A method and system for estimating cluster baseline load

Abstract

This invention relates to a cluster baseline load estimation method and system, relating to the field of power system demand response. The method includes: acquiring electricity load data of a residential user cluster at a set time; the set time includes a demand response day and a set number of similar days prior to the demand response day; the similar days are historical dates with the same day type as the demand response day; performing first-level clustering and accumulation on all users of the residential user cluster to obtain first-level clusters; performing second-level clustering and adjustment on the first-level clusters to obtain the optimal cluster of the user cluster; and estimating the electricity load based on the optimal cluster of the user cluster and the electricity load data to obtain the estimated baseline load value of the residential user cluster. This invention can improve the accuracy of cluster baseline load estimation.

Description

Technical Field

[0001] This invention relates to the field of power system demand response, and in particular to a method and system for estimating cluster baseline load. Background Technology

[0002] With the continuous construction of new power systems, Demand Response (DR) has gained widespread attention due to its advantages in improving grid flexibility and facilitating the integration of renewable energy. Currently, DR is mainly divided into two types: price-based and incentive-based. In incentive-based DR, system operators purchase DR response services from virtual power plants (VPS) or load aggregators. VPS or load aggregators incentivize users to change their electricity consumption behavior within a specific time period by paying compensation, thereby meeting the system operator's requirements. When system operators transact with VPS or load aggregators, the user clusters represented by the VPS or load aggregators participate as a whole. Therefore, it is necessary to estimate the baseline load of the user clusters as a basis for measuring the effectiveness of DR implementation and calculating the DR response compensation accordingly. In summary, accurate baseline load estimation of user clusters is crucial for system operators, VPS, and load aggregators.

[0003] Currently, there is no separate method for estimating cluster baseline load; the sum of individual baseline loads is typically used as the cluster baseline load. However, individual baseline estimations are often ineffective, and using a simple summation method to calculate the cluster baseline load leads to the continuous accumulation of individual baseline estimation errors, resulting in a large error in the final estimation result. To improve the accuracy of user cluster baseline load estimation, it is urgent to propose a scientific and effective method for estimating cluster baseline load. Summary of the Invention

[0004] The purpose of this invention is to provide a cluster baseline load estimation method and system to improve the accuracy of cluster baseline load estimation.

[0005] To achieve the above objectives, the present invention provides the following solution:

[0006] A cluster baseline load estimation method includes:

[0007] Obtain electricity load data for a residential user cluster at a set time; the set time includes the demand response date and a set number of similar days prior to the demand response date; the similar days are historical dates with the same day type as the demand response date;

[0008] Perform first-level clustering and summation on all users in the aforementioned resident user cluster to obtain first-level clusters;

[0009] The first-level clusters are then subjected to second-level clustering and adjustment to obtain the optimal clustering of the user clusters.

[0010] Based on the optimal clustering of the user cluster and the electricity load data, the electricity load is estimated to obtain the baseline load estimate of the residential user cluster.

[0011] Optionally, the step of performing first-level clustering and summation on all users of the resident user cluster to obtain first-level clusters specifically includes:

[0012] Based on the set number of first-level clusters, all users in the resident user cluster are clustered into first-level clusters to obtain the first-level clustering results;

[0013] The user load data in all clusters of the first-level clustering result are accumulated to obtain the first-level clusters.

[0014] Optionally, the step of performing secondary clustering and adjustment on the first-level clusters to obtain the optimal clustering of the user cluster specifically includes:

[0015] The primary clusters are then subjected to secondary clustering based on a set number of secondary clusters to obtain the secondary clustering results.

[0016] Based on each secondary cluster in the secondary clustering results, load estimation is performed for the non-demand response period on the demand response day to obtain the load estimate value for the non-demand response period on the demand response day for each secondary cluster.

[0017] The load estimates for the non-demand response period on the demand response day of each of the secondary clusters are summed to obtain the estimated electricity load of the residential user cluster during the non-demand response period on the demand response day.

[0018] The secondary clustering results are adjusted and iterated based on the estimated electricity load of the residential user cluster during the non-demand response period on the demand response day to obtain the optimal clustering of the user cluster.

[0019] Optionally, the step of estimating the electricity load based on the optimal clustering of the user cluster and the electricity load data to obtain the baseline load estimate of the residential user cluster specifically includes:

[0020] The electricity load data of all users in all secondary clusters of the optimal cluster of the user cluster at the set time are summed to obtain the electricity load curve of the user cluster.

[0021] Based on the power load curve of the user cluster, a baseline estimate is performed to obtain the baseline load estimate for the demand response period on the demand response day;

[0022] The baseline load estimates for the demand response period on all the demand response days of the aforementioned user clusters are summed to obtain the baseline load estimate for the residential user clusters.

[0023] Optionally, adjusting and iterating the secondary clustering results based on the estimated electricity load of the residential user cluster during the non-demand response period on the demand response day to obtain the optimal clustering of the user cluster specifically includes:

[0024] Based on the electricity load estimation results of the residential user cluster during the non-demand response period on the demand response day, determine the initial average absolute percentage error and the initial average percentage deviation of the electricity load estimation of the residential user cluster during the non-demand response period on the demand response day.

[0025] Any user in the secondary cluster of the secondary clustering result is moved to another secondary cluster to obtain a new secondary clustering result; the other secondary clusters are secondary clusters other than the secondary cluster where the user was originally located;

[0026] Based on the new secondary clustering results, the adjusted average absolute percentage error and the adjusted average percentage deviation of the electricity load estimation for residential user clusters during the non-demand response period on the demand response day are determined.

[0027] Determine whether the mean absolute percentage error after the adjusted clustering is less than the mean absolute percentage error, and whether the mean percentage deviation after the adjusted clustering is less than the initial mean percentage deviation, to obtain a first determination result;

[0028] If the first judgment result is yes, then the initial average absolute percentage error and the initial average percentage deviation are updated using the adjusted average absolute percentage error and the adjusted average percentage deviation, and the new secondary clustering result is taken as the current optimal clustering; if the first judgment result is no, then the secondary clustering result before the shift is determined as the current optimal clustering; until all the secondary clusters are traversed;

[0029] Replace the user in the secondary cluster of the secondary clustering result, and return to the step "move any user in the secondary cluster of the secondary clustering result into another secondary cluster to obtain a new secondary clustering result; the other secondary clusters are secondary clusters other than the secondary cluster where the user was before being moved", until all users in all the secondary clusters are traversed;

[0030] Determine whether the termination condition has been met to obtain a second determination result. If the second determination result is yes, then determine that the secondary clustering result of the current iteration is the optimal clustering of the user cluster.

[0031] If the second judgment result is negative, then the secondary cluster of the secondary clustering result is replaced;

[0032] Select any user from the replaced secondary cluster and return to the step "Move any user from the secondary cluster of the secondary clustering result into another secondary cluster to obtain a new secondary clustering result".

[0033] Optionally, the termination condition is:

[0034]

[0035] Or r = r max

[0036] Where ξ1 is the first set threshold, ξ2 is the second set threshold, and r is the current number of iterations for cluster adjustment. max ε is the upper limit of the number of iterations. MAPE (r) represents the mean absolute percentage error in the r-th cycle, ε MPB λ(r) represents the average percentage deviation in the r-th cycle, λ1(r) represents the absolute value of the change in the average percentage error between the r-th cycle and the previous cycle, λ2(r) represents the absolute value of the change in the average percentage deviation between the r-th cycle and the previous cycle, and ε MAPE (r-1) represents the mean absolute percentage error in the (r-1)th cycle, ε MPB (r-1) represents the average percentage deviation in the r-th cycle.

[0037] A cluster baseline load estimation system includes:

[0038] The acquisition module is used to acquire electricity load data of a residential user cluster at a set time; the set time includes the demand response day and a set number of similar days before the demand response day; the similar days are dates in historical dates that have the same day type as the demand response day;

[0039] The first-level clustering and accumulation module is used to perform first-level clustering and accumulation on all users of the resident user cluster to obtain first-level clusters;

[0040] The secondary clustering and adjustment module is used to perform secondary clustering and adjustment on the primary clusters to obtain the optimal clustering of the user clusters;

[0041] The electricity load estimation module is used to estimate the electricity load based on the optimal clustering of the user cluster and the electricity load data, so as to obtain the baseline load estimate of the residential user cluster.

[0042] Optionally, the first-level clustering and accumulation module specifically includes:

[0043] A first-level clustering unit is used to perform first-level clustering on all users of the resident user cluster according to a set number of first-level clusters to obtain the first-level clustering result;

[0044] The accumulation unit is used to accumulate user load data from all clusters in the first-level clustering result to obtain the first-level clusters.

[0045] Optionally, the secondary clustering and adjustment module specifically includes:

[0046] The secondary clustering unit is used to perform secondary clustering on the primary clusters according to a set number of secondary clusters to obtain the secondary clustering results;

[0047] The non-demand response period load estimation unit on the demand response day is used to estimate the non-demand response period load on the demand response day based on each secondary cluster in the secondary clustering results, and obtain the non-demand response period load estimate value for each secondary cluster on the demand response day.

[0048] The summation unit is used to sum the load estimates for the non-demand response period on the demand response day for each of the secondary clusters, so as to obtain the estimated electricity load of the residential user cluster during the non-demand response period on the demand response day;

[0049] The adjustment and iteration unit is used to adjust and iterate the secondary clustering results based on the estimated electricity load of the residential user cluster during the non-demand response period on the demand response day, so as to obtain the optimal clustering of the user cluster.

[0050] Optionally, the power load estimation module specifically includes:

[0051] The user cluster power load curve determination unit is used to accumulate all power load data of all users in all secondary clusters of the optimal cluster of the user cluster at the set time to obtain the user cluster power load curve.

[0052] The baseline estimation unit is used to perform baseline estimation based on the power load curve of the user cluster to obtain the baseline load estimate value for the demand response period on the demand response day;

[0053] The residential user cluster baseline load estimation unit is used to accumulate the baseline load estimates of all user clusters during the demand response period of the demand response day to obtain the residential user cluster baseline load estimate.

[0054] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0055] This invention acquires electricity load data for a residential user cluster at a set time. The set time includes the demand response date and a set number of similar days prior to the demand response date. The similar days are historical dates with the same day type as the demand response date. First-level clustering and accumulation are performed on all users in the residential user cluster to obtain first-level clusters. Second-level clustering and adjustment are then performed on the first-level clusters to obtain the optimal cluster for the user cluster. Based on the optimal cluster and the electricity load data, electricity load estimation is performed to obtain the baseline load estimate for the residential user cluster. This invention, through the linkage of clustering and estimation, obtains the user clustering scheme most conducive to reducing estimation errors, which can effectively improve the accuracy of cluster baseline load estimation. Attached Figure Description

[0056] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0057] Figure 1 The flowchart of the cluster baseline load estimation method provided by the present invention is shown below;

[0058] Figure 2 This is a schematic diagram of the cluster baseline load estimation system provided by the present invention;

[0059] Figure 3 The flowchart shows the cluster baseline load estimation method provided by this invention in a specific application. Detailed Implementation

[0060] 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 embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0061] The purpose of this invention is to provide a cluster baseline load estimation method and system to improve the accuracy of cluster baseline load estimation.

[0062] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0063] like Figure 1 As shown, the present invention provides a cluster baseline load estimation method, comprising:

[0064] Step 101: Obtain the electricity load data of the residential user cluster at a set time; the set time includes the demand response day and a set number of similar days before the demand response day; the similar days are dates in the historical dates that have the same day type as the demand response day.

[0065] Step 102: Perform first-level clustering and summation on all users in the resident user cluster to obtain first-level clusters. Step 102 specifically includes:

[0066] The residents in the user cluster are clustered into primary clusters according to the set primary cluster number to obtain the primary clustering results; the user load data in all clusters in the primary clustering results are accumulated to obtain the primary clusters.

[0067] Step 103: Perform secondary clustering and adjustment on the first-level clusters to obtain the optimal clustering of the user clusters.

[0068] Step 103 specifically includes:

[0069] The primary clusters are clustered into secondary clusters according to a set number of secondary clusters to obtain secondary clustering results. Load estimation for the non-demand response period on the demand response day is performed for each secondary cluster in the secondary clustering results to obtain the load estimate for the non-demand response period on the demand response day for each secondary cluster. The load estimates for the non-demand response period on the demand response day for each secondary cluster are summed to obtain the estimated electricity load of the residential user cluster during the non-demand response period on the demand response day. The secondary clustering results are adjusted and iterated based on the estimated electricity load of the residential user cluster during the non-demand response period on the demand response day to obtain the optimal clustering of the user cluster.

[0070] Specifically, adjusting and iterating the secondary clustering results based on the estimated electricity load of the residential user cluster during the non-demand response period on the demand response day to obtain the optimal clustering of the user cluster includes:

[0071] Based on the electricity load estimation results of the residential user cluster during the non-demand response period on the demand response day, determine the initial average absolute percentage error and the initial average percentage deviation of the electricity load estimation of the residential user cluster during the non-demand response period on the demand response day.

[0072] Any user in the secondary cluster of the secondary clustering result is moved to another secondary cluster to obtain a new secondary clustering result; the other secondary clusters are secondary clusters other than the secondary cluster where the user was originally located;

[0073] Based on the new secondary clustering results, the adjusted average absolute percentage error and the adjusted average percentage deviation of the electricity load estimation for residential user clusters during the non-demand response period on the demand response day are determined.

[0074] Determine whether the mean absolute percentage error after the adjusted clustering is less than the mean absolute percentage error, and whether the mean percentage deviation after the adjusted clustering is less than the initial mean percentage deviation, to obtain a first determination result;

[0075] If the first judgment result is yes, then the initial average absolute percentage error and the initial average percentage deviation are updated using the adjusted average absolute percentage error and the adjusted average percentage deviation, and the new secondary clustering result is taken as the current optimal clustering; if the first judgment result is no, then the secondary clustering result before the shift is determined as the current optimal clustering; until all the secondary clusters are traversed;

[0076] Replace the user in the secondary cluster of the secondary clustering result, and return to the step "move any user in the secondary cluster of the secondary clustering result into another secondary cluster to obtain a new secondary clustering result; the other secondary clusters are secondary clusters other than the secondary cluster where the user was before being moved", until all users in all the secondary clusters are traversed;

[0077] Determine whether the termination condition has been met to obtain a second determination result. If the second determination result is yes, then determine that the secondary clustering result of the current iteration is the optimal clustering of the user cluster.

[0078] If the second judgment result is negative, then the secondary cluster of the secondary clustering result is replaced;

[0079] Select any user from the replaced secondary cluster and return to the step "Move any user from the secondary cluster of the secondary clustering result into another secondary cluster to obtain a new secondary clustering result".

[0080] The termination condition is as follows:

[0081]

[0082] Or r = r max

[0083] Where ξ1 is the first set threshold, ξ2 is the second set threshold, and r is the current number of iterations for cluster adjustment. max ε is the upper limit of the number of iterations. MAPE (r) represents the mean absolute percentage error in the r-th cycle, ε MPBλ(r) represents the average percentage deviation in the r-th cycle, λ1(r) represents the absolute value of the change in the average percentage error between the r-th cycle and the previous cycle, λ2(r) represents the absolute value of the change in the average percentage deviation between the r-th cycle and the previous cycle, and ε MAPE (r-1) represents the mean absolute percentage error in the (r-1)th cycle, ε MPB (r-1) represents the average percentage deviation in the r-th cycle.

[0084] Step 104: Based on the optimal clustering of the user cluster and the electricity load data, perform electricity load estimation to obtain the baseline load estimate of the residential user cluster.

[0085] Step 104 specifically includes:

[0086] The electricity load data of all users in all secondary clusters of the optimal cluster of the user cluster at the set time are accumulated to obtain the electricity load curve of the user cluster; a baseline estimate is performed based on the electricity load curve of the user cluster to obtain the baseline load estimate of the demand response period on the demand response day; the baseline load estimates of the demand response period on the demand response day of all user clusters are accumulated to obtain the baseline load estimate of the residential user cluster.

[0087] This invention aims to improve the accuracy of existing cluster baseline estimation methods. The invention performs first-level clustering and summation of user clusters, followed by second-level clustering. By adjusting the clustering scheme based on load estimation results during non-DR periods on DR days, the optimal clustering of user clusters is obtained. Finally, the baseline load of each cluster is estimated, and these are summed to obtain the final user baseline load estimation result. This invention, through the linkage of clustering and estimation, obtains the user clustering scheme most conducive to reducing estimation errors, which can effectively improve the accuracy of cluster baseline load estimation. This, in turn, improves the settlement accuracy of virtual power plants and load aggregators participating in incentive-based DR, and is of great significance to the further development and future promotion of DR.

[0088] like Figure 3 As shown, this invention proposes a user cluster baseline load estimation method with a workflow in practical applications. The method includes the following steps:

[0089] 1) Determine the day type, similar days to the DR day, the DR period of the DR day, and the non-DR period of the DR day for each date; the specific steps are as follows:

[0090] 1-1) According to the legal arrangement, the dates are classified into working days and non-working days to determine the day type of each date.

[0091] 1-2) Select dates in the historical dates that have the same day type as the DR date as similar dates to the DR date.

[0092] 1-3) During a DR day, the period during which DR is performed is designated as the DR period; the period during which no DR is performed is designated as the non-DR period.

[0093] 2) For a residential user cluster containing M (M≥100) users, for each user m, obtain the electricity load data of the day on the DR date and the D nearest similar days before the DR date by reading smart meter data or other means.

[0094] 3) Perform first-level clustering and summation on all users in the residential user cluster; the specific steps are as follows:

[0095] 3-1) Set the number of first-level clusters K1, K1≤M.

[0096] 3-2) Using any clustering method, such as the K-means method, cluster M individual users into K1 user clusters. Let the first-order clustering result be... Among them, C 1 The groups formed by this clustering, This is the cluster consisting of all users of the k1th class in this clustering.

[0097] 3-3) For all clusters, the load of all users contained within them is summed up. After summing, each cluster forms a load matrix, which becomes a "large user". All clusters generate a total of K1 "large users". The "large user" cluster is a first-level cluster.

[0098] 4) Perform secondary clustering and adjustment on the "large user" cluster obtained in step 3) to obtain the optimal clustering of user clusters; the specific steps are as follows:

[0099] 4-1) Set the number of secondary clusters K2, which can be 5.

[0100] 4-2) Using any clustering method, such as K-means, cluster the K1 "large users" into K2 user clusters. Let the result of the second-order clustering be... Among them, C 2 The groups formed by this clustering, This is a secondary cluster composed of all users of the k2th class in this clustering.

[0101] 4-3) For the results of secondary clustering Each secondary cluster is treated as a whole, and the non-DR period load on the DR day under that cluster is estimated to obtain the corresponding non-DR period load estimate for the DR day. Any existing estimation method can be used; the secondary clusters are estimated using the method specified in the current Chinese national standard "Technical Requirements for Measurement and Verification of Power Saving in Demand Response of Electricity Users GB / T 37016-2018". Taking the electricity load during non-DR hours on a DR day as an example:

[0102]

[0103] In the formula, τ represents the set of non-DR time periods on a DR day. Representing a cluster Estimated electricity load during non-DR periods on DR day; GB(N,min,max) represents the N-2 days remaining after removing the two days with the highest and lowest load from the N similar days before DR day, where N can be 7; Representing a cluster The electricity load during the τ period on day d; K is a correction factor, which is the ratio of the estimated load value to the actual load value for the first two hours of the DR period, and its value range is limited to 0.8 to 1.2.

[0104] 4-4) Using the results of 4-3), sum the electricity load values ​​for each secondary cluster during the non-DR period on the DR day to obtain the estimated electricity load of the user cluster during the non-DR period on the DR day:

[0105]

[0106] in, This is the estimated electricity load of the user cluster during the non-DR period on DR days.

[0107] 4-5) Based on the results of 4-4), calculate the average absolute percentage error ε of the power load estimation during non-DR periods on the DR day for the user cluster. MAPE and average percentage deviation ε MPB If we let the time series L = {l1, l2, ..., l...} n} represents the actual value of the electricity load to be estimated, and the corresponding estimated value is represented by a time series. The formulas for calculating the mean absolute percentage error and the mean percentage deviation are as follows:

[0108]

[0109]

[0110] Where n represents the number of sampling points in the time series.

[0111] 4-6) Select a specific "large user" in a certain secondary cluster, move it to a new secondary cluster, and adjust the clustering results; then calculate the mean absolute percentage error after adjustment according to equation (3), denoted as ε. MAPE * Calculate the average percentage deviation after clustering according to equation (4), denoted as ε. MPB *

[0112] Note that when performing this step again, the new second-level cluster to which the specific "large user" is about to be moved must not be the same as any second-level cluster that the "large user" has previously existed.

[0113] 4-7) The ε obtained in step 4-5) MAPE ε MPB The ε obtained in steps 4-6) MAPE *、ε MPB * Compare, if ε MAPE *<ε MAPE And ε MPB *<ε MPB If the adjusted clustering method is proven to be better, then it is taken as the current optimal clustering; otherwise, the result of the adjusted clustering is discarded, and the previous clustering method is considered to be the current optimal clustering.

[0114] 4-8) Let ε be the mean absolute percentage error corresponding to the current best cluster. MAPE The average percentage deviation is denoted as ε. MPB .

[0115] 4-9) Execute steps 4-6), 4-7), and 4-8 in sequence until all K2 second-level clusters are traversed.

[0116] 4-10) Select a new "large user" as the specific "large user", and execute steps 4-6), 4-7), 4-8), and 4-9 in sequence until all K1 large users are traversed.

[0117] 4-11) Execute steps 4-6), 4-7), 4-8), 4-9), and 4-10 sequentially until the termination condition is met. The clustering method at this point is the optimal clustering of the user cluster. The termination condition is as follows:

[0118]

[0119] In the formula: ξ1 and ξ2 are manually set thresholds, which can be 0.001, 0.0005, or other values ​​respectively; r is the current number of iterations for cluster adjustment; r max This is the maximum number of loop iterations, which can be 5 or other values; it can be set according to requirements in practical applications.

[0120] The following is a specific example to illustrate the user cluster optimization process:

[0121] For the first type The first "large user" in the class is moved from its current class to a new class, for example, the large user is placed in... In the middle, the cluster baseline load estimation results on the validation set under the current clustering method are calculated to obtain ε. MAPE* and ε MPB *. If ε MAPE *<ε MAPE And ε MPB *<ε MPB If it is proven that the adjusted clustering method is better, then it is retained as the current optimal clustering method; otherwise, the result of this adjustment is discarded, and the previous clustering method is considered to be the current optimal clustering method.

[0122] The "large user" is then placed into another class to continue comparing the cluster baseline estimation results on the validation set, until all K2 classes have been traversed. The same steps are then performed on the next "large user," continuing until all K1 "large users" have been traversed. The first round of complete cluster adjustment is now finished, yielding the clustering method that most accurately estimates the resident user cluster baseline after the first round of adjustment.

[0123] Cluster adjustments will be repeated multiple times until the termination condition is met.

[0124] 5) Under the optimal clustering of user clusters obtained in step 4), estimate the electricity load during the DR period on the DR day to obtain the final baseline load estimate for the user clusters; the specific steps are as follows:

[0125] 5-1) Under the optimal clustering of user clusters, for all K2 secondary clusters, sum up the electricity load of all users contained therein to obtain the electricity load curves of K2 sub-user clusters.

[0126] 5-2) For the electricity load curves of all K2 sub-user clusters, use any existing baseline estimation method, such as the method specified in the current Chinese national standard "Technical Requirements for Measurement and Verification of Power Saving in Demand Response of Electricity Users GB / T 37016-2018", to obtain the baseline load estimate for each user during the DR period on the DR day.

[0127] 5-3) Sum the baseline load estimates of all K2 sub-user clusters to obtain the final user cluster baseline load estimate.

[0128] 1. This invention employs a two-level clustering model. First, the user cluster is clustered into several sub-clusters. The user load in each sub-cluster is summed as a whole for secondary clustering. Next, the clustering scheme is adjusted based on the load estimation performance of the user cluster during non-DR periods on DR days. Then, the load of all users in each secondary cluster is summed as a whole for estimation. Finally, the estimates of all secondary clusters are summed to obtain the final baseline load estimate of the user cluster. This method significantly improves the effectiveness of user cluster baseline load estimation.

[0129] 2. By using the load estimation performance of user clusters during non-DR periods on DR days to adjust the clustering scheme, the estimation accuracy can be used as a guide to adjust user clustering, realize the linkage between clustering and estimation, unify the goals of clustering and estimation, and obtain the user clustering results that are most conducive to improving the estimation accuracy.

[0130] 3. By using both mean absolute percentage error and mean percentage deviation as error evaluation indicators for the estimation effect, the accuracy of single-point load estimation and the overall load curve estimation deviation can be considered together, and the estimation results can be optimized from two perspectives, effectively improving the accuracy of user cluster baseline load estimation.

[0131] 4. This invention can significantly improve the effectiveness of user cluster baseline load estimation, thereby improving the settlement accuracy of power saving compensation in DR for virtual power plants and load aggregators, helping to increase the willingness of virtual power plants and load aggregators to participate in DR, and also benefiting the further development and promotion of DR.

[0132] like Figure 2 As shown, the present invention provides a cluster baseline load estimation system, comprising:

[0133] The acquisition module 201 is used to acquire the electricity load data of a residential user cluster at a set time; the set time includes the demand response day and a set number of similar days before the demand response day; the similar days are dates in the historical dates that have the same day type as the demand response day.

[0134] The first-level clustering and accumulation module 202 is used to perform first-level clustering and accumulation on all users of the resident user cluster to obtain first-level clusters.

[0135] The primary clustering and accumulation module 202 specifically includes: a primary clustering unit, used to perform primary clustering on all users of the resident user cluster according to a set primary clustering number, to obtain a primary clustering result; and an accumulation unit, used to accumulate user load data in all clusters in the primary clustering result to obtain a primary cluster.

[0136] The secondary clustering and adjustment module 203 is used to perform secondary clustering and adjustment on the primary clusters to obtain the optimal clustering of the user cluster.

[0137] The secondary clustering and adjustment module 203 specifically includes: a secondary clustering unit, used to perform secondary clustering on the primary clusters according to a set number of secondary clusters, to obtain secondary clustering results; a load estimation unit for the non-demand response period on the demand response day, used to estimate the load for the non-demand response period on the demand response day based on each secondary cluster in the secondary clustering results, to obtain the load estimate value for the non-demand response period on the demand response day for each secondary cluster; a summation unit, used to sum the load estimate values ​​for the non-demand response period on the demand response day for each secondary cluster, to obtain the estimated electricity load of the residential user cluster during the non-demand response period on the demand response day; and an adjustment and iteration unit, used to adjust and iterate the secondary clustering results based on the estimated electricity load of the residential user cluster during the non-demand response period on the demand response day, to obtain the optimal clustering of the user cluster.

[0138] The electricity load estimation module 204 is used to estimate the electricity load based on the optimal clustering of the user cluster and the electricity load data, so as to obtain the baseline load estimate of the residential user cluster.

[0139] The electricity load estimation module 204 specifically includes: a user cluster electricity load curve determination unit, used to accumulate all electricity load data of all users in all secondary clusters of the optimal cluster of the user cluster at the set time to obtain the user cluster electricity load curve; a baseline estimation unit, used to perform baseline estimation based on the user cluster electricity load curve to obtain the baseline load estimate value for the demand response period on the demand response day; and a residential user cluster baseline load estimate value determination unit, used to accumulate the baseline load estimates value for the demand response period on the demand response day of all user clusters to obtain the residential user cluster baseline load estimate value.

[0140] This invention uses estimation accuracy as a guide for adjusting user clustering. First, it performs a primary clustering of all users, summing the loads of all users in the resulting clusters as input for a secondary clustering. Second, it performs secondary clustering, adjusting the clustering scheme based on load estimation results during non-DR periods on DR days, thus linking clustering and estimation to obtain the optimal clustering of user groups. Finally, it estimates the baseline load of each cluster and sums them to obtain the final user baseline load estimation result. This invention significantly improves the accuracy of cluster baseline load estimation, which is beneficial to the development and promotion of incentive-based demand response.

[0141] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0142] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A cluster baseline load estimation method, characterized in that, include: Obtain electricity load data for a residential user cluster at a set time; the set time includes the demand response day and a set number of similar days prior to the demand response day; The similar date is a date in historical dates that has the same day type as the demand response date; the day type is either a weekday or a non-working day; Perform first-level clustering and summation on all users in the residential user cluster to obtain first-level clusters; specifically, performing first-level clustering and summation on all users in the residential user cluster to obtain first-level clusters includes: performing first-level clustering on all users in the residential user cluster according to a set number of first-level clusters to obtain first-level clustering results; and summing the user load data in all clusters in the first-level clustering results to obtain first-level clusters. The optimal clustering of user clusters is obtained by performing secondary clustering and adjustment on the primary clusters. Specifically, this process includes: performing secondary clustering on the primary clusters according to a set number of secondary clusters to obtain secondary clustering results; estimating the load during the non-demand response period on the demand response day for each secondary cluster in the secondary clustering results to obtain the estimated load value for the non-demand response period on the demand response day for each secondary cluster; summing the estimated load values ​​for the non-demand response period on the demand response day for each secondary cluster to obtain the estimated electricity load of the residential user cluster during the non-demand response period on the demand response day; and adjusting and iterating the secondary clustering results based on the estimated electricity load of the residential user cluster during the non-demand response period on the demand response day to obtain the optimal clustering of user clusters. Based on the optimal clustering of the user clusters and the electricity load data, an electricity load estimation is performed to obtain a baseline load estimate for the residential user clusters. Specifically, this includes: summing up all electricity load data of all users in all secondary clusters of the optimal clustering of the user clusters at a set time to obtain a user cluster electricity load curve; performing baseline estimation based on the user cluster electricity load curve to obtain a baseline load estimate for the demand response period on the demand response day; and summing up the baseline load estimates for the demand response period on the demand response day of all user clusters to obtain the baseline load estimate for the residential user clusters.

2. The cluster baseline load estimation method according to claim 1, characterized in that, The step of adjusting and iterating the secondary clustering results based on the estimated electricity load of the residential user cluster during the non-demand response period on the demand response day to obtain the optimal clustering of the user cluster specifically includes: Based on the electricity load estimation results of the residential user cluster during the non-demand response period on the demand response day, determine the initial average absolute percentage error and the initial average percentage deviation of the electricity load estimation of the residential user cluster during the non-demand response period on the demand response day. Any user in the secondary cluster of the secondary clustering result is moved to another secondary cluster to obtain a new secondary clustering result; the other secondary clusters are secondary clusters other than the secondary cluster where the user was originally located; Based on the new secondary clustering results, the adjusted average absolute percentage error and the adjusted average percentage deviation of the electricity load estimation for residential user clusters during the non-demand response period on the demand response day are determined. Determine whether the mean absolute percentage error after the adjusted clustering is less than the mean absolute percentage error, and whether the mean percentage deviation after the adjusted clustering is less than the initial mean percentage deviation, to obtain a first determination result; If the first judgment result is yes, then the initial average absolute percentage error and the initial average percentage deviation are updated using the adjusted average absolute percentage error and the adjusted average percentage deviation, and the new secondary clustering result is taken as the current optimal clustering; if the first judgment result is no, then the secondary clustering result before the shift is determined as the current optimal clustering; until all the secondary clusters are traversed; Replace the user in the secondary cluster of the secondary clustering result, and return to the step "move any user in the secondary cluster of the secondary clustering result into another secondary cluster to obtain a new secondary clustering result; the other secondary clusters are secondary clusters other than the secondary cluster where the user was before being moved", until all users in all the secondary clusters are traversed; Determine whether the termination condition has been met to obtain a second determination result. If the second determination result is yes, then determine that the secondary clustering result of the current iteration is the optimal clustering of the user cluster. If the second judgment result is negative, then the secondary cluster of the secondary clustering result is replaced; Select any user from the replaced secondary cluster and return to the step "Move any user from the secondary cluster of the secondary clustering result into another secondary cluster to obtain a new secondary clustering result".

3. The cluster baseline load estimation method according to claim 2, characterized in that, The termination condition is: Or r＝r max Where ξ1 is the first set threshold, ξ2 is the second set threshold, and r is the current number of iterations for cluster adjustment. max ε is the upper limit of the number of iterations. MAPE (r) represents the mean absolute percentage error in the r-th cycle, ε MPB λ(r) represents the average percentage deviation in the r-th cycle, λ1(r) represents the absolute value of the change in the average percentage error between the r-th cycle and the previous cycle, λ2(r) represents the absolute value of the change in the average percentage deviation between the r-th cycle and the previous cycle, and ε MAPE (r-1) represents the mean absolute percentage error in the (r-1)th cycle, ε MPB (r-1) represents the average percentage deviation in the r-th cycle.

4. A cluster baseline load estimation system, characterized in that, include: The acquisition module is used to acquire electricity load data of a residential user cluster at a set time; the set time includes the demand response date and a set number of similar days before the demand response date; the similar days are dates in historical dates with the same day type as the demand response date; the day type is a weekday or a non-working day; The first-level clustering and accumulation module is used to perform first-level clustering and accumulation on all users of the resident user cluster to obtain first-level clusters. The first-level clustering and accumulation module specifically includes: a first-level clustering unit, used to perform first-level clustering on all users of the resident user cluster according to a set number of first-level clusters to obtain first-level clustering results; and an accumulation unit, used to accumulate user load data in all clusters of the first-level clustering results to obtain first-level clusters. A secondary clustering and adjustment module is used to perform secondary clustering and adjustment on the primary clusters to obtain the optimal clustering of the user cluster. Specifically, the secondary clustering and adjustment module includes: a secondary clustering unit, used to perform secondary clustering on the primary clusters according to a set number of secondary clusters to obtain secondary clustering results; a non-demand response period load estimation unit for the demand response day, used to estimate the non-demand response period load for each secondary cluster in the secondary clustering results to obtain the estimated non-demand response period load value for each secondary cluster on the demand response day; a summation unit, used to sum the estimated non-demand response period load values ​​for each secondary cluster to obtain the estimated electricity load of the residential user cluster during the non-demand response period on the demand response day; and an adjustment and iteration unit, used to adjust and iterate the secondary clustering results based on the estimated electricity load of the residential user cluster during the non-demand response period on the demand response day to obtain the optimal clustering of the user cluster. The electricity load estimation module is used to estimate the electricity load based on the optimal clustering of the user cluster and the electricity load data to obtain the baseline load estimate of the residential user cluster. Specifically, the electricity load estimation module includes: a user cluster electricity load curve determination unit, used to accumulate all electricity load data of all users in all secondary clusters of the optimal clustering of the user cluster at a set time to obtain the user cluster electricity load curve; a baseline estimation unit, used to perform baseline estimation based on the user cluster electricity load curve to obtain the baseline load estimate of the demand response period on the demand response day; and a residential user cluster baseline load estimate determination unit, used to accumulate the baseline load estimates of the demand response period on the demand response day of all user clusters to obtain the baseline load estimate of the residential user cluster.

Images