Virtual power plant adjustable resource dynamic scheduling method and system

Abstract

This invention provides a method and system for dynamic scheduling of adjustable resources in a virtual power plant, relating to the field of power dispatching technology. The method includes: collecting distributed energy node data to determine output constraint boundaries, calculating segmented adjustment capacity, and statistically analyzing response attenuation coefficients; identifying capacity complementarity relationships, calculating collaborative reliability, and forming resource aggregation units; calculating target output values ​​based on power demand commands and decomposing them into allocated output values; pre-compensating for the target output values ​​and adjusting reserve capacity; and updating the response attenuation coefficients and collaborative reliability in real time to complete dynamic reconfiguration. This invention improves the collaborative reliability and scheduling accuracy of distributed energy resources.

Description

Technical Field

[0001] This invention relates to the field of power dispatching technology, and in particular to a method and system for dynamic dispatching of adjustable resources in a virtual power plant. Background Technology

[0002] With the reform of the electricity market and the increase in the proportion of renewable energy, the share of distributed energy in the power system continues to rise. To better manage and optimize these dispersed energy resources, virtual power plants have emerged as a new energy management model. Virtual power plants integrate and coordinate geographically dispersed and diverse distributed energy resources through information technology, forming a controllable and dispatchable energy aggregate to provide flexible power balancing services to the power grid.

[0003] Currently, virtual power plant technology has been applied in many countries, mainly by aggregating resources such as distributed generation, controllable loads, and energy storage devices to participate in electricity market transactions and ancillary services. However, with the expansion of the scale of virtual power plants and the diversification of resource types, their dispatch and management face many challenges.

[0004] Existing virtual power plant dispatching technologies still have some defects and shortcomings. They lack precision in assessing the adjustability of distributed energy resources, making it difficult to accurately grasp the adjustment capabilities of various resources at different time scales. Static assessment methods are often used, failing to reflect the dynamic changes in resources and leading to response deviations during actual dispatching. Furthermore, the lack of effective mechanisms for identifying and utilizing complementary relationships between distributed energy resources prevents the realization of resource synergy, resulting in the overall dispatching capacity of the virtual power plant being limited by the characteristics of a single resource type, hindering synergistic effects and leading to low overall system dispatching efficiency. Existing dispatching methods also underutilize historical response data, failing to effectively learn and adapt to the response characteristics of distributed energy resources. Particularly when facing response decay and deviation, the lack of corresponding pre-compensation and dynamic adjustment mechanisms causes the response accuracy of the virtual power plant to gradually decrease during continuous dispatching tasks, affecting the system's reliability and stability. Summary of the Invention

[0005] This invention provides a method and system for dynamic scheduling of adjustable resources in a virtual power plant, which can solve the problems in the prior art.

[0006] A first aspect of this invention provides a method for dynamic scheduling of adjustable resources in a virtual power plant, comprising:

[0007] Collect operational status data, historical response data, and power demand commands from each distributed energy node;

[0008] Based on the operational status data, the output constraint boundary of each distributed energy node is determined, the segmented adjustment capacity under different scheduling durations is calculated, and the response attenuation coefficient is statistically calculated based on historical response data.

[0009] Based on the differences in the distribution of segmented adjustment capacity over time, the capacity complementarity relationship between distributed energy nodes is identified. The collaborative reliability is calculated by combining the response decay coefficient. Distributed energy nodes with capacity complementarity and collaborative reliability that meet the preset reliability threshold are grouped into resource aggregation units.

[0010] Based on the distributed energy nodes and corresponding segmented regulation capacity contained in the resource aggregation unit, the target output value is calculated in conjunction with the power demand command, and the target output value is checked against the output constraint boundary. The output value is then decomposed into allocated output values ​​according to the segmented regulation capacity ratio.

[0011] Extract historical response deviations, establish a mapping relationship with the target output value, pre-compensate for the target output value, and adjust the reserve capacity according to the magnitude of the response deviation.

[0012] The power sequence of the resource aggregation unit is collected to update the response attenuation coefficient, the capacity complementarity relationship is identified to calculate the cooperative reliability, and dynamic reorganization is completed according to the cooperative reliability.

[0013] In one optional embodiment, the output constraint boundary of each distributed energy node is determined based on the operating status data, the segmented adjustment capacity under different scheduling durations is calculated, and the response attenuation coefficient is statistically calculated based on historical response data, including:

[0014] The operating parameter time series of each distributed energy node is extracted from the operating status data. The operating parameter time series is reconstructed into a high-dimensional phase space. The attractor boundary of the operating status trajectory is identified in the high-dimensional phase space. The power value corresponding to the projection of the current operating status on the attractor boundary is determined as the output constraint boundary.

[0015] The response speed of each distributed energy node under different initial power and different adjustment range in historical scheduling is extracted. The response surface of the response speed with respect to the initial power and adjustment range is constructed. The adjustment path from the current operating state to the output constraint boundary is calculated. The response time is obtained by integrating the response surface along the adjustment path. When the response time is less than the scheduling duration, the difference between the output constraint boundary and the current power is used as the segmented adjustment capacity. When the response time exceeds the scheduling duration, the maximum power under the scheduling duration constraint is solved iteratively. The difference between the maximum power and the current power is used as the segmented adjustment capacity.

[0016] The deviation sequence between scheduling instructions and actual execution power is extracted from historical response data. Autoregressive spectral analysis is performed on the deviation sequence to obtain the spectral distribution. The frequency component with the highest energy is extracted from the spectral distribution, and the energy attenuation index of the frequency component with the highest energy is calculated as the response attenuation coefficient.

[0017] In one optional embodiment, based on the distribution differences of segmented adjustment capacity in the time dimension, the capacity complementarity relationship between distributed energy nodes is identified, and the cooperative reliability is calculated in combination with the response attenuation coefficient. Distributed energy nodes with capacity complementarity and cooperative reliability meeting a preset reliability threshold are grouped into resource aggregation units, including:

[0018] The segmented adjustment capacity of each distributed energy node under different scheduling durations is arranged according to the duration dimension to form a capacity distribution sequence. Any two distributed energy nodes are selected to determine the first node and the second node, resulting in a node pair. Cross-correlation analysis is performed on the capacity distribution sequence of the node pair. The correlation coefficient under different time offsets is calculated through a sliding time window. The time offset that makes the correlation coefficient reach a negative minimum value is recorded to determine the phase offset. The frequency of the occurrence of the peak time of the first node corresponding to the valley time of the second node is counted to determine the peak-valley misalignment degree. When the phase offset reaches half a cycle and the peak-valley misalignment degree exceeds the preset misalignment threshold, it is determined that the first node and the second node have a capacity complementary relationship.

[0019] A capacity complementarity network is constructed using each distributed energy node as a network node. Network edges are established between node pairs with capacity complementarity. The reciprocal of the response attenuation coefficient is used as the node weight, and the peak-valley misalignment degree is used as the edge weight. The capacity complementarity network is divided into multiple connected subgraphs. The harmonic average of the node weights in each connected subgraph is calculated to determine the collaborative reliability.

[0020] All distributed energy nodes in a connected subgraph whose collaborative reliability meets a preset reliability threshold are grouped into the same resource aggregation unit.

[0021] In one alternative embodiment, segmenting the capacity complementarity network to obtain multiple connected subgraphs includes:

[0022] Select the node with the largest node weight from the capacity complementary relationship network to determine the initial seed node, and include all nodes that are directly connected to the initial seed node through network edges into the first candidate node set;

[0023] Calculate the product of the edge weight between each node in the first candidate node set and the initial seed node and the node weight of each node, determine the aggregation gain value of the corresponding node, and absorb the nodes in the first candidate node set into the first connected subgraph where the initial seed node is located in descending order of aggregation gain value.

[0024] After absorbing a node, the unabsorbed nodes that are directly connected to the absorbed node through network edges are added as supplementary nodes to the first candidate node set, and the aggregation gain value of the supplementary nodes is recalculated. When the largest aggregation gain value in the first candidate node set is lower than the preset gain threshold, the absorption process stops and the construction of the first connected subgraph is completed.

[0025] From the remaining unassigned nodes in the capacity complementarity network, select the node with the largest node weight to obtain a new seed node, and repeat the above construction process until all nodes in the capacity complementarity network are assigned to the corresponding connected subgraph.

[0026] In one optional embodiment, extracting historical response deviations, establishing a mapping relationship with the target output value, pre-compensating for the target output value, and adjusting the reserve capacity according to the magnitude of the response deviation include:

[0027] Extract the actual output power and target output power of each resource aggregation unit from historical response data, calculate the difference between the actual output power and the target output power to obtain the actual response deviation, perform sliding window processing on the actual response deviation, calculate the mean value within each window to obtain the trend deviation, and calculate the residual between the actual response deviation and the trend deviation to obtain the fluctuation deviation.

[0028] Using the target output value as the independent variable and the trend deviation as the dependent variable, a polynomial fitting is performed to determine the fitting curve. The target output value is divided into multiple intervals, and the extreme range of the fluctuation deviation in each interval is statistically analyzed to form a range table. The fitting curve and the range table are used together as a mapping relationship.

[0029] Substitute the current target output value into the fitted curve to obtain the predicted trend deviation. Add the predicted trend deviation to the current target output value to obtain the pre-compensation target output value. Query the extreme value range of the interval corresponding to the current target output value from the range table. Calculate the span of the extreme value range to determine the response deviation amplitude. Reserve reserve capacity from the regulating capacity according to the ratio of the response deviation amplitude to the pre-compensation target output value.

[0030] In one optional embodiment, the process of acquiring the power sequence update response attenuation coefficient of the resource aggregation unit, identifying capacity complementarity relationships to calculate cooperative reliability, and performing dynamic reorganization based on cooperative reliability includes:

[0031] Collect the actual output power sequence and scheduling instruction power sequence of each resource aggregation unit in the current scheduling cycle, calculate the corresponding root mean square error, determine the actual execution deviation, and calculate the ratio of the actual execution deviation to the reciprocal of the historical response attenuation coefficient to obtain the attenuation update factor. Multiply the historical response attenuation coefficient by the attenuation update factor to obtain the updated response attenuation coefficient.

[0032] Extract the output fluctuation sequence of each resource aggregation unit, calculate the convolution operation result of the output fluctuation sequences of any two resource aggregation units in the time domain, and determine that the resource aggregation units have a capacity complementary relationship and form a complementary combination when the peak position of the convolution operation result corresponds to a negative value.

[0033] For each complementary combination, the update response decay coefficient and segmented adjustment capacity of each resource aggregation unit within the combination are extracted. The combined response index is obtained by harmonic averaging the update response decay coefficient and the combined adjustment capacity is obtained by summing the segmented adjustment capacity. The product of the combined response index and the combined adjustment capacity is calculated to determine the collaborative reliability.

[0034] All complementary combinations are sorted in descending order of collaborative reliability. A corresponding number of complementary combinations are selected according to a preset optimal ratio. The resource aggregation units within the selected complementary combinations are merged to complete the dynamic reorganization.

[0035] A second aspect of the present invention provides a dynamic scheduling system for adjustable resources in a virtual power plant, comprising:

[0036] The data acquisition module is used to collect the operating status data, historical response data and power demand commands of each distributed energy node;

[0037] The capacity assessment module is used to determine the output constraint boundary of each distributed energy node based on the operating status data, calculate the segmented adjustment capacity under different scheduling durations, and calculate the response attenuation coefficient based on historical response data.

[0038] The resource aggregation module is used to identify the capacity complementarity between distributed energy nodes based on the distribution differences of segmented adjustment capacity in the time dimension, calculate the collaborative reliability in combination with the response attenuation coefficient, and group distributed energy nodes with capacity complementarity and collaborative reliability that meet the preset reliability threshold into resource aggregation units.

[0039] The power allocation module is used to calculate the target power output value based on the distributed energy nodes and corresponding segmented regulation capacity contained in the resource aggregation unit, combined with the power demand command, and to perform boundary verification on the target power output value according to the power output constraint boundary, and decompose it into allocated power output value according to the segmented regulation capacity ratio.

[0040] The deviation compensation module is used to extract historical response deviations, establish a mapping relationship with the target output value, pre-compensate the target output value, and adjust the reserve capacity according to the magnitude of the response deviation.

[0041] The dynamic reorganization module is used to collect the power sequence update response attenuation coefficient of the resource aggregation unit, identify the capacity complementarity relationship, calculate the cooperative reliability, and complete the dynamic reorganization according to the cooperative reliability.

[0042] A third aspect of the present invention provides an electronic device, comprising:

[0043] processor;

[0044] Memory used to store processor-executable instructions;

[0045] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0046] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0047] In this embodiment of the invention, capacity complementarity is identified based on the distribution differences of segmented regulation capacity in the time dimension, and collaborative reliability is calculated in conjunction with the response attenuation coefficient, making the combination of resource aggregation units more scientific and reasonable, and improving the overall response reliability. The target output value is decomposed into the allocated output value according to the segmented regulation capacity ratio, which allocates the load more reasonably and avoids the problem of some nodes exceeding the regulation capacity caused by the traditional average allocation method. By extracting historical response deviations and establishing a mapping relationship with the target output value for pre-compensation, and dynamically adjusting the reserve capacity according to the response deviation amplitude, the deviation between the actual response and the expected target is effectively reduced. The response attenuation coefficient is updated based on the real-time power sequence, and the collaborative reliability is re-evaluated accordingly for dynamic resource reorganization, enabling the resource configuration of the virtual power plant to adaptively adjust with changes in environment and load, thereby improving the overall flexibility and adaptability of the system. Attached Figure Description

[0048] Figure 1 This is a flowchart illustrating the dynamic scheduling method for adjustable resources in a virtual power plant according to an embodiment of the present invention.

[0049] Figure 2 This is a flowchart for assessing the regulation capacity of distributed energy resources. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0051] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0052] Figure 1 This is a flowchart illustrating the dynamic scheduling method for adjustable resources in a virtual power plant according to an embodiment of the present invention. Figure 1 As shown, the method includes:

[0053] Collect operational status data, historical response data, and power demand commands from each distributed energy node;

[0054] Based on the operational status data, the output constraint boundary of each distributed energy node is determined, the segmented adjustment capacity under different scheduling durations is calculated, and the response attenuation coefficient is statistically calculated based on historical response data.

[0055] Based on the differences in the distribution of segmented adjustment capacity over time, the capacity complementarity relationship between distributed energy nodes is identified. The collaborative reliability is calculated by combining the response decay coefficient. Distributed energy nodes with capacity complementarity and collaborative reliability that meet the preset reliability threshold are grouped into resource aggregation units.

[0056] Based on the distributed energy nodes and corresponding segmented regulation capacity contained in the resource aggregation unit, the target output value is calculated in conjunction with the power demand command, and the target output value is checked against the output constraint boundary. The output value is then decomposed into allocated output values ​​according to the segmented regulation capacity ratio.

[0057] Extract historical response deviations, establish a mapping relationship with the target output value, pre-compensate for the target output value, and adjust the reserve capacity according to the magnitude of the response deviation.

[0058] The power sequence of the resource aggregation unit is collected to update the response attenuation coefficient, the capacity complementarity relationship is identified to calculate the cooperative reliability, and dynamic reorganization is completed according to the cooperative reliability.

[0059] In one optional implementation, the output constraint boundary of each distributed energy node is determined based on the operating status data, the segmented adjustment capacity under different scheduling durations is calculated, and the response attenuation coefficient is statistically analyzed based on historical response data, including:

[0060] The operating parameter time series of each distributed energy node is extracted from the operating status data. The operating parameter time series is reconstructed into a high-dimensional phase space. The attractor boundary of the operating status trajectory is identified in the high-dimensional phase space. The power value corresponding to the projection of the current operating status on the attractor boundary is determined as the output constraint boundary.

[0061] The response speed of each distributed energy node under different initial power and different adjustment range in historical scheduling is extracted. The response surface of the response speed with respect to the initial power and adjustment range is constructed. The adjustment path from the current operating state to the output constraint boundary is calculated. The response time is obtained by integrating the response surface along the adjustment path. When the response time is less than the scheduling duration, the difference between the output constraint boundary and the current power is used as the segmented adjustment capacity. When the response time exceeds the scheduling duration, the maximum power under the scheduling duration constraint is solved iteratively. The difference between the maximum power and the current power is used as the segmented adjustment capacity.

[0062] The deviation sequence between scheduling instructions and actual execution power is extracted from historical response data. Autoregressive spectral analysis is performed on the deviation sequence to obtain the spectral distribution. The frequency component with the highest energy is extracted from the spectral distribution, and the energy attenuation index of the frequency component with the highest energy is calculated as the response attenuation coefficient.

[0063] In one specific implementation, time series of operating parameters for each distributed energy node are extracted from operational status data. These parameters include key indicators such as power output, energy availability, and environmental factors. Taking a photovoltaic power generation system as an example, the extracted parameters may include irradiance, temperature, and active power output; for a wind power system, they may include wind speed, wind direction, and power generation. The extracted time series records the operational trajectory of the distributed energy nodes, providing data support for subsequent analysis.

[0064] Phase space reconstruction is performed on the time series of operating parameters, transforming the one-dimensional time series into a high-dimensional phase space representation. Phase space reconstruction employs the delayed coordinate method, selecting an embedding dimension *m* and a delay time *τ* to map the original time series *x(t)* into an *m*-dimensional vector [x(t), x(t+τ), x(t+2τ), ..., x(t+(m-1)τ)]. The embedding dimension *m* is determined using a pseudo nearest neighbor algorithm, and the delay time *τ* is determined by minimizing the mutual information function. After phase space reconstruction, the operating state of the energy nodes is mapped to a series of points in the high-dimensional phase space, forming the trajectory of the system's dynamic evolution.

[0065] In high-dimensional phase space, attractor boundaries for the operating state trajectory are identified. An attractor represents a state region where the system tends to stabilize over long-term operation. Density clustering methods, such as the DBSCAN algorithm, are used to cluster points in the phase space, identifying high-density regions as attractor cores. The distance distribution from the core region to the peripheral points is calculated, and a boundary threshold is determined. The set of points with distances less than the threshold is defined as the attractor boundary. The boundary characterizes the state constraint range of the distributed energy node under normal operating conditions.

[0066] The power value corresponding to the projection of the current operating state onto the attractor boundary is determined as the output constraint boundary. The distance from the current state point to each point on the attractor boundary is calculated, and the n boundary points with the smallest distances are selected. Based on these n nearest boundary points, the projected coordinates in each dimension are calculated by weighting the average distance to the current point. The corresponding power components are extracted from the projected coordinates and determined as the output constraint boundary of this distributed energy node. For example, for a photovoltaic system, the maximum possible output power and minimum stable power under current weather conditions can be obtained.

[0067] The response speeds of each distributed energy node under different initial power and adjustment ranges in historical scheduling are extracted. By analyzing historical scheduling commands and actual power change data, the time required for the power to reach the target value from the issuance of the scheduling command is recorded, and the average response speed during the adjustment process is calculated. A bivariate function relationship is constructed with initial power and adjustment range as independent variables and response speed as the dependent variable.

[0068] A response surface is constructed to describe the response speed with respect to initial power and regulation range. A radial basis function (RBF) neural network is used to interpolate and fit the scattered data, generating a continuous response speed function surface. The response surface describes the expected response speed of the distributed energy node under arbitrary initial power and regulation range conditions, providing a foundation for subsequent regulation capacity calculations.

[0069] Calculate the adjustment path from the current operating state to the output constraint boundary. Using the current power as the starting point and the output constraint boundary as the ending point, construct the power adjustment trajectory. Considering the adjustment characteristics of distributed energy resources, the path can be designed as a piecewise linear or curved form to ensure a smooth and effective adjustment process.

[0070] The response time is obtained by integrating the response surface along the adjustment path. The adjustment path is discretized into multiple segments, each corresponding to an initial power and adjustment amplitude. The response velocity of each segment is obtained by querying the response surface, the time required for each segment is calculated, and the total response time from the current state to the output constraint boundary is obtained by summing them.

[0071] When the response time is less than the scheduling duration, it indicates that the distributed energy node can reach the output constraint boundary within the given scheduling time. At this time, the difference between the output constraint boundary and the current power is used as the segmented regulation capacity, representing the maximum regulation capability of the node under this scheduling duration.

[0072] When the response time exceeds the scheduling duration, the maximum power under the scheduling duration constraint is iteratively solved. An initial guess value is set between the current power and the output constraint boundary, and the response time to that power point is calculated. Based on the difference between the response time and the scheduling duration, the guess value is adjusted, and the calculation process is repeated until the maximum power point achievable within the scheduling duration is found. The difference between this maximum power and the current power is used as the segmented regulation capacity.

[0073] Extract the deviation sequence between scheduling instructions and actual executed power from historical response data. Calculate the time series difference between actual power output and scheduling instructions during each scheduling process to form a deviation sequence, which is used to analyze the response stability and sustainability of distributed energy nodes.

[0074] Autoregressive spectral analysis is performed on the deviation sequence to obtain its spectral distribution. Parametric methods such as autoregressive (AR) models are used to estimate the power spectral density function, or non-parametric methods such as periodogram methods are used to directly calculate the spectrum. The spectral distribution reveals the periodic components contained in the deviation sequence and their intensity.

[0075] The frequency component with the highest energy is extracted from the spectral distribution, as this component typically reflects the most significant fluctuation characteristic of the system response. The energy decay trend of this frequency component over time is calculated, and an exponential decay model e is fitted. -αt α is the energy decay exponent, which serves as the response decay coefficient. The decay coefficient reflects the stability of the distributed energy node after executing scheduling commands; the smaller the value, the more stable the system response.

[0076] The above methods can be used to comprehensively evaluate the output constraints and adjustment capabilities of distributed energy nodes under different scheduling conditions, providing reliable technical support for energy scheduling optimization.

[0077] like Figure 2 The diagram shown illustrates the flowchart for assessing the regulation capacity of distributed energy resources.

[0078] In one optional implementation, based on the distribution differences of segmented adjustment capacity over time, the capacity complementarity relationship between distributed energy nodes is identified. The cooperative reliability is calculated using the response decay coefficient. Distributed energy nodes with capacity complementarity and cooperative reliability meeting a preset reliability threshold are grouped into resource aggregation units, including:

[0079] The segmented adjustment capacity of each distributed energy node under different scheduling durations is arranged according to the duration dimension to form a capacity distribution sequence. Any two distributed energy nodes are selected to determine the first node and the second node, resulting in a node pair. Cross-correlation analysis is performed on the capacity distribution sequence of the node pair. The correlation coefficient under different time offsets is calculated through a sliding time window. The time offset that makes the correlation coefficient reach a negative minimum value is recorded to determine the phase offset. The frequency of the occurrence of the peak time of the first node corresponding to the valley time of the second node is counted to determine the peak-valley misalignment degree. When the phase offset reaches half a cycle and the peak-valley misalignment degree exceeds the preset misalignment threshold, it is determined that the first node and the second node have a capacity complementary relationship.

[0080] A capacity complementarity network is constructed using each distributed energy node as a network node. Network edges are established between node pairs with capacity complementarity. The reciprocal of the response attenuation coefficient is used as the node weight, and the peak-valley misalignment degree is used as the edge weight. The capacity complementarity network is divided into multiple connected subgraphs. The harmonic average of the node weights in each connected subgraph is calculated to determine the collaborative reliability.

[0081] All distributed energy nodes in a connected subgraph whose collaborative reliability meets a preset reliability threshold are grouped into the same resource aggregation unit.

[0082] In one specific implementation, the dynamic scheduling method for adjustable resources in a virtual power plant requires analyzing the segmented adjustment capacity of each distributed energy node under different scheduling durations. This involves acquiring segmented adjustment capacity data for each distributed energy node, including adjustment capacity values ​​for multiple time dimensions such as 5 minutes, 15 minutes, 30 minutes, 1 hour, 2 hours, and 4 hours. Taking a photovoltaic power station as an example, its 5-minute adjustment capacity is 100 kW, 15-minute adjustment capacity is 250 kW, 30-minute adjustment capacity is 400 kW, 1-hour adjustment capacity is 600 kW, 2-hour adjustment capacity is 800 kW, and 4-hour adjustment capacity is 1000 kW. These data are arranged according to the duration dimension to form the capacity distribution sequence of the node.

[0083] From a large pool of distributed energy nodes, any two nodes are selected for analysis, designated as the first and second nodes, forming a node pair. For example, a photovoltaic power station is selected as the first node and a wind farm as the second node within a virtual power plant. Cross-correlation analysis is performed on the capacity distribution sequences of this node pair, calculating the cross-correlation coefficient at different time offsets within a 24-hour sliding time window. Specifically, the capacity distribution sequence of the first node is kept fixed, while the capacity distribution sequence of the second node is shifted along the time axis with a shift step of 10 minutes, ranging from -12 hours to +12 hours. The Pearson correlation coefficient between the two sequences is calculated at each shift point. The time offset that causes the correlation coefficient to reach its negative minimum, i.e., the offset corresponding to the maximum negative absolute value of the correlation coefficient, is recorded as the phase offset between the two nodes.

[0084] In the example of the photovoltaic power plant and wind farm mentioned above, when the capacity distribution sequence of the second node is shifted forward by 6 hours, the correlation coefficient reaches -0.85, and the phase shift is recorded as 6 hours. Further analysis is needed to determine the frequency of the peak times of the first node corresponding to the valley times of the second node. Specifically, all peak points (local maxima) are identified from the capacity distribution sequence of the first node, and the corresponding time points are recorded. For each peak time point, it is checked whether the second node experiences a valley (local minimum) at the corresponding time point (considering the phase shift). The number of time points meeting the condition is counted and divided by the total number of peak points of the first node to obtain the peak-valley misalignment. If the photovoltaic power plant has 24 peak points, and 20 of these peak points correspond to valley points of the wind farm, then the peak-valley misalignment is 83.3%.

[0085] When the calculated phase offset is close to half a cycle (e.g., the offset within a 24-hour cycle is 10 to 14 hours), and the peak-valley misalignment exceeds a preset misalignment threshold (e.g., 75%), it is determined that the pair of nodes have a capacity complementary relationship. In the example above, the phase offset is 6 hours (close to half a 12-hour cycle), and the peak-valley misalignment is 83.3%, exceeding the preset threshold of 75%. Therefore, it is determined that the photovoltaic power station and the wind farm have a capacity complementary relationship.

[0086] After determining the complementarity of all node pairs, a capacity complementarity network is constructed. Each distributed energy node is treated as a network node, and network edges are established between node pairs with capacity complementarity. Each node is assigned a weight, which is the reciprocal of its response attenuation coefficient. For example, if the response attenuation coefficient of an energy storage power station is 0.85, then its node weight is 1.176. Each edge is also assigned a weight, which is the peak-valley misalignment between the two connected nodes. For example, if the peak-valley misalignment between a photovoltaic power station and a wind farm is 83.3%, then the edge weight between them is 0.833.

[0087] A community detection algorithm is used to analyze the constructed capacity complementarity network, dividing the network into multiple connected subgraphs. The Louvain algorithm is employed for network segmentation, iteratively assigning nodes to different communities by optimizing the modularity index. For each resulting connected subgraph, the harmonic mean of the weights of all nodes within it is calculated as the cooperative reliability of that subgraph. The harmonic mean is calculated by dividing the total number of nodes by the sum of the reciprocals of the weights of each node. For example, if a connected subgraph contains 5 nodes with weights of 1.176, 1.25, 1.111, 1.333, and 1.429, its harmonic mean is 1.243, meaning its cooperative reliability is 1.243.

[0088] A preset reliability threshold of 1.2 is set, and the cooperative reliability of each connected subgraph is checked to see if it meets the threshold requirement. For connected subgraphs with a cooperative reliability greater than or equal to 1.2, all distributed energy nodes contained therein are merged into the same resource aggregation unit. For example, if the connected subgraph with a cooperative reliability of 1.243 meets the reliability threshold requirement, then the 5 nodes in this subgraph (including photovoltaic power plants, wind farms, and energy storage power plants, etc.) are merged into resource aggregation unit A. In this way, multiple distributed energy nodes within the virtual power plant are divided into several resource aggregation units.

[0089] Taking a real virtual power plant as an example, it contains 20 distributed energy nodes (8 photovoltaic power plants, 6 wind farms, 4 energy storage power plants, and 2 controllable loads). After analysis using the above method, three resource aggregation units are obtained. Aggregation unit 1 contains 3 photovoltaic power plants, 2 wind farms, and 1 energy storage power plant, with a collaborative reliability of 1.35; aggregation unit 2 contains 2 photovoltaic power plants, 3 wind farms, 2 energy storage power plants, and 1 controllable load, with a collaborative reliability of 1.28; aggregation unit 3 contains 3 photovoltaic power plants, 1 wind farm, 1 energy storage power plant, and 1 controllable load, with a collaborative reliability of 1.22.

[0090] In one alternative implementation, segmenting the capacity complementary relationship network to obtain multiple connected subgraphs includes:

[0091] Select the node with the largest node weight from the capacity complementary relationship network to determine the initial seed node, and include all nodes that are directly connected to the initial seed node through network edges into the first candidate node set;

[0092] Calculate the product of the edge weight between each node in the first candidate node set and the initial seed node and the node weight of each node, determine the aggregation gain value of the corresponding node, and absorb the nodes in the first candidate node set into the first connected subgraph where the initial seed node is located in descending order of aggregation gain value.

[0093] After absorbing a node, the unabsorbed nodes that are directly connected to the absorbed node through network edges are added as supplementary nodes to the first candidate node set, and the aggregation gain value of the supplementary nodes is recalculated. When the largest aggregation gain value in the first candidate node set is lower than the preset gain threshold, the absorption process stops and the construction of the first connected subgraph is completed.

[0094] From the remaining unassigned nodes in the capacity complementarity network, select the node with the largest node weight to obtain a new seed node, and repeat the above construction process until all nodes in the capacity complementarity network are assigned to the corresponding connected subgraph.

[0095] In one specific implementation, the dynamic scheduling method for adjustable resources in a virtual power plant requires the rational segmentation of the constructed capacity complementarity network to form efficient resource aggregation units. The segmentation process employs a clustering method based on seed node expansion, iteratively absorbing nodes with high aggregation gains to form multiple connected subgraphs.

[0096] The capacity complementarity network is traversed to find the node with the largest node weight, which is then selected as the initial seed node. In a practical application scenario, assume the network contains 20 distributed energy nodes, with node weights as follows: photovoltaic power station A has a weight of 1.37, photovoltaic power station B has a weight of 1.25, wind farm C has a weight of 1.43, wind farm D has a weight of 1.32, and energy storage station E has a weight of 1.52, which is the largest, and therefore it is determined as the initial seed node.

[0097] Identify all nodes directly connected to the initial seed node, energy storage station E, via network edges, and include these nodes in the first candidate node set. Assume that the nodes directly connected to energy storage station E are photovoltaic power station A, photovoltaic power station B, wind farm C, and controllable load F, with node weights of 1.37, 1.25, 1.43, and 1.18, respectively, and edge weights (i.e., peak-valley misalignment) of 0.85, 0.78, 0.92, and 0.73, respectively.

[0098] Calculate the aggregate gain value of each node in the first candidate node set. The aggregate gain value is calculated by multiplying the edge weight between the node and the initial seed node by the node's own weight. For the four candidate nodes mentioned above, the calculated aggregate gain values ​​are: 0.85 × 1.37 = 1.1645 for photovoltaic power station A, 0.78 × 1.25 = 0.975 for photovoltaic power station B, 0.92 × 1.43 = 1.3156 for wind farm C, and 0.73 × 1.18 = 0.8614 for controllable load F.

[0099] Based on the aggregation gain values ​​sorted from largest to smallest, the order is: Wind Farm C (1.3156), Photovoltaic Power Plant A (1.1645), Photovoltaic Power Plant B (0.975), and Controllable Load F (0.8614). Following this order, candidate nodes are sequentially absorbed into the first connected subgraph containing the initial seed node.

[0100] After absorbing the first candidate node, wind farm C, check the nodes directly connected to wind farm C that have not yet been absorbed, and add these nodes as supplementary nodes to the first candidate node set. Assume wind farm C is also connected to wind farm G (node ​​weight 1.28) and energy storage station H (node ​​weight 1.45), with edge weights of 0.81 and 0.88 respectively. Calculate the aggregation gain values ​​of these two supplementary nodes: the aggregation gain value of wind farm G is 0.81 × 1.28 = 1.0368, and the aggregation gain value of energy storage station H is 0.88 × 1.45 = 1.276.

[0101] The aggregation gain values ​​of the first candidate node set are updated and sorted. The new order is: Energy Storage Station H (1.276), Photovoltaic Power Station A (1.1645), Wind Farm G (1.0368), Photovoltaic Power Station B (0.975), and Controllable Load F (0.8614). The node with the highest ranking, Energy Storage Station H, is then absorbed.

[0102] After absorbing energy storage station H, check the nodes connected to energy storage station H that have not yet been absorbed. Assume there is photovoltaic power station I (node ​​weight 1.32) and the edge weight is 0.76. Calculate its aggregate gain value as 0.76 × 1.32 = 1.0032, and update the ranking of the candidate node set.

[0103] Repeat the absorption and update process described above until the largest aggregate gain value in the candidate node set is lower than the preset gain threshold. Assuming the preset gain threshold is 0.95, after several rounds of absorption, the largest aggregate gain value in the candidate node set is 0.93 (lower than the threshold of 0.95). At this point, the absorption process stops, and the construction of the first connected subgraph is complete. The first connected subgraph includes the initial seed node, energy storage station E, and the absorbed nodes: wind farm C, energy storage station H, photovoltaic power station A, wind farm G, and photovoltaic power station I.

[0104] From the remaining unassigned nodes in the capacity complementarity network, the node with the largest node weight is selected again as the new seed node. Assuming that the node with the largest node weight among the remaining nodes is wind farm J (node ​​weight 1.48), it is selected as the second seed node, and the above construction process is repeated to form the second connected subgraph.

[0105] Continue this process until all nodes in the capacity complementarity network have been assigned to their corresponding connected subgraphs. Ultimately, the entire network is divided into multiple connected subgraphs, each representing a resource aggregation unit.

[0106] In practical applications, a virtual power plant in a certain region contains 25 distributed energy nodes, which are divided into four connected subgraphs using the method described above. The first connected subgraph contains 7 nodes (2 photovoltaic power plants, 3 wind farms, and 2 energy storage stations), with an average aggregation gain of 1.27; the second connected subgraph contains 8 nodes (3 photovoltaic power plants, 2 wind farms, 1 energy storage station, and 2 controllable loads), with an average aggregation gain of 1.18; the third connected subgraph contains 6 nodes (2 photovoltaic power plants, 1 wind farm, 2 energy storage stations, and 1 controllable load), with an average aggregation gain of 1.13; and the fourth connected subgraph contains 4 nodes (1 photovoltaic power plant, 1 wind farm, 1 energy storage station, and 1 controllable load), with an average aggregation gain of 1.05.

[0107] The method in this embodiment is applicable to the collaborative aggregation of various types of distributed energy resources, including distributed photovoltaics, distributed wind power, distributed energy storage, and controllable loads. It can dynamically form the optimal aggregation unit according to the actual operating characteristics, providing technical support for the efficient operation of virtual power plants.

[0108] In one optional implementation, extracting historical response deviations, establishing a mapping relationship with the target output value, pre-compensating for the target output value, and adjusting the reserve capacity according to the magnitude of the response deviation include:

[0109] Extract the actual output power and target output power of each resource aggregation unit from historical response data, calculate the difference between the actual output power and the target output power to obtain the actual response deviation, perform sliding window processing on the actual response deviation, calculate the mean value within each window to obtain the trend deviation, and calculate the residual between the actual response deviation and the trend deviation to obtain the fluctuation deviation.

[0110] Using the target output value as the independent variable and the trend deviation as the dependent variable, a polynomial fitting is performed to determine the fitting curve. The target output value is divided into multiple intervals, and the extreme range of the fluctuation deviation in each interval is statistically analyzed to form a range table. The fitting curve and the range table are used together as a mapping relationship.

[0111] Substitute the current target output value into the fitted curve to obtain the predicted trend deviation. Add the predicted trend deviation to the current target output value to obtain the pre-compensation target output value. Query the extreme value range of the interval corresponding to the current target output value from the range table. Calculate the span of the extreme value range to determine the response deviation amplitude. Reserve reserve capacity from the regulating capacity according to the ratio of the response deviation amplitude to the pre-compensation target output value.

[0112] In one specific implementation, in the dynamic scheduling method for adjustable resources in a virtual power plant, accurate modeling of the response characteristics of resource aggregation units is crucial for achieving efficient scheduling. This requires extracting the actual output power and target output value of each resource aggregation unit from historical response data, establishing a response deviation model, and performing pre-compensation scheduling accordingly.

[0113] Scheduling data for each resource aggregation unit over the past 30 days was collected, including target output and actual output power every 5 minutes. Taking a wind-solar-storage hybrid aggregation unit as an example, its target output sequence and actual output power sequence were extracted, with 288 sampling points per day, totaling 8640 data points. The difference between the actual output power and the target output value was calculated to obtain the actual response deviation sequence. For example, if the target output value was 5000 kW and the actual output power was 4850 kW at a certain moment, the actual response deviation would be -150 kW.

[0114] A sliding window process is applied to the actual response deviation sequence, with a window length of 12 points (i.e., 1 hour) and a sliding step size of 1 point (i.e., 5 minutes). The average actual response deviation is calculated within each window to obtain the trend deviation. For example, if the average actual response deviation over 30 minutes before and after time t (a total of 12 sampling points) is -120 kW, then the trend deviation at time t is -120 kW. The difference between the actual response deviation and the trend deviation is calculated to obtain the fluctuation deviation. If, at the aforementioned time, the actual response deviation is -150 kW and the trend deviation is -120 kW, then the fluctuation deviation is -30 kW.

[0115] A polynomial fit is performed with the target output value as the independent variable and the trend deviation as the dependent variable to determine the mapping relationship between the two. Different orders of polynomials are tried during the fit, and the most suitable order is selected by comparing goodness-of-fit indices. In this example, a third-order polynomial is used for fitting, with fitting coefficients of -0.0000025, 0.0218, -42.5, and 15.2. These coefficients can express the non-linear relationship between the trend deviation and the target output value. For example, when the target output value is 4000 kW, the trend deviation calculated by the polynomial is -105 kW; when the target output value is 6000 kW, the trend deviation is -155 kW.

[0116] The target output value is divided into multiple intervals, with interval boundaries determined based on the capacity characteristics of the aggregation unit. For an aggregation unit with a rated capacity of 10,000 kW, the target output value can be divided into 10 equally wide intervals: 0-1000 kW, 1000-2000 kW, 2000-3000 kW, etc., up to 9000-10000 kW. The maximum and minimum values ​​of the fluctuation deviation within each interval are statistically analyzed to construct a fluctuation deviation extreme value range table. For example, in the 4000-5000 kW interval, the maximum fluctuation deviation is 85 kW, and the minimum is -95 kW; in the 5000-6000 kW interval, the maximum fluctuation deviation is 110 kW, and the minimum is -120 kW. The fitted polynomial curve and the fluctuation deviation extreme value range table are used together as the response deviation mapping relationship for this resource aggregation unit.

[0117] In actual dispatching, the current target output value is substituted into a polynomial fitting curve to calculate the predicted trend deviation. Assuming the current target output value is 5500 kW, substituting it into a third-order polynomial yields a predicted trend deviation of -135 kW. This predicted trend deviation is then added to the current target output value to obtain the pre-compensated target output value. In this example, the pre-compensated target output value is 5500 + (-135) = 5365 kW. The extreme value range (5000-6000 kW) corresponding to the current target output value is found in the extreme value range table. The maximum fluctuation deviation in this range is 110 kW, and the minimum is -120 kW. The span of the extreme value range is calculated as the absolute value of the maximum minus the minimum, yielding the response deviation amplitude. In this example, the response deviation amplitude is |110 - (-120)| = 230 kW.

[0118] The ratio of the response deviation to the pre-compensation target output value is calculated to obtain the deviation ratio. In this example, the deviation ratio is 230 / 5365 = 4.29%. Reserve capacity is then reserved from the regulating capacity of the aggregation unit based on this ratio. If the current adjustable capacity of the aggregation unit is 800 kW and the adjustable capacity is 600 kW, then the reserved reserve capacity is: adjustable reserve capacity = 800 × 4.29% = 34.32 kW (rounded to 35 kW), adjustable reserve capacity = 600 × 4.29% = 25.74 kW (rounded to 26 kW). During dispatch, the actual available adjustable capacity is 800 - 35 = 765 kW and the adjustable capacity is 600 - 26 = 574 kW.

[0119] Experimental results show that the pre-compensation model established using historical response data can effectively improve the dispatch accuracy and reliability of virtual power plants. Especially when renewable energy output fluctuates significantly, reserving appropriate backup capacity can significantly reduce dispatch risks and enhance system stability. This method is applicable not only to normal operating conditions but also to emergency dispatch needs under extreme weather and unforeseen events, providing strong support for the precise dispatch and efficient operation of virtual power plants.

[0120] In one optional implementation, the process of acquiring the power sequence update response attenuation coefficient of the resource aggregation unit, identifying capacity complementarity relationships, calculating cooperative reliability, and performing dynamic reorganization based on cooperative reliability includes:

[0121] Collect the actual output power sequence and scheduling instruction power sequence of each resource aggregation unit in the current scheduling cycle, calculate the corresponding root mean square error, determine the actual execution deviation, and calculate the ratio of the actual execution deviation to the reciprocal of the historical response attenuation coefficient to obtain the attenuation update factor. Multiply the historical response attenuation coefficient by the attenuation update factor to obtain the updated response attenuation coefficient.

[0122] Extract the output fluctuation sequence of each resource aggregation unit, calculate the convolution operation result of the output fluctuation sequences of any two resource aggregation units in the time domain, and determine that the resource aggregation units have a capacity complementary relationship and form a complementary combination when the peak position of the convolution operation result corresponds to a negative value.

[0123] For each complementary combination, the update response decay coefficient and segmented adjustment capacity of each resource aggregation unit within the combination are extracted. The combined response index is obtained by harmonic averaging the update response decay coefficient and the combined adjustment capacity is obtained by summing the segmented adjustment capacity. The product of the combined response index and the combined adjustment capacity is calculated to determine the collaborative reliability.

[0124] All complementary combinations are sorted in descending order of collaborative reliability. A corresponding number of complementary combinations are selected according to a preset optimal ratio. The resource aggregation units within the selected complementary combinations are merged to complete the dynamic reorganization.

[0125] In one specific implementation, the dynamic scheduling method for adjustable resources in a virtual power plant requires real-time updates to the response characteristics of each resource aggregation unit and dynamic reorganization based on complementary characteristics to improve the overall scheduling effect. The actual output power sequence and scheduling command power sequence of each resource aggregation unit within the current scheduling cycle are collected, with a sampling interval of 5 minutes and a sampling duration of 4 hours, resulting in 48 sampling points. Taking a wind-solar-storage hybrid aggregation unit as an example, the average actual output power within the current scheduling cycle is 4200 kW with a standard deviation of 320 kW; the average scheduling command power is 4300 kW with a standard deviation of 350 kW. The root mean square error between the actual output power sequence and the scheduling command power sequence is calculated to obtain the actual execution deviation. Specifically, the square of the difference between corresponding points in the two sequences is first calculated, then the average is calculated, and finally the square root is taken. In this example, the calculated actual execution deviation is 180 kW.

[0126] The attenuation update factor is obtained by comparing the actual performance deviation with the reciprocal of the historical response attenuation coefficient. Assuming the historical response attenuation coefficient of this aggregation unit is 0.85, its reciprocal is 1.1765. The calculated attenuation update factor is 180 / 1.1765 = 153.0. Dividing this value by the actual performance deviation of the previous cycle (assumed to be 165 kW) yields a ratio of 0.927. Multiplying the historical response attenuation coefficient by the attenuation update factor, i.e., 0.85 × 0.927 = 0.788, gives the updated response attenuation coefficient. This method allows for dynamic adjustment of the response attenuation coefficient based on the actual response of the aggregation unit, making the evaluation indicators more closely reflect actual operating characteristics.

[0127] Output fluctuation sequences of each resource aggregation unit are extracted and obtained by calculating the first difference of the actual output power. Output fluctuation data are recorded continuously for 4 hours at 5-minute intervals, resulting in a sequence of length 47. The convolution operation of the output fluctuation sequences of any two resource aggregation units in the time domain is calculated. The convolution operation is achieved by sliding one sequence, multiplying it correspondingly with the other sequence, and then summing the results. The result reflects the similarity between the two sequences at different time offsets. Assuming that the output fluctuation sequences of aggregation units A and B are both 47 in length, the length of the convolution result is 93. When the value corresponding to the peak position in the convolution result is negative, it indicates that there is an inverse correlation between the output fluctuations of the two aggregation units. That is, when the output of one aggregation unit increases, the output of the other aggregation unit tends to decrease. This indicates that the two aggregation units have a capacity complementarity relationship and can form a complementary combination.

[0128] In practical applications, a virtual power plant comprises 10 resource aggregation units. By calculating the convolution results of the power fluctuation sequences pairwise, 12 complementary combinations with complementary capacity relationships are identified: aggregation unit A and aggregation unit C, aggregation unit A and aggregation unit H, aggregation unit B and aggregation unit E, etc. For each complementary combination, the update response attenuation coefficient and segmented regulation capacity of each resource aggregation unit within the combination are extracted. For example, in the complementary combination of aggregation unit A and aggregation unit C, the update response attenuation coefficient of aggregation unit A is 0.788, and the segmented regulation capacity is 1500 kW increase and 1200 kW decrease; the update response attenuation coefficient of aggregation unit C is 0.823, and the segmented regulation capacity is 1800 kW increase and 1400 kW decrease.

[0129] The harmonic average of the update response attenuation coefficients of each resource aggregation unit within the complementary combination is used to obtain the combined response index. The harmonic average is calculated by dividing the number of aggregation units within the combination by the sum of the reciprocals of the response attenuation coefficients of each aggregation unit. In the complementary combination of aggregation unit A and aggregation unit C, the harmonic average is 2 / ((1 / 0.788)+(1 / 0.823))=0.805, i.e., the combined response index is 0.805. The combined regulation capacity is obtained by summing the segmented regulation capacities of each resource aggregation unit within the complementary combination. In the above complementary combination, the combined regulation capacity is 3300 kW for upward regulation and 2600 kW for downward regulation. The product of the combined response index and the combined regulation capacity is calculated to determine the coordinated reliability. In specific calculations, the average of the upward and downward regulation capacities can be multiplied by the combined response index. In this embodiment, the average combined regulation capacity is (3300+2600) / 2=2950 kW, and the cooperative reliability is 0.805×2950=2375.

[0130] All complementary combinations are arranged in descending order of collaborative reliability. Assume the collaborative reliability of the 12 complementary combinations, from highest to lowest, is: Aggregator A and Aggregator C (2375), Aggregator B and Aggregator E (2210), Aggregator D and Aggregator G (2180), etc. Setting the preset selection ratio at 30%, the number of complementary combinations to be selected is calculated to be 12 × 30% = 3.6, rounded down to 3. The top 3 complementary combinations with the highest collaborative reliability are selected: Aggregator A and Aggregator C, Aggregator B and Aggregator E, and Aggregator D and Aggregator G. The resource aggregation units within these complementary combinations are then merged to complete dynamic reorganization.

[0131] The newly formed aggregation units through dynamic reorganization exhibit higher synergy and reliability. Experimental data shows that the reorganized aggregation units demonstrate improved response accuracy and shorter response time when addressing grid frequency regulation demands. When handling load surge events, the reorganized aggregation units can complete power adjustment within 2.5 minutes, while unreorganized aggregation units require 3.8 minutes to achieve the same level. Long-term operational data indicates that the dynamic reorganization mechanism effectively adapts to changes in the output characteristics of distributed energy resources, improves the overall regulation flexibility and stability of the virtual power plant, reduces dispatch deviations, decreases reserve capacity requirements, and enhances economic efficiency. The method in this embodiment is particularly suitable for virtual power plant systems with a high proportion of renewable energy, effectively addressing output fluctuations caused by weather changes and providing reliable support for the stable operation of the power system.

[0132] The virtual power plant adjustable resource dynamic scheduling system of this invention includes:

[0133] The data acquisition module is used to collect the operating status data, historical response data and power demand commands of each distributed energy node;

[0134] The capacity assessment module is used to determine the output constraint boundary of each distributed energy node based on the operating status data, calculate the segmented adjustment capacity under different scheduling durations, and calculate the response attenuation coefficient based on historical response data.

[0135] The resource aggregation module is used to identify the capacity complementarity between distributed energy nodes based on the distribution differences of segmented adjustment capacity in the time dimension, calculate the collaborative reliability in combination with the response attenuation coefficient, and group distributed energy nodes with capacity complementarity and collaborative reliability that meet the preset reliability threshold into resource aggregation units.

[0136] The power allocation module is used to calculate the target power output value based on the distributed energy nodes and corresponding segmented regulation capacity contained in the resource aggregation unit, combined with the power demand command, and to perform boundary verification on the target power output value according to the power output constraint boundary, and decompose it into allocated power output value according to the segmented regulation capacity ratio.

[0137] The deviation compensation module is used to extract historical response deviations, establish a mapping relationship with the target output value, pre-compensate the target output value, and adjust the reserve capacity according to the magnitude of the response deviation.

[0138] The dynamic reorganization module is used to collect the power sequence update response attenuation coefficient of the resource aggregation unit, identify the capacity complementarity relationship, calculate the cooperative reliability, and complete the dynamic reorganization according to the cooperative reliability.

[0139] A third aspect of the present invention provides an electronic device, comprising:

[0140] processor;

[0141] Memory used to store processor-executable instructions;

[0142] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0143] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0144] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0145] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for dynamic scheduling of adjustable resources in a virtual power plant, characterized in that, include: Collect operational status data, historical response data, and power demand commands from each distributed energy node; Based on the operational status data, the output constraint boundary of each distributed energy node is determined, the segmented adjustment capacity under different scheduling durations is calculated, and the response attenuation coefficient is statistically calculated based on historical response data. Based on the differences in the distribution of segmented adjustment capacity over time, the capacity complementarity relationship between distributed energy nodes is identified. The collaborative reliability is calculated by combining the response decay coefficient. Distributed energy nodes with capacity complementarity and collaborative reliability that meet the preset reliability threshold are grouped into resource aggregation units. Based on the distributed energy nodes and corresponding segmented regulation capacity contained in the resource aggregation unit, the target output value is calculated in conjunction with the power demand command, and the target output value is checked against the output constraint boundary. The output value is then decomposed into allocated output values ​​according to the segmented regulation capacity ratio. Extract historical response deviations, establish a mapping relationship with the target output value, pre-compensate for the target output value, and adjust the reserve capacity according to the magnitude of the response deviation. The power sequence of the resource aggregation unit is collected to update the response attenuation coefficient, the capacity complementarity relationship is identified to calculate the cooperative reliability, and dynamic reorganization is completed according to the cooperative reliability.

2. The method of claim 1, wherein, Based on the operational status data, the output constraint boundaries of each distributed energy node are determined, the segmented adjustment capacity under different scheduling durations is calculated, and the response attenuation coefficient is statistically analyzed based on historical response data, including: The operating parameter time series of each distributed energy node is extracted from the operating status data. The operating parameter time series is reconstructed into a high-dimensional phase space. The attractor boundary of the operating status trajectory is identified in the high-dimensional phase space. The power value corresponding to the projection of the current operating status on the attractor boundary is determined as the output constraint boundary. The response speed of each distributed energy node under different initial power and different adjustment range in historical scheduling is extracted. The response surface of the response speed with respect to the initial power and adjustment range is constructed. The adjustment path from the current operating state to the output constraint boundary is calculated. The response time is obtained by integrating the response surface along the adjustment path. When the response time is less than the scheduling duration, the difference between the output constraint boundary and the current power is used as the segmented adjustment capacity. When the response time exceeds the scheduling duration, the maximum power under the scheduling duration constraint is solved iteratively. The difference between the maximum power and the current power is used as the segmented adjustment capacity. The deviation sequence between scheduling instructions and actual execution power is extracted from historical response data. Autoregressive spectral analysis is performed on the deviation sequence to obtain the spectral distribution. The frequency component with the highest energy is extracted from the spectral distribution, and the energy attenuation index of the frequency component with the highest energy is calculated as the response attenuation coefficient.

3. The method of claim 1, wherein, Based on the differences in the distribution of segmented regulation capacity over time, the capacity complementarity relationship between distributed energy nodes is identified. Combined with the response decay coefficient, the cooperative reliability is calculated. Distributed energy nodes with capacity complementarity and cooperative reliability meeting a preset reliability threshold are grouped into resource aggregation units, including: The segmented adjustment capacity of each distributed energy node under different scheduling durations is arranged according to the duration dimension to form a capacity distribution sequence. Any two distributed energy nodes are selected to determine the first node and the second node, resulting in a node pair. Cross-correlation analysis is performed on the capacity distribution sequence of the node pair. The correlation coefficient under different time offsets is calculated through a sliding time window. The time offset that makes the correlation coefficient reach a negative minimum value is recorded to determine the phase offset. The frequency of the occurrence of the peak time of the first node corresponding to the valley time of the second node is counted to determine the peak-valley misalignment degree. When the phase offset reaches half a cycle and the peak-valley misalignment degree exceeds the preset misalignment threshold, it is determined that the first node and the second node have a capacity complementary relationship. A capacity complementarity network is constructed using each distributed energy node as a network node. Network edges are established between node pairs with capacity complementarity. The reciprocal of the response attenuation coefficient is used as the node weight, and the peak-valley misalignment degree is used as the edge weight. The capacity complementarity network is divided into multiple connected subgraphs. The harmonic average of the node weights in each connected subgraph is calculated to determine the collaborative reliability. All distributed energy nodes in a connected subgraph whose collaborative reliability meets a preset reliability threshold are grouped into the same resource aggregation unit.

4. The method of claim 3, wherein, Segmenting the capacity-complementary relationship network yields multiple connected subgraphs, including: Select the node with the largest node weight from the capacity complementary relationship network to determine the initial seed node, and include all nodes that are directly connected to the initial seed node through network edges into the first candidate node set; Calculate the product of the edge weight between each node in the first candidate node set and the initial seed node and the node weight of each node, determine the aggregation gain value of the corresponding node, and absorb the nodes in the first candidate node set into the first connected subgraph where the initial seed node is located in descending order of aggregation gain value. After absorbing a node, the unabsorbed nodes that are directly connected to the absorbed node through network edges are added as supplementary nodes to the first candidate node set, and the aggregation gain value of the supplementary nodes is recalculated. When the largest aggregation gain value in the first candidate node set is lower than the preset gain threshold, the absorption process stops and the construction of the first connected subgraph is completed. From the remaining unassigned nodes in the capacity complementarity network, select the node with the largest node weight to obtain a new seed node, and repeat the above construction process until all nodes in the capacity complementarity network are assigned to the corresponding connected subgraph.

5. The method of claim 1, wherein, Extracting historical response deviations, establishing a mapping relationship with target output values, pre-compensating for target output values, and adjusting reserve capacity based on the magnitude of response deviations include: Extract the actual output power and target output power of each resource aggregation unit from historical response data, calculate the difference between the actual output power and the target output power to obtain the actual response deviation, perform sliding window processing on the actual response deviation, calculate the mean value within each window to obtain the trend deviation, and calculate the residual between the actual response deviation and the trend deviation to obtain the fluctuation deviation. Using the target output value as the independent variable and the trend deviation as the dependent variable, a polynomial fitting is performed to determine the fitting curve. The target output value is divided into multiple intervals, and the extreme range of the fluctuation deviation in each interval is statistically analyzed to form a range table. The fitting curve and the range table are used together as a mapping relationship. Substitute the current target output value into the fitted curve to obtain the predicted trend deviation. Add the predicted trend deviation to the current target output value to obtain the pre-compensation target output value. Query the extreme value range of the interval corresponding to the current target output value from the range table. Calculate the span of the extreme value range to determine the response deviation amplitude. Reserve reserve capacity from the regulating capacity according to the ratio of the response deviation amplitude to the pre-compensation target output value.

6. The method of claim 1, wherein, The process of updating the response attenuation coefficient of the power sequence of the resource aggregation unit, identifying capacity complementarity relationships, calculating cooperative reliability, and performing dynamic reorganization based on cooperative reliability includes: Collect the actual output power sequence and scheduling instruction power sequence of each resource aggregation unit in the current scheduling cycle, calculate the corresponding root mean square error, determine the actual execution deviation, and calculate the ratio of the actual execution deviation to the reciprocal of the historical response attenuation coefficient to obtain the attenuation update factor. Multiply the historical response attenuation coefficient by the attenuation update factor to obtain the updated response attenuation coefficient. Extract the output fluctuation sequence of each resource aggregation unit, calculate the convolution operation result of the output fluctuation sequences of any two resource aggregation units in the time domain, and determine that the resource aggregation units have a capacity complementary relationship and form a complementary combination when the peak position of the convolution operation result corresponds to a negative value. For each complementary combination, the update response decay coefficient and segmented adjustment capacity of each resource aggregation unit within the combination are extracted. The combined response index is obtained by harmonic averaging the update response decay coefficient and the combined adjustment capacity is obtained by summing the segmented adjustment capacity. The product of the combined response index and the combined adjustment capacity is calculated to determine the collaborative reliability. All complementary combinations are sorted in descending order of collaborative reliability. A corresponding number of complementary combinations are selected according to a preset optimal ratio. The resource aggregation units within the selected complementary combinations are merged to complete the dynamic reorganization.

7. A virtual power plant adjustable resource dynamic scheduling system for implementing the method of any one of the preceding claims 1-6, characterized in that, include: The data acquisition module is used to collect the operating status data, historical response data and power demand commands of each distributed energy node; The capacity assessment module is used to determine the output constraint boundary of each distributed energy node based on the operating status data, calculate the segmented adjustment capacity under different scheduling durations, and calculate the response attenuation coefficient based on historical response data. The resource aggregation module is used to identify the capacity complementarity between distributed energy nodes based on the distribution differences of segmented adjustment capacity in the time dimension, calculate the collaborative reliability in combination with the response attenuation coefficient, and group distributed energy nodes with capacity complementarity and collaborative reliability that meet the preset reliability threshold into resource aggregation units. The power allocation module is used to calculate the target power output value based on the distributed energy nodes and corresponding segmented regulation capacity contained in the resource aggregation unit, combined with the power demand command, and to perform boundary verification on the target power output value according to the power output constraint boundary, and decompose it into allocated power output value according to the segmented regulation capacity ratio. The deviation compensation module is used to extract historical response deviations, establish a mapping relationship with the target output value, pre-compensate the target output value, and adjust the reserve capacity according to the magnitude of the response deviation. The dynamic reorganization module is used to collect the power sequence update response attenuation coefficient of the resource aggregation unit, identify the capacity complementarity relationship, calculate the cooperative reliability, and complete the dynamic reorganization according to the cooperative reliability.

8. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 6.

Images