described in further detail with reference to the accompanying drawings.

[0053] Example:

[0054] This invention provides a method for extracting extreme wind-light-load scenes at multiple time scales, such as... Figure 1 As shown, the specific steps include the following:

[0055] S1. Obtain hourly time-series data of wind-solar-load at each node in the power grid in a horizontal year as scene data to form a time-series scene;

[0056] S2. Cluster the time-series scenarios in the time dimension to obtain multiple typical wind-solar-load operation scenarios based on a combination of time correlation and geographical distribution characteristics;

[0057] S3. Based on each typical wind-solar-load operation scenario, the time series scenario that is farthest from the cluster center corresponding to the typical wind-solar-load operation scenario is taken as the extreme scenario corresponding to the typical wind-solar-load operation scenario.

[0058] In this embodiment, existing methods for extracting extreme wind-solar-load scenarios mostly focus on one type of wind, solar, or load, rarely studying all three simultaneously. Furthermore, the influencing factors considered during time-series generation are not comprehensive enough, such as season, date type, and meteorological factors. This embodiment fully considers the correlation between wind, solar, and load, and takes into account multiple influencing factors, including meteorology, to generate multi-timescale wind-solar-load extreme scenario extractions. This significantly improves the accuracy of extreme scenario extraction, serving power system operation and planning.

[0059] This invention provides a method for extracting extreme wind-solar-load scenes at multiple time scales based on weighted clustering, specifically including the following steps:

[0060] S1. Obtain hourly time-series data of each node in the power grid during the horizontal year (wind-solar-load) as scene data to construct a time-series scene set; that is:

[0061] S11. Construct a set of combined wind-solar-load operation scenarios based on time correlation and geographical distribution characteristics.

[0062] Specifically, S11 involves constructing a set of combined wind-solar-load operation scenarios based on time correlation and geographical distribution characteristics, which can utilize time-series data of annual and hourly levels of node wind power, node photovoltaic, and node load.

[0063] The time-series scene variable matrix is ​​constructed as follows:

[0064] P j =(P j1 P j2 ..., P jT );

[0065] In the formula, the matrix represents the scene P. j Let N be the wind and solar power output or load of the j-th node, with a total of N nodes, where j = 1, 2, 3, ..., N, and T be the total number of scenes. Generally, one scene is defined as 1 hour. Therefore, in the study year, T = 8760 scenes, and P jT This refers to the wind and solar power output or load of the j-th node in the T-th scenario.

[0066] S2. Cluster the aforementioned time-series scenarios along the time dimension to obtain multiple typical wind-solar-load operation scenarios based on a combination of time correlation and geographical distribution characteristics, such as... Figure 2 As shown, it includes:

[0067] S21. Based on the time-series data of each wind-solar-load variable in the time-series scenario, assign weights to the wind-solar-load variables, wherein the wind-solar-load variables include wind power, photovoltaic power, and load at each node in the power grid; step S21 specifically includes:

[0068] S210. Based on the time series data of each wind-solar-load variable in the time series scenario, the wind-solar-load variable is weighted using the entropy weight method to obtain the objective weight of each variable.

[0069] S211. Based on the capacity of each wind-solar-load variable, calculate the subjective weight of each variable;

[0070] S212. The objective weights and subjective weights of each variable are weighted and averaged to obtain the combined weights of each variable, and the combined weights are used as the weights of each wind-solar-load variable.

[0071] In this embodiment, the clustering variables exhibit significant differences in type, dimension, and importance, necessitating the setting of clustering weights to reflect the characteristics of each variable. Traditional Euclidean distance fails to adequately reflect the importance of variables across different dimensions. For example, the contribution of wind and solar power generation variables at different locations to extreme scenarios varies depending on their impact on the research question. Therefore, during clustering, the varying importance of variables necessitates consideration of both subjective and objective weights. Regarding objective weights, variables with significant temporal fluctuations tend to have a greater impact on the extraction of extreme scenarios. The greater the variable fluctuation, the farther the data edge points are from the cluster centers, and the higher the likelihood of them contributing to extreme edge points. Regarding subjective weights, it is necessary to analyze the impact of each variable on the research question and assign greater weights to variables with greater influence.

[0072] This embodiment employs the entropy weighting method to assign weights to the wind-solar-load variables, obtaining objective weights for each variable. Its basic principle is to measure the differences between variables using entropy values ​​and assign weights to each variable. It only reflects the volatility of variable data and is unrelated to the type and relationship of the variables. Based on the characteristics of the entropy weighting method, entropy can be used to determine the degree of dispersion of variables. Based on this principle, calculating the Euclidean distance between scenes with entropy-weighted variables can yield more accurate results.

[0073] In step S210, based on the time-series data of each wind-solar-load variable in the time-series scenario, the entropy weight method is used to determine the weights of the wind-solar-load variables. The specific method for obtaining the objective weights of each variable is as follows:

[0074] Suppose there are T time-series scene objects to be clustered along the time dimension, and each time-series scene contains N... WD +N Load +N PV The system uses 124 dimensional variables, defines each variable as a feature index, calculates the entropy value of each feature index, and determines the objective weight of each variable based on the entropy value of each feature index. The method for obtaining the objective weight of each variable is as follows:

[0075]

[0076] In the formula, H i Let f be the entropy value of the i-th feature index, n be the total dimension of the temporal variables for each scene, T be the total number of objects in the temporal scene, and f be the entropy value of the i-th feature index. ij Let s be the proportion of the j-th temporal scene object at the i-th feature index. ij For the data of the variable, p i N is the entropy weight of the i-th feature index. WD N Load N PV These are the dimensional variables for each time series scenario.

[0077] In practical implementation, the data is compressed to the interval [0,1] using the following formula:

[0078]

[0079] In the formula, P ij This refers to the wind and solar power output or load of the i-th node in the j-th scenario.

[0080] When the variable has low volatility and H i When the value is large, its clustering effect is small, and its contribution to the selection of extreme scenarios is small. The greater the difference in variable values ​​in the scenario, the smaller the H value. i The smaller the value, the greater the clustering influence of the variable, and the greater its contribution to the selection of extreme scenarios. When all data for a certain variable in a scenario are equal, H... i =H max =1, its clustering influence is zero.

[0081] In step S211, the subjective weights of each variable are calculated based on the capacity of each wind-solar-load variable; that is, the wind power and load capacity ratio is used as an alternative to subjective weighting.

[0082] In step S212, the objective and subjective weights of each variable are weighted and averaged to obtain the combined weight of each variable. The specific method for using the combined weight as the weight of each wind-solar-load variable is as follows:

[0083]

[0084] In the formula, w i p represents the combined weights of the variables. i q is the objective weight of the i-th feature index determined by the entropy weight method. i The subjective weight of the determined i-th feature index.

[0085] S22. Based on the weights of the wind-solar-load variables, cluster the time-series scenarios using weighted Euclidean distance to obtain multiple typical wind-solar-load operation scenarios based on a combination of time correlation and geographical distribution characteristics.

[0086] Step S22 specifically involves using a weighted clustering method to obtain multiple typical wind-solar-load operation scenarios by performing weighted clustering on time-series scenarios in the time dimension.

[0087] In this embodiment, the K-means clustering algorithm is used to pre-assign data objects to the nearest cluster based on the principle of minimizing the distance to the selected k initial cluster centers. During the iteration process, the dataset is divided into different categories to optimize the criterion function for evaluating clustering performance.

[0088] Specifically, based on the time-series scene matrix dataset x = {x i |x i ∈R h Let h = 1, 2, ..., t, where h is the dimension of the data variable and i is the number of data points in the dataset. Select k initial cluster centers c1, c2, ..., ct. k ;

[0089] Based on the principle of minimizing the distance between each data object and the selected k initial cluster centers, data objects are pre-assigned to the nearest cluster. The formula for calculating the weighted Euclidean distance is as follows:

[0090]

[0091] In the formula, x i Given the i-th element in the dataset, x j Given the j-th element in the dataset, d(x) i ,x j ) is x i and x j The weighted Euclidean distance between them, w v For the weights of the v-th dimension variable in the time-series scene data, in this specific implementation, w v It is generally set to 1, where h is the dimension of the time-series scene data, and x is the dimension of the time-series scene data. iv Let x be the value of the i-th data point on the v-th variable. jv Let j be the value of the j-th data point in the v-th variable;

[0092] Recalculate the cluster centers, defined as follows:

[0093]

[0094] In the formula, N j It is a class C j The number of samples in the dataset, C1, C2, ..., C k Cluster centers c1, c2, ..., c k The data collection of the class to which it belongs;

[0095] The clustering problem is transformed into an optimization problem to obtain the optimal clustering result. The objective function is calculated as follows:

[0096]

[0097] S3. Based on each typical wind-solar-load operation scenario, the time-series scenario farthest from the cluster center corresponding to the typical wind-solar-load operation scenario is taken as the extreme scenario corresponding to the typical wind-solar-load operation scenario; that is:

[0098] S31. Extract the edge running points far from the cluster center as extreme scenarios.

[0099] Specifically, S31 involves clustering wind-solar-load time-series scenarios to generate several typical wind-solar-load operation scenarios along the time dimension, serving as cluster centers. Within each cluster, the edge points furthest from the cluster center are designated as the extreme values ​​corresponding to each typical wind-solar-load operation scenario. Multiple typical wind-solar-load operation scenarios correspond to multiple extreme scenarios. The extreme scenario extraction method is as follows: Figure 3 As shown, the extreme operating scenario is represented as follows:

[0100]

[0101] In the formula, i is the number of classes, e i It is an edge point associated with the i-th class.

[0102] When each category has n edge points, meaning each typical scenario corresponds to n extreme scenarios, a total of n×k extreme operation scenarios will occur, defined as:

[0103]

[0104] In the formula, m represents the extreme scenario of the m-th layer corresponding to each typical scenario, m = 1, 2, ..., n, e i m Let m be the extreme scenario at the m-th level corresponding to the i-th typical scenario.

[0105] The edge points of the first layer represent the worst-case scenario in each category. Due to the clustering method used, the extreme scenarios are largely dependent on the typical scenarios. These extreme scenarios may cover extremes of scenery and load, or even uniform conditions, meaning that extreme scenarios are not only located outside the sample data points but may also lie within the regions enclosed by the data points, depending on the number of typical scenarios selected.

[0106] To improve the efficiency of the extraction method, this embodiment uses a clustering evaluation index to determine the optimal number of clusters. Specifically, it calculates the sum of weighted Euclidean distances between extreme points and their corresponding cluster centers. This sum is called the extreme point squared error, denoted as:

[0107]

[0108] In the formula, J e E is a clustering evaluation metric. i Let k be the set of extreme points corresponding to the i-th cluster center, and k be the number of clusters.

[0109] Secondly, the present invention provides a multi-timescale wind-solar-load extreme scene extraction system, characterized in that it includes: a data acquisition module, a typical scene module, and an extreme scene module;

[0110] The data acquisition module is used to acquire hourly time-series data of wind-solar-load at each node in the power grid in a horizontal year as scene data to form a time-series scene;

[0111] The typical scenario module is used to cluster the time-series scenarios in the time dimension to obtain multiple typical wind-solar-load operation scenarios based on a combination of time correlation and geographical distribution characteristics.

[0112] The extreme scenario module is used to select the time series scenario that is farthest from the cluster center corresponding to each typical wind-solar-load operation scenario as the extreme scenario corresponding to the typical wind-solar-load operation scenario.

[0113] In practical implementation, the typical scenario module includes: the weighting unit and the typical scenario unit.

[0114] The weighting unit in the typical scenario module is used to assign weights to the wind-solar-load variables based on the time-series data of each wind-solar-load variable in the time-series scenario. The wind-solar-load variables include wind power, photovoltaic power and load at each node in the power grid.

[0115] In practice, the empowerment unit includes: objective empowerment sub-unit, subjective empowerment sub-unit, and combined empowerment sub-unit;

[0116] The objective weighting subunit is used to assign weights to the wind-solar-load variables based on the time series data of each wind-solar-load variable in the time series scenario using the entropy weight method, so as to obtain the objective weights of each variable. The wind-solar-load variables include wind power, photovoltaic power and load at each node in the power grid.

[0117] The subjective weighting subunit is used to calculate the subjective weight of each variable based on the capacity of each wind-solar-load variable;

[0118] The combined weighting subunit is used to perform a weighted average of the objective weighting subunit and the subjective weighting subunit to obtain the combined weight of each variable, and the combined weight is used as the weight of each wind-solar-load variable.

[0119] In this embodiment, the clustering variables exhibit significant differences in type, dimension, and importance, necessitating the setting of clustering weights to reflect the characteristics of each variable. Traditional Euclidean distance fails to adequately reflect the importance of variables across different dimensions. For example, the contribution of wind and solar power generation variables at different locations to extreme scenarios varies depending on their impact on the research question. Therefore, during clustering, the varying importance of variables necessitates consideration of both subjective and objective weights. Regarding objective weights, variables with significant temporal fluctuations tend to have a greater impact on the extraction of extreme scenarios. The greater the variable fluctuation, the farther the data edge points are from the cluster centers, and the higher the likelihood of them contributing to extreme edge points. Regarding subjective weights, it is necessary to analyze the impact of each variable on the research question and assign greater weights to variables with greater influence.

[0120] This embodiment employs the entropy weighting method to assign weights to the wind-solar-load variables, obtaining objective weights for each variable. Its basic principle is to measure the differences between variables using entropy values ​​and assign weights to each variable. It only reflects the volatility of variable data and is unrelated to the type and relationship of the variables. Based on the characteristics of the entropy weighting method, entropy can be used to determine the degree of dispersion of variables. Based on this principle, calculating the Euclidean distance between scenes with entropy-weighted variables can yield more accurate results.

[0121] Specifically, in the objective weighting sub-unit, the weights of the wind-solar-load variables are determined using the entropy weight method based on the time-series data of each wind-solar-load variable in the time-series scenario. The specific method for obtaining the objective weights of each variable is as follows:

[0122] Suppose there are T time-series scene objects to be clustered along the time dimension, and each time-series scene contains N... WD +N Load +N PV The system uses 124 dimensional variables, defines each variable as a feature index, calculates the entropy value of each feature index, and determines the objective weight of each variable based on the entropy value of each feature index. The method for obtaining the objective weight of each variable is as follows:

[0123]

[0124] In the formula, H i Let f be the entropy value of the i-th feature index, n be the total dimension of the temporal variables for each scene, T be the total number of objects in the temporal scene, and f be the entropy value of the i-th feature index. ij Let s be the proportion of the j-th temporal scene object at the i-th feature index. ij For the data of the variable, p i N is the entropy weight of the i-th feature index. WD N Load N PV These are the dimensional variables for each time series scenario.

[0125] In practical implementation, the data is compressed to the interval [0,1] using the following formula:

[0126]

[0127] In the formula, P ij This refers to the wind and solar power output or load of the i-th node in the j-th scenario.

[0128] When the variable has low volatility and H i When the value is large, its clustering effect is small, and its contribution to the selection of extreme scenarios is small. The greater the difference in variable values ​​in the scenario, the smaller the H value. i The smaller the value, the greater the clustering influence of the variable, and the greater its contribution to the selection of extreme scenarios. When all data for a certain variable in a scenario are equal, H... i =H max =1, its clustering influence is zero.

[0129] Specifically, in the subjective weighting sub-unit, the subjective weight of each variable is calculated based on the capacity of each wind-solar-load variable, that is, the wind power and load capacity ratio is used as an alternative to subjective weighting.

[0130] Specifically, in the combined weighting subunit, the objective weights and subjective weights of each variable are weighted and averaged to obtain the combined weight of each variable. The specific method for using the combined weight as the weight of each wind-solar-load variable is as follows:

[0131]

[0132] In the formula, w i p represents the combined weights of the variables. i q is the objective weight of the i-th feature index determined by the entropy weight method. i The subjective weight of the determined i-th feature index.

[0133] In specific implementation, the typical scenario unit in the typical scenario module is used to cluster the time-series scenarios based on the weighted Euclidean distance according to the weights of the wind-solar-load variables, so as to obtain multiple typical wind-solar-load operation scenarios based on the combination of time correlation and geographical distribution characteristics.

[0134] Specifically, in typical scenario units, a weighted clustering method is adopted to obtain multiple typical wind-solar-load operation scenarios by weighted clustering of time-series scenarios in the time dimension.

[0135] In this embodiment, the K-means clustering algorithm is used to pre-assign data objects to the nearest cluster based on the principle of minimizing the distance to the selected k initial cluster centers. During the iteration process, the dataset is divided into different categories to optimize the criterion function for evaluating clustering performance.

[0136] Specifically, based on the time-series scene matrix dataset x = {x i |x i ∈R h Let h = 1, 2, ..., t, where h is the dimension of the data variable and i is the number of data points in the dataset. Select k initial cluster centers c1, c2, ..., ct. k ;

[0137] Based on the principle of minimizing the distance between each data object and the selected k initial cluster centers, data objects are pre-assigned to the nearest cluster. The formula for calculating the weighted Euclidean distance is as follows:

[0138]

[0139] In the formula, x i Given the i-th element in the dataset, x j Given the j-th element in the dataset, d(x) i ,x j ) is x i and x j The weighted Euclidean distance between them, w v For the weights of the v-th dimension variable in the time-series scene data, in this specific implementation, w v It is generally set to 1, where h is the dimension of the time-series scene data, and x is the dimension of the time-series scene data. iv Let x be the value of the i-th data point on the v-th variable. jv Let j be the value of the j-th data point in the v-th variable;

[0140] Recalculate the cluster centers, defined as follows:

[0141]

[0142] In the formula, N j It is a class C j The number of samples in the dataset, C1, C2, ..., C k Cluster centers c1, c2, ..., c k The data collection of the class to which it belongs;

[0143] The clustering problem is transformed into an optimization problem to obtain the optimal clustering result. The objective function is calculated as follows:

[0144]

[0145] In specific implementation, the extreme scenario module is to generate several typical wind-solar-load operation scenarios by clustering them in the time dimension based on the wind-solar-load time sequence scenario, which serve as cluster centers.

[0146] Within each cluster, the edge points furthest from the cluster center are taken as the extreme values ​​corresponding to each typical wind-solar-load operation scenario. Multiple typical wind-solar-load operation scenarios correspond to multiple extreme scenarios. The extreme scenario extraction method is as follows: Figure 3 As shown, the extreme operating scenario is represented as follows:

[0147]

[0148] In the formula, i is the number of classes, e i It is an edge point associated with the i-th class.

[0149] When each category has n edge points, meaning each typical scenario corresponds to n extreme scenarios, a total of n×k extreme operation scenarios will occur, defined as:

[0150]

[0151] In the formula, m represents the extreme scenario of the m-th layer corresponding to each typical scenario, where m = 1, 2, ..., n. Let m be the extreme scenario at the m-th level corresponding to the i-th typical scenario.

[0152] The edge points of the first layer represent the worst-case scenario in each category. Due to the clustering method used, the extreme scenarios are largely dependent on the typical scenarios. These extreme scenarios may cover extremes of scenery and load, or even uniform conditions, meaning that extreme scenarios are not only located outside the sample data points but may also lie within the regions enclosed by the data points, depending on the number of typical scenarios selected.

[0153] To improve the efficiency of the extraction method, this embodiment uses a clustering evaluation index to determine the optimal number of clusters. Specifically, it calculates the sum of weighted Euclidean distances between extreme points and their corresponding cluster centers. This sum is called the extreme point squared error, denoted as:

[0154]

[0155] In the formula, J e E is a clustering evaluation metric. i Let k be the set of extreme points corresponding to the i-th cluster center, and k be the number of clusters.

[0156] Thirdly, the present invention also provides a storage medium containing a computer-executable program, which, when executed by a computer processor, is used to perform the above-described multi-timescale wind-light-load extreme scene extraction method of the present invention.

[0157] The storage medium can be any combination of one or more computer-readable media. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. A computer-readable storage medium can include: an electrical connection having one or more wires, a portable computer disk, a hard disk, a random access memory, a read-only memory, an erasable programmable read-only memory, an optical fiber, a portable compact disk read-only memory, an optical storage device, a magnetic storage device, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0158] The code for a computer-executable program that performs the operations of this invention can be written in one or more programming languages ​​or a combination thereof. Programming languages ​​include object-oriented programming languages ​​such as Java, Smalltalk, and ++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer can be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer.

[0159] To more clearly and intuitively illustrate the effectiveness of this method, the following verification explanation is provided.

[0160] One wind farm, one solar power plant, and one node load were selected as the main variables. The capacities of the wind farm and the solar power plant were 100MW and 50MW, respectively, and the maximum load of the node was 150MW.

[0161] In traditional power transmission network planning, the annual maximum load scenario is typically considered the riskiest scenario to verify the safety of the planning scheme. This is usually the operating scenario with the greatest network pressure. However, with the large-scale grid connection of intermittent renewable energy sources, the annual maximum load scenario is no longer the worst-case scenario for the planning scheme. Figure 4 As shown, the worst-case extreme wind-solar-load scenario extracted by clustering is no longer the scenario with the maximum or minimum load, but appears at 4618 hours and 8151 hours. Grid connection of wind power has led to changes in the system's extreme operating scenarios.

[0162] This model uses the entropy weight method and the capacity method to determine the ratio of objective and subjective weights for variables and combines them. In Table 1, the original weights of the three variables are set to 0.328. Although load power has a high value and the largest subjective weight among the three variables, its volatility is still less than that of wind and solar power output. This indicates that load power has a weaker impact on extreme scenarios, hence its clustering combination weight remains relatively small.

[0163] Table 1. Clustering Variable Weights

[0164]

[0165] To construct the wind-solar-load scenario, wind farms and solar power fields are treated as a single variable, without considering their location. By setting the cluster size to 3, the worst-case scenarios numbered [4618, 8759, 5656] were obtained. By considering their different geographical locations, wind and solar power were assumed to be separate clustering variables, and the extreme scenarios obtained through clustering were numbered [4618, 8151, 8656]. The extreme cases changed, and the corresponding wind-solar-load situations are shown in Table 2.

[0166] Table 2 Comparison of Extreme Scenarios with Distributed and Centralized Clustering

[0167]

[0168] Depend on Figure 5 It can be seen that, considering the geographical distribution of wind and solar power, the wind-solar-load operation scenario obtained by clustering is more extreme at the global level than that obtained by centralized clustering. This scenario also has a higher load level and a lower wind and solar power output level. Therefore, the different geographical locations of wind and solar power generation must be considered in the clustering method to extract more extreme scenarios.

[0169] As the number of clusters increases, a greater number of extreme scenarios with more comprehensive information can be derived. However, the increase in the number of clusters also increases computational demands and leads to the extraction of redundant extreme scenarios. Figure 6 It can be seen that for cluster numbers greater than 3, the standard of EPSE does not change much, meaning that increasing the number of clusters does not produce positive feedback.

[0170] In summary, this invention proposes a method for extracting extreme operating scenarios of wind-solar-load based on the concept of weighted clustering. This method selects time-series wind-solar-load data as clustering objects and identifies the operating point furthest from the weighted Euclidean center of the cluster as the extreme scenario. With this contribution, large-scale wind and solar power can be identified during the planning stage, and stability assessments and power flow analyses can be performed on weak links in power grid operation. Results demonstrate the effectiveness of the new method. It enables planners to quickly assess the operational safety of system planning schemes and improves the rationality and scientific rigor of large-scale power grid stability scenario analysis.

[0171] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described with reference to preferred embodiments, those skilled in the art should understand that various changes in form and detail can be made without departing from the spirit and scope of the invention as defined in the appended claims.

Claims

1. A method for extracting extreme wind-light-load scenes at multiple time scales, characterized in that, Includes the following steps: The time-series data of wind-solar-load at each node in the power grid in a horizontal year are obtained as scene data to form a time-series scene; Clustering the aforementioned time-series scenarios along the time dimension yields several typical wind-solar-load operation scenarios based on a combination of temporal correlation and geographical distribution characteristics, specifically including: Based on the time-series data of each wind-solar-load variable in the time-series scenario, the wind-solar-load variable is weighted, and the wind-solar-load variable includes wind power, photovoltaic power and load at each node in the power grid; Based on the weights of the wind-solar-load variables, the time-series scenarios are clustered using weighted Euclidean distance to obtain multiple typical wind-solar-load operation scenarios based on a combination of time correlation and geographical distribution characteristics. Specifically, assigning weights to the wind-solar-load variables based on the time-series data of each wind-solar-load variable in the time-series scenario includes: Based on the time series data of each wind-solar-load variable in the time series scenario, the wind-solar-load variable is weighted using the entropy weight method to obtain the objective weight of each variable; Based on the capacity of each wind-solar-load variable, the subjective weights of each variable are calculated. The objective and subjective weights of each variable are weighted and averaged to obtain the combined weight of each variable, and the combined weight is used as the weight of each wind-solar-load variable. Based on various typical wind-solar-load operation scenarios, the time series scenario that is farthest from the cluster center corresponding to the typical wind-solar-load operation scenario is taken as the extreme scenario corresponding to the typical wind-solar-load operation scenario.

2. The method for extracting extreme wind-light-load scenes at multiple time scales according to claim 1, characterized in that, The specific method for determining the weights of the wind-solar-load variables using the entropy weight method based on the time-series data of each wind-solar-load variable in the time-series scenario, and obtaining the objective weights of each variable, is as follows: Assuming there is a time dimension Clustering is performed on each time-series scene object, and each time-series scene includes The system uses 124 dimensional variables, defines each variable as a feature index, calculates the entropy value of each feature index, and determines the objective weight of each variable based on the entropy value of each feature index. The method for obtaining the objective weight of each variable is as follows: ； ； ； In the formula, For the first The entropy value of each feature index, where n is the total dimension of the time-series variables for each scene. This represents the total number of objects in the time-series scene. For the first The first time-series scene object in the... The proportion at each characteristic index For data of variables, For the first Entropy weights of each characteristic index These are the dimensional variables for each time series scenario.

3. The method for extracting extreme wind-light-load scenes at multiple time scales according to claim 2, characterized in that, The specific method for obtaining the combined weight of each variable by weighting the objective and subjective weights is as follows: ； In the formula, The combined weights of the variables, The first one determined by the entropy weight method The objective weights of each characteristic index, For the determined first Subjective weights of each feature index.

4. The method for extracting extreme wind-light-load scenes at multiple time scales according to claim 1, characterized in that, The formula for calculating the weighted Euclidean distance is as follows: ； In the formula, The first in the time series scene matrix dataset One element, For the time series scene dataset, the first One element, for and The weighted Euclidean distance between them For time series scene data Weights of dimensional variables, For time-series scenario data, For the first The data in the first The values ​​of each variable For the first The data in the first The values ​​of each variable.

5. The method for extracting extreme wind-light-load scenes at multiple time scales according to claim 1, characterized in that, The phrase "based on various typical wind-solar-load operation scenarios, taking the time-series scenario farthest from the cluster center corresponding to each typical wind-solar-load operation scenario as the extreme scenario corresponding to the typical wind-solar-load operation scenario" specifically includes: For each typical wind-solar-load operation scenario, based on a preset number of required scenarios, from the classes corresponding to the typical wind-solar-load operation scenarios, the time series scenarios with the longest weighted Euclidean distance are selected as the extreme scenarios corresponding to the typical wind-solar-load operation scenarios.

6. A multi-timescale wind-solar-load extreme scene extraction system, characterized in that, include: Data acquisition module, typical scenario module, and extreme scenario module; The data acquisition module is used to acquire hourly time-series data of wind-solar-load at each node in the power grid in a horizontal year as scene data to form a time-series scene; The typical scenario module is used to cluster the time-series scenarios in the time dimension to obtain multiple typical wind-solar-load operation scenarios based on a combination of time correlation and geographical distribution characteristics. The typical scenario module includes: a weighting unit and a typical scenario unit; The weighting unit is used to assign weights to the wind-solar-load variables based on the time-series data of each wind-solar-load variable in the time-series scenario. The wind-solar-load variables include wind power, photovoltaic power, and load at each node in the power grid. The weighting unit includes: an objective weighting subunit, a subjective weighting subunit, and a combined weighting subunit. The objective weighting subunit is used to assign weights to the wind-solar-load variables based on the time series data of each wind-solar-load variable in the time series scenario using the entropy weight method, so as to obtain the objective weights of each variable. The wind-solar-load variables include wind power, photovoltaic power and load at each node in the power grid. The subjective weighting subunit is used to calculate the subjective weight of each variable based on the capacity of each wind-solar-load variable; The combined weighting subunit is used to perform a weighted average of the objective weighting subunit and the subjective weighting subunit to obtain the combined weight of each variable. The typical scenario unit is used to cluster the time-series scenarios based on the weights of the wind-solar-load variables and the weighted Euclidean distance, to obtain multiple typical wind-solar-load operation scenarios based on a combination of time correlation and geographical distribution characteristics. The extreme scenario module is used to select the time series scenario that is farthest from the cluster center corresponding to each typical wind-solar-load operation scenario as the extreme scenario corresponding to the typical wind-solar-load operation scenario.

7. A storage medium containing a computer-executable program, characterized in that, When executed by a computer processor, the computer executable program is used to perform the multi-timescale wind-light-load extreme scene extraction method as described in any one of claims 1 to 5.

Images