A virtual power plant distributed resource self-adaptive scheduling instruction decomposition method

Abstract

The application discloses a kind of virtual power plant distributed resource adaptive scheduling instruction decomposition method.The method is based on the scheduling priority relationship of each distributed resource, initializes the power adjustment parameter corresponding to each distributed resource;According to the power adjustment parameter, the output power of each distributed resource is iteratively adjusted, and the power deviation between the aggregated output power and the scheduling capacity is determined;When the power deviation direction of adjacent iteration changes, according to the output power change of each distributed resource between adjacent iterations and the scheduling priority relationship, the power adjustment parameter is updated;According to the updated power adjustment parameter, continue iteration until the power deviation meets the preset condition, and output the target output power of each distributed resource.The application can reflect the scheduling priority difference of heterogeneous distributed resources, reduce the power reciprocating change near the scheduling capacity, and reduce the dependence of the initial power adjustment parameter setting result on the scheduling instruction decomposition process.

Description

Technical Field

[0001] This invention relates to the field of virtual power plant scheduling and control technology, and specifically to a method for decomposing distributed resource adaptive scheduling instructions in a virtual power plant. Background Technology

[0002] Virtual power plants aggregate distributed resources such as distributed power sources, energy storage devices, and controllable loads to form a controllable entity capable of participating in grid dispatch in a unified manner. The dispatch instructions received by a virtual power plant typically correspond to the dispatch capacity at the aggregation level, while actual power regulation needs to be completed jointly by multiple internal distributed resources. Therefore, it is necessary to decompose the aggregated dispatch instructions into the target output power corresponding to each distributed resource.

[0003] In existing technologies, there are already schemes that use iterative methods to adjust the power of multiple resources. For example, WO2024167468A1 discloses a power grid regulation method based on additive increase and multiplicative decrease algorithms, which increases power by an additive constant when the power is below the limit and reduces power by a multiplicative factor when the power exceeds the limit; CN113191680B discloses a distributed economic dispatch method for virtual power plants, which compares the total output of each distributed resource with the output target of the virtual power plant, iteratively adjusts the unified regulation signal according to the output error between the two, and determines the iteration step size according to the responsive resources and their output sensitivity, until the error between the total output and the output target meets the requirements.

[0004] However, the aforementioned existing technologies still have the following problems: 1. Existing solutions mostly use uniform or preset power regulation parameters, which are difficult to reflect the differences in regulation capabilities and response characteristics of different distributed resources, and are prone to causing a mismatch between scheduling power allocation and actual resource performance.

[0005] 2. Fixed power adjustment parameters are difficult to adapt to the constantly changing power deviations during the iteration process. If the adjustment range is too large, the aggregated output power will easily exceed the scheduling capacity and oscillate repeatedly. If the adjustment range is too small, the convergence will be slow.

[0006] Therefore, a scheduling instruction decomposition method is needed that can take into account the performance differences of distributed resources and the actual power changes during the iteration process, and adaptively determine and update the power adjustment parameters of each resource. Summary of the Invention

[0007] In order to overcome the above-mentioned defects of the prior art, the purpose of this invention is to provide a virtual power plant distributed resource adaptive scheduling instruction decomposition method to solve the problem that the existing virtual power plant scheduling instruction decomposition method uses uniform or pre-set power adjustment parameters, which is difficult to adapt to the performance differences of different distributed resources and the dynamic changes of power deviation during the iteration process, resulting in unreasonable scheduling power allocation and easy oscillation or slow convergence near the scheduling capacity.

[0008] To achieve the above objectives, the present invention provides a method for decomposing distributed resource adaptive scheduling instructions in a virtual power plant, comprising: Based on the scheduling priority relationship of each distributed resource, initialize the power adjustment parameters corresponding to each distributed resource. The output power of each distributed resource is iteratively adjusted based on the power adjustment parameters, and the power deviation characterizing the difference between the aggregated output power and the scheduling capacity is determined. When the power deviation direction changes between adjacent iterations, the power adjustment parameters are updated according to the output power changes of each distributed resource and the scheduling priority relationship between adjacent iterations. The process continues iterating based on the updated power adjustment parameters until the power deviation meets the preset conditions, at which point the target output power of each distributed resource is output.

[0009] Specifically, after receiving the scheduling capacity from the aggregation layer, the virtual power plant first determines the scheduling priority relationship between the distributed resources and initializes the corresponding power adjustment parameters for each distributed resource according to the scheduling priority relationship. In each iteration, based on the current power adjustment parameters of each distributed resource, the output power of the previous iteration is adjusted to obtain the output power of the current iteration. Subsequently, the output power of each distributed resource in the current iteration is aggregated to obtain the aggregated output power of the current iteration, and the aggregated output power is compared with the scheduling capacity to obtain the power deviation of the current iteration.

[0010] When the power deviation directions of adjacent iterations are the same, it indicates that the aggregated output power is still on the same side of the scheduling capacity. In this case, the output power of each distributed resource continues to be adjusted according to the current power adjustment parameters, so that the aggregated output power continues to approach the scheduling capacity. When the power deviation direction of adjacent iterations changes, it indicates that the aggregated output power has changed from one side of the scheduling capacity to the other side, that is, the adjustment in this iteration has exceeded the scheduling capacity.

[0011] After the power deviation direction changes, the output power change of each distributed resource between adjacent iterations is determined, and the power adjustment parameters corresponding to each distributed resource are updated based on the scheduling priority relationship of each distributed resource. This ensures that the power adjustment amplitude used in subsequent iterations reflects both the scheduling priority of each distributed resource and the actual power change in the previous stage, rather than continuing to use the power adjustment parameters from before exceeding the scheduling capacity.

[0012] Subsequently, the output power of each distributed resource is adjusted based on the updated power regulation parameters, and the power deviation between the aggregated output power and the scheduling capacity is redefined. This process of output power adjustment, power deviation determination, and power regulation parameter update is repeated until the power deviation meets the preset conditions. At this point, the output power of each distributed resource in the current iteration is determined as its corresponding target output power, thereby decomposing the scheduling capacity received by the virtual power plant into the resource-level target output power of each distributed resource.

[0013] Beneficial effects: Compared with the prior art, the present invention initializes the corresponding power adjustment parameters according to the scheduling priority relationship of each distributed resource, so that different distributed resources have a power adjustment range that matches their scheduling priority during the decomposition of scheduling instructions, thereby avoiding the use of the same adjustment method for resources with different adjustment capabilities and response characteristics, and improving the rationality of the allocation of scheduling capacity among distributed resources.

[0014] During the iterative adjustment process, when the power deviation direction changes between adjacent iterations, the power adjustment parameters are updated based on the actual output power changes of each distributed resource between adjacent iterations and their scheduling priority relationships. This allows the subsequent adjustment range to be adaptively adjusted in combination with the actual adjustment results of the resources and the changes in the power deviation direction, reducing over-adjustment or under-adjustment caused by continuing to use the original parameters, and reducing the possibility of the aggregated output power repeatedly exceeding the limit near the scheduling capacity.

[0015] Therefore, this invention can take into account the rationality of power allocation among heterogeneous resources, as well as the stability and convergence speed of iterative adjustment, while the aggregated output power gradually approaches the scheduling capacity. When the power deviation meets the preset conditions, it can obtain the target output power corresponding to each distributed resource, thereby realizing the adaptive decomposition of aggregated scheduling instructions into resource-level target output power. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0017] Figure 1 This is a schematic diagram of the overall process of a distributed resource adaptive scheduling instruction decomposition method for virtual power plants. Figure 2 Flowchart of the adaptive update process for power regulation parameters; Figure 3 This is a comparison diagram of the aggregated output power iteration process of the method of the present invention and the fixed parameter AIMD method. Detailed Implementation

[0018] To enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings.

[0019] This embodiment provides a method for decomposing distributed resource adaptive scheduling instructions in a virtual power plant. The specific process is as follows: Figure 1 As shown, the aggregation-level scheduling capacity issued by the superior scheduling agency to the virtual power plant is decomposed into the target output power to be executed by each distributed resource within the virtual power plant.

[0020] In this embodiment, the virtual power plant aggregates various types of distributed resources, including one or more of distributed photovoltaic, distributed wind power, energy storage devices, electric vehicles, temperature-controlled loads, small hydropower, and distributed generators. Different distributed resources differ in operating costs, adjustable capacity, response speed, ramp-up capability, continuous operation capability, and command tracking accuracy. After receiving the dispatch capacity, the virtual power plant needs to convert it into target output power for each distributed resource to execute, and then aggregate the output power formed by these target output powers to reach the dispatch capacity. If the allocation is only proportionally based on the rated capacity of each distributed resource, or if the same power regulation parameters are used for different distributed resources, it is difficult to reflect the differences in economy, regulation performance, and response capability among the resources. Therefore, this embodiment first determines the dispatch priority relationship based on multiple evaluation indicators of each distributed resource, then sets the corresponding power regulation parameters for each distributed resource according to the dispatch priority relationship, and iteratively adjusts the parameters based on the difference between the aggregate output power and the dispatch capacity. Thus, the comprehensive resource evaluation results are transmitted to the resource-level power regulation process, enabling different distributed resources to have corresponding adjustment ranges during the dispatch command decomposition process.

[0021] Before executing the decomposition of dispatch instructions, the dispatch capacity corresponding to the virtual power plant, the current output power of each distributed resource, and evaluation index data for determining dispatch priorities are obtained. The dispatch capacity represents the aggregate output power target that the virtual power plant needs to achieve in this dispatch task. The current output power of each distributed resource is used as the initial output power for iterative adjustment. For different types of distributed resources, their current operating status can be converted into corresponding output power data according to the unified power caliber adopted by the virtual power plant control platform. In one embodiment, the allowable output power range of each distributed resource can also be obtained. The allowable output power range can be determined based on the equipment's rated parameters, current operating status, and adjustability. For energy storage devices and electric vehicles, their allowable output power range can be determined by combining the state of charge, allowable charging and discharging power, and sustainable response time; for distributed photovoltaic and distributed wind power, their allowable output power range can be determined by combining the current measured power and available output; for controllable loads, their equivalent output power range can be determined based on the load baseline power and allowable adjustment range.

[0022] To unify the power scale for subsequent iterations, this embodiment uses the same baseline power to standardize the output power, upper and lower limits of output power, and scheduling capacity of each distributed resource. Subsequent calculations involving the above parameters will all use the corresponding standard values.

[0023] Let the total number of distributed resources participating in this scheduling instruction decomposition be . , It is an integer not less than 2; using Represents the distributed resource sequence number, and . No. The minimum and maximum allowable output power of a distributed resource are respectively expressed as: and To ensure that scheduling capacity can be achieved through the distributed resources participating in scheduling, in one implementation, the scheduling capacity satisfies:

[0024] In the formula, This represents the dispatch capacity received by the virtual power plant, and ; This represents the sum of the minimum allowable output power for all distributed resources participating in the scheduling; This represents the sum of the maximum allowable output power of all distributed resources participating in the scheduling. This condition indicates that the scheduling capacity lies within the adjustable output power range formed by all participating resources. If the scheduling capacity exceeds this range, it means that the current resource set cannot directly complete the scheduling task. In this case, distributed resources participating in the scheduling can be reselected, or the actual regulating capacity that the virtual power plant can provide can be determined based on the current availability of each resource.

[0025] This embodiment adopts Representing the iteration number, using Indicates the first The distributed resource in the first Output power in the nth iteration. Each distributed resource in the nth iteration... The aggregated output power in the next iteration is expressed as: The method for determining it is as follows:

[0026] By comparing the aggregated output power with the scheduling capacity, it can be determined whether the power provided by all distributed resources meets the requirements of this scheduling task. If the aggregated output power is lower than the scheduling capacity, it indicates that the output power of the corresponding distributed resources needs to be increased; if the aggregated output power is higher than the scheduling capacity, it indicates that the output power of the corresponding distributed resources needs to be reduced.

[0027] The power regulation parameters in this embodiment include the addition step size and the multiplication factor. Indicates the first The distributed resource in the first The corresponding addition step size in the next iteration is adopted This represents the multiplication factor corresponding to the distributed resource. The addition step size is used to determine the power increase of the corresponding distributed resource in the next iteration when the aggregated output power is lower than the scheduling capacity. The multiplication factor is used to determine the power retention ratio of the corresponding distributed resource in the next iteration when the aggregated output power is higher than the scheduling capacity. The multiplication factor is typically set in a range greater than zero and not greater than one. The closer the multiplication factor is to one, the smaller the power reduction of the corresponding distributed resource; the smaller the multiplication factor, the larger the power reduction of the corresponding distributed resource.

[0028] Different distributed resources are assigned corresponding addition step sizes and multiplication factors, resulting in different power change amplitudes during the additive increase and multiplicative decrease phases. Resources with higher scheduling priority can bear larger power increases when increasing output power and retain more output power when decreasing output power; resources with lower scheduling priority bear smaller power increases or larger power decreases accordingly. This allows the comprehensive regulation performance of different resources to be incorporated into the scheduling instruction decomposition process, avoiding the need to determine the power change amplitude of each resource solely based on resource type, rated capacity, or uniform regulation parameters.

[0029] Based on the above application scenarios, input data, and parameter definitions, this embodiment sequentially determines the scheduling priority relationship of each distributed resource, initializes the power adjustment parameters corresponding to each distributed resource, and iteratively adjusts the output power of each distributed resource according to the power deviation; when the power deviation direction of adjacent iterations changes, the power adjustment parameters are updated until the power deviation meets the preset conditions, and the current output power of each distributed resource is determined as the corresponding target output power.

[0030] S1. Based on the scheduling priority relationship of each distributed resource, initialize the power adjustment parameters corresponding to each distributed resource; Because the distributed resources aggregated within a virtual power plant differ in terms of operating costs, adjustability, response performance, and command tracking accuracy, applying the same power regulation parameters to each distributed resource makes it difficult to ensure that the power allocation results match the actual performance of each resource. Therefore, this embodiment first determines the corresponding scheduling priority relationship based on the evaluation index data of each distributed resource, and then initializes the power regulation parameters of each distributed resource according to the scheduling priority relationship.

[0031] S11. Obtain the evaluation index data of each distributed resource and determine the index weight corresponding to each evaluation index. S111. Determine the composition, technical meaning, and positive / negative attributes of the evaluation index data; This embodiment acquires the generation cost, generation capacity, regulation capacity, response time, ramp rate, generation duration, and regulation deviation of each distributed resource, and uses this data as evaluation indicators to determine the scheduling priority relationship of each distributed resource. The evaluation indicators can be obtained based on the equipment parameters, historical operating data, and current operating status of each distributed resource. Specifically, generation cost reflects the economic cost required for a distributed resource to participate in scheduling; generation capacity reflects the power scale that a distributed resource can provide; regulation capacity reflects the range of power that a distributed resource can increase or decrease relative to its current operating state; response time reflects the time required for a distributed resource to change its output power from receiving a scheduling command; ramp rate reflects the magnitude of power change that a distributed resource can achieve per unit time; generation duration reflects the ability of a distributed resource to maintain its corresponding output power; and regulation deviation reflects the tracking accuracy of the distributed resource's actual output power to the scheduling command.

[0032] Since smaller values ​​for generation cost, response time, and regulation deviation indicate better economic efficiency, timely response, or regulation accuracy of the corresponding distributed resource, generation cost, response time, and regulation deviation are designated as negative evaluation indicators. Conversely, larger values ​​for generation capacity, regulation capacity, ramp rate, and generation duration indicate stronger power scale or regulation capability of the corresponding distributed resource; therefore, generation capacity, regulation capacity, ramp rate, and generation duration are designated as positive evaluation indicators.

[0033] Let the total number of evaluation indicators be ,use Indicates the evaluation indicator number, and ,use Indicates the first The distributed resource in the first The original evaluation data under each evaluation indicator. An original evaluation matrix is ​​constructed based on the original evaluation data of each distributed resource:

[0034] In the original evaluation matrix, each row corresponds to a distributed resource, and each column corresponds to an evaluation metric. In this embodiment, the total number of evaluation metrics... It is 7.

[0035] S112. Perform homogenization and normalization processing on the evaluation index data, and determine the index weights corresponding to each evaluation index based on information entropy. Since negative evaluation indicators such as power generation cost, response time, and regulation deviation have opposite evaluation directions to positive evaluation indicators such as power generation capacity and regulation capacity, and different evaluation indicators differ in dimensions and numerical ranges, this embodiment first performs homogenization processing on the evaluation indicator data, and then normalizes the homogenized evaluation data so that the larger the processed evaluation value, the better the performance of the corresponding distributed resource under the evaluation indicator.

[0036] For positive evaluation indicators such as generation capacity, regulation capacity, ramp-up rate, and generation duration, the larger the original evaluation data, the better the performance of the corresponding distributed resource. Therefore, the original evaluation data is retained, and the corresponding homogenized evaluation value is determined as follows:

[0037] In the formula, Indicates the first The distributed resource in the first The same evaluation value under each evaluation indicator.

[0038] For negative evaluation indicators such as generation cost, response time, and regulation deviation, the smaller the original evaluation data, the better the performance of the corresponding distributed resource. To ensure that negative and positive evaluation indicators have the same evaluation direction, the first... The maximum original evaluation value of each evaluation indicator across all distributed resources. and minimum original evaluation value And perform a linear inverse transformation on the negative evaluation indicators:

[0039] Among them, the The maximum and minimum original evaluation values ​​of each evaluation indicator across all distributed resources are as follows:

[0040]

[0041] In the formula, Indicates the first The maximum original evaluation value of each evaluation indicator across all distributed resources; Indicates the first The minimum original evaluation value of each evaluation indicator among all distributed resources; This represents the resource index used when traversing all distributed resources, and ; Indicates the first The distributed resource in the first The original evaluation data under each evaluation indicator.

[0042] After adopting the above linear inverse transformation, the smaller the original evaluation data of the negative evaluation index, the larger its homogenized evaluation value, so that both the positive and negative evaluation indices satisfy the evaluation direction that the larger the homogenized evaluation value, the better the performance of the corresponding distributed resource; at the same time, this linear inverse transformation preserves the relative differences between different distributed resources under the same evaluation index.

[0043] Based on the homogeneous evaluation values ​​of each distributed resource under each evaluation index, a homogeneous evaluation matrix is ​​constructed:

[0044] In the formula, This represents the homogeneous evaluation matrix. Each row in the homogeneous evaluation matrix corresponds to a distributed resource, and each column corresponds to an evaluation index. Each evaluation index in the matrix has an evaluation direction where the larger the homogeneous evaluation value, the better the performance of the corresponding distributed resource.

[0045] After completing the homogenization process, to eliminate the differences in dimensions and orders of magnitude between different evaluation indicators, the homogenized evaluation values ​​of each distributed resource under the same evaluation indicator are normalized to determine the corresponding normalized evaluation value:

[0046] In the formula, Indicates the first The distributed resource in the first Normalized evaluation values ​​under each evaluation indicator; Indicates the first The distributed resource in the first The normalized evaluation values ​​under each evaluation metric. For any evaluation metric, the normalized evaluation values ​​corresponding to each distributed resource satisfy:

[0047] Therefore, the normalized evaluation value can reflect the relative performance of the corresponding distributed resource under the evaluation index, and enable evaluation indices of different dimensions to be included in subsequent unified calculations.

[0048] If the first The sum of all the homogenized evaluation values ​​corresponding to each evaluation indicator is zero, that is:

[0049] Then, normalization cannot be performed according to the above formula. In this case, it indicates that the evaluation index does not provide effective distinguishing information among the current distributed resources. Therefore, the normalized evaluation value for each distributed resource under this evaluation index is uniformly set as follows:

[0050] Based on the normalized evaluation values ​​of each distributed resource under each evaluation index, a normalized evaluation matrix is ​​constructed:

[0051] In the formula, This represents the normalized evaluation matrix. Each row in the normalized evaluation matrix corresponds to a distributed resource, and each column corresponds to an evaluation index. The normalized evaluation matrix preserves the relative differences between different distributed resources under the same evaluation index and eliminates the differences in dimensions and orders of magnitude between different evaluation indices.

[0052] Complete the homogenization and normalization processes and construct the normalized evaluation matrix. Subsequently, this embodiment uses the entropy weight method to determine the weight of each evaluation indicator. The entropy weight method measures the amount of information contained in the same evaluation indicator based on the degree of data dispersion among different distributed resources: the more obvious the difference between the evaluation indicator and different distributed resources, the smaller its information entropy, the stronger its distinguishing effect on the comprehensive performance of resources, and the greater the corresponding indicator weight; conversely, the smaller the difference between the evaluation indicator and different distributed resources, the greater its information entropy, and the smaller the corresponding indicator weight.

[0053] Due to normalized evaluation value Simultaneously indicates the first The distributed resource in the first The relative data proportion under the evaluation indicator, therefore, the first The information entropy of each evaluation indicator is determined as follows:

[0054] In the formula, Indicates the first The information entropy of each evaluation indicator, and .when At that time, according to the limit relationship:

[0055] When the normalized evaluation values ​​of various distributed resources are relatively close under a certain evaluation index, the information entropy of the evaluation index is large, indicating that its ability to distinguish between different distributed resources is weak; when the normalized evaluation values ​​of various distributed resources are significantly different, the information entropy of the evaluation index is small, indicating that the evaluation index can provide more information for distinguishing resource performance.

[0056] Based on the degree of difference in indicators reflected by information entropy, determine the first... The degree of difference among the evaluation indicators:

[0057] In the formula, Indicates the first The degree of difference among the evaluation indicators, and The smaller the information entropy, the greater the difference in the corresponding evaluation indicators; the larger the information entropy, the smaller the difference in the corresponding evaluation indicators.

[0058] Further, based on the degree of difference among the various evaluation indicators, determine the first... The weights of each evaluation indicator:

[0059] In the formula, Indicates the first The weights of each evaluation indicator; This indicates the index number used when traversing all evaluation indicators, and ; Indicates the first The degree of difference among the evaluation indicators. The weights of each evaluation indicator satisfy the following:

[0060] Therefore, evaluation indicators that show significant differences between distributed resources and have a strong differentiating effect receive larger weights, while evaluation indicators that have relatively similar values ​​for each resource receive smaller weights.

[0061] If the sum of the differences in all evaluation indicators is zero, that is:

[0062] This indicates that none of the evaluation indicators can effectively distinguish the distributed resources currently participating in the evaluation. In this case, the weights of each evaluation indicator will be uniformly set as follows:

[0063] In summary, the obtained normalized evaluation matrix and the corresponding weights of each evaluation indicator This is used to construct a weighted decision matrix using the TOPSIS method, and to further determine the relative proximity and scheduling priority of each distributed resource.

[0064] S12. Based on the evaluation index data and the corresponding index weights, determine the relative proximity of each distributed resource, and determine the scheduling priority relationship of each distributed resource according to the relative proximity. After processing by S11, the normalized evaluation matrix has been obtained. and the corresponding weights of each evaluation indicator. The normalized evaluation matrix reflects the relative performance of each distributed resource under different evaluation indicators, while the indicator weights reflect the degree to which the corresponding evaluation indicators distinguish the overall performance of the current distributed resources.

[0065] Based on this, this embodiment employs the Top-Approximation-Ideal-Solution (TOPSIS) method to evaluate the overall scheduling performance of each distributed resource. TOPSIS constructs positive ideal solutions with superior overall performance and negative ideal solutions with inferior overall performance, and compares the distances between each distributed resource and these solutions to determine the degree to which each resource approaches a better state and deviates from a worse state, thereby determining its relative proximity. Compared to ranking based solely on a single evaluation metric or the simple summation of various evaluation metrics, TOPSIS comprehensively considers various evaluation metrics within the same evaluation space and simultaneously compares each distributed resource using both positive and negative ideal solutions, avoiding the direct determination of resources that perform well only under a few evaluation metrics as high-priority resources.

[0066] S121. Construct a weighted decision matrix based on the evaluation index data and index weights after normalization and homogenization, and determine the positive ideal solution and the negative ideal solution; To incorporate the weights of each evaluation metric into the comprehensive performance evaluation of distributed resources and to avoid treating different evaluation metrics equally in subsequent distance calculations, a diagonal weight matrix is ​​first constructed based on the order of the evaluation metrics in the normalized evaluation matrix and their corresponding weights:

[0067] In the formula, Represents the diagonal weight matrix; This indicates that a diagonal matrix is ​​constructed using the weights of each indicator within the parentheses as the main diagonal elements, with the remaining positions set to zero. Using this diagonal matrix form ensures that each indicator weight only affects the corresponding column of the normalized evaluation matrix, without altering the original correspondence between different evaluation indicators.

[0068] Based on the aforementioned diagonal weight matrix, the data corresponding to each evaluation index in the normalized evaluation matrix are weighted to obtain the weighted decision matrix:

[0069] In the formula, Represents the weighted decision matrix; Indicates the first The distributed resource in the first The weighted evaluation values ​​under each evaluation indicator are as follows:

[0070] Therefore, the relative performance of each distributed resource under the same evaluation index is still determined by the normalized evaluation value. The degree to which the corresponding evaluation index plays a role in the comprehensive performance evaluation is determined by the index weight. Adjustments were made. Evaluation indicators with larger weights have a relatively larger impact on the subsequent calculation of the ideal solution distance, while evaluation indicators with smaller weights have a relatively smaller impact. This ensures that the weighted decision matrix can simultaneously reflect the performance of each distributed resource and the differentiating role of each evaluation indicator.

[0071] Since the aforementioned unification process has already ensured that all evaluation indicators share a unified evaluation direction—the larger the evaluation value, the better the performance of the corresponding distributed resources—this step does not require further differentiation between positive and negative evaluation indicators. Regarding the first... The evaluation index is used to determine the maximum value among the weighted evaluation values ​​corresponding to each distributed resource as the positive ideal value, and the minimum value among them as the negative ideal value: Positive ideal value:

[0072] Negative ideal value:

[0073] In the formula, Indicates the first The positive ideal value corresponding to each evaluation indicator; Indicates the first The negative ideal value corresponding to each evaluation indicator; Indicates the first The distributed resource in the first The weighted evaluation value under each evaluation indicator.

[0074] According to the order of the evaluation indicators, the positive ideal values ​​corresponding to each evaluation indicator are combined to form a positive ideal solution, and the negative ideal values ​​corresponding to each evaluation indicator are combined to form a negative ideal solution: Positive ideal solution:

[0075] Negative ideal solution:

[0076] In the formula, It represents the positive ideal solution, used to characterize the overall better reference state that the distributed resources currently participating in the evaluation can form under various evaluation indicators; This represents the negative ideal solution, used to characterize the corresponding poorly synthesized reference state.

[0077] Both the positive and negative ideal solutions are evaluation benchmarks constructed based on the current resource set, and do not require that a certain distributed resource actually exists that simultaneously possesses all the index values ​​in either the positive or negative ideal solution. By constructing these two evaluation benchmarks, the comprehensive performance of different distributed resources can be converted into the distance relationship between them and the better and worse states within the same multidimensional evaluation space.

[0078] S122. Determine the corresponding relative proximity based on the distance between each distributed resource and the positive and negative ideal solutions; After obtaining the positive and negative ideal solutions, this embodiment uses Euclidean distance to measure the proximity of each distributed resource to the two ideal solutions. For the first... The distance between a distributed resource and the ideal solution is determined as follows:

[0079] In the formula, Indicates the first The Euclidean distance between a distributed resource and a positive ideal solution. The smaller the value, the closer the overall performance of the distributed resource under various evaluation metrics is to the optimal reference state in the current resource set. Correspondingly, the... The distance between each distributed resource and the negative ideal solution is determined as follows:

[0080] In the formula, Indicates the first The Euclidean distance between a distributed resource and a negative ideal solution. The larger the value, the further the overall performance of the distributed resource is from the worst reference state in the current resource set.

[0081] However, evaluating solely based on the distance to the positive ideal solution cannot fully reflect the extent to which a distributed resource is far from a poor state; evaluating solely based on the distance to the negative ideal solution also cannot fully reflect its extent to which it approaches a better state. Therefore, this embodiment comprehensively utilizes both distances to determine the... The relative proximity of distributed resources:

[0082] In the formula, Indicates the first The relative proximity of distributed resources, when At that time, the relative closeness is satisfied .

[0083] A higher relative proximity indicates that the corresponding distributed resource is closer to the positive ideal solution and further away from the negative ideal solution, resulting in better overall performance in terms of generation cost, regulation capacity, response performance, continuous operation capability, and regulation accuracy. Conversely, a lower relative proximity indicates that the corresponding distributed resource has relatively lower overall performance. If the positive and negative ideal solutions coincide, making the corresponding distributed resources... and If all values ​​are zero, it indicates that the current evaluation metrics cannot distinguish the overall performance of each distributed resource. In this case, the relative proximity of each distributed resource is uniformly set to a preset median value, and each distributed resource is determined to have the same scheduling priority, without being sorted according to relative proximity.

[0084] S123. Determine the corresponding scheduling priority relationship according to the relative proximity of each distributed resource; Distributed resources are sorted from highest to lowest relative proximity. Resources with higher relative proximity are assigned higher scheduling priority, while those with lower relative proximity are assigned lower priority. Resources with the same relative proximity are assigned the same scheduling priority. A resource with higher relative proximity does not necessarily perform better only under a single evaluation metric; rather, its weighted evaluation result is closer to the positive ideal solution and further away from the negative ideal solution. Therefore, determining scheduling priority based on relative proximity avoids relying solely on single evaluation metrics such as generation capacity and response time to arrange resource adjustment order, allowing the scheduling priority to reflect the comprehensive applicability of each distributed resource for scheduling.

[0085] The scheduling priority relationship is not used to directly exclude distributed resources with lower scheduling priority, but rather to differentiate the power regulation parameters for each distributed resource in subsequent settings. Specifically, when it is necessary to increase the output power of the virtual power plant, the distributed resources with higher scheduling priority are given a relatively larger power increase; when it is necessary to reduce the output power of the virtual power plant, the distributed resources with higher scheduling priority are given a relatively smaller power reduction, so that the distributed resources with better overall performance bear a relatively larger amount of regulation power in the scheduling instruction decomposition results.

[0086] Through the above process, the index weights obtained by the entropy weight method and the resource proximity obtained by the TOPSIS method are converted into scheduling priority relationships between distributed resources. Subsequently, based on the relationship between the initial aggregate output power and scheduling capacity of each distributed resource, the addition step size and multiplication factor corresponding to each distributed resource are initialized.

[0087] S13. Based on the relationship between the initial aggregated output power and the scheduling capacity, initialize the addition step size and multiplication factor corresponding to each distributed resource; After the above processing, the relative proximity of each distributed resource is obtained, and the corresponding scheduling priority relationship is determined according to the relative proximity. The relative proximity reflects the comprehensive performance of the distributed resource under various evaluation indicators. The larger the relative proximity, the closer the corresponding distributed resource is to the comprehensive optimal state, and the higher its priority in participating in the virtual power plant scheduling.

[0088] To incorporate the comprehensive resource performance evaluation results into the power adjustment process, this embodiment employs an additive increase and multiplicative decrease algorithm, namely the AIMD algorithm, and initializes the corresponding addition step size and multiplication factor based on the relative proximity of each distributed resource. The addition step size determines the increase in output power of each distributed resource when the aggregated output power is lower than the scheduling capacity; the multiplication factor determines the retention ratio of each distributed resource relative to the output power of the previous iteration when the aggregated output power is higher than the scheduling capacity. In an implementation using per-unit output power for AIMD iteration, the addition step size represents the dimensionless relative power adjustment magnitude, and the relative proximity is also a dimensionless value. Therefore, the relative proximity can be mapped to the initial addition step size or initial multiplication factor of the corresponding distributed resource based on the current power adjustment direction.

[0089] S131. When the initial aggregated output power is less than the scheduling capacity, determine the initial addition step size for each distributed resource according to the scheduling priority relationship, and set the corresponding initial multiplication factor to a unit value. In this embodiment, the initial aggregated output power is expressed as: Then, when the initial aggregated output power satisfies:

[0090] When the current aggregated output power of the virtual power plant is lower than the dispatch capacity, it indicates that the output power of each distributed resource needs to be increased. At this time, the additive increase phase of the AIMD algorithm is entered.

[0091] According to the The relative proximity of distributed resources determines their initial addition step size:

[0092] In the formula, Indicates the first The initial addition step size corresponds to each distributed resource. The greater the relative proximity, the larger the corresponding initial addition step size, allowing distributed resources with higher scheduling priority to have a larger increase in output power during the additive addition phase; the smaller the relative proximity, the smaller the corresponding initial addition step size, allowing distributed resources with lower scheduling priority to have a smaller increase in output power.

[0093] Since the current adjustment direction is to increase output power, there is no need to multiplicatively reduce the output power of each distributed resource. Therefore, the initial multiplication factor corresponding to each distributed resource is set to a unit value.

[0094] In the formula, Indicates the first The initial multiplication factor corresponds to each distributed resource. When the relative proximity is 0, in order to avoid the initial addition step size of the corresponding distributed resource being 0 and unable to participate in subsequent additive increases, its initial addition step size can be set to a positive value close to 0 within a preset effective range.

[0095] It should be noted that this boundary treatment does not change the correspondence that the greater the relative proximity, the larger the initial addition step size.

[0096] S132. When the initial aggregated output power is greater than the scheduling capacity, the initial addition step size corresponding to each distributed resource is set to zero, and the corresponding initial multiplication factor is determined according to the scheduling priority relationship. When the initial aggregated output power satisfies:

[0097] When the output power of the virtual power plant is higher than the scheduling capacity, it indicates that the output power of each distributed resource needs to be reduced, and the multiplicative reduction phase of the AIMD algorithm is entered.

[0098] Since there is no need to further increase the output power at present, the initial addition step size for each distributed resource is set as follows:

[0099] The multiplication factor represents the proportion of output power retained after a multiplicative reduction of the corresponding distributed resource. The larger the multiplication factor, the higher the proportion of output power retained, and the smaller the corresponding power reduction; the smaller the multiplication factor, the lower the proportion of output power retained, and the larger the corresponding power reduction.

[0100] When the When the relative proximity of a distributed resource is within the range of greater than 0 and less than 1, the initial multiplication factor is determined based on its relative proximity:

[0101] Therefore, distributed resources with higher relative proximity have larger initial multiplication factors, retaining more output power during the multiplicative reduction phase; distributed resources with lower relative proximity have smaller initial multiplication factors, bearing a larger power reduction. When the relative proximity is 0 or 1, the corresponding initial multiplication factor is set to a nearby value within a preset effective range greater than 0 and less than 1, to avoid directly reducing the output power of the corresponding resource to zero when the multiplication factor is 0, or failing to produce actual power reduction when the multiplication factor is 1. This boundary handling does not change the correspondence that the greater the relative proximity, the larger the initial multiplication factor.

[0102] When the initial aggregated output power satisfies:

[0103] When the current aggregated output power has reached the scheduling capacity, the initial output power of each distributed resource can be determined as the corresponding target output power, and there is no need to continue to perform power adjustment iteration.

[0104] Through the above processing, the resource comprehensive performance and scheduling priority relationship represented by the relative proximity are passed to the initial parameters of the AIMD algorithm, so that the distributed resources with higher scheduling priority have a larger increase in power during the power increase phase and a smaller decrease in power during the power reduction phase. Thus, different distributed resources participate in the decomposition of scheduling instructions according to their corresponding scheduling priority relationship from the first power adjustment.

[0105] S2. Iteratively adjust the output power of each distributed resource according to the power adjustment parameters, and determine the power deviation that characterizes the difference between the aggregated output power and the scheduling capacity; After initializing the power regulation parameters for each distributed resource, this embodiment uses the AIMD algorithm to iteratively adjust the output power of each distributed resource. The AIMD algorithm uses the aggregated output power status of the virtual power plant as feedback. Each distributed resource completes its current iteration based on this feedback and its own current output power and power regulation parameters, without needing to obtain the specific output status of other distributed resources. This reduces the dependence of the scheduling instruction decomposition process on the global resource model and information interaction between resources.

[0106] To characterize the difference between aggregated output power and dispatch capacity, the first The power deviation corresponding to the next iteration is determined as follows:

[0107] In the formula, Indicates the first The power deviation corresponds to the next iteration. A negative power deviation indicates that the current aggregated output power is lower than the scheduling capacity; a positive power deviation indicates that the current aggregated output power is higher than the scheduling capacity. Based on the sign of the power deviation from the previous iteration, it is determined whether the current iteration will use an additive increase method or a multiplicative decrease method.

[0108] S21. When the power deviation of the previous iteration is negative, add the addition step size to the output power of the previous iteration of the corresponding distributed resource to determine the output power of the current iteration. When the power deviation of the previous iteration is negative, it indicates that the aggregated output power of the previous iteration has not yet met the scheduling capacity, and the aggregated output power needs to be further increased. In this case, an additive increase method is adopted, and the... The distributed resource in the first The output power in the next iteration is determined as follows:

[0109] In the formula, Indicates the first The distributed resource in the first The addition step size used after the next iteration.

[0110] If the output power determined in the above manner exceeds the first If the maximum allowed output power of a distributed resource is [amount], then the output power of its current iteration is limited to [amount]. Otherwise, the calculated output power is determined as the output power of the current iteration. Thus, within the resource-allowed output range, the power gap between the aggregated output power and the scheduling capacity is gradually replenished.

[0111] S22. When the power deviation of the previous iteration is positive, determine the output power of the current iteration based on the multiplication factor and the output power of the corresponding distributed resource in the previous iteration. When the power deviation of the previous iteration is positive, it indicates that the aggregated output power of the previous iteration has exceeded the scheduling capacity, and the aggregated output power needs to be reduced. In this case, a multiplicative reduction method is used. The distributed resource in the first The output power in the next iteration is determined as follows:

[0112] In the formula, Indicates the first The distributed resource in the first The multiplication factor used after the next iteration.

[0113] If the output power determined in the above manner is lower than the first The minimum allowed output power of a distributed resource is then used to limit the output power of its current iteration to . Otherwise, the calculated output power is determined as the output power of the current iteration. Thus, within the resource-allowed output range, the aggregated output power exceeding the scheduling capacity is reduced.

[0114] S23. Determine the power deviation corresponding to the current iteration based on the updated output power of each distributed resource; Complete the first After updating the output power of each distributed resource in each iteration, the aggregated output power of the current iteration is obtained according to the aforementioned method for determining the aggregated output power. Substituting this into the power deviation determination relation, we obtain the power deviation corresponding to the current iteration. .

[0115] The power deviation of the current iteration reflects the position and degree of difference of the aggregated output power relative to the scheduling capacity after this power adjustment, and serves as feedback information for subsequent judgments on whether the direction of the power deviation has changed.

[0116] S3. When the power deviation direction changes between adjacent iterations, update the power adjustment parameters according to the output power changes of each distributed resource and the scheduling priority relationship between adjacent iterations. In the process of iteratively adjusting the output power of each distributed resource using the AIMD algorithm, the addition step size and multiplication factor determine the power adjustment range of each distributed resource in a single iteration. If the power adjustment range is large, the aggregated output power may cross from one side of the scheduling capacity to the other after one iteration; if the original power adjustment parameters are continued, the aggregated output power is prone to significant back-and-forth changes on both sides of the scheduling capacity. If the power adjustment range is small, the number of iterations required for the aggregated output power to approach the scheduling capacity increases accordingly. Therefore, this embodiment does not update the power adjustment parameters after each iteration, but determines whether the aggregated output power has exceeded the scheduling capacity based on whether the power deviation direction of adjacent iterations has changed. When the power deviation direction changes, the parameter update amount corresponding to each distributed resource is determined based on the actual output power change of each distributed resource in this iteration, the scheduling priority relationship of each distributed resource, and the degree of deviation after the aggregated output power exceeds the scheduling capacity, and the addition step size or multiplication factor used in the next stage is updated accordingly.

[0117] Through the above processing, the update of power adjustment parameters not only depends on the overall deviation between aggregated output power and scheduling capacity, but also reflects the actual changes of different distributed resources in this power adjustment, and maintains the guiding role of the aforementioned scheduling priority relationship in the subsequent power allocation process. The specific process is as follows: Figure 2 As shown.

[0118] S31. Determine whether the power deviation direction changes between adjacent iterations; The sign of the power deviation is used to indicate whether the aggregated output power is above or below the scheduling capacity. To determine whether the aggregated output power exceeds the scheduling capacity between adjacent iterations, this embodiment uses the product of the current iteration power deviation and the previous iteration power deviation.

[0119] When the following conditions are met:

[0120] When the power deviation direction changes between two adjacent iterations, it is determined that the direction of the power deviation changes.

[0121] Wherein, when the following conditions are met:

[0122]

[0123] When, it indicates the first The aggregated output power corresponding to the second iteration is lower than the scheduling capacity. After additive increase, the power of the second iteration is lower than the scheduling capacity. The aggregated output power corresponding to the next iteration has exceeded the scheduling capacity, that is, the aggregated output power has moved from below the scheduling capacity to above the scheduling capacity.

[0124] When the following conditions are met:

[0125]

[0126] When, it indicates the first The aggregated output power corresponding to the second iteration is higher than the scheduling capacity. After multiplicative reduction, the power of the third iteration is... The aggregated output power corresponding to the next iteration is already lower than the scheduling capacity, that is, the aggregated output power has moved from above the scheduling capacity to below the scheduling capacity.

[0127] The change in the direction of power deviation indicates that the current power adjustment parameters have caused the aggregated output power to exceed the scheduling capacity. The power adjustment direction in the next iteration will change accordingly. Therefore, it is necessary to redetermine the power adjustment parameters to be used in the next stage based on the actual power adjustment results of this iteration.

[0128] When the following conditions are met:

[0129] When the power deviation direction between two adjacent iterations remains unchanged, it is determined that the power deviation direction has not changed. Specifically, when both the power deviation of the previous iteration and the power deviation of the current iteration are less than zero, it indicates that the aggregated output power corresponding to the two adjacent iterations is lower than the scheduling capacity, and the aggregated output power, after additive increase, has not yet exceeded the scheduling capacity. When both the power deviation of the previous iteration and the power deviation of the current iteration are greater than zero, it indicates that the aggregated output power corresponding to the two adjacent iterations is higher than the scheduling capacity, and the aggregated output power, after multiplicative decrease, has not yet exceeded the scheduling capacity. In both of these cases, the current power adjustment direction can still bring the aggregated output power closer to the scheduling capacity. Therefore, the power adjustment parameter update is not triggered, the currently used addition step size and multiplication factor are maintained, and the next iteration continues according to the corresponding power adjustment direction.

[0130] When the current iteration power deviation is equal to zero, or the current iteration power deviation meets the preset conditions, it indicates that the current aggregated output power has reached the scheduling capacity or is within the allowable power deviation range, and the power adjustment parameter update in this step will no longer be performed.

[0131] S32. Determine the corresponding change in output power based on the difference in output power of each distributed resource in adjacent iterations; After determining that the power deviation direction has changed, further determine the actual output power change of each distributed resource during the process of the aggregated output power exceeding the scheduling capacity.

[0132] No. The change in output power of a distributed resource between adjacent iterations is determined as follows:

[0133] In the formula, Indicates the first The distributed resource is provided by the first The iteration continues to the... The change in output power in each iteration. In the additive increase phase, the change in output power is usually positive, and its value reflects the actual increase in output power of the corresponding distributed resource; in the multiplicative decrease phase, the change in output power is usually negative, and its absolute value reflects the actual decrease in output power of the corresponding distributed resource.

[0134] This embodiment uses the actual output power change between adjacent iterations, rather than directly using a preset addition step size or multiplication factor. This is because the output power of distributed resources may be limited by the maximum allowable output power, the minimum allowable output power, and the actual response state. The theoretical adjustment range calculated according to the power adjustment parameters may differ from the actual power change achieved by the resource. Using the actual output power change allows subsequent parameter updates to be based on the power adjustment results already achieved by each distributed resource in this iteration.

[0135] S33. Normalize the absolute value of the output power change of each distributed resource to obtain the corresponding power change percentage. Different distributed resources may have different capacity levels and actual power change rates. Directly using the output power change of each distributed resource makes it difficult to compare its relative change rate within the overall power change. Therefore, this embodiment normalizes the absolute value of the output power change of each distributed resource to obtain the corresponding power change percentage:

[0136] In the formula, Indicates the first The distributed resource in the first The percentage of power change corresponding to each iteration; Indicates the first The distributed resource is provided by the first The iteration continues to the... The output power change in each iteration. The sum of the power change percentages corresponding to each distributed resource is 1. The absolute value of the output power change is taken to eliminate the sign effect caused by the different directions of power change in the additive increase and multiplicative decrease stages, so that the actual power change amplitudes of each distributed resource in different adjustment stages can be compared using a unified standard.

[0137] After normalization, the resulting power change percentage reflects the proportion of the actual output power change of the corresponding distributed resource to the total actual power change of all distributed resources. A larger power change percentage indicates a larger relative power change magnitude for the corresponding distributed resource in this iteration; a smaller power change percentage indicates a smaller relative power change magnitude. Therefore, using the power change percentage to determine the parameter update amount allows the parameter update magnitude for different distributed resources to be matched with their actual power change in this iteration.

[0138] Since the aggregated output power has actually changed between adjacent iterations when the power deviation direction changes, and the output power change of at least one distributed resource is not zero, the denominator of the above normalization calculation is greater than zero.

[0139] S34. Determine the scaling factor based on the ratio of the absolute value of the current power deviation to the scheduling capacity; A change in the direction of power deviation indicates that the aggregated output power has exceeded the scheduling capacity, but it does not directly reflect the degree of deviation after exceeding the capacity. For example, the aggregated output power may only slightly exceed the scheduling capacity, or it may result in a large power excess or power deficit after one iteration. If the same parameter update magnitude is used for different degrees of deviation, it is difficult to make subsequent power adjustments adapt to the current aggregated output power state.

[0140] Therefore, in this embodiment, the scaling factor corresponding to the current iteration is determined based on the ratio of the absolute value of the current power deviation to the scheduling capacity.

[0141] In the formula, Indicates the first The scaling factor corresponding to the next iteration.

[0142] The absolute value of the current power deviation reflects the degree of deviation of the aggregated output power after exceeding the scheduling capacity. Normalizing the current power deviation using the scheduling capacity can reduce the impact of different scheduling task capacity levels on the parameter update amplitude, making the scaling factor reflect the proportion of the current power deviation relative to the current scheduling capacity. When the deviation of the aggregated output power after exceeding the scheduling capacity is large, the scaling factor is large, resulting in a relatively large update amplitude for subsequent power adjustment parameters; when the aggregated output power only slightly exceeds the scheduling capacity, the scaling factor is small, making subsequent power adjustment more gradual and avoiding the use of a large parameter update amplitude when the scheduling capacity is already close.

[0143] In one implementation, the scaling factor can be limited according to a preset scaling range to avoid excessive parameter updates caused by abnormal power deviations. The limiting process does not alter the correlation between the scaling factor and the increase in the current power deviation ratio.

[0144] S35. Determine the parameter update amount for each distributed resource based on the power change ratio, scheduling priority relationship and scaling factor; After obtaining the power change percentage for each distributed resource and the scaling factor for the current iteration, this embodiment further determines the corresponding parameter update amount by combining the scheduling priority relationship of each distributed resource. The power change percentage reflects the relative magnitude of the corresponding distributed resource in the current actual power change; the scheduling priority relationship reflects the scheduling priority differences between different distributed resources; and the scaling factor reflects the degree of deviation after the aggregated output power exceeds the scheduling capacity. By using these three pieces of information to determine the parameter update amount, it avoids applying the same parameter correction magnitude to all distributed resources based solely on the overall power deviation.

[0145] In one specific implementation, relative proximity is used. As the first A numerical representation of the scheduling priority relationship of distributed resources. When the power deviation changes from negative to positive, the aggregated output power moves from below the scheduling capacity to above it, and the output power needs to be reduced in the next stage. To ensure that distributed resources with higher scheduling priority retain relatively more output power, the inverse vector of relative proximity is used to determine the parameter update amount:

[0146] When the power deviation changes from positive to negative, the aggregated output power moves from above the scheduling capacity to below it, requiring an increase in output power in the next stage. To ensure that distributed resources with higher scheduling priority bear relatively larger power increments, relative proximity is used to determine the parameter update amount:

[0147] In the formula, Indicates the first The distributed resource in the first The parameter update amount corresponding to each iteration.

[0148] Among these, under the condition that the power deviation changes from negative to positive and the proportion of power change is the same, the greater the relative closeness, the better. The smaller the value, the smaller the parameter update amount for distributed resources with higher scheduling priority; the smaller the relative proximity, The larger the relative proximity, the greater the parameter update amount for distributed resources with lower scheduling priority; under the condition that the power deviation changes from positive to negative and the proportion of power change is the same, the relative proximity... The larger the value, the greater the corresponding parameter update amount, allowing distributed resources with higher scheduling priority to have a relatively larger power increase in the next additive increase phase; relative proximity The smaller the value, the smaller the corresponding parameter update amount.

[0149] A larger percentage of power change indicates a greater proportion of the actual power change of the corresponding distributed resource in this iteration, resulting in a more significant update to its power adjustment parameters. A larger scaling factor indicates a greater deviation of the aggregated output power after exceeding the scheduling capacity, leading to a corresponding increase in the parameter update magnitude for each distributed resource. Therefore, the parameter update amount simultaneously reflects the actual power change of each distributed resource, scheduling priority relationships, and the current aggregated output power status, and is used to determine the power adjustment parameters to be used in the next power adjustment phase.

[0150] S36. When the power deviation changes from negative to positive, reduce the multiplication factor of the corresponding distributed resource according to the parameter update amount. When the power deviation of the previous iteration is negative and the power deviation of the current iteration is positive, it indicates that the aggregated output power after additive increase has exceeded the scheduling capacity, and the next iteration needs to switch from the additive increase phase to the multiplicative decrease phase. At this time, according to the... The parameter update amount corresponding to each distributed resource is reduced by decreasing its multiplication factor:

[0151] Since the multiplication factor in the additive addition phase is set to a unit value, the above update causes the multiplication factor to change from a unit value to less than 1. Simultaneously, the first... The addition step size corresponding to each distributed resource is set to zero, so that the next iteration performs multiplicative reduction according to the updated multiplication factor.

[0152] The larger the parameter update amount, the greater the decrease in the multiplication factor, and the greater the reduction in the output power of the distributed resource in the next iteration; the smaller the parameter update amount, the closer the multiplication factor is to a unit value, and the more output power the distributed resource retains. When the multiplication factor determined in the above manner is not... When the value is within the valid range, the multiplication factor is limited to ensure that the updated multiplication factor is within the valid range.

[0153] By combining the power change ratio, the inverse vector of relative proximity, and the scaling factor to reduce the multiplication factor, distributed resources that had a large actual power change and low scheduling priority in the previous additive increase phase will bear a relatively large power reduction in the subsequent multiplicative decrease phase.

[0154] S37. When the power deviation changes from positive to negative, increase the addition step size of the corresponding distributed resource according to the parameter update amount. When the power deviation in the previous iteration is positive and the power deviation in the current iteration is negative, it indicates that the aggregated output power after multiplicative reduction is already lower than the scheduling capacity, and the next iteration needs to switch from the multiplicative reduction phase to the additive increase phase. At this time, according to the... To increase the parameter update amount for each distributed resource, increase its addition step size:

[0155] Since the addition step size in the multiplicative decrease phase is set to zero, the above update determines the addition step size for the next additive increase phase based on the parameter update amount determined in this update. Simultaneously, the... The multiplication factor for each distributed resource is set to a unit value, causing the next iteration to perform additive increases according to the updated addition step size. The larger the parameter update amount, the greater the increase in the output power of the corresponding distributed resource in the next iteration; the smaller the parameter update amount, the smoother the increase in the output power of the corresponding distributed resource.

[0156] By combining the power change ratio, relative proximity, and scaling factor with the increase in the addition step size, distributed resources with larger actual power changes and higher scheduling priority in the previous multiplicative reduction phase will bear a relatively larger power increment in the subsequent additive increase phase.

[0157] Through the above parameter update process, whenever the aggregated output power exceeds the scheduling capacity between adjacent iterations, the power adjustment parameters for the next stage are redefined based on the actual power change, the scheduling priority of each distributed resource, and the degree of power deviation after the exceedance. As the aggregated output power gradually approaches the scheduling capacity, the current power deviation and corresponding scaling factor gradually decrease, and the parameter update magnitude for each distributed resource decreases accordingly, thereby gradually narrowing the range of variation of the aggregated output power on both sides of the scheduling capacity. The updated addition step size and multiplication factor are used for subsequent power iterations until the power deviation meets the preset conditions.

[0158] S4. Continue iterating based on the updated power adjustment parameters until the power deviation meets the preset conditions, and output the target output power of each distributed resource. After adjusting the output power for the current iteration, it is determined whether the current power deviation meets the preset conditions. If the power deviation does not meet the preset conditions, and the direction of the current power deviation changes from that of the previous iteration, the updated power adjustment parameters from S3 are used for the next iteration; otherwise, the current power adjustment parameters are maintained for continued iteration. Thus, the power adjustment parameters are only corrected when the aggregated output power exceeds the scheduling capacity, and the existing parameters are maintained as the aggregated output power continues to approach the scheduling capacity.

[0159] S41. Continue to adjust the output power of each distributed resource based on the current power deviation and the corresponding power adjustment parameters; When the current power deviation is negative, it indicates that the aggregated output power is lower than the scheduling capacity. The next iteration will continue to use the additive increase method and adjust the output power according to the current addition step size of each distributed resource. When the current power deviation is positive, it indicates that the aggregated output power is higher than the scheduling capacity. The next iteration will continue to use the multiplicative decrease method and adjust the output power according to the current multiplication factor of each distributed resource. After completing the next iteration in the manner described in S2, the aggregated output power and power deviation corresponding to the next iteration will be re-determined, and it will continue to be determined whether the power deviation meets the preset conditions.

[0160] S42. Determine whether the power deviation of the current iteration meets the preset conditions; To determine whether the current aggregated output power has met the requirements for dispatch command decomposition, this embodiment sets an allowable relative deviation threshold based on the ratio of the current power deviation to the dispatch capacity. When the following conditions are met:

[0161] When the power deviation of the current iteration meets the preset condition, it is determined that the power deviation of the current iteration meets the preset condition.

[0162] In the formula, This indicates the allowable relative deviation threshold, and The allowable relative deviation threshold can be determined based on the power point tracking accuracy of the virtual power plant, the power regulation resolution of distributed resources, and the dispatch response requirements. A smaller allowable relative deviation threshold results in a smaller permissible difference between the aggregated output power and the dispatch capacity, but the number of iterations required to reach the preset condition may increase; a larger allowable relative deviation threshold may reduce the number of iterations required to reach the preset condition, but the permissible power difference increases accordingly. In another embodiment, the absolute value of the power deviation not exceeding a preset allowable power deviation can also be used as the preset condition.

[0163] S43, Output the target output power of each distributed resource; When the power deviation of the current iteration meets the preset condition, the power iteration stops, and the output power of each distributed resource in the current iteration is determined as the corresponding target output power:

[0164] In the formula, Indicates the first The target output power corresponds to each distributed resource. Since each distributed resource is subject to corresponding minimum and maximum allowable output power limitations during the iteration process, the obtained target output power is within the allowable output power range of the corresponding distributed resource. The difference between the aggregated output power formed by the target output power of each distributed resource and the scheduling capacity satisfies the preset condition. If the aforementioned iteration process uses per-unit values ​​for calculation, the target output power of each distributed resource is converted into an actual power value according to the corresponding reference power, and the converted target output power is output as the decomposition result of the virtual power plant scheduling instruction.

[0165] To verify the power iteration effect of the method of the present invention, 50 representative distributed resources were selected to construct a simulation scenario. Based on the power generation cost, power generation capacity, regulation capacity, response time, ramp rate, power generation duration, and regulation deviation of each distributed resource, the relative proximity of each distributed resource was determined according to the method described in S1. And scheduling priority relationship.

[0166] This simulation sets the scheduling capacity to 40kW, the initial output power of each distributed resource to 0, and the allowable power deviation threshold to 50W. Following the aforementioned per-unit processing method, the scheduling capacity, the initial output power of each distributed resource, and the allowable power deviation threshold are converted to their corresponding per-unit values ​​before AIMD iteration is performed. For the method of this invention, the first... The initial addition step size and initial multiplication factor for each distributed resource are set as follows:

[0167]

[0168] During the iteration process, when the power deviation direction of adjacent iterations changes, the corresponding power adjustment parameters are updated according to the power change ratio of each distributed resource, scheduling priority relationship and current power deviation, as described in S3.

[0169] For comparison, the fixed-parameter AIMD method was used to iterate the power of the same distributed resource. The addition step size and multiplication factor of the fixed-parameter AIMD method were set as follows:

[0170]

[0171] And it remains unchanged during the iteration process. Except for the setting and updating method of power regulation parameters, the two methods use the same distributed resources, initial output power, scheduling capacity, and allowable power deviation threshold.

[0172] It should be noted that the fixed-parameter AIMD method still uses the relative proximity of each distributed resource to set the corresponding addition step size and multiplication factor, ensuring it adopts the same scheduling priority relationship as the method of this invention. This eliminates the influence of different scheduling priority relationships between the two methods on the comparison results. This setting is not intended to indicate that the relative proximity is the optimal fixed parameter, but rather to make the main difference between the two methods focus on whether the power adjustment parameter is adaptively updated according to the iteration state. The fixed-parameter AIMD method keeps the addition step size and multiplication factor unchanged during the iteration process. Therefore, even if the aggregated output power is close to the scheduling capacity, the subsequent additive increase and multiplicative reduction ratio remain at the original level, which easily causes the aggregated output power to repeatedly exceed the scheduling capacity. The method of this invention, on the other hand, updates the power adjustment parameter according to the current power deviation, the power change ratio of each distributed resource, and the scheduling priority relationship when the power deviation direction changes. This ensures that the subsequent power adjustment amplitude is adapted to the current deviation level, thereby gradually reducing the range of repeated changes in aggregated output power relative to the scheduling capacity.

[0173] To compare the effects of fixed power regulation parameters and adaptively updated power regulation parameters on the aggregation output power variation process, Figure 3 The aggregated output power iteration trajectory of the two methods is shown continuously. During the actual execution of the scheduling instruction decomposition, iteration stops when the power deviation meets the preset condition described in S42.

[0174] like Figure 3 As shown, in the initial iteration phase, both methods gradually bring the aggregated output power closer to the 40kW scheduling capacity. The fixed-parameter AIMD method uses a constant addition step size and multiplication factor, causing the aggregated output power to be multiplicatively reduced after exceeding the scheduling capacity and then additively increased again after falling below it, resulting in a significant back-and-forth variation around the scheduling capacity. The power variation near the scheduling capacity in the fixed-parameter AIMD method depends on the numerical matching relationship between the fixed power adjustment parameters and the current scheduling conditions. In contrast, the method of this invention updates the power adjustment parameters corresponding to each distributed resource when the power deviation direction changes, causing the difference between the aggregated output power and the scheduling capacity to gradually decrease during the iteration process.

[0175] The simulation results show that the method of the present invention can adaptively update the power adjustment parameters according to the iterative state of the aggregated output power and the actual power changes of each distributed resource, reduce the power fluctuations near the scheduling capacity caused by fixed parameter settings, and reduce the dependence of the scheduling instruction decomposition process on the initial power adjustment parameter tuning results.

[0176] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for decomposing distributed resource adaptive scheduling instructions in a virtual power plant, characterized in that, include: Based on the scheduling priority relationship of each distributed resource, initialize the power adjustment parameters corresponding to each distributed resource. The output power of each distributed resource is iteratively adjusted based on the power adjustment parameters, and the power deviation characterizing the difference between the aggregated output power and the scheduling capacity is determined. When the power deviation direction changes between adjacent iterations, the power adjustment parameters are updated according to the output power changes of each distributed resource and the scheduling priority relationship between adjacent iterations. The process continues iterating based on the updated power adjustment parameters until the power deviation meets the preset conditions, at which point the target output power of each distributed resource is output.

2. The virtual power plant distributed resource adaptive scheduling instruction decomposition method according to claim 1, characterized in that, The determination of the scheduling priority relationship includes: Obtain the evaluation index data for each distributed resource and determine the index weight corresponding to each evaluation index; Based on the evaluation index data and the corresponding index weights, the relative proximity of each distributed resource is determined, and the scheduling priority relationship of each distributed resource is determined according to the relative proximity.

3. The virtual power plant distributed resource adaptive scheduling instruction decomposition method according to claim 2, characterized in that, The evaluation index data includes power generation cost, power generation capacity, regulation capacity, response time, ramp rate, power generation duration, and regulation deviation; Among them, the power generation cost, response time and regulation deviation are negative evaluation indicators, while the power generation capacity, regulation capacity, ramp rate and power generation duration are positive evaluation indicators.

4. The virtual power plant distributed resource adaptive scheduling instruction decomposition method according to claim 2, characterized in that, The determination of the indicator weights includes: The evaluation index data is subjected to homogenization and normalization processing. The information entropy of each evaluation index is determined based on the processed evaluation index data. The corresponding index weight is determined based on the degree of index difference represented by the information entropy.

5. The virtual power plant distributed resource adaptive scheduling instruction decomposition method according to claim 4, characterized in that, The determination of the relative proximity includes: Based on the evaluation index data after normalization and homogenization and the index weights, a weighted decision matrix is ​​constructed to determine the positive ideal solution and the negative ideal solution. The relative proximity is determined based on the distance between each distributed resource and the positive ideal solution and the negative ideal solution, respectively.

6. The method for decomposing distributed resource adaptive scheduling instructions for virtual power plants according to claim 1, characterized in that, The power adjustment parameters include the addition step size and the multiplication factor; When the power deviation of the previous iteration is negative, the addition step size is added to the previous iteration output power of the corresponding distributed resource to determine the current iteration output power; When the power deviation of the previous iteration is positive, the output power of the current iteration is determined by multiplying the multiplication factor with the output power of the corresponding distributed resource in the previous iteration.

7. The virtual power plant distributed resource adaptive scheduling instruction decomposition method according to claim 6, characterized in that, The initialization of power adjustment parameters corresponding to each distributed resource includes: When the initial aggregated output power is less than the scheduling capacity, the initial addition step size corresponding to each distributed resource is determined according to the scheduling priority relationship, and the corresponding initial multiplication factor is set to a unit value. When the initial aggregated output power is greater than the scheduling capacity, the initial addition step size corresponding to each distributed resource is set to zero, and the corresponding initial multiplication factor is determined according to the scheduling priority relationship.

8. The virtual power plant distributed resource adaptive scheduling instruction decomposition method according to claim 1, characterized in that, When the product of the power deviations between two consecutive iterations is less than zero, it is determined that the direction of the power deviation has changed between the adjacent iterations. The change in output power for each distributed resource is determined based on the difference in output power between adjacent iterations.

9. The virtual power plant distributed resource adaptive scheduling instruction decomposition method according to claim 8, characterized in that, The updated power adjustment parameters include: Normalize the absolute value of the output power change of each distributed resource to obtain the power change ratio of each distributed resource. The scaling factor is determined based on the ratio of the absolute value of the current power deviation to the scheduling capacity; The parameter update amount for each distributed resource is determined based on the power change ratio, the scheduling priority relationship, and the scaling factor.

10. The method for decomposing distributed resource adaptive scheduling instructions for virtual power plants according to claim 9, characterized in that, The power regulation parameters include an addition step size for increasing output power and a multiplication factor for reducing output power. When the power deviation changes from negative to positive, the multiplication factor of the corresponding distributed resource is reduced according to the parameter update amount; When the power deviation changes from positive to negative, the addition step size of the corresponding distributed resource is increased according to the parameter update amount.

Images