Flexible load regulation capability assessment method and system based on integrated clustering algorithm

Abstract

The invention relates to the technical field of power distribution network flexible resource regulation capability evaluation, and provides a flexible load regulation capability evaluation method and system based on an integrated clustering algorithm, and the method comprises the steps: collecting a historical daily load sample set of a power user and a daily load sample set of a day before a regulation day; dividing the sample set into K class clusters through a clustering and integrated clustering algorithm so as to obtain the class cluster to which the baseline load of the regulation and control day belongs, obtaining the regulation and control day baseline load of all the power users participating in the demand response and the upward and downward regulation quantity limit values thereof through the class cluster samples, and calculating the upward and downward regulation quantity limit values of the load of the power distribution network node; constructing a multi-dimensional evaluation model of the flexible load system-level regulation capability based on the power flow constraint of the power distribution network; and quantizing the output regulation capability. Through the method, the accuracy of power distribution network flexible load regulation capability evaluation is improved, and the diversity of evaluation dimensions is increased.

Description

Technical Field

[0001] This invention relates to the field of flexible resource regulation capability assessment technology for distribution networks, specifically to a method and system for assessing flexible load regulation capability based on an integrated clustering algorithm. Background Technology

[0002] Currently, the installed capacity of new energy sources increased by 360 million kilowatts in 2024, bringing the total to 1.41 billion kilowatts (890 million kilowatts of photovoltaic power and 520 million kilowatts of wind power). It is expected to exceed 1.7 billion kilowatts in 2025 and reach about 3 billion kilowatts in 2030.

[0003] However, the inherent volatility and uncontrollability of new energy sources pose a severe challenge to the safe and stable operation of the power system. Their randomness and intermittency lead to increased fluctuations in system net load and a significant increase in peak-valley differences, resulting in a surge in system peak-shaving pressure. Simultaneously, the high degree of uncertainty in the spatiotemporal distribution of new energy generation further exacerbates wind and solar curtailment. Therefore, it is urgent to deeply explore the regulation potential of controllable resources in the power system to improve the safe and stable operation of the power grid.

[0004] As a crucial component of controllable resources in the power system, flexible loads possess unique advantages in enhancing system regulation capabilities due to their flexible response characteristics. Compared to traditional generator units, flexible loads not only achieve more refined power regulation but also exhibit significant advantages such as lower regulation costs, stronger scalability, and bidirectional regulation capabilities. Scientifically assessing the regulation capabilities of flexible loads is a vital technical foundation for fully unleashing their regulation potential and improving the safe and stable operation of the power system.

[0005] In assessing the adaptability of flexible loads, multiple factors such as load characteristics, response speed, external environmental constraints, and user willingness to participate must be considered comprehensively. Therefore, how to comprehensively, accurately, and quickly assess the adaptability of flexible loads has become an urgent problem to be solved, requiring consideration of both the practicality and feasibility of the assessment method, as well as its comprehensiveness. Traditional methods for assessing the adaptability of flexible loads mainly fall into two categories: one relies on complex physical modeling, typically requiring detailed equipment-level modeling and collecting multi-dimensional data such as equipment parameters, user behavior characteristics, and environmental parameters for various types of flexible loads, achieving adaptability assessment through complex calculations. Although this method theoretically has an advantage in accuracy, in practical scenarios involving large-scale heterogeneous flexible load integration, the difficulty of multi-dimensional data collection and high computational complexity lead to a significant increase in engineering implementation costs and significantly limited operability. The second type of method uses clustering algorithms to analyze historical load data to assess the user's flexible adaptability, avoiding complex physical modeling. For example, the Chinese invention patent CN114219205B, "A Method for Calculating the Reliable Capacity of Flexible Loads for Power Grid Planning," extracts features using the K-means clustering algorithm based on historical electricity load data of the regional distribution network. Within the scheduling cycle, with the goal of minimizing the operating cost of the regional distribution network, an optimization model of the regional distribution network considering flexible loads is established, constraints are configured in the model, and iterative solutions are obtained to obtain typical daily demand response values. Based on the calculated typical daily demand response values, combined with the flexible load response, the response amount and certainty of users under the corresponding incentive level are calculated. However, the above method uses a single clustering algorithm, which is difficult to fully adapt to the temporal, high-dimensional, and uncertain nature of the load data, making the clustering effect susceptible to interference and affecting the accuracy and stability of subsequent evaluation results.

[0006] Meanwhile, like the methods mentioned above, existing research methods suffer from insufficient evaluation dimensions and a lack of systematic assessment. Current research mostly focuses on basic parameters such as the adjustable power range and response time of flexible loads, resulting in overly simplistic evaluation dimensions and a lack of system-level indicators such as load curve smoothness, peak load reduction, and renewable energy consumption contribution rate. This single-dimensional evaluation model fails to establish a multi-level evaluation framework from users to nodes to the distribution system, making it difficult to comprehensively evaluate the global regulation capability of flexible loads from macroscopic perspectives such as grid security and optimized dispatch. Summary of the Invention

[0007] The technical problem to be solved by this invention is how to improve the accuracy of the assessment of the flexible load regulation capability of the distribution network and increase the diversity of assessment dimensions.

[0008] The present invention solves the above-mentioned technical problems through the following technical means: This invention provides a method for evaluating flexible load regulation capability based on an ensemble clustering algorithm, comprising the following steps: S1. Obtain the historical daily load sample set of any power user to be evaluated in the distribution network. and regulation days Daily load sample set ; S2, Calculate the daily baseline load for electricity user regulation. ; S3, Historical daily load sample set of electricity users conduct Clustering yields the optimal number of clusters K; S4. Based on the optimal clustering number K obtained in step S3, use an ensemble clustering algorithm to cluster the sample set. Divided into K clusters ,in ; S5. Calculate and regulate daily baseline load. For each cluster membership degree Obtain the baseline load on the control day. To which the cluster belongs ; S6, Baseline load based on the control date To which category Based on the daily load sample data, calculate the upward and downward adjustment limits of the daily baseline load for power user regulation; S7. Repeat steps S1-S6 to obtain the baseline load of all power users participating in demand response in the distribution network and their upward and downward adjustment limits, and calculate the upward and downward adjustment limits of the distribution network nodes. S8. Construct a multi-dimensional evaluation model for the system-level regulation capability of flexible loads in distribution networks based on power flow constraints. S9. Based on the multi-dimensional evaluation model constructed in step S8, quantitatively output the flexible load regulation capability of the distribution network system.

[0009] Furthermore, the historical daily load sample set of electricity users mentioned in step S1 As shown in the following formula:

[0010]

[0011] in, Indicates the electricity user number Daily load data for the day; Indicates the electricity user number Heavenly t Load data sampled at various times; The number of days for load data; The number of sampling times for the daily load; The date of the regulation Daily load sample set As shown in the following formula:

[0012]

[0013] in, This indicates that electricity users are on the [number]th day before the control date. Daily load data for the day; This indicates that electricity users are on the [number]th day before the control date. Heavenly t Load data sampled at various times; The number of sampling times for the daily load; Step S2 describes the daily baseline load for the regulation of electricity users. The specific calculation is as follows: .

[0014] Further, step S3 includes the following steps: S31. Determine the number of clusters based on sample characteristics and sample size. k Range, traversal k Historical daily load sample set conduct Clustering; S32. Calculate the number of clusters. k DBI indicator The The calculation is as follows:

[0015] in, Representing the Each category / cluster; Representing the The average intra-class distance of each cluster; Representing the j The average intra-class distance of each cluster; Representing the The first cluster and the first j The distance between the center points of each cluster is calculated using the following formula:

[0016]

[0017] in, Representative clusters The number of samples within; Representative clusters The sample center is given by the following formula:

[0018] in, Representative clusters Daily load data samples.

[0019] S33. Based on the calculation results of step S32, with the minimum Corresponding k Value as the optimal cluster number .

[0020] Further, step S4 includes the following steps: S41. Use Bootstrap resampling technology to sample the data set. conduct Sub-independent sampling with replacement, each time drawing Each sample forms a new sample set, as shown in the following formula:

[0021] in, Representing the The sample set obtained by resampling; Representing the The second resampling sample set Daily load data for the day; S42, with the optimal number of clusters As a uniform parameter, for each Bootstrap sample set conduct Clustering, generating the base clustering results, as shown in the following formula:

[0022]

[0023] in, Represents the base cluster set; Representing the Base clustering based on secondary resampling; Representative base clustering The first in Each category / cluster; S43. Constructing a consensus matrix , dimension consensus matrix elements Indicates sample and samples In all base cluster sets The frequency of being assigned to the same cluster is as follows:

[0024]

[0025] S44. Based on the consensus matrix A, hierarchical clustering is used to integrate the base clusters to obtain the final clustering result. The specific execution method is as follows: (1) The sample set Each sample data point is considered an independent cluster. Based on the consensus matrix obtained in step S43, the consensus matrix is ​​calculated sequentially for any two clusters. Similarity between As shown in the following formula

[0026] in, and Clusters The number of samples in the sample; Clusters The set of sample subscripts in; (2) Merge the two clusters with the highest similarity until the number of clusters is [number missing]. The final clustering result is then obtained, as shown in the following formula:

[0027] in, For clusters The first in Daily load data; For clusters The number of samples.

[0028] Furthermore, the membership degree described in step S5 The calculation is as follows:

[0029] in, and Representing clusters and clusters The sample center; Adjusting the daily baseline load Classified into the cluster with the highest corresponding membership degree As shown in the following formula:

[0030] in, For clusters The number of samples.

[0031] Further, step S6 includes the following steps: S61, by cluster Power matrix constructed from sample data As shown in the following formula:

[0032] in, Representing a cluster The Middle A sample, as shown in the following formula:

[0033] Each column in Indicates the same sampling time Load data; S62. Set the significance level Calculate the load of electricity users at each sampling time. quantiles and Quantiles, specifically: For sampling time The load, i.e. The first in The data was analyzed using the quantile estimation method to obtain the electricity user load at the sampling time. of quantiles and quantiles ; S63. Calculate the upward adjustment limit for the daily load control respectively. and downward adjustment limit As shown in the following formula:

[0034] .

[0035] Further, step S7 includes the following steps: S71. Repeat steps S1 to S6 to obtain the baseline load, upward adjustment limit, and downward adjustment limit of all power users participating in demand response in the distribution network, denoted as follows: , and , ;in, The number of electricity users participating in demand response; The number of sampling times for the daily load; S72, Calculate distribution network nodes i Baseline load of electricity users As shown in the following formula:

[0036] in, It is a node The set of electricity users on the platform; It represents the number of nodes in the distribution network. The number of sampling times for the daily load; S73, Calculate distribution network nodes i Upward adjustment limit for electricity users and downward adjustment limit As shown in the following formula:

[0037]

[0038] Further, step S8 includes the following steps: S81. Set objective functions from three evaluation dimensions: smoothing the system load curve, increasing the amount of renewable energy absorbed, and reducing peak-shaving pressure, as follows:

[0039]

[0040]

[0041] in, For the smoothness of the load curve; This refers to the amount of new energy consumed; This refers to the peak-to-valley difference in system load. This represents the number of nodes in the distribution network. The number of sampling times for the daily load; and Representing distribution network nodes Maximum and minimum daily load power; Indicates distribution network node exist t Load value at any given time; Indicates distribution network node The average daily load; Indicates distribution network node New energy sources have contributed significantly; S82. Set the distribution network security constraints for the distribution network-level flexible load regulation capability assessment model. The constraint conditions are as follows:

[0042]

[0043]

[0044]

[0045]

[0046]

[0047]

[0048] in, Indicated by The set of the starting nodes of the branches of the terminal nodes; Indicated by The set of end nodes of the branches of the first node; and They represent Flowing through the side road Active and reactive power; and They represent Flowing through the side road The active and reactive power; express Time Node The voltage amplitude; Indicates a branch The resistance; Indicates a branch The reactance; Indicates a branch impedance; These represent the upper and lower limits of the voltage, respectively. express Time Branch The carrying capacity; Indicates a branch Safe current; for Time Node Load regulation capability; For nodes The load at The power factor angle at time; for Time Node New energy sources have contributed significantly; for Time Node The predicted value of new energy power generation.

[0049] Further, step S9 includes the following steps: S91. Based on the objective function and constraints constructed in step S8, solve the three optimization models to obtain the optimal system operating state under each optimization objective. Specifically, output the following optimal indices: (1) Minimum value of load curve smoothness ; (2) Maximum value of new energy consumption ; (3) Minimum value of system load peak-valley difference ; S92. Based on the optimal indicators obtained in step S91, calculate the adjustment rates of flexible load on load curve smoothness, renewable energy absorption, and system load peak-valley difference, as shown in the following formulas:

[0050] in, These are the adjustment rates of flexible load on the smoothness of the load curve, the amount of new energy absorbed, and the peak-valley difference of the system load, respectively. These are the smoothness of the load curve, the amount of new energy consumption, and the peak-valley difference of the system load before the implementation of flexible load control.

[0051] This invention also provides a flexible load adjustment capability assessment system based on an ensemble clustering algorithm. The system operates using the above-mentioned method and includes the following modules: The data acquisition module is used to obtain the historical daily load sample set of any power user to be evaluated in the distribution network. and regulation days Daily load sample set ; The control baseline load calculation module is used to calculate the control baseline load of electricity users. ; The clustering module is used to analyze historical daily load samples from electricity users. conduct Clustering yields the optimal number of clusters K; The ensemble clustering module is used to cluster the sample set based on the optimal number of clusters K obtained from the clustering module. Divided into K clusters ,in ; The classification module is used to calculate and regulate the daily baseline load. For each cluster membership degree Obtain the baseline load on the control day. To which the cluster belongs ; User-level adjustment calculation module for baseline load based on the adjustment day. To which category Based on the daily load sample data, calculate the upward and downward adjustment limits of the daily baseline load for power user regulation; The node-level regulation calculation module is used to repeatedly execute the above modules in sequence to obtain the baseline load of all power users participating in demand response in the distribution network and their upward and downward regulation limits, and to obtain the upward and downward regulation limits of system nodes. The system-level multidimensional evaluation model construction module is used to construct a multidimensional evaluation model of the system-level regulation capability of flexible loads in distribution networks based on distribution network power flow constraints. The output module is used to quantify the flexible load regulation capability of the distribution network system based on the multi-dimensional evaluation model constructed by the multi-dimensional evaluation model construction module.

[0052] The advantages of this invention are: (1) This invention uses an integrated clustering algorithm to mine features of historical load data of power users. Multiple base clusters are generated by Bootstrap resampling, and hierarchical clustering integration is performed based on the consensus matrix. This effectively overcomes the problem of single clustering algorithms being sensitive to noise and outliers, and significantly improves the stability and accuracy of load clustering, thus laying the foundation for the accurate quantification of user-level regulation capabilities.

[0053] (2) This invention proposes a three-tier evaluation system from the user level, node level to the system level. In the system-level evaluation, multiple objectives such as smoothing the system load curve, improving the renewable energy absorption capacity and reducing peak-shaving pressure are comprehensively considered. An optimization model that considers the power flow and security constraints of the distribution network is established to achieve a comprehensive and systematic evaluation of the flexible load regulation capacity, providing multi-dimensional decision support for power grid dispatch and renewable energy absorption. Attached Figure Description

[0054] Figure 1 This is a flowchart illustrating the flexible load adjustment capability assessment method based on an integrated clustering algorithm according to an embodiment of the present invention. Figure 2 This is a schematic diagram of the modified IEEE 33-node power distribution system used in the simulation experiment of this invention. Figure 3 This is a schematic diagram illustrating the upward and downward adjustment limits of the power load for a power user in a simulation experiment of an embodiment of the present invention; Figure 4 This is a schematic diagram of the upward and downward adjustment limits of the power load at node 9 in the simulation experiment of this embodiment of the invention; Figure 5 This is a schematic diagram of the system-level regulation capability index of flexible load control in the simulation experiment of an embodiment of the present invention. Detailed Implementation

[0055] 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 in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0056] Example 1 This embodiment provides a flexible load regulation capacity assessment method based on an integrated clustering algorithm. Through multi-level, systematic design and structured coupling, a progressive collaborative optimization assessment architecture is constructed. Specifically, the user-level assessment first relies on the integrated clustering algorithm to accurately cluster historical flexible load data. This avoids the problem that a single algorithm cannot fully adapt to the temporal, high-dimensional, and uncertain nature of load data, leading to easily interfered clustering results and affecting the accuracy of subsequent assessments. Based on this, the user-level quantitative results are integrated through node-level assessment to form the accurate input parameters required for system-level optimization. Finally, in the system-level assessment, an optimization model considering distribution network power flow and security constraints is established. Starting from three objectives—smoothing the load curve, improving renewable energy absorption capacity, and reducing the system peak-valley difference—the adjustment rate of flexible load regulation on load curve smoothness, renewable energy absorption, and system load peak-valley difference is calculated, achieving a multi-dimensional quantitative assessment of system regulation capacity and support for scheduling decisions.

[0057] Specifically, the evaluation method and process are as follows: Figure 1 As shown, it includes the following steps: S1. Obtain the historical daily load sample set of any power user to be evaluated in the distribution network. and regulation days Daily load sample set In this embodiment, the historical daily load samples of the power users participating in flexible load regulation use their historical load data from the most recent year. The sampling interval for the daily load data is 15 minutes, meaning that 96 data points are sampled each day. The formula is as follows:

[0058]

[0059] in, Indicates the electricity user number Daily load data for the day; Indicates the electricity user number Heavenly t Load data sampled at specific times; 365 represents the number of days for daily load data; 96 represents the number of sampling times for daily load. In this embodiment, "control day" refers to the specific date during power system operation on which flexible loads are planned to be adjusted to respond to system demand, i.e., the target day for load adjustment; the date is selected from the range of one week prior to the control day, i.e., the daily load sample set of the 7 days prior to the control day. As shown in the following formula:

[0060]

[0061] in, This indicates that electricity users are on the [number]th day before the control date. Daily load data for the day; This indicates that electricity users are on the [number]th day before the control date. Heavenly t Load data sampled at specific times; 96 represents the number of sampling times for the daily load. S2, Calculate the daily baseline load for electricity user regulation. ; as shown in the following formula:

[0062] S3, Historical daily load sample set of electricity users conduct Clustering yields the optimal number of clusters K; the specific implementation includes the following steps: S31. Determine the number of clusters based on sample characteristics and sample size. k Range, traversal Historical daily load sample set conduct Clustering; based on the sample size in this embodiment, k The initial value range is 2-10; S32. Calculate the number of clusters. k DBI indicator The The calculation is as follows:

[0063] in, Representing the Each category / cluster; Representing the The average intra-class distance of each cluster; Representing the j The average intra-class distance of each cluster; Representing the The first cluster and the first j The distance between the center points of each cluster is calculated using the following formula:

[0064]

[0065] in, Representative clusters The number of samples within; Representative clusters The sample center is taken as its mean in this embodiment, as shown in the following formula:

[0066] in, Representative clusters Daily load data samples.

[0067] S33. Based on the calculation results of step S32, with the minimum Corresponding k Value as the optimal cluster number .

[0068] S4. Based on the optimal clustering number K obtained in step S3, use an ensemble clustering algorithm to cluster the sample set. Divided into K clusters ,in The specific implementation method includes the following steps: S41. In order to reduce K To address the instability of random initialization in the Means clustering algorithm, Bootstrap resampling is employed for the sample set. Perform 100 independent samplings with replacement, drawing 365 samples each time to form a new sample set. The probability of each sample being selected is 1 / 2. As shown in the following formula:

[0069] in, Representing the The sample set obtained by resampling; Representing the The second resampling sample set Daily load data for the day; S42, with the optimal number of clusters As a uniform parameter, for each Bootstrap sample set conduct Clustering, generating the base clustering results, as shown in the following formula:

[0070]

[0071] in, Represents the base cluster set; Representing the Base clustering based on secondary resampling; Representative base clustering The first in Each category / cluster; S43. Constructing a consensus matrix , dimension consensus matrix elements Indicates sample and samples In all base cluster sets The frequency of samples being classified into the same cluster, where samples and samples The first one obtained in step S1 day and day The daily load data is as follows:

[0072]

[0073] S44. Based on the consensus matrix A, hierarchical clustering is used to integrate the base clusters to obtain the final clustering result. The specific execution method is as follows: (1) The sample set Each sample data point is considered an independent cluster. Based on the consensus matrix obtained in step S43, the consensus matrix is ​​calculated sequentially for any two clusters. Similarity between As shown in the following formula

[0074] in, and Clusters The number of samples in the sample; Clusters The set of sample subscripts in; (2) Merge the two clusters with the highest similarity until the number of clusters is [number missing]. The final clustering result is then obtained, as shown in the following formula:

[0075] in, For clusters The first in Daily load data; For clusters The number of samples.

[0076] Thus, the instability of a simple clustering algorithm has been avoided, and the sample set has been completely clustered. Divided Each category is a cluster.

[0077] S5. Calculate and regulate daily baseline load. For each cluster membership degree Obtain the baseline load on the control day. To which the cluster belongs The membership degree The calculation is as follows:

[0078] in, and Representing clusters and clusters The sample center; Adjusting the daily baseline load Classified into the cluster with the highest corresponding membership degree As shown in the following formula:

[0079] in, For clusters The number of samples.

[0080] S6, Baseline load based on the control date To which category Based on daily load sample data, calculate the upward and downward adjustment limits for the daily baseline load of power users; the specific implementation method includes the following steps: S61, by cluster Power matrix constructed from sample data As shown in the following formula:

[0081] in, Representing a cluster The Middle A sample, as shown in the following formula:

[0082] Each column in Indicates the same sampling time Load data; S62. Set the significance level In this embodiment, Calculate the load of electricity users at each sampling time. quantiles and Quantiles, specifically: For sampling time The load, i.e. The first in The data was analyzed using the quantile estimation method to obtain the electricity user load at the sampling time. of quantiles and 0.95 quantile ; S63. Calculate the upward adjustment limit for the daily load control respectively. and downward adjustment limit As shown in the following formula:

[0083]

[0084] S7. Repeat steps S1-S6 to obtain the baseline load and upward and downward adjustment limits of all power users participating in demand response in the distribution network, and calculate the upward and downward adjustment limits of the distribution network nodes; the specific implementation includes the following steps: S71. Repeat steps S1 to S6 to obtain the baseline load, upward adjustment limit, and downward adjustment limit of all power users participating in demand response in the distribution network, denoted as follows: , and , ;in, The number of electricity users participating in demand response; S72, Calculate distribution network nodes i Baseline load of electricity users As shown in the following formula:

[0085] in, It is a node The set of electricity users on the platform; It represents the number of nodes in the distribution network. S73, Calculate distribution network nodes i Upward adjustment limit for electricity users and downward adjustment limit As shown in the following formula:

[0086]

[0087] S8. Construct a multi-dimensional evaluation model for the system-level regulation capability of flexible loads in a distribution network based on power flow constraints; the specific implementation includes the following steps: S81. Set objective functions from three evaluation dimensions: smoothing the system load curve, increasing the amount of renewable energy absorbed, and reducing peak-shaving pressure, as follows:

[0088]

[0089]

[0090] in, For the smoothness of the load curve; This refers to the amount of new energy consumed; This refers to the peak-to-valley difference in system load. This represents the number of nodes in the distribution network. The number of sampling times for the daily load; and Representing distribution network nodes Maximum and minimum daily load power; Indicates distribution network node exist t Load value at any given time; Indicates distribution network node The average daily load; Indicates distribution network node New energy sources have contributed significantly; S82. Set distribution network security constraints for the distribution network-level flexible load regulation capability assessment model. To ensure the consistency and comparability of the assessment results of each model, all three optimization models adopt the same constraints, including power flow balance constraints, grid security constraints, and flexible load regulation limit constraints. The specific constraints are as follows:

[0091]

[0092]

[0093]

[0094]

[0095]

[0096]

[0097] in, Indicated by The set of the starting nodes of the branches of the terminal nodes; Indicated by The set of end nodes of the branches of the first node; and They represent Flowing through the side road Active and reactive power; and They represent Flowing through the side road The active and reactive power; express Time Node The voltage amplitude; Indicates a branch The resistance; Indicates a branch The reactance; Indicates a branch impedance; These represent the upper and lower limits of the voltage, respectively. express Time Branch The carrying capacity; Indicates a branch Safe current; for Time Node Load regulation capability; For nodes The load at The power factor angle at time; for Time Node New energy sources have contributed significantly; for Time Node The predicted value of new energy power generation.

[0098] S9. Based on the multi-dimensional evaluation model constructed in step S8, quantify and output the flexible load regulation capability at the distribution network system level. The specific implementation includes the following steps: S91. Based on the objective function and constraints constructed in step S8, the Gurobi optimizer is used to solve the three optimization models to obtain the optimal system running state under each optimization objective. The following optimal indices are specifically output: (1) Minimum value of load curve smoothness ; (2) Maximum value of new energy consumption ; (3) Minimum value of system load peak-valley difference ; S92. Based on the optimal indicators obtained in step S91, calculate the adjustment rates of flexible load on load curve smoothness, renewable energy absorption, and system load peak-valley difference, as shown in the following formulas:

[0099] in, These are the adjustment rates of flexible load on the smoothness of the load curve, the amount of new energy absorbed, and the peak-valley difference of the system load, respectively. These are the smoothness of the load curve, the amount of new energy consumption, and the peak-valley difference of the system load before the implementation of flexible load control.

[0100] This embodiment also provides simulation experiments applying the above method. The simulation experiments are based on MATLAB R2024b, using the Gurobi optimization solver, and take a modified IEEE 33-bus distribution system as an example. Figure 2 As shown, the distributed power supply connection is as follows: Node 5 is connected to a distributed wind power source with a rated capacity of 1.7MVA, and Node 13 is connected to a distributed wind power source with a rated capacity of 2MVA. The flexible load connection is as follows: Node 9 is connected to 11 flexible power user loads, Node 11 to 9 flexible power user loads, Node 16 to 12 flexible power user loads, Node 20 to 10 flexible power user loads, Node 24 to 11 flexible power user loads, and Node 29 to 9 flexible power user loads; the user loads of the remaining nodes are rigid loads.

[0101] The user-level flexible load adjustment capability assessment based on the ensemble clustering algorithm described in this embodiment is illustrated using the load of a specific electricity user at node 9. Electricity load data for this user throughout 2023 (365 days) was collected, with a time resolution of 15 minutes (i.e., 96 sampling points per day). Partial data is shown in Table 1. Table 1. Partial Data on Electricity Load of a Certain Electricity User in 2023

[0102] Through steps S1-S6 described in the embodiment, the user's baseline load and regulation capacity are obtained, such as... Figure 3 As shown.

[0103] The regulation capacity of other flexible power loads on node 9 is obtained using the same method, thereby obtaining the power load regulation capacity of node 9 (i.e., according to step S7), such as... Figure 4 As shown.

[0104] Similarly, the power load regulation capacity of other nodes can be obtained. Finally, the optimal system indicators (minimum load curve smoothness, maximum renewable energy absorption, and minimum system load peak-valley difference) are calculated according to steps S8-S9, and the regulation capacity indicators of flexible load control (the regulation rate of flexible load on load curve smoothness, renewable energy absorption, and system load peak-valley difference) are obtained, such as... Figure 5 As shown.

[0105] In particular, to further verify the stability of the ensemble clustering algorithm, taking the aforementioned electricity user load data as an example, the ensemble clustering algorithm and individual clustering algorithms were respectively used. The clustering algorithm performed 10 clustering calculations on each user load data and calculated the DBI index of each clustering result. The results are shown in Table 2.

[0106] Table 2 Comparison of DBI index of clustering results

[0107] As shown in Table 2, the DBI index of the K-Means clustering algorithm alone fluctuates significantly across 10 clustering iterations. For example, the DBI indexes of the 4th and 9th clustering results reach 0.403 and 0.481 respectively (a larger DBI value indicates a worse clustering effect), reflecting that its clustering results are significantly affected by random initial centers and have poor stability. In contrast, the DBI index of the ensemble clustering algorithm remains stable between 0.352 and 0.354, with minimal variation, indicating that it has good robustness and effectively overcomes the inconsistency problem of clustering results caused by initialization sensitivity in traditional clustering algorithms.

[0108] Example 2 It should be further explained that, based on the same inventive concept, this embodiment provides a flexible load adjustment capability assessment system based on an integrated clustering algorithm. The system operates using the method described in Embodiment 1, and includes the following modules: The data acquisition module is used to obtain the historical daily load sample set of any power user to be evaluated in the distribution network. and regulation days Daily load sample set ; The control baseline load calculation module is used to calculate the control baseline load of electricity users. ; The clustering module is used to analyze historical daily load samples from electricity users. conduct Clustering yields the optimal number of clusters K; The ensemble clustering module is used to cluster the sample set based on the optimal number of clusters K obtained from the clustering module. Divided into K clusters ,in ; The classification module is used to calculate and regulate the daily baseline load. For each cluster membership degree Obtain the baseline load on the control day. To which the cluster belongs ; User-level adjustment calculation module for baseline load based on the adjustment day. To which category Based on the daily load sample data, calculate the upward and downward adjustment limits of the daily baseline load for power user regulation; The node-level regulation calculation module is used to repeatedly execute the above modules in sequence to obtain the baseline load of all power users participating in demand response in the distribution network and their upward and downward regulation limits, and to obtain the upward and downward regulation limits of system nodes. The system-level multidimensional evaluation model construction module is used to construct a multidimensional evaluation model of the system-level regulation capability of flexible loads in distribution networks based on distribution network power flow constraints. The output module is used to quantify the flexible load regulation capability of the distribution network system based on the multi-dimensional evaluation model constructed by the multi-dimensional evaluation model construction module.

[0109] 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 in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for evaluating flexible load adjustment capability based on ensemble clustering algorithm, characterized in that, Includes the following steps: S1. Obtain the historical daily load sample set of any power user to be evaluated in the distribution network. and regulation days Daily load sample set ; S2, Calculate the daily baseline load for electricity user regulation. ; S3, Historical daily load sample set of electricity users conduct Clustering yields the optimal number of clusters K; S4. Based on the optimal clustering number K obtained in step S3, use an ensemble clustering algorithm to cluster the sample set. Divided into K clusters ,in ; S5. Calculate and regulate daily baseline load. For each cluster membership degree Obtain the baseline load on the control day. To which the cluster belongs ; S6, Baseline load based on the control date To which category Based on the daily load sample data, calculate the upward and downward adjustment limits of the daily baseline load for power user regulation; S7. Repeat steps S1-S6 to obtain the baseline load of all power users participating in demand response in the distribution network and their upward and downward adjustment limits, and calculate the upward and downward adjustment limits of the distribution network nodes. S8. Construct a multi-dimensional evaluation model for the system-level regulation capability of flexible loads in distribution networks based on power flow constraints. S9. Based on the multi-dimensional evaluation model constructed in step S8, quantitatively output the flexible load regulation capability of the distribution network system.

2. The method for evaluating flexible load adjustment capability based on ensemble clustering algorithm according to claim 1, characterized in that, The historical daily load sample set of electricity users mentioned in step S1 As shown in the following formula: in, Indicates the electricity user number Daily load data for the day; Indicates the electricity user number Heavenly t Load data sampled at various times; The number of days for daily load data; The number of sampling times for the daily load; The date of the regulation Daily load sample set As shown in the following formula: in, This indicates that electricity users are on the [number]th day before the control date. Daily load data for the day; This indicates that electricity users are on the [number]th day before the control date. Heavenly t Load data sampled at various times; The number of sampling times for the daily load; Step S2 describes the daily baseline load for the regulation of electricity users. The specific calculation is as follows: 。 3. The method for evaluating flexible load adjustment capability based on ensemble clustering algorithm according to claim 2, characterized in that, Step S3 includes the following steps: S31. Determine the number of clusters based on sample characteristics and sample size. k Range, traversal k Historical daily load sample set conduct Clustering; S32. Calculate the number of clusters. k DBI indicator The The calculation is as follows: in, Representing the Each category / cluster; Representing the The average intra-class distance of each cluster; Representing the j The average intra-class distance of each cluster; Representing the The first cluster and the first j The distance between the center points of each cluster is calculated using the following formula: in, Representative clusters The number of samples within; Representative clusters The sample center is given by the following formula: in, Representative clusters Daily load data samples; S33. Based on the calculation results of step S32, with the minimum Corresponding k Value as the optimal cluster number .

4. The method for evaluating flexible load adjustment capability based on ensemble clustering algorithm according to claim 2, characterized in that, Step S4 includes the following steps: S41. Use Bootstrap resampling technology to sample the data set. conduct Sub-independent sampling with replacement, each time drawing Each sample forms a new sample set, as shown in the following formula: in, Representing the The sample set obtained by resampling; Representing the The second resampling sample set Daily load data for the day; S42, with the optimal number of clusters As a uniform parameter, for each Bootstrap sample set conduct Clustering, generating the base clustering results, as shown in the following formula: in, Represents the base cluster set; Representing the Base clustering based on secondary resampling; Representative base clustering The first in Each category / cluster; S43. Constructing a consensus matrix , dimension consensus matrix elements Indicates sample and samples In all base cluster sets The frequency of being assigned to the same cluster is as follows: S44. Based on the consensus matrix A, hierarchical clustering is used to integrate the base clusters to obtain the final clustering result. The specific execution method is as follows: (1) The sample set Each sample data point is considered an independent cluster. Based on the consensus matrix obtained in step S43, the consensus matrix is ​​calculated sequentially for any two clusters. Similarity between As shown in the following formula in, and Clusters The number of samples in the sample; Clusters The set of sample subscripts in; (2) Merge the two clusters with the highest similarity until the number of clusters is [number missing]. The final clustering result is then obtained, as shown in the following formula: in, For clusters The first in Daily load data; For clusters The number of samples.

5. The method for evaluating flexible load adjustment capability based on ensemble clustering algorithm according to claim 4, characterized in that, The membership degree described in step S5 The calculation is as follows: in, and Representing clusters and clusters The sample center; Adjusting the daily baseline load Classified into the cluster with the highest corresponding membership degree As shown in the following formula: in, For clusters The number of samples.

6. The method for evaluating flexible load adjustment capability based on ensemble clustering algorithm according to claim 5, characterized in that, Step S6 includes the following steps: S61, by cluster Power matrix constructed from sample data As shown in the following formula: in, Representing a cluster The Middle A sample, as shown in the following formula: Each column in Indicates the same sampling time Load data; S62. Set the significance level Calculate the load of electricity users at each sampling time. quantiles and Quantiles, specifically: For sampling time The load, i.e. The first in The data was analyzed using the quantile estimation method to obtain the electricity user load at the sampling time. of quantiles and quantiles ; S63. Calculate the upper and lower adjustment limits for the daily load control respectively. and As shown in the following formula: 。 7. The method for evaluating flexible load adjustment capability based on ensemble clustering algorithm according to claim 6, characterized in that, Step S7 includes the following steps: S71. Repeat steps S1 to S6 to obtain the baseline load, upward adjustment limit, and downward adjustment limit of all power users participating in demand response in the distribution network, denoted as follows: , and , ;in, The number of electricity users participating in demand response; The number of sampling times for the daily load; S72, Calculate distribution network nodes i Baseline load of electricity users As shown in the following formula: in, It is a node The set of electricity users on the platform; It represents the number of nodes in the distribution network. The number of sampling times for the daily load; S73, Calculate distribution network nodes i Upward adjustment limit for electricity users and downward adjustment limit As shown in the following formula: 。 8. The method for evaluating flexible load adjustment capability based on ensemble clustering algorithm according to claim 7, characterized in that, Step S8 includes the following steps: S81. Set objective functions from three evaluation dimensions: smoothing the system load curve, increasing the amount of renewable energy absorbed, and reducing peak-shaving pressure, as follows: in, For the smoothness of the load curve; This refers to the amount of new energy consumed; This refers to the peak-to-valley difference in system load. This represents the number of nodes in the distribution network. The number of sampling times for the daily load; and Representing distribution network nodes Maximum and minimum daily load power; Indicates distribution network node exist Load value at any given time; Indicates distribution network node The average daily load; Indicates distribution network node New energy sources have contributed significantly; S82. Set the distribution network security constraints for the distribution network-level flexible load regulation capability assessment model. The constraint conditions are as follows: in, Indicated by The set of the starting nodes of the branches of the terminal nodes; Indicated by The set of end nodes of the branches of the first node; and They represent Flowing through the side road Active and reactive power; and They represent Flowing through the side road The active and reactive power; express Time Node The voltage amplitude; Indicates a branch The resistance; Indicates a branch The reactance; Indicates a branch impedance; These represent the upper and lower limits of the voltage, respectively. express Time Branch The carrying capacity; Indicates a branch Safe current; for Time Node Load regulation capability; For nodes The load at The power factor angle at time; for Time Node New energy sources have contributed significantly; for Time Node The predicted value of new energy power generation.

9. The method for evaluating flexible load adjustment capability based on ensemble clustering algorithm according to claim 8, characterized in that, Step S9 includes the following steps: S91. Based on the objective function and constraints constructed in step S8, solve the three optimization models to obtain the optimal system operating state under each optimization objective. Specifically, output the following optimal indices: (1) Minimum value of load curve smoothness ; (2) Maximum value of new energy consumption ; (3) Minimum value of system load peak-valley difference ; S92. Based on the optimal indicators obtained in step S91, calculate the adjustment rates of flexible load on load curve smoothness, renewable energy absorption, and system load peak-valley difference, as shown in the following formulas: in, These are respectively the smoothness of the load curve of flexible load, the amount of renewable energy absorbed, and the peak load of the system. The adjustment rate of the trough difference; These are the smoothness of the load curve, the amount of new energy consumption, and the peak-valley difference of the system load before the implementation of flexible load control.

10. A flexible load regulation capacity assessment system based on ensemble clustering algorithm, characterized in that, Includes the following modules: The data acquisition module is used to obtain the historical daily load sample set of any power user to be evaluated in the distribution network. and regulation days Daily load sample set ; The control baseline load calculation module is used to calculate the control baseline load of electricity users. ; The clustering module is used to analyze historical daily load samples from electricity users. conduct Clustering yields the optimal number of clusters K; The ensemble clustering module is used to cluster the sample set based on the optimal number of clusters K obtained from the clustering module. Divided into K clusters ,in ; The classification module is used to calculate and regulate the daily baseline load. For each cluster membership degree Obtain the baseline load on the control day. To which the cluster belongs ; User-level adjustment calculation module for baseline load based on the adjustment day. To which category Based on the daily load sample data, calculate the upward and downward adjustment limits of the daily baseline load for power user regulation; The node-level regulation calculation module is used to repeatedly execute the above modules in sequence to obtain the baseline load of all power users participating in demand response in the distribution network and their upward and downward regulation limits, and to obtain the upward and downward regulation limits of system nodes. The system-level multidimensional evaluation model construction module is used to construct a multidimensional evaluation model of the system-level regulation capability of flexible loads in distribution networks based on distribution network power flow constraints. The output module is used to quantify the flexible load regulation capability of the distribution network system based on the multi-dimensional evaluation model constructed by the multi-dimensional evaluation model construction module.

Images