Virtual power plant resource dynamic coordination configuration method and system
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- TAIYUAN UNIVERSITY OF TECHNOLOGY
- Filing Date
- 2026-03-17
- Publication Date
- 2026-05-29
Smart Images
Figure CN121863425B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of virtual power plant resource allocation technology, and in particular to a method and system for dynamic coordination and allocation of virtual power plant resources. Background Technology
[0002] Virtual power plants, as a key technology capable of aggregating and coordinating diverse resources such as distributed energy, energy storage systems, and flexible loads, are becoming a core means of improving the flexibility, reliability, and economy of power systems. Through advanced information and communication technologies and control strategies, they integrate massive, dispersed, and diverse resources into a controllable whole, participating in grid dispatch and market transactions, effectively addressing the dual challenges of renewable energy volatility and uncertain electricity demand.
[0003] However, with the significant increase in the types of access resources and the growing complexity of the power grid operating environment, a fundamental technical bottleneck has become increasingly prominent: how to deeply coordinate the response of resources with vastly different characteristics to cope with the rapidly changing dynamic demands of the power grid. Existing mainstream solutions mostly focus on the optimized scheduling of single-type resources (such as energy storage or interruptible loads) or use relatively simplified aggregation models to control multiple resources as a homogeneous whole. While these methods are effective in specific scenarios, they ignore the essential differences in core operational characteristics such as physical response speed, adjustment duration, and geographic spatial distribution among different types of resources. For example, electric vehicle charging stations can respond to power adjustment commands in seconds, while temperature-controlled loads require several minutes or even longer to complete state transitions and generate continuous power regulation effects. This inherent difference in response speed can lead to a situation where some fast-response resources are overloaded while slower-response resources are unable to function effectively when dealing with sudden frequency fluctuations or rapid power shortages in the power grid, resulting in a coordinated failure that severely restricts the full realization of the overall flexible adjustment capabilities of the virtual power plant.
[0004] More importantly, existing technological solutions often treat the external power grid environment (such as load forecasting errors, renewable energy output fluctuations, and network congestion signals) as static or isolated inputs, lacking a collaborative decision-making framework capable of sensing the intensity of external demand in real time and accurately matching it with the dynamic characteristics of internal resources. As a result, when facing peak loads or complex events, the scheduling schemes of virtual power plants are often lagging and rigid, making it difficult to achieve dynamic optimal configuration across resources and time and space. This not only fails to maximize resource utilization efficiency but also harbors risks that may affect the stable operation of the local power grid. Summary of the Invention
[0005] Therefore, the technical problem to be solved by the present invention is to overcome the problem that the scheduling scheme of virtual power plants in the prior art is often lagging and rigid, and it is difficult to achieve dynamic optimal configuration across resources and time and space. The invention provides a method and system for dynamic coordination and configuration of virtual power plant resources, and constructs a complete technical path from accurate perception and difference assessment to dynamic optimization. Through the synergistic effect of multi-level technical features, the invention fundamentally solves the problem of matching resource characteristic differences with dynamic needs.
[0006] To address the aforementioned technical problems, this invention provides a method for dynamic coordination and allocation of virtual power plant resources, comprising the following steps:
[0007] Real-time acquisition of operational data of various resources within the virtual power plant and external power grid environment data; based on the operational data, extraction of operational characteristic parameters reflecting the differences in response speed of various resources.
[0008] Based on operational characteristic parameters, cluster analysis is performed on various resources to form resource clusters with different response speed levels. Based on external power grid environment data, the intensity of the power grid's coordinated dispatch demand in the current period is assessed.
[0009] Based on historical data and external power grid environment data, predict the power grid load balance status for the target period in the future; combine the response speed level and coordination demand intensity of resource clusters to determine the available response capacity of each resource cluster for the target period.
[0010] Based on the callable response capability, resource area distribution information, and power grid load balance status, multi-objective optimization calculations are performed to meet power grid stability constraints, generating dynamic coordination configuration schemes for various resources within the target time period. This is used to improve the overall resource utilization efficiency of the virtual power plant and ensure power grid load balance.
[0011] In one embodiment of the present invention, the operating data includes: a time series of power regulation command values received by various resources, and a corresponding time series of actual output power values; and command response delay time and power stabilization adjustment time calculated based on the power regulation command value time series and the actual output power value time series, and a power change capability value per unit time obtained by analyzing the actual output power value time series as operating characteristic parameters.
[0012] In one embodiment of the present invention, cluster analysis is performed on multiple resources based on operational characteristic parameters to form resource clusters with different response speed levels, including:
[0013] The runtime characteristic parameter vectors of all resources are mapped to a multi-dimensional feature space. Dimensional correlation analysis is performed to identify the core dimension subset that is directly and strongly correlated with response speed, and the projected coordinates of each resource on the core speed dimension are generated.
[0014] In the feature subspace formed by the core dimension subset, calculate the local density of each resource projection coordinate point and its neighboring points, and simultaneously calculate the minimum distance from each point to a point with a higher local density; select several points with the highest product of local density and minimum distance as initial clustering anchor points;
[0015] With each initial cluster anchor point as the center, region growth is performed along the local density gradient descent direction in the feature subspace until a boundary formed by the growth of other anchor points is encountered or a preset feature difference threshold is reached. Each continuous region formed in this way is defined as a candidate range of a resource cluster. All resources in the feature subspace are initially assigned to clusters based on the candidate range to which their projected coordinate points belong.
[0016] For each resource cluster initially divided, calculate the numerical dispersion of all resources within it on the preset key dimension of response speed; if the dispersion of a resource cluster exceeds the preset threshold, then perform a secondary split operation based on feature similarity on the cluster.
[0017] Based on the overall average level of each resource cluster in the key speed dimension, they are sorted from high to low and labeled as the first response speed level, the second response speed level, etc., thus forming resource clusters with clear speed level labels.
[0018] In one embodiment of the present invention, the intensity of the coordinated dispatch demand of the power grid in the current period is assessed based on external power grid environmental data, specifically including:
[0019] Based on real-time frequency deviation, power flow at key sections, and load forecasting error data from external power grid environmental data, the instantaneous state anomaly index of power grid operation and the power regulation demand in the near future are calculated.
[0020] Based on the response speed level of the resource clusters, obtain the total adjustable power capacity and average adjustment rate of each cluster in the current time period.
[0021] Based on the power regulation requirements for regulation speed and duration, and the regulation rate of each resource cluster, the priority order for calling resource clusters that meet the requirements is determined.
[0022] Determine the processing priority for different power regulation requirements based on the degree of abnormality.
[0023] Simulated scheduling is performed according to the scenario processing priority and the priority order of resource cluster invocation. The proportion of demand allocated to each resource cluster is calculated, and the intensity of the collaborative allocation demand for each resource cluster is evaluated based on the demand proportion.
[0024] In one embodiment of the present invention, predicting the power grid load balance state for a future target period based on historical data and external power grid environment data includes:
[0025] Extract typical load change patterns and periodic regularities associated with the target time period from historical data; at the same time, identify known planned event signals and network topology constraint change information that affect the load from external power grid environment data;
[0026] Based on typical load change patterns and periodicity, a baseline load forecast curve for the target period is generated; and based on known planned event signals, the deterministic superposition or reduction impact on the baseline load is assessed to form a load forecast curve that takes the events into account.
[0027] Based on the load forecast curve that takes events into account, a load envelope representing the possible upper limit of the forecast and a load envelope representing the possible lower limit of the forecast are generated.
[0028] The net load demand for the target period and its upper and lower fluctuation envelopes together constitute the power grid load balance state.
[0029] In one embodiment of the present invention, the available response capability of each resource cluster during a target time period is determined by combining the response speed level of the resource cluster with the intensity of collaborative allocation requirements, including:
[0030] Based on the real-time operating status of all resources in each resource cluster, the maximum increase and decrease in output that each resource cluster can physically allow at the start of the target time period are aggregated and calculated as its theoretical adjustment potential.
[0031] Obtain the intensity of coordinated dispatch demand to indicate the degree of differentiated demand of the power grid for resources at different speed levels; among them: for resource clusters with demand intensity indicating high urgency, confirm the availability improvement of their theoretical regulation potential based on the availability confidence and regulation reliability of their internal resources; for resource clusters with demand intensity indicating normal or low demand, consider their response willingness and economic threshold, and perform conservative available capacity calculation.
[0032] Based on the response speed level of each resource cluster, the corrected available adjustment potential is encapsulated into a set of callable capability parameters that change over time within the target period. Specifically, for clusters with high response speed levels, the parameter set highlights their fast-arriving power value, shortest duration, and maximum adjustment frequency; for clusters with low response speed levels, the parameter set highlights the gradual change in adjustable power, longest duration, and shortest warning time required for startup.
[0033] In one embodiment of the present invention, the process of determining the callable response capabilities of each resource cluster during a target time period also includes optimization of capability transfer and complementarity between clusters, including:
[0034] Analyze the complementary relationship between resource clusters of different response speed levels in terms of adjustment timing;
[0035] Based on the complementary relationship, the callable capability parameter sets of different clusters with time sequence are virtually aggregated in the scheduling logic to form a virtual collaborative resource unit with better comprehensive response characteristics.
[0036] The callable capabilities of virtual collaborative resource units and each independent cluster are verified against disturbances, and the final set of callable response capabilities with robustness confirmed is output.
[0037] In one embodiment of the present invention, based on callable response capabilities, resource area distribution information, and power grid load balance status, multi-objective optimization calculations are performed to satisfy power grid stability constraints, generating a dynamic coordination and allocation scheme for various resources within a target time period, including:
[0038] The available response capability parameter set of each resource cluster is converted into a discrete sequence of power adjustable quantities in preset time intervals within the target time period; the resource area distribution information is converted into the coupling relationship and limit of power transmission between areas; and the power grid load balance status is converted into the minimum regulating power target required for each time section within the target time period and its allowable regulating error band.
[0039] The first optimization objective is to minimize the overall deviation between the actual regulating power and the minimum regulating power target at each time segment; the second optimization objective is to maximize the average utilization efficiency of all resource clusters throughout the target period; and the grid stability constraints are integrated, including at least the power balance constraints of each region, the power limit constraints of key transmission sections, and the power change rate and duration constraints of each resource cluster itself.
[0040] Based on the optimization elements, objectives, and constraints, a unified optimization problem is constructed; by interactively adjusting the emphasis on the first and second optimization objectives, multiple rounds of solution are performed to generate a sequence of candidate configuration schemes with different trade-offs between adjustment accuracy and resource efficiency.
[0041] For each of the candidate configuration schemes, its execution process during the target time period is simulated, and its ability to maintain power grid stability under preset typical disturbance scenarios is evaluated. Based on the adjustment accuracy, resource efficiency and stability performance of each scheme, one scheme is selected as the dynamic coordination configuration scheme.
[0042] To address the aforementioned technical problems, this invention also provides a virtual power plant resource dynamic coordination and allocation system for implementing the above method, comprising:
[0043] The data acquisition and feature extraction module is used to collect real-time operational data of various resources inside the virtual power plant and external power grid environment data, and extract operational feature parameters reflecting the differences in response speed of various resources based on the operational data.
[0044] The resource clustering and demand assessment module is used to perform clustering analysis on various resources based on the operational characteristic parameters, form resource clusters with different response speed levels, and assess the intensity of the grid's coordinated dispatch demand in the current period based on the external power grid environment data.
[0045] The load forecasting and capacity matching module is used to predict the power grid load balance status for a future target period based on historical data and the external power grid environment data, and to determine the available response capacity of each resource cluster for the target period by combining the response speed level of the resource cluster and the intensity of the coordinated dispatch demand.
[0046] The multi-objective optimization and scheme generation module is used to perform multi-objective optimization calculations that satisfy power grid stability constraints based on the callable response capability, resource area distribution information and the power grid load balance status, and generate a dynamic coordination configuration scheme for the various resources within the target time period.
[0047] The technical solution of the present invention has the following advantages compared with the prior art:
[0048] The virtual power plant resource dynamic coordination and configuration method described in this invention improves the accuracy and coordination of the virtual power plant in responding to the dynamic demands of the power grid. Because the scheme can identify and utilize the differences in the response speed of resources and match them with the intensity of external demand dynamically assessed, it can direct fast resources to respond to instantaneous fluctuations and arrange slow resources for continuous adjustment, realizing the tiered utilization and seamless coordination of resources, thereby enhancing the adjustment flexibility and response capability of the entire virtual power plant.
[0049] Secondly, this method significantly improves the overall efficiency of resource utilization because its optimized allocation scheme is generated based on a full understanding of the spatiotemporal distribution and capacity characteristics of resources, with the goal of meeting the actual stability needs of the power grid. This minimizes the ineffective use of resources or local congestion, enabling all types of resources to play their maximum value at the right time and in the right place. Attached Figure Description
[0050] To make the content of this invention easier to understand, the invention will be further described in detail below with reference to specific embodiments and accompanying drawings, wherein:
[0051] Figure 1 This is a flowchart of the steps of the virtual power plant resource dynamic coordination and allocation method of the present invention;
[0052] Figure 2 This is a flowchart of the steps in this invention to form resource clusters with different response speed levels;
[0053] Figure 3This is a flowchart of the steps in the present invention to assess the intensity of the coordinated dispatch demand of the power grid in the current period;
[0054] Figure 4 This is a flowchart of the steps in the present invention to predict the power grid load balance state for a future target period;
[0055] Figure 5 This is a flowchart of the steps in the present invention to virtually determine the callable response capabilities of each resource cluster during a target time period;
[0056] Figure 6 It is a flowchart of the steps to generate a dynamic coordination and configuration scheme for multiple resources within a target time period;
[0057] Figure 7 This is a flowchart of the steps of the virtual power plant resource dynamic coordination and configuration system of the present invention. Detailed Implementation
[0058] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, so that those skilled in the art can better understand and implement the present invention. However, the embodiments described are not intended to limit the present invention.
[0059] Reference Figure 1 As shown, the virtual power plant resource dynamic coordination and configuration method of the present invention constructs a complete technical path from accurate perception and difference assessment to dynamic optimization. Its principle lies in fundamentally solving the matching problem between resource characteristic differences and dynamic demands through the synergistic effect of multi-level technical features. The method includes the following steps: real-time collection of operational data of various resources within the virtual power plant and external power grid environment data; extraction of operational characteristic parameters reflecting the differences in response speed of various resources based on the operational data; cluster analysis of various resources based on the operational characteristic parameters to form resource clusters with different response speed levels; assessment of the intensity of coordinated dispatch demand of the power grid in the current period based on external power grid environment data; prediction of the power grid load balance state in the future target period based on historical data and external power grid environment data; determination of the callable response capability of each resource cluster in the target period by combining the response speed level of the resource clusters and the intensity of coordinated dispatch demand; and multi-objective optimization calculation satisfying power grid stability constraints based on the callable response capability, resource regional distribution information, and power grid load balance state to generate a dynamic coordination and configuration scheme for various resources in the target period, used to improve the overall resource utilization efficiency of the virtual power plant and ensure power grid load balance.
[0060] Specifically, the technical solution of this invention first extracts operational characteristic parameters reflecting differences in response speed and performs cluster analysis to form resource clusters with different response speed levels. This method fundamentally identifies and structurally characterizes the heterogeneity of resources within a virtual power plant. This step overcomes the shortcomings of existing technologies that treat multiple resources as homogeneous, laying the foundation for differentiated and precise control. Based on this, by combining the assessment of external power grid environmental data to obtain the intensity of coordinated dispatch demand, the system can dynamically perceive the real-time pressure status of the power grid, thereby establishing a quantifiable correlation between internal resource capacity classification and external dynamic demand.
[0061] Secondly, based on the predicted grid load balance during the target period, the available response capacity is determined by combining the speed level of resource clusters with demand intensity. This step achieves a forward-looking match between supply and demand in the time dimension. The principle behind this is that abstract demand intensity is transformed into a quantifiable value of availability for each resource cluster in a specific future period, avoiding the problem of scheduling instructions being disconnected from the actual dynamic characteristics of resources in traditional methods. This ensures that subsequent optimization decisions are based on reliable capacity pre-assessment, rather than blind global instruction allocation.
[0062] Furthermore, the core step in achieving synergistic effects is to perform multi-objective optimization calculations based on callable response capabilities, resource regional distribution information, and load balance status to meet grid stability constraints. This step integrates the structured information output from the preceding steps—including tiered resource response potential, resource geographical location constraints, and predicted grid load targets—into a unified optimization framework. By solving this optimization problem, the generated dynamic coordinated configuration scheme can simultaneously optimize resource utilization efficiency and load balance targets while meeting the grid's hard constraints such as line capacity and voltage safety. Its technical contribution lies in transforming the traditional coordination challenge of resource response speed differences into a finely adjustable differential variable in the optimization model, thereby achieving a scientific and coordinated arrangement of fast and slow resources in both time and space.
[0063] In summary, this method, through the organic combination of the aforementioned technical features, achieves a complete technical closed loop from resource characteristic recognition to dynamic supply and demand matching and then to global optimization decision-making. The beneficial effects are: significantly improving the resource coordination accuracy and response adaptability of virtual power plants in responding to dynamic grid demands. Because dispatch instructions are generated based on the inherent response characteristics of resources and the matching of real-time demand intensity, they can effectively coordinate fast and slow resources, forming a tiered adjustment capability; simultaneously improving the overall utilization efficiency of resources within the virtual power plant, as the optimization model fully considers the differences in resource capabilities and spatiotemporal distribution, avoiding ineffective resource allocation or local overload, enabling various resources to maximize their value within the overall optimal framework of the system; and finally, enhancing the ability of virtual power plants to support the stable operation of the grid. By providing more accurate, reliable, and efficient flexible resource adjustment schemes, it helps the grid smooth fluctuations and alleviate congestion, providing effective technical support for the safe and stable operation of new power systems.
[0064] Specifically, in this embodiment, the operational data includes: time series of power regulation command values received by various resources, and corresponding time series of actual output power values; command response delay time and power stabilization adjustment time calculated based on the power regulation command value time series and the actual output power value time series, and the power change capability value per unit time obtained from the analysis of the actual output power value time series, as operational characteristic parameters. By collecting command-response data pairs and extracting the above three parameters, a set of standardized quantitative indicators (delay, duration, rate) can accurately describe its core response characteristics, which allows fast resources (such as flywheel energy storage, fast gas turbines) and slow resources (such as temperature-controlled loads, certain coal-fired units) to be clearly distinguished at the data level. Based on the delay time and adjustment time, clustering can be performed according to response speed; combined with the power change capability value, the strength of the regulation capability can be further distinguished within the same speed level.
[0065] When performing cluster analysis on multiple resources, traditional clustering typically treats all feature dimensions equally. However, resource feature vectors often contain dimensions weakly correlated with the core response speed objective, such as geographical coordinates or economic costs. These noisy dimensions severely dilute the clustering's focus on response characteristics, causing the classification results to deviate from the core engineering goal of "speed." More problematic is that the implicit spherical cluster assumption in most algorithms is incompatible with the arbitrary shape distribution of resources in complex multidimensional feature spaces, leading to ambiguity in determining the affiliation of resources in transitional zones. Even after grouping, existing technologies lack quantitative verification mechanisms for intra-cluster consistency, and the process of mapping abstract clustering results to specific response speed levels relies heavily on subjective experience, lacking objective and unified quantitative standards. This results in a weak foundation for the constructed resource hierarchy, failing to provide a reliable basis for subsequent refined scheduling.
[0066] To address the aforementioned systemic deficiencies, refer to Figure 2 As shown, this invention constructs a hierarchical clustering mechanism. First, through intelligent dimensional correlation analysis, it automatically filters out a subset of core dimensions directly and strongly correlated with response speed from all features, and projects the resource feature vectors into this dimensionality-reduced space. This filters out irrelevant noise at the source, allowing subsequent analysis to focus on the essential attributes that determine the dynamic performance of resources. Next, the scheme abandons the traditional random initialization and proposes a robust anchor point discovery method based on density peak detection: it simultaneously calculates the local density of each data point and its minimum distance to higher density points in the feature subspace, selecting the point with the highest product of these two as the initial clustering anchor point. This design automatically locates natural density peaks in the feature space; these points are both typical representatives of potential categories and well-separated from each other, fundamentally ensuring the objectivity and optimality of the clustering starting point and guaranteeing high reproducibility of the results.
[0067] After determining the core anchor points, the scheme employs an adaptive shape cluster boundary growth strategy to define resource clusters. This method uses each anchor point as the center and intelligently expands the region along the direction of local density gradient descent until it meets the growth boundaries of other clusters or reaches a preset feature difference threshold. This growth method can naturally fit any actual distribution pattern of resources in the feature space, forming irregularly shaped clusters and clearly defining inter-class boundaries, thereby accurately handling resources that are in the blind spots of traditional methods. After the initial partitioning is completed, a strict consistency verification step is further introduced: the numerical dispersion of each cluster on the speed-critical dimension is calculated. If it exceeds the threshold, a secondary split operation based on feature similarity is triggered to ensure that each final cluster has a high degree of homogeneity in response characteristics.
[0068] Finally, the solution establishes a scientific hierarchical mapping system. By objectively ranking the overall mean level of each resource cluster in its key speed dimension from high to low, and assigning clear labels such as first response speed level, second response speed level, etc., the data-driven clustering results are seamlessly transformed into a clear, quantifiable, and directly applicable engineering performance hierarchy for scheduling decisions. This complete technical chain, from intelligent dimensionality reduction and focusing, robust center discovery, and adaptive shape clustering to consistency verification and objective calibration, is interconnected and jointly achieves stable, accurate, and interpretable speed classification of distributed resources with huge intrinsic differences in the virtual power plant. This lays a solid and reliable data foundation for the dynamic and coordinated configuration of the entire virtual power plant.
[0069] In virtual power plant operation decision-making, transforming abstract grid demand into precise utilization of different types of resources has always been a technical challenge. Existing technologies often employ simple or static assessment methods, which have significant drawbacks. A common approach is to allocate resources evenly or in fixed proportions based solely on the overall power deficit, completely ignoring the dynamic changes in grid urgency and the fundamental differences in response speed and duration among different resources. Another approach, while considering resource differences, involves isolated assessments, such as separately assessing frequency regulation and peak shaving demands and then simply overlaying them. This fragments the overall grid demand, failing to prioritize and coordinate multiple demand scenarios with varying urgency levels. This may result in resources being prematurely occupied by secondary demands, leaving resources unable to cope with sudden, more critical events.
[0070] To fundamentally solve the above problems, refer to Figure 3 As shown, when assessing the intensity of coordinated dispatch demand of the power grid in the current period based on external power grid environmental data, this invention constructs a precise demand intensity assessment system based on dynamic matching and simulation. First, the scheme extracts core indicators characterizing the urgency of the power grid from multi-source data. By comprehensively considering real-time frequency deviation, critical section over-limit risks, and load forecasting errors, it calculates a quantified instantaneous state anomaly index and power regulation demand. This achieves the transformation from raw data to a structured, computable "demand scenario." Second, the demand assessment is linked to the established resource velocity level system to obtain key capability parameters such as the total adjustable capacity and average regulation rate of each velocity level cluster at the current moment, thereby establishing the assessment on a reliable capability baseline.
[0071] Furthermore, the scheme implements a two-dimensional sorting and matching logic: On the one hand, for each specific power regulation demand, its specific requirements for regulation speed and duration are analyzed, and the matching degree is calculated with the inherent regulation rate of each resource cluster. Based on this, a priority order for resource clusters to be called to solve the demand is generated, which reflects the optimal matching principle. On the other hand, according to the urgency reflected by the state anomaly index, different power regulation demands (such as instantaneous frequency support and continuous overload mitigation) are prioritized for scenario processing, clarifying the strategic trade-offs when resources are limited. Finally, the scheme introduces a pre-decision-making stage of simulated scheduling: according to the scheduled scenario priorities and the resource calling order under each scenario, all demands are virtually attempted to be met, and in this process, the proportion of demand allocated to each resource cluster is accurately counted. This proportion is not an artificially set weight, but a result that naturally evolves based on the actual matching logic and resource constraints. It directly and quantitatively reflects the system's real and differentiated demand for resource clusters with different response speeds under the current power grid situation, that is, the intensity of collaborative allocation demand.
[0072] By implementing this technical solution, the ambiguity and static nature of demand assessment have been fundamentally changed, achieving a leap from total quantity assessment to structured and differentiated assessment. The resulting demand intensity index is no longer a single value, but an intensity distribution vector for resource clusters at different speed levels. This vector clearly indicates which clusters should enter a high-alert state and the approximate proportion to be deployed. This provides precise input guidance for subsequent steps to determine the available response capabilities of each resource cluster, enabling resource capacity preparation and allocation to closely align with the actual and dynamic needs of the power grid. This fundamentally improves the synergy, agility, and economy of the overall response strategy of the virtual power plant, building a forward-looking decision-making advantage for dealing with complex power grid conditions. In specific implementation, when the system detects a sudden frequency drop (high state anomaly index), it quickly identifies this as a high-priority rapid power support scenario and immediately prioritizes high-speed response clusters (such as flywheels and rapid energy storage) in the call order based on matching degree. Simulated scheduling will prioritize allocating this portion of demand to these clusters, thus giving them extremely high demand intensity values. For concurrent, low-priority intraday peak-shaving demands, the invocation order tends to favor medium- and low-speed clusters, and the demand intensity values of these clusters are correspondingly lower after simulation allocation.
[0073] In the dispatching decisions of virtual power plants, forecasting the future grid load state is the starting point for formulating all action plans. However, existing forecasting technologies often suffer from overly rigid and simplistic forecast results, making it difficult to meet the grid's needs for risk prediction and flexible decision-making under high-proportion renewable energy integration. Common methods often provide a single, deterministic load forecast curve. While this point forecast provides a most probable value, it completely masks the inherent uncertainty of the forecast itself. When the actual load deviates from this forecast value due to sudden weather changes, random events, or unpredictable behavior, the dispatching plan based on this rigid forecast will immediately face risks, potentially leading to insufficient or excessive resource allocation, causing grid security problems or economic losses.
[0074] To overcome the above-mentioned shortcomings, in the process of predicting the power grid load balance state for future target periods, reference is made to... Figure 4 As shown, this invention proposes a load balance state prediction method based on hierarchical fusion and explicit characterization of uncertainty: First, the scheme performs structured preprocessing on the input data, and captures the inherent inertia of the load on daily, weekly, and seasonal scales by extracting typical load change patterns and periodic laws from historical data; at the same time, it identifies known planned event signals (such as factory maintenance plans and schedules of major sports events) and network topology constraint change information (such as line commissioning and decommissioning) from external power grid environment data, thereby separating the predictable factors affecting the load from random factors at the source.
[0075] In the core forecasting phase, the scheme employs a layered synthesis strategy: First, a baseline load forecast curve is generated based on extracted typical patterns, reflecting the normal load trend under conditions without special events. Second, instead of vaguely incorporating the impact of events into the model, the scheme independently assesses the deterministic superposition or reduction of the impact based on known planned event signals, and explicitly applies this impact to the baseline forecast, thus forming a load forecast curve that takes events into account. This two-step approach of baseline + event correction significantly improves the accuracy of forecasts during special periods and makes the forecast results interpretable—dispatchers can clearly understand the composition of the forecast values.
[0076] Furthermore, a load envelope representing the upper limit of the forecast and a load envelope representing the lower limit of the forecast are actively generated. These two envelopes are not arbitrarily set, but are calculated by analyzing the statistical characteristics of historical forecast errors, considering the forecast uncertainties of key driving factors such as weather, and combining the inherent random fluctuation range of the load, using a specific fluctuation model (such as probability forecasting based on quantiles). Together, they define a reasonable fluctuation range for the forecast value. Finally, the scheme packages and outputs the net load demand for the target period (the difference between the predicted load value taking into account events and the predicted output of renewable energy) and its upper and lower fluctuation envelopes as the grid load balance state.
[0077] Implementing this scheme transforms the virtual power plant's decision-making system from facing a certain but fragile future to facing a future with clearly defined risk boundaries. This information-rich description of the load balance state provides crucial risk-aware input for all subsequent steps. For example, during multi-objective optimization calculations, the optimization model can no longer merely pursue satisfaction of a single "point," but can be designed to achieve economic optimality under normal scenarios while ensuring grid security constraints are met under any fluctuation envelope, thereby automatically generating inherently robust dispatch schemes. When assessing the intensity of coordinated dispatch demand, the fluctuation ceiling can be combined to evaluate demand under worst-case conditions, allowing for the preparation of sufficient response resources in advance. This essentially transforms the uncertainty of prediction from a threat that needs to be prevented into a design parameter that can be proactively managed, greatly enhancing the virtual power plant's decision-making resilience and proactive defense capabilities in the face of complex and uncertain environments, providing key technical support for building a highly resilient power grid.
[0078] In practice, for the next scheduling day, the system generates a baseline based on historical data of the same type, overlays it with the known load increment from a large exhibition hall event, and obtains a predicted curve taking the event into account. Simultaneously, based on the uncertainty of the weather forecast for that day (e.g., temperature forecast error ±2℃), and combined with historical similar day error analysis, it calculates the upper and lower envelopes for possible load fluctuations of 3% upward and 2.5% downward. Finally, combined with the predicted photovoltaic output, it outputs the net load demand value for the next 24 hours at 15-minute intervals, along with its upper and lower bounds, serving as the basis for the entire dynamic coordination and configuration process.
[0079] In resource scheduling of virtual power plants, a critical gap exists between owning resources and being able to reliably use them. Existing technologies typically treat resource capacity assessment as a static, isolated step, which has two significant drawbacks. First, capacity assessment often deviates from actual demand scenarios. A common practice is to simply aggregate the nameplate capacity or maximum technical output of resources and treat it as a fixed, unchanging available capacity. This assessment completely ignores the dynamic impact of the urgency of grid demand on the priority and confidence of different resource calls. For example, in grid emergencies, it is essential to be 100% certain that some resources are available; while during routine regulation, a certain risk of call failure is acceptable for economic reasons. Existing static assessments cannot achieve this scenario-based, flexible capacity management. Second, the assessment process views the capacity of each resource or cluster in isolation, ignoring the inherent complementary and collaborative potential of different types of resources over time. For example, there is a natural synergy between a fast but short-duration resource (such as energy storage) and a slow-start but long-lasting resource (such as a gas turbine). Existing technologies lack mechanisms to identify and utilize this synergy inherent in differences in resource characteristics, resulting in the overall callable capacity failing to fully tap the aggregation potential of virtual power plants.
[0080] In response to the above problems, refer to Figure 5As shown, when determining the callable response capabilities of each resource cluster during the target time period, this invention constructs a refined capability assessment system that is demand-driven, dynamically corrected, and hierarchically encapsulated. This scheme first starts from physical limits, calculating the theoretical adjustment potential of each cluster based on the real-time status of resources, thus establishing an objective upper limit of capability. Instead of directly using theoretical potential, the scheme introduces the intensity of coordinated allocation demand as a key correction factor. For clusters marked as highly urgent by demand intensity (e.g., those urgently requiring rapid response from the power grid), it deeply assesses the availability confidence and adjustment reliability of each resource within them (e.g., whether the energy storage charge status is sufficient, and whether the performance history of controllable load users is good), and confirms the availability improvement of the theoretical potential accordingly. This essentially filters and locks in the most reliable portion of capability for critical tasks. Conversely, for clusters with normal or low demand intensity, a conservative calculation is performed using response willingness and economic thresholds (e.g., user bids for demand response, and compensation thresholds for interruptible loads). This reflects the trade-off between the economics and success rate of dispatching in non-emergency situations. Through this mechanism, resource capacity is no longer a fixed value, but dynamically evolves into available adjustable capacity based on its role and importance in the current power grid operation plan, thus achieving a precise match between resource value and power grid demand.
[0081] Subsequently, the solution encapsulates the abstract capacity figures into a set of callable capability parameters for scheduling operations, based on the cluster's response speed level. This is not a simple formatting, but rather a translation of capabilities into a scheduling language: for high-speed clusters, the parameter set emphasizes the power value that arrives quickly, the shortest possible duration, and the maximum adjustment frequency, which are key to executing second-level frequency adjustment commands; for low-speed clusters, it describes the gradual change process of its adjustable power, the longest possible duration, and the start-up warning time, which is suitable for hourly peak shaving.
[0082] Building upon the above embodiments, this embodiment further addresses the issue of isolated resource evaluation by proposing a complementary optimization strategy. First, by analyzing and adjusting the complementary relationships in timing, it intelligently identifies cluster combinations that can naturally relay or cover each other. For example, it identifies the potential for timing integration between A (fast-charging and discharging energy storage, fast speed but short duration) and B (slow-tuning gas turbine, slow start-up but long duration). Based on this, the scheme performs virtual aggregation: without changing physical connections and ownership, it merges the callable capability parameter sets of A and B at the scheduling logic level to generate a virtual collaborative resource unit. This virtual unit may possess superior comprehensive characteristics of fast start-up (inherited from A) + long duration (inherited from B), which is impossible for a single cluster. Finally, the scheme performs anti-disturbance verification on all independent capabilities and virtual collaborative units, simulating the robustness of the entire capability supply system under adverse conditions such as sudden failure of some resources, ensuring that the final output set of callable response capabilities is not only large in quantity but also of high quality and reliability.
[0083] In a specific scenario, when the system predicts a 30-minute load spike one hour later (high demand intensity indicates peak shaving), while the current frequency experiences slight fluctuations (medium demand intensity indicates frequency regulation), the system rigorously verifies the real-time state of charge of its energy storage and the health of its power conversion system for high-speed frequency regulation clusters, confirming their reliable capacity for rapid frequency regulation. For medium-speed peak shaving clusters (such as those capable of adjusting industrial loads), the system calculates a probable available capacity based on the contract terms with users (economic threshold) and the current production plan. Simultaneously, the system identifies the temporal complementarity between rapid energy storage clusters (capable of providing 5 minutes of support quickly) and gas turbine clusters (requiring 10 minutes of startup but capable of long-term operation), logically aggregating them virtually into a single unit that can start quickly and maintain operation for extended periods. Finally, the system simulates and verifies whether the gas turbine can independently take over in the event of a sudden energy storage failure. The final output is a rich set of capabilities, including the independent capabilities of each cluster confirmed by demand and the verified capabilities of the virtual collaborative unit, for use by the optimization and configuration module.
[0084] In the stage of generating the final dispatch scheme in a virtual power plant, existing optimization methods often pursue the optimal solution for a single objective. For example, they may take minimizing the regulation deviation as the absolute objective, mobilizing all resources at all costs to achieve precise tracking, which may lead to resource abuse and high costs; or they may take economy (efficiency) as the sole guide, which may jeopardize grid security due to insufficient regulation accuracy. This "either / or" optimization cannot meet the complex requirements of modern power grids for the comprehensive optimization of security and economy. More importantly, most optimization models treat complex resource capabilities, network constraints, and demand targets as static, deterministic inputs, outputting the optimal solution after a one-time solution. Such solutions are often extremely vulnerable to real-world disturbances such as load forecasting errors, renewable energy fluctuations, and random equipment failures.
[0085] To overcome the above challenges, when generating the final scheduling scheme, reference was made to... Figure 6As shown, this invention further constructs a two-layer dynamic optimization decision-making framework based on multi-objective trade-offs and multi-scenario robustness verification. This scheme acknowledges the non-existence of perfect solutions and instead uses a systematic approach to generate and select a satisfactory solution that exhibits optimal overall performance and strong resilience under current cognitive conditions. Its implementation begins with a comprehensive data standardization and structuring process. The scheme unifies the diverse key information input from upstream into a standardized language that the optimization engine can process: it decomposes and reassembles the set of callable response capability parameters characterizing resource dynamic capabilities into a discrete sequence of power adjustable quantities with fixed time steps, enabling precise issuance of scheduling instructions; it extracts regional distribution information describing resource spatial locations into inter-regional power transmission coupling relationships and limits characterizing electrical connections, transforming geographical distribution into network constraints influencing decision-making; and crucially, it analyzes the grid load balance state, rich in uncertain information, into two parts: first, the minimum regulating power target required at each time segment (i.e., demand based on the most probable predicted value); and second, the allowable regulation error band around this target (i.e., the acceptable range based on the fluctuation envelope).
[0086] In defining the problem, the solution clearly establishes two optimization objectives: the first objective is to minimize the overall deviation between the actual regulation power and the target, safeguarding the safety and accuracy of grid operation; the second objective is to maximize the average utilization efficiency of all resource clusters, pursuing operational economy. Simultaneously, the solution systematically integrates grid stability constraints, including regional power balance constraints to ensure local power balance, critical section power limit constraints to prevent line overload, and power change rate and duration constraints that respect the physical limits of resources.
[0087] This solution's solution and decision-making mechanism abandons the traditional approach of seeking a single, absolutely optimal solution. Instead, it uses interactive adjustments to the emphasis on two optimization objectives, performing multiple rounds of solutions. For example, one round assigns extremely high weight to adjustment accuracy, generating a near-perfect tracking solution that may be costly; the next round increases the weight to resource efficiency, generating a more economical solution with slightly looser tracking. In this way, a sequence of candidate configurations with different trade-offs between adjustment accuracy and resource efficiency can be generated, i.e., a "Pareto optimal frontier" solution set. This provides a rich spectrum of optional strategies for the final decision.
[0088] The final decision is not based on the numerical value of a single objective, but rather introduces a higher-dimensional evaluation standard oriented towards the real operating environment: the ability to maintain grid stability. For each candidate scheme, its execution process is simulated within the target time period and subjected to stress tests under preset typical disturbance scenarios (such as positive or negative deviations in load forecasting, or the sudden withdrawal of a key resource) to evaluate its ability to maintain grid stability. Finally, decision-makers can comprehensively consider the scores of each scheme across three dimensions: regulation accuracy, resource efficiency, and stability performance, and select the scheme with the best overall performance under a specific operational orientation (such as prioritizing safety under extreme weather conditions, or pursuing economic efficiency under normal conditions) as the final dynamically coordinated configuration scheme.
[0089] Implementing this scheme elevates the generation of scheduling plans from a deterministic, single-objective mathematical computation process to a robust, intelligent decision-making process based on uncertainty, multiple objective trade-offs, and robust verification. Its output is not a solution, but a verified optimal choice. This ensures that the final scheme is not only mathematically optimal under ideal models, but also engineering-optimal in approximating real-world complex environments.
[0090] Reference Figure 7 As shown, in order to implement the above method, the present invention also proposes a virtual power plant resource dynamic coordination and allocation system, including:
[0091] The data acquisition and feature extraction module is used to collect real-time operational data of various resources inside the virtual power plant and external power grid environment data, and extract operational feature parameters reflecting the differences in response speed of various resources based on the operational data.
[0092] The resource clustering and demand assessment module is used to perform clustering analysis on various resources based on the operational characteristic parameters, form resource clusters with different response speed levels, and assess the intensity of the grid's coordinated dispatch demand in the current period based on the external power grid environment data.
[0093] The load forecasting and capacity matching module is used to predict the power grid load balance status for a future target period based on historical data and the external power grid environment data, and to determine the available response capacity of each resource cluster for the target period by combining the response speed level of the resource cluster and the intensity of the coordinated dispatch demand.
[0094] The multi-objective optimization and scheme generation module is used to perform multi-objective optimization calculations that satisfy power grid stability constraints based on the callable response capability, resource area distribution information and the power grid load balance status, and generate a dynamic coordination configuration scheme for the various resources within the target time period.
[0095] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.
Claims
1. A method for dynamic coordination and allocation of virtual power plant resources, characterized in that, Includes the following steps: Real-time acquisition of operational data of various resources within the virtual power plant and external power grid environment data; based on the operational data, extraction of operational characteristic parameters reflecting the differences in response speed of various resources. Based on operational characteristic parameters, cluster analysis is performed on various resources to form resource clusters with different response speed levels. Then, based on external power grid environmental data, the intensity of coordinated dispatch demand for the power grid in the current period is assessed. Specifically, this includes: calculating the instantaneous state anomaly index of the power grid operation and the power regulation demand in the short term based on real-time frequency deviation, power flow at key sections, and load forecasting error data from the external power grid environmental data; obtaining the total adjustable power capacity and average regulation rate of each cluster in the current period based on the response speed level of the resource clusters; determining the priority order for calling resource clusters that meet the demand based on the requirements for regulation speed and duration, and the regulation rate of each resource cluster; determining the scenario processing priority for different power regulation demands based on the degree of state anomaly; performing simulated scheduling according to the scenario processing priority and the priority order for calling resource clusters, and statistically analyzing the demand ratio allocated to each resource cluster to assess the intensity of coordinated dispatch demand for each resource cluster based on the demand ratio. Based on historical data and external power grid environment data, predict the power grid load balance status for the target period in the future; combine the response speed level and coordination demand intensity of resource clusters to determine the available response capacity of each resource cluster for the target period. Based on the callable response capability, resource area distribution information, and power grid load balance status, multi-objective optimization calculations are performed to meet power grid stability constraints, generating dynamic coordination configuration schemes for various resources within the target time period. This is used to improve the overall resource utilization efficiency of the virtual power plant and ensure power grid load balance.
2. The method for dynamic coordination and allocation of virtual power plant resources according to claim 1, characterized in that: The operational data includes: time series of power regulation command values received by various resources, and corresponding time series of actual output power values; command response delay time and power stabilization adjustment time calculated based on the power regulation command value time series and the actual output power value time series, and power change capability per unit time calculated based on the actual output power value time series as operational characteristic parameters.
3. The method for dynamic coordination and allocation of virtual power plant resources according to claim 1, characterized in that: Based on operational characteristic parameters, cluster analysis is performed on various resources to form resource clusters with different response speed levels, including: The runtime characteristic parameter vectors of all resources are mapped to a multi-dimensional feature space. Dimensional correlation analysis is performed to identify the core dimension subset that is directly and strongly correlated with response speed, and the projected coordinates of each resource on the core speed dimension are generated. In the feature subspace formed by the core dimension subset, calculate the local density of each resource projection coordinate point and its neighboring points, and simultaneously calculate the minimum distance from each point to a point with a higher local density; select several points with the highest product of local density and minimum distance as initial clustering anchor points; With each initial cluster anchor point as the center, region growth is performed along the local density gradient descent direction in the feature subspace until a boundary formed by the growth of other anchor points is encountered or a preset feature difference threshold is reached. Each continuous region formed in this way is defined as a candidate range of a resource cluster. All resources in the feature subspace are initially assigned to clusters based on the candidate range to which their projected coordinate points belong. For each resource cluster initially divided, calculate the numerical dispersion of all resources within it on the preset key dimension of response speed; if the dispersion of a resource cluster exceeds the preset threshold, then perform a secondary split operation based on feature similarity on the cluster. Based on the overall average level of each resource cluster in the key speed dimension, they are sorted from high to low and labeled as the first response speed level, the second response speed level, etc., thus forming resource clusters with clear speed level labels.
4. The method for dynamic coordination and allocation of virtual power plant resources according to claim 1, characterized in that: Based on historical data and external power grid environment data, predict the power grid load balance status for future target periods, including: Extract typical load change patterns and periodic regularities associated with the target time period from historical data; at the same time, identify known planned event signals and network topology constraint change information that affect the load from external power grid environment data; Based on typical load change patterns and periodicity, a baseline load forecast curve for the target period is generated; and based on known planned event signals, the deterministic superposition or reduction impact on the baseline load is assessed to form a load forecast curve that takes the events into account. Based on the load forecast curve that takes events into account, a load envelope representing the possible upper limit of the forecast and a load envelope representing the possible lower limit of the forecast are generated. The net load demand for the target period and its upper and lower fluctuation envelopes together constitute the power grid load balance state.
5. The method for dynamic coordination and allocation of virtual power plant resources according to claim 1, characterized in that: Based on the response speed level and coordination demand intensity of resource clusters, determine the available response capacity of each resource cluster during the target time period, including: Based on the real-time operating status of all resources in each resource cluster, the maximum increase and decrease in output that each resource cluster can physically allow at the start of the target time period are aggregated and calculated as its theoretical adjustment potential. Obtain the intensity of coordinated dispatch demand to indicate the degree of differentiated demand of the power grid for resources at different speed levels; among them: for resource clusters with demand intensity indicating high urgency, confirm the availability improvement of their theoretical regulation potential based on the availability confidence and regulation reliability of their internal resources; for resource clusters with demand intensity indicating normal or low demand, consider their response willingness and economic threshold, and perform conservative available capacity calculation. Based on the response speed level of each resource cluster, the corrected available adjustment potential is encapsulated into a set of callable capability parameters that change over time within the target period. Specifically, for clusters with high response speed levels, the parameter set highlights their fast-arriving power value, shortest duration, and maximum adjustment frequency; for clusters with low response speed levels, the parameter set highlights the gradual change in adjustable power, longest duration, and shortest warning time required for startup.
6. The method for dynamic coordination and allocation of virtual power plant resources according to claim 5, characterized in that: The process of determining the available response capabilities of each resource cluster during the target time period also includes optimizing the transfer and complementarity of capabilities between clusters, including: Analyze the complementary relationship between resource clusters of different response speed levels in terms of adjustment timing; Based on the complementary relationship, the callable capability parameter sets of different clusters with time sequence are virtually aggregated in the scheduling logic to form a virtual collaborative resource unit with better comprehensive response characteristics. The callable capabilities of virtual collaborative resource units and each independent cluster are verified against disturbances, and the final set of callable response capabilities with robustness confirmed is output.
7. The method for dynamic coordination and allocation of virtual power plant resources according to claim 1, characterized in that: Based on callable response capabilities, resource distribution information, and grid load balance status, multi-objective optimization calculations are performed to satisfy grid stability constraints, generating dynamic coordination and allocation schemes for various resources within the target time period, including: The available response capability parameter set of each resource cluster is converted into a discrete sequence of power adjustable quantities in preset time intervals within the target time period; the resource area distribution information is converted into the coupling relationship and limit of power transmission between areas; and the power grid load balance status is converted into the minimum regulating power target required for each time section within the target time period and its allowable regulating error band. The first optimization objective is to minimize the overall deviation between the actual regulating power and the minimum regulating power target at each time segment; the second optimization objective is to maximize the average utilization efficiency of all resource clusters throughout the target period; and the grid stability constraints are integrated, including at least the power balance constraints of each region, the power limit constraints of key transmission sections, and the power change rate and duration constraints of each resource cluster itself. Based on the optimization elements, objectives, and constraints, a unified optimization problem is constructed; by interactively adjusting the emphasis on the first and second optimization objectives, multiple rounds of solution are performed to generate a sequence of candidate configuration schemes with different trade-offs between adjustment accuracy and resource efficiency. For each of the candidate configuration schemes, its execution process during the target time period is simulated, and its ability to maintain power grid stability under preset typical disturbance scenarios is evaluated. Based on the adjustment accuracy, resource efficiency and stability performance of each scheme, one scheme is selected as the dynamic coordination configuration scheme.
8. A virtual power plant resource dynamic coordination and allocation system, used to implement the method described in any one of claims 1 to 7, characterized in that: include: The data acquisition and feature extraction module is used to collect real-time operational data of various resources inside the virtual power plant and external power grid environment data, and extract operational feature parameters reflecting the differences in response speed of various resources based on the operational data. The resource clustering and demand assessment module is used to perform clustering analysis on various resources based on the operational characteristic parameters, form resource clusters with different response speed levels, and assess the intensity of the grid's coordinated dispatch demand in the current period based on the external power grid environment data. The load forecasting and capacity matching module is used to predict the power grid load balance status for a future target period based on historical data and the external power grid environment data, and to determine the available response capacity of each resource cluster for the target period by combining the response speed level of the resource cluster and the intensity of the coordinated dispatch demand. The multi-objective optimization and scheme generation module is used to perform multi-objective optimization calculations that satisfy power grid stability constraints based on the callable response capability, resource area distribution information and the power grid load balance status, and generate a dynamic coordination configuration scheme for the various resources within the target time period.