Virtual power plant participation demand response optimization scheduling method and system
By clustering the power load data of agent users of virtual power plants and identifying different types of users, the problem of poor efficiency and effectiveness of virtual power plants in demand response is solved, and the success rate and strategy accuracy are improved.
Patent Information
- Application Number
- CN202410965732.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-18
- Publication Date
- 2024-11-12
AI Technical Summary
When virtual power plants participate in demand response, due to the complex and changeable load characteristics of the user side, the demand response implementation efficiency and poor results.
The K-means clustering algorithm based on Canopy is used to cluster and analyze the power load data of agent users, identify different types of users, and then formulate accurate demand response strategies to invite corresponding types of users to participate in demand response.
It improves the success rate and efficiency of virtual power plants participating in demand response, helps virtual power plants to formulate accurate and effective demand response strategies, reduces workload, and improves the grasp of the changing trends of power load.
Smart Images

Figure CN118941008A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of electric power technology, and in particular to a method and system for a virtual power plant to participate in demand response optimization scheduling. Background Art
[0002] Virtual power plant is an important direction for the deep integration of energy and information technology. It aggregates one or more resources such as adjustable loads, energy storage and power supply in different spaces to achieve autonomous coordinated optimization control and participate in power system operation and power market transactions. It is a cross-space wide-area source-grid-load-storage integrator. Demand response means that power users can voluntarily respond to the market dynamic price signals and incentive mechanisms, as well as the power supplier's demand for load adjustment. On the premise of meeting the basic power demand of users, the load adjustment demand is achieved by changing the original power usage mode, thereby achieving stable operation of the power system.
[0003] The application of virtual power plant technology to proxy users to participate in demand response can achieve load control of power-consuming equipment, effectively shaving peaks and filling valleys, and improve the flexibility of grid operation. At the same time, based on the settlement and profit sharing signed with proxy users, a profit model for virtual power plants can be created to realize commercial operation of virtual power plants.
[0004] However, the current load characteristics on the user side and the demand side are complex and changeable, and the electricity consumption habits of different types of users vary greatly. This has brought difficulties to the design and implementation of demand response-related strategies for virtual power plants, resulting in low efficiency and poor results in the current demand response implementation of virtual power plants. Summary of the invention
[0005] Purpose of the invention: The first purpose of the present invention is to provide an efficient and effective method for optimizing the scheduling of virtual power plants participating in demand response based on clustering of user electricity consumption characteristics, which can effectively improve the success rate of virtual power plants participating in demand response; the second purpose is to provide a virtual power plant participating in demand response optimizing scheduling system corresponding to the above-mentioned method for optimizing the scheduling of virtual power plants participating in demand response.
[0006] Technical solution: A method for optimizing the dispatching of a virtual power plant in response to demand, comprising the following steps:
[0007] (1) Determine the start-up conditions for demand response based on the power supply and demand situation;
[0008] (2) Publish demand response information for the next day and sign a demand response agreement with the virtual power plant;
[0009] (3) The virtual power plant collects power load data of proxy users;
[0010] (4) Performing cluster analysis on the power load data of proxy users using a clustering algorithm to obtain power load data of different types of proxy users;
[0011] (5) The virtual power plant invites corresponding types of proxy users to participate in demand response based on the electricity load data of different types of proxy users and the demand response information.
[0012] Specifically, demand response includes valley filling demand response and peak shaving demand response. Demand response is used to solve the problem of grid power surplus caused by the large-scale development of new energy such as wind, solar, and hydropower, or grid power shortage caused by peak power loads in winter and summer.
[0013] Specifically, the demand response information includes the demand response implementation time period, the demand response scale and the feedback deadline.
[0014] Specifically, in step (4), the clustering algorithm is a K-means clustering algorithm improved based on Canopy; step (4) includes the following sub-steps:
[0015] (41) The power load data of the proxy user is normalized by a linear function to obtain a data set, and the initial distance thresholds T1 and T2 of the Canopy algorithm are given, and T1>T2;
[0016] (42) Randomly select a sample point from the data set as the first centroid of the Canopy cluster, generate the first Canopy subset for the centroid, and delete the sample point from the data set;
[0017] (43) Randomly select a sample point from the data set again and calculate its distance D to the first centroid of Canopy;
[0018] If D≤T1, the sample point is weakly marked and put into the first Canopy subset;
[0019] If D≤T2, the sample point is strongly marked and put into the first Canopy subset;
[0020] If D>T1, a new Canopy subset is generated for the sample point;
[0021] After completing the above processing, the sample point is deleted from the data set;
[0022] Repeat this step until the number of elements in the data set is zero. In each Canopy subset, the center position of all strongly labeled samples is the centroid of the Canopy subset. A total of K Canopy subsets are obtained, corresponding to K centroids.
[0023] (44) The K centroids obtained after Canopy clustering are set as the initial cluster centers of the K-means clustering algorithm;
[0024] (45) Calculate the similarity between the data points other than the K cluster centers and the K cluster centers respectively, divide the data points into the class with the largest similarity, and complete the first clustering;
[0025] (46) Based on the results of the first clustering, the arithmetic mean of the corresponding dimensions of all data points in each category is calculated as the new cluster center;
[0026] (47) Repeat steps (45) and (46) until the termination condition set by the K-means clustering algorithm is reached, and K different clustering data sets are obtained, corresponding to the power load data of different types of proxy users.
[0027] Specifically, in step (45), the similarity is the Euclidean distance.
[0028] Specifically, in step (47), the termination condition is that the sum of squared errors is locally minimized or the number of iterations reaches a set number.
[0029] Specifically, step (4) also includes: pre-processing the power load data of the proxy user, eliminating invalid data samples with missing data greater than 10%, and then performing linear function normalization processing.
[0030] Specifically, step (5) includes the following sub-steps:
[0031] (51) The virtual power plant confirms its participation in this demand response and uploads the quantity and price information of this demand response;
[0032] (52) The virtual power plant receives the clearing curve for this demand response, performs strategy decomposition on the clearing curve based on the power load data of different types of proxy users, and invites corresponding types of proxy users to participate in this demand response;
[0033] (53) The virtual power plant receives the settlement information of this demand response.
[0034] The present invention also provides a virtual power plant participating in demand response optimization scheduling system, comprising:
[0035] Demand response start module: used to determine the start conditions that meet the demand response according to the power supply and demand conditions;
[0036] Demand response receiving module: used to publish demand response information for the next day and sign a demand response agreement with the virtual power plant;
[0037] Data collection module: used for the virtual power plant to collect power load data of proxy users;
[0038] Cluster analysis module: used to perform cluster analysis on the power load data of proxy users using clustering algorithms to obtain power load data of different types of proxy users;
[0039] Demand response invitation module: used by virtual power plants to invite corresponding types of proxy users to participate in demand response based on the electricity load data of different types of proxy users and demand response information.
[0040] Beneficial effect: Compared with the prior art, the present invention has the following significant effects: the present invention performs cluster analysis on the power load data of the proxy users who have signed contracts with the virtual power plant through the K-means clustering algorithm improved based on Canopy, realizes unsupervised classification of all proxy users, and then mines the power consumption behavior of each category of users, thereby helping the virtual power plant to grasp the changing trends and characteristics of the power load of the proxy users, and screen out the type of proxy users corresponding to the demand response before inviting them, thereby reducing the workload of invitations and improving the efficiency of invitations, as well as improving the success rate of the virtual power plant's participation in demand response, providing theoretical support for the virtual power plant to formulate accurate and effective demand response strategies, and having good interpretability. BRIEF DESCRIPTION OF THE DRAWINGS
[0041] Figure 1 It is a flow chart of the method of embodiment 1 of the present invention.
[0042] Figure 2 This is a flow chart of the K-means clustering algorithm improved based on Canopy in Example 1 of the present invention.
[0043] Figure 3 This is a power load curve diagram of the category 1 proxy user in embodiment 2 of the present invention.
[0044] Figure 4 This is a power load curve diagram of the proxy user of category 2 in embodiment 2 of the present invention.
[0045] Figure 5 This is a power load curve diagram of the category 3 proxy user in embodiment 2 of the present invention.
[0046] Figure 6 This is a power load curve diagram of the category 4 proxy user in embodiment 2 of the present invention. DETAILED DESCRIPTION
[0047] The present invention is further explained below in conjunction with the accompanying drawings and specific implementation methods.
[0048] Example 1
[0049] See also Figure 1 As shown, this embodiment provides a method for a virtual power plant to participate in demand response optimization scheduling, including the following steps:
[0050] (1) The power grid company determines the conditions for starting demand response based on the power supply and demand situation. Demand response includes valley filling demand response and peak shaving demand response. Demand response is used to solve the problem of power grid overcapacity caused by the large-scale development of new energy sources such as wind, solar, and hydropower, or power grid shortage caused by peak power loads in winter and summer.
[0051] (2) The power grid company publishes the demand response information for the next day and signs a demand response agreement with the virtual power plant; the demand response information includes the demand response implementation time period, demand response scale and feedback deadline.
[0052] (3) The virtual power plant collects the power load data of the proxy users. The user load data includes historical operation curves, adjustable capacity, historical participation evaluation and other information.
[0053] (4) Use clustering algorithm to perform cluster analysis on the power load data of proxy users to obtain the power load data of different types of proxy users.
[0054] Please refer to Figure 2 As shown, in this embodiment, the clustering algorithm is a K-means clustering algorithm improved based on Canopy; the Canopy clustering algorithm is a fast pre-clustering algorithm based on density clustering, and the K-means clustering algorithm is an unsupervised clustering algorithm. Through the Canopy clustering algorithm, the initial clustering center of the K-means clustering algorithm can be quickly obtained, which specifically includes the following steps:
[0055] (41) Preprocessing the power load data of proxy users and eliminating invalid data samples with missing data greater than 10%;
[0056] (42) The preprocessed power load data is normalized by a linear function so that the load values of different user types are between 0 and 1, and the data set List = [x1, x2, ... x n ], the cross-validation method is used to determine the initial distance thresholds T1 and T2 of the Canopy algorithm, and T1>T2;
[0057] (43) From the data set List = [x1, x2, ... x n ] randomly selects a sample point, denoted as P, as the first centroid of the Canopy cluster, and generates the first Canopy subset for the centroid P, while deleting the sample point P from the data set;
[0058] (43) From the data set List = [x1, x2, ... x n ] again randomly selects a sample point, denoted as Q, and calculates its distance D to the first centroid P of Canopy;
[0059] If D≤T1, Q is weakly marked and put into the first Canopy subset;
[0060] If D≤T2, Q is strongly marked and put into the first Canopy subset;
[0061] If D>T1, a new Canopy subset is generated for Q;
[0062] After completing the above processing, delete Q from the data set List;
[0063] Repeat this step until the number of elements in the dataset List is zero. In each Canopy subset, the center position of all strongly labeled samples is the centroid of the Canopy subset. A total of K Canopy subsets are obtained, corresponding to K centroids.
[0064] (44) The K centroids obtained after Canopy clustering are set as the initial cluster centers of the K-means clustering algorithm;
[0065] (45) Calculate the similarity between the data points of the data set List except the K cluster centers and the K cluster centers respectively, divide the data points into the class with the largest similarity, and complete the first clustering; in this embodiment, the similarity is the Euclidean distance, and the smaller the Euclidean distance, the greater the similarity between the two points;
[0066] (46) Based on the results of the first clustering, calculate the arithmetic mean of the corresponding dimensions of all data points in each category as the new K cluster centers;
[0067] (47) Repeat steps (45) and (46) until the termination condition set by the K-means clustering algorithm is reached, and K different clustering data sets are obtained, corresponding to the power load data of different types of proxy users; in this embodiment, the termination condition is that the sum of squared errors is locally minimized or the number of iterations reaches a set number.
[0068] (5) The virtual power plant invites corresponding types of proxy users to participate in demand response based on the power load data of different types of proxy users and the demand response information; specifically, the following steps are included:
[0069] (51) The virtual power plant confirms its participation in this demand response and uploads the quantity and price information of this demand response to the power grid company;
[0070] (52) The virtual power plant receives the clearing curve for this demand response issued by the power grid company, performs strategy decomposition on the clearing curve based on the power load data of different types of proxy users, and invites corresponding types of proxy users to participate in this demand response;
[0071] (53) The virtual power plant receives the settlement information of this demand response issued by the power grid company.
[0072] The above method is explained below in conjunction with a specific application scenario.
[0073] Example 2
[0074] With respect to the method for optimizing scheduling of a virtual power plant participating in demand response provided in Example 1, this embodiment provides an application example of the method in a real-life scenario.
[0075] (1) Based on recent power operation data, the power grid company predicted that there would be a power shortage at noon the next day. In order to ensure the safe and stable operation of the power grid, it was decided to initiate peak demand response after analysis and judgment.
[0076] (2) The relevant departments of the power grid company issue a demand response demand for peak shaving the next day to the virtual power plant, where the demand response start and end time is 12:00-14:00, the virtual power plant's control capacity is 10MW, and the feedback deadline is before 14:00; and sign a demand response agreement with the virtual power plant.
[0077] (3) The virtual power plant collects relevant information such as the operating load curves of proxy users on a daily basis. Taking a virtual power plant operator as an example, its proxy users include commercial buildings, industrial loads, residential users and other resources. After receiving the demand response information released by the power grid company, the virtual power plant operator can analyze the load curves of proxy users according to categories such as working days, non-working days, and holidays based on the actual situation of the next day.
[0078] (4) Taking working days as an example, the daily load curve data of all proxy users on working days is used for explanation. Each curve contains 96 data points (the sampling interval is 15 minutes).
[0079] First, the data is preprocessed to eliminate invalid data samples with missing data greater than 10%, and the preprocessed data is normalized by linear function so that the load values of different user types are between 0 and 1 to obtain a data set. Then, the virtual power plant operator applies the Canopy-based improved K-means clustering algorithm provided in Example 1 to perform clustering analysis on the data set. In this embodiment, the values of T1 and T2 of the Canopy algorithm are determined by cross-validation. After several iterative experiments, the values of T1 and T2 are determined to be 6.2 and 6 respectively, and the initial clustering K value is 4. Finally, 4 different types of clustering results are obtained, corresponding to the electricity load data of 4 categories of users. The electricity load data of the above 4 categories of users are plotted into an electricity load curve.
[0080] The following is an analysis of the power load curve: Please refer to Figure 3The figure shows the power load curve of the proxy user of category 1. The load peak occurs at around 19:00. The load of this category of users is relatively low during the rest of the time. The user profile of this category is a residential user who works during the day. Please refer to Figure 4 The figure shows the power load curve of the proxy user of category 2. The load peak occurs between 11:00 and 20:00, and the power load drops rapidly after 21:00. The user profile of this category is a large commercial user; please refer to Figure 5 As shown in the figure, it is the power load curve of the proxy user of category 3. The power consumption changes smoothly throughout the day and the load is low. It can be judged that the potential of this category of users to participate in demand response is small; please refer to Figure 6 As shown, this is the electricity load curve of the proxy users of category 4. The load peaks occur around 11 o'clock and 19 o'clock, and there is a continuous high electricity consumption in the middle. The user profile of this category is a residential user who stays at home during the day.
[0081] (5) Based on the time period of this demand response (12:00-14:00), and according to the analysis results of the electricity load curve in step (4), the virtual power plant finds that the proxy users of category 1 and category 3 are obviously not suitable for participating in this demand response. The virtual power plant invites the proxy users of category 2 and category 4 to participate in this demand response until the regulation capacity of 10MW for this demand response is met.
[0082] Example 3
[0083] This embodiment provides a virtual power plant participating in demand response optimization scheduling system corresponding to a virtual power plant participating in demand response optimization scheduling method provided in Embodiment 1, the system comprising:
[0084] Demand response start module: used by power grid companies to determine the start conditions for demand response based on power supply and demand conditions;
[0085] Demand response receiving module: used by the power grid company to publish demand response information for the next day and sign a demand response agreement with the virtual power plant;
[0086] Data collection module: used for the virtual power plant to collect power load data of proxy users;
[0087] Cluster analysis module: used to perform cluster analysis on the power load data of proxy users using clustering algorithms to obtain power load data of different types of proxy users;
[0088] Demand response invitation module: used by virtual power plants to invite corresponding types of proxy users to participate in demand response based on the electricity load data of different types of proxy users and demand response information.
[0089] In summary, the present invention uses a clustering algorithm to perform cluster analysis on the power load data of proxy users who have signed contracts with virtual power plants, and obtains the power load data of various types of proxy users, so as to invite corresponding proxy users to participate in demand response according to demand response information. The clustering algorithm part uses the Canopy clustering algorithm for pre-processing, which effectively solves the problem of uncertainty of the initial cluster center and K value of the K-means clustering algorithm, and improves the overall calculation efficiency of the algorithm. The present invention simplifies the strategy of virtual power plants participating in demand response and improves the success rate of virtual power plants participating in demand response.
Claims
1. A method for optimizing the scheduling of virtual power plants in response to demand, characterized in that: The following steps are involved: (1) Determine the start-up conditions for demand response based on the power supply and demand situation; (2) Based on the demand response information of the next day, a demand response agreement is signed with the virtual power plant; (3) The virtual power plant collects power load data of proxy users; (4) Performing cluster analysis on the power load data of proxy users using a clustering algorithm to obtain power load data of different types of proxy users; (5) The virtual power plant invites corresponding types of proxy users to participate in demand response based on the electricity load data of different types of proxy users and the demand response information.
2. The method for optimizing the scheduling of virtual power plants participating in demand response according to claim 1 is characterized in that: The demand response includes valley filling demand response and peak shaving demand response.
3. The method for optimizing the scheduling of virtual power plants participating in demand response according to claim 1 is characterized in that: The demand response information includes the demand response implementation time period, the demand response scale and the feedback deadline.
4. The method for optimizing the scheduling of virtual power plants participating in demand response according to claim 1 is characterized in that: In the step (4), the clustering algorithm is a K-means clustering algorithm improved based on Canopy.
5. The method for optimizing the scheduling of virtual power plants participating in demand response according to claim 4 is characterized in that: The step (4) comprises the following sub-steps: (41) The power load data of the proxy user is normalized by a linear function to obtain a data set, and the initial distance thresholds T1 and T2 of the Canopy algorithm are given, and T1>T2; (42) Randomly select a sample point from the data set as the first centroid of the Canopy cluster, generate the first Canopy subset for the centroid, and delete the sample point from the data set; (43) Randomly select a sample point from the data set again and calculate its distance D to the first centroid of Canopy; If D≤T1, the sample point is weakly marked and put into the first Canopy subset; If D≤T2, the sample point is strongly marked and put into the first Canopy subset; If D>T1, a new Canopy subset is generated for the sample point; After completing the above processing, the sample point is deleted from the data set; Repeat this step until the number of elements in the data set is zero. In each Canopy subset, the center position of all strongly labeled samples is the centroid of the Canopy subset. A total of K Canopy subsets are obtained, corresponding to K centroids. (44) The K centroids obtained after Canopy clustering are set as the initial cluster centers of the K-means clustering algorithm; (45) Calculate the similarity between the data points other than the K cluster centers and the K cluster centers respectively, divide the data points into the class with the largest similarity, and complete the first clustering; (46) Based on the results of the first clustering, the arithmetic mean of the corresponding dimensions of all data points in each category is calculated as the new cluster center; (47) Repeat steps (45) and (46) until the termination condition set by the K-means clustering algorithm is reached, and K different clustering data sets are obtained, corresponding to the power load data of different types of proxy users.
6. The method for optimizing the scheduling of virtual power plants participating in demand response according to claim 5 is characterized in that: In the step (45), the similarity is the Euclidean distance.
7. The method for optimizing the scheduling of virtual power plants participating in demand response according to claim 5 is characterized by: In the step (47), the termination condition is that the sum of squared errors is locally minimized or the number of iterations reaches a set number.
8. The method for optimizing the scheduling of virtual power plants participating in demand response according to claim 5 is characterized by: The step (4) further includes: pre-processing the power load data of the proxy user, eliminating invalid data samples with missing data greater than 10%, and then performing linear function normalization processing.
9. The method for optimizing the scheduling of virtual power plants participating in demand response according to claim 1 is characterized by: The step (5) comprises the following sub-steps: (51) The virtual power plant confirms its participation in this demand response and uploads the quantity and price information of this demand response; (52) The virtual power plant receives the clearing curve of this demand response, performs strategy decomposition on the clearing curve according to the power load data of different types of proxy users, and invites corresponding types of proxy users to participate in this demand response; (53) The virtual power plant receives the settlement information of this demand response.
10. A virtual power plant participating in demand response optimization scheduling system, characterized in that: include: Demand response start module: used to determine the start conditions that meet the demand response according to the power supply and demand conditions; Demand response receiving module: used to sign a demand response agreement with the virtual power plant based on the demand response information of the next day; Data collection module: used for the virtual power plant to collect power load data of proxy users; Cluster analysis module: used to perform cluster analysis on the power load data of proxy users using clustering algorithms to obtain power load data of different types of proxy users; Demand response invitation module: used by virtual power plants to invite corresponding types of proxy users to participate in demand response based on the electricity load data of different types of proxy users and demand response information.
Citation Information
Cited By
Differentiation load directrix generation method based on load response characteristic clustering
CN122136922A
