Classification aggregation identification method based on adjustable load response performance

Abstract

The invention discloses a classification aggregation identification method based on adjustable load response performance, and the method comprises the steps: constructing index order-based classification according to a integrating degree index and an access standard, and dividing adjustable load resources managed by a virtual power plant into a main energy load resource set, a frequency modulation load resource set, a standby load resource set and a comprehensive load resource set; and aggregating the classified resources by using a self-organizing mapping (SOM) clustering method based on feature vectors, and performing external characteristic identification analysis on an aggregate based on maximum flexibility. According to the method, the adaptation degree of the adjustable load resources and the power transaction products is fully considered, corresponding integrating degree indexes are set, classification is carried out in a bit order sorting mode, and the adjustable load resources in different categories are aggregated by adopting a self-organizing mapping neural network (SOM) clustering method based on feature vectors, so that the power transaction products are classified. And finally, evaluating the external characteristics of the polymer by adopting an external characteristic identification mode with maximized flexibility.

Description

Technical Field

[0001] The present invention relates to a technology in the field of virtual power plants, specifically a classification aggregation identification method based on adjustable load response performance. Background Art

[0002] Due to the diversity and flexibility of resources within virtual power plants, their classification and aggregation methods are extremely random. Relying solely on traditional resource classification and aggregation methods is no longer sufficient to meet the unique needs of a variety of power trading products, resulting in a waste of adjustable load regulation capabilities and an inability to maximize resource utilization. In addition, the external characteristics of the aggregate are not a simple superposition of internal resources, but are limited by its operating structure. Summary of the invention

[0003] In view of the deficiency that the prior art does not calculate the external characteristics for the goal of maximizing the flexibility of the aggregate, resulting in the inability to classify the demand for adjustable loads of power trading products, the present invention proposes a classification aggregation identification method based on the adjustable load response performance, fully considers the degree of adaptation of adjustable load resources and power trading products, sets corresponding fit indicators, classifies them by ranking, and adopts a self-organizing map neural network (SOM) clustering method based on eigenvectors to aggregate adjustable load resources under different categories, and finally adopts an external characteristic identification method that maximizes flexibility to evaluate the external characteristics of the aggregate.

[0004] The present invention is achieved through the following technical solutions:

[0005] The present invention relates to a classification aggregation identification method based on adjustable load response performance. According to the fit index and the access standard, an index ranking classification is constructed, and the adjustable load resources managed by the virtual power plant are divided into: a main energy load resource set, a frequency regulation load resource set, a standby load resource set and a comprehensive load resource set; a self-organizing map (SOM) clustering method based on a feature vector is used to aggregate the classified resources and an external characteristic identification analysis is performed on the aggregate based on maximum flexibility. BRIEF DESCRIPTION OF THE DRAWINGS

[0006] Figure 1 It is a flow chart of the present invention;

[0007] Figure 2 It is a diagram of the classification and aggregation regulatory framework;

[0008] Figure 3 It is the power baseline of the main energy aggregate and the upper and lower limits of the overall operating power;

[0009] Figure 4 It is a graph of the maximum upward and downward regulating capabilities of the main energy aggregate;

[0010] Figure 5 A graph of the upward and downward regulation rates of the main energy aggregate. DETAILED DESCRIPTION

[0011] like Figure 1 As shown, this embodiment relates to a classification aggregation identification method based on adjustable load response performance, including:

[0012] Step 1) According to the fit index and the admission standard, the index ranking classification is constructed to divide the adjustable load resources managed by the virtual power plant into: main energy load resource set, frequency regulation load resource set, standby load resource set and comprehensive load resource set, specifically including:

[0013] 1.1) Extract the response performance parameters from the adjustable load equipment managed by the virtual power plant, normalize them and calculate the compatibility index F of each adjustable load with different power trading products set =[F E ,F A ,F R ], specifically: Main energy compatibility index: FM fit index: Alternative fit indicators: Where: The normalized parameter of the response delay of the i-th adjustable load resource is T delay.i is the longest time required for the ith adjustable load resource to fully reach the required output level after receiving the command issued by the system, N is the total number of adjustable load resources in the virtual power plant, and the response duration normalization parameter of the ith adjustable load resource is T last.i is the shortest time that the ith adjustable load resource maintains the output level required by the command after responding to the system command. The response adjustable frequency normalization parameter of the ith adjustable load resource is f adjust.i is the number of times the i-th adjustable load resource can be called within a certain period of time.

[0014] The access rules include: delay ≤T delay,permitted , T last ≥T last,permitted , f adjust ≥f adjust,permitted , where: T delay,permitted is the maximum response delay allowed by the corresponding power trading product, T last,permitted is the minimum response duration allowed by the corresponding power trading product, f adjust,permitted The minimum response adjustable frequency allowed for the corresponding power trading product.

[0015] The response performance parameters of the adjustable load include: response delay, response sustainable time and response adjustable frequency.

[0016] 1.2) According to the fit index F E 、F A and F R Sort the adjustable loads in descending order to obtain the corresponding load sequence, and use λ E , A and λ R Record the corresponding positions of the adjustable load in the main energy, frequency regulation and standby sequences respectively, and combine them into rank = [λ E ,λ A ,λ R ];

[0017] 1.3) Comparison of λ E , A and λ R The size of each adjustable load is obtained by obtaining the minimum ranking index I. If λ E minimum, then I=1, if λ A minimum, then I=2, if λ R Minimum, then I = 3, that is

[0018] 1.4) Define the ranking difference index And calculate the rank difference index α1 and α2 of each adjustable load;

[0019] 1.5) To avoid the situation where the fit between a load and different target requirements is not significantly different when the ranking is close, resulting in the set division being too absolute, a performance difference margin is set. Where: γ H is the performance difference margin coefficient, which ranges from 0 to 1 and has a symbol To round up;

[0020] 1.6) Considering the case of I = 1, the ranking difference indexes α1 and α2 are compared with the performance difference margin H, and the following four results will appear:

[0021] i. When α1>H and α2>H, the access conditions of the main energy are further checked. If they are met, the load is included in the main energy load resource set;

[0022] ii. When α1≤H and α2>H, the access conditions of main energy and frequency modulation are further checked. If the access conditions of main energy are met but the access conditions of frequency modulation are not met, the load is classified into the main energy load resource set. If the access conditions of frequency modulation are met but the access conditions of main energy are not met, the load is classified into the frequency modulation load resource set. If the access conditions of both main energy and frequency modulation are met, the load is classified into the comprehensive load resource set.

[0023] iii. When α1>H and α2≤H, the access conditions of the main energy and the backup are further checked. If the access conditions of the main energy are met but the access conditions of the backup are not met, the load is classified into the main energy load resource set. If the access conditions of the backup are met but the access conditions of the main energy are not met, the load is classified into the backup load resource set. If the access conditions of both the main energy and the backup are met, the load is classified into the comprehensive load resource set.

[0024] iv. When α1≤H and α2≤H, the access conditions of main energy, frequency regulation and standby are further checked. If two or more of the access conditions are met, the load is classified into the comprehensive load resource set. Otherwise, if the access condition of main energy is met, the load is classified into the main energy load resource set. If the access condition of frequency regulation is met, the load is classified into the frequency regulation load resource set. If the access condition of standby is met, the load is classified into the standby load resource set.

[0025] The case where I=2 or I=3 is similar to the case where I=1 and will not be described in detail.

[0026] Adjustable loads that meet two or more target demands are included in the comprehensive load resource set.

[0027] 1.7) If there is an adjustable load that does not meet the above conditions, it will be eliminated and no longer considered.

[0028] Step 2) aggregating the resources classified in step 1 using a self-organizing map (SOM) clustering method based on feature vectors, specifically including:

[0029] 2.1) Initialization: Set the connection weights of the SOM network to random values, determine the initial learning value, the initial value of the input vector and the output neurons, that is, the number of clusters.

[0030] 2.2) Confirm the Best Matching Unit (BMU): Calculate the Euclidean distance between domain nodes, find the minimum distance by comparing the Euclidean distance between domain nodes, and confirm the winning neuron. Specifically: in: is the standardized load power characteristic of load i (i=1,2,...,N) and is the normalized load power characteristic of load j (j=1,2,...,N).

[0031] 2.3) Weight update adjustment: Update the node neuron weight within the BMU radius, and use the Gaussian function to shrink the range of each iteration. The neighborhood radius is Among them: r(n) is the neighborhood radius updated for the nth time, r0 is the initial radius, and a is a constant.

[0032] 2.4) Repeat steps 2.2) to 2.3) until the maximum number of iterations is reached and output the result.

[0033] 2.5) Calculate the Davidson-Boulding Index (DBI), obtain the number of clusters, and select the cluster with the smallest DBI as the final result of clustering. The type of each aggregate is determined by the type of the selected set K, K∈{main energy load resource set, frequency regulation load resource set, backup load resource set, comprehensive load resource set}. For the main energy load resource set, it is the main energy aggregate; for the frequency regulation load resource set, it is the frequency regulation aggregate; for the backup load resource set, it is the backup aggregate; for the comprehensive load resource set, it is the comprehensive aggregate.

[0034] Step 3) Perform external characteristic recognition analysis on the polymer based on maximum flexibility, specifically including:

[0035] 3.1) According to the operating constraints of the adjustable load resources in the aggregate, the corresponding operating power feasible domain is obtained to form a set of scheduling solutions for the adjustable load resources within the aggregate. Where: P i,t is the operational power of adjustable load i at time t, P i,t,min is the minimum technical output of adjustable load i at time t, P i,t,max is the maximum technical output of adjustable load i at time t, R i,min R is the minimum climbing rate allowed when the adjustable load i is running. i,max It is the maximum climbing rate allowed when the adjustable load i is running.

[0036] 3.2) Considering that maximizing the flexibility of the aggregate is beneficial for the virtual power plant to deal with the uncertainty problem when participating in the demand-side response, the adjustable range is used as the core quantitative indicator to evaluate the flexibility of the aggregate, and the objective function is set to maximize the adjustable range, that is, in: is the upper limit of the overall operating power of the aggregate, is the lower limit of the overall operating power of the aggregate, and is the decision variable of the objective function.

[0037] According to the existing electricity market access rules, the access requirements for adjustable load resources for different electricity trading products are set as shown in Table 1, and according to the access requirements for different electricity trading products, the time domain response performance parameters of 10 types of typical adjustable load resources are set as shown in Table 2.

[0038] Table 1 Market access requirements for different power trading products

[0039] Table 2 Typical adjustable load resource time domain response performance parameters

[0040] The time domain response performance parameters of 90 adjustable load resources managed by a power plant in Shanghai are selected. The time domain response performance parameter values ​​of the first 10 adjustable loads are listed here, and their main energy fit index, frequency regulation fit index and standby fit index are calculated, as shown in Table 3.

[0041] Table 3 Performance parameters of the first 10 adjustable loads

[0042] The time domain response performance data of the above 90 adjustable load resources are processed and analyzed, and according to the established index ranking-based classification algorithm, the 90 adjustable load resources are divided into four sets. The final classification results are shown in Table 4.

[0043] Table 4 Adjustable load classification results

[0044] The power data obtained by hourly monitoring of 90 adjustable load resources managed by a power plant in Shanghai on a certain day is used as the power characteristic curve of the 90 adjustable load resources in the example, and the SOM clustering method based on the eigenvector is used for clustering. Finally, the aggregation results under the corresponding set are shown in Table 5.

[0045] Table 5 Adjustable load aggregation results

[0046] Taking the three main energy aggregates in the main energy load resource set as an example, according to the proposed external characteristic identification model based on maximum flexibility, the power baseline, maximum upward adjustment capability, maximum downward adjustment capability, upward adjustment rate and downward adjustment rate external characteristic indicators of the corresponding aggregates can be obtained, such as Figure 3-Figure 5 As shown, it can provide technical support for the subsequent dispatching process of the virtual power plant.

[0047] Compared with the prior art, the present invention forms corresponding aggregates for different power trading products, which is more in line with the technical characteristic requirements of different power trading products for adjustable load resources. The external characteristic identification model based on maximum flexibility fully considers the time coupling constraints between the various flexibility resources within the aggregate, thereby achieving a more precise definition of the feasible domain of the load aggregate.

[0048] The above-mentioned specific implementation can be partially adjusted in different ways by those skilled in the art without departing from the principle and purpose of the present invention. The protection scope of the present invention shall be based on the claims and shall not be limited by the above-mentioned specific implementation. Each implementation scheme within its scope shall be subject to the constraints of the present invention.

Claims

1. A classification and aggregation identification method based on adjustable load response performance, characterized in that: According to the fit index and access standard, a classification based on the index ranking is constructed, and the adjustable load resources managed by the virtual power plant are divided into: main energy load resource set, frequency regulation load resource set, backup load resource set and comprehensive load resource set; the classified resources are aggregated using the self-organizing map (SOM) clustering method based on feature vectors, and the external characteristics of the aggregates are identified and analyzed based on maximum flexibility.

2. The classification and aggregation identification method based on adjustable load response performance according to claim 1 is characterized in that: The adjustable load resource division specifically includes: 1.1) Extract response performance parameters from the adjustable load devices managed by the virtual power plant, normalize them and calculate the compatibility index of each adjustable load with different power trading products; 1.2) According to the fit index F E 、F A and F R Sort the adjustable loads in descending order to obtain the corresponding load sequence, and use λ E , A and λ R Record the corresponding positions of the adjustable load in the main energy, frequency regulation and standby sequences respectively, and combine them into rank = [λ E ,λ A ,λ R ]; 1.3) Comparison of λ E , A and λ R The size of each adjustable load is obtained by obtaining the minimum ranking index I. If λ E minimum, then I=1, if λ A minimum, then I=2, if λ R Minimum, then I = 3, that is 1.4) Define the ranking difference index And calculate the rank difference index α1 and α2 of each adjustable load; 1.5) To avoid the situation where the fit between a load and different target requirements is not significantly different when the ranking is close, resulting in the set division being too absolute, a performance difference margin is set. Where: γ H is the performance difference margin coefficient, which ranges from 0 to 1 and has a symbol To round up; 1.6) Considering the case of I=1, the ranking difference indexes α1 and α2 are compared with the performance difference margin H respectively; 1.7) If there is an adjustable load that does not meet the above conditions, it will be eliminated and no longer considered.

3. The classification and aggregation identification method based on adjustable load response performance according to claim 2 is characterized in that: The fit index F set =[F E ,F A ,F R ], specifically: Main energy compatibility index: FM fit index: Alternative fit indicators: Where: The normalized parameter of the response delay of the i-th adjustable load resource is T delay.i is the longest time required for the ith adjustable load resource to fully reach the required output level after receiving the command issued by the system, N is the total number of adjustable load resources in the virtual power plant, and the response duration normalization parameter of the ith adjustable load resource is T last.i is the shortest time that the ith adjustable load resource maintains the output level required by the command after responding to the system command. The response adjustable frequency normalization parameter of the ith adjustable load resource is f adjust.i is the number of times the i-th adjustable load resource is called within a certain period of time.

4. The classification and aggregation identification method based on adjustable load response performance according to claim 2 is characterized in that: The response performance parameters of the adjustable load include: response delay, response sustainable time and response adjustable frequency.

5. The classification and aggregation identification method based on adjustable load response performance according to claim 2 is characterized in that: The access rules include: delay ≤T delay,permitted , T last ≥T last,permitted , f adjust ≥f adjust,permitted , where: T delay,permitted is the maximum response delay allowed by the corresponding power trading product, T last,permitted is the minimum response duration allowed by the corresponding power trading product, f adjust,permitted The minimum response adjustable frequency allowed for the corresponding power trading product.

6. The classification and aggregation identification method based on adjustable load response performance according to claim 2 is characterized in that: The comparison specifically includes: i. When α1>H and α2>H, the access conditions of the main energy are further checked. If they are met, the load is included in the main energy load resource set; ii. When α1≤H and α2>H, the access conditions of main energy and frequency modulation are further checked. If the access conditions of main energy are met but the access conditions of frequency modulation are not met, the load is classified into the main energy load resource set. If the access conditions of frequency modulation are met but the access conditions of main energy are not met, the load is classified into the frequency modulation load resource set. If the access conditions of both main energy and frequency modulation are met, the load is classified into the comprehensive load resource set. iii. When α1>H and α2≤H, the access conditions of the main energy and the backup are further checked. If the access conditions of the main energy are met but the access conditions of the backup are not met, the load is classified into the main energy load resource set. If the access conditions of the backup are met but the access conditions of the main energy are not met, the load is classified into the backup load resource set. If the access conditions of both the main energy and the backup are met, the load is classified into the comprehensive load resource set. iv. When α1≤H and α2≤H, the access conditions of main energy, frequency regulation and standby are further checked. If two or more of the access conditions are met, the load is classified into the comprehensive load resource set. Otherwise, if the access condition of main energy is met, the load is classified into the main energy load resource set. If the access condition of frequency regulation is met, the load is classified into the frequency regulation load resource set. If the access condition of standby is met, the load is classified into the standby load resource set. The case of I=2 or I=3 is similar to the case of I=1; Adjustable loads that meet two or more target demands are included in the comprehensive load resource set.

7. The classification and aggregation identification method based on adjustable load response performance according to claim 1 is characterized in that: The self-organizing map clustering method based on feature vectors specifically includes: 2.1) Initialization: Set the SOM network connection weights to random values, determine the initial learning value, the initial input vector value and the output neurons, that is, the number of clusters; 2.2) Confirm the best matching unit (BMU): Calculate the Euclidean distance between domain nodes, find the minimum distance by comparing the Euclidean distance between domain nodes, and confirm the winning neuron; 2.3) Weight update adjustment: Update the node neuron weight within the BMU radius, and use the Gaussian function to shrink the range of each iteration. The neighborhood radius is Where: r(n) is the neighborhood radius updated for the nth time, r0 is the initial radius, and a is a constant; 2.4) Repeat steps 2.2) to 2.3) until the maximum number of iterations is reached, and output the result; 2.5) Calculate the Davidson-Boulding Index (DBI) and obtain the number of clusters. Select the cluster with the smallest DBI as the final result of clustering. The type of each aggregate is determined by the type of the selected set K, K∈{main energy load resource set, frequency regulation load resource set, backup load resource set, comprehensive load resource set}. For the main energy load resource set, it is the main energy aggregate; for the frequency regulation load resource set, it is the frequency regulation aggregate; for the backup load resource set, it is the backup aggregate; for the comprehensive load resource set, it is the comprehensive aggregate.

8. The classification and aggregation identification method based on adjustable load response performance according to claim 7 is characterized in that: The step 2.2 is specifically as follows: in: is the standardized load power characteristic of load i (i=1,2,...,N) and is the normalized load power characteristic of load j (j=1,2,...,N).

9. The classification and aggregation identification method based on adjustable load response performance according to claim 1 is characterized in that: The external characteristic recognition analysis specifically includes: 3.1) According to the operating constraints of the adjustable load resources in the aggregate, the corresponding operating power feasible domain is obtained to form a set of scheduling schemes for the adjustable load resources within the aggregate; 3.2) Considering that maximizing the flexibility of the aggregate is beneficial for the virtual power plant to deal with the uncertainty problem when participating in the demand-side response, the adjustable range is used as the core quantitative indicator to evaluate the flexibility of the aggregate, and the objective function is set to maximize the adjustable range, that is, in: is the upper limit of the overall operating power of the aggregate, is the lower limit of the overall operating power of the aggregate, and is the decision variable of the objective function.

10. The classification and aggregation identification method based on adjustable load response performance according to claim 9 is characterized in that: The aggregate internal adjustable load resource scheduling scheme set is specifically: Where: P i,t is the operational power of adjustable load i at time t, P i,t,min is the minimum technical output of adjustable load i at time t, P i,t,max is the maximum technical output of adjustable load i at time t, R i,min R is the minimum climbing rate allowed when the adjustable load i is running. i,max It is the maximum climbing rate allowed when the adjustable load i is running.

Images