A virtual power plant multi-time scale adaptive demand response control method and system

Abstract

The application relates to the technical field of artificial intelligence, and discloses a virtual power plant multi-time scale adaptive demand response control method and system.The method comprises the following steps: constructing a time characteristic embedding vector according to operation data of a target virtual power plant; performing cross-scale similarity analysis on the time characteristic embedding vector to obtain correlation intensity; performing distributed clock synchronization and delay compensation on local clock stamps of initial resource nodes; performing multi-time scale analysis on adjustable capacity of the synchronized resource nodes; performing adaptive updating on the multi-time scale analysis results to obtain target adjustable capacity; performing distributed collaborative optimization on the target adjustable capacity and a preset target response amount to generate a coordinated optimization control strategy; and controlling demand response of each synchronized resource node of the target virtual power plant to obtain a synchronized response control result.The application effectively realizes unified clock and synchronized feedback, reduces prediction error and response lag, and improves stability and control precision.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a method and system for multi-timescale adaptive demand response control of a virtual power plant. Background Technology

[0002] In modern power systems, with the large-scale integration of renewable energy and the increase in flexible resources on the load side, virtual power plants, as a distributed control and aggregation mechanism, can uniformly schedule multiple types of distributed resources, thereby enhancing system flexibility and demand response management. Traditional virtual power plant demand response control methods mostly rely on optimization at a single time scale, such as day-ahead, intraday, or real-time scheduling strategies, which are difficult to fully coordinate resource responses at different time scales.

[0003] However, most existing methods have significant drawbacks: due to the distributed autonomous structure of virtual power plants, resource nodes are geographically dispersed, communication delays and data asynchrony are unavoidable, and a unified clock and synchronization feedback mechanism are lacking. Compared to centralized control systems, virtual power plants cannot guarantee the immediate availability of the global state, leading to lags or error accumulation in the actual operation of AI-based prediction and optimization strategies, thus reducing response accuracy and stability.

[0004] Secondly, most existing methods optimize only within a single time scale, lacking an adaptive coordination mechanism across time scales. The operating status of resource nodes and the power system load vary significantly across different time scales, and fixed-time-scale optimization cannot dynamically adjust strategies based on real-time operating conditions. This fragmentation of time scales makes real-time scheduling strategies prone to response lags under rapidly fluctuating loads and renewable output conditions, affecting operational safety and economy.

[0005] In summary, existing technologies urgently need a multi-timescale adaptive demand response control method that can simultaneously overcome the distributed asynchronous characteristics of nodes and the problem of time scale fragmentation. This method should be able to achieve dynamic coordinated control across time scales while ensuring the consistency of distributed node states and prediction accuracy, thereby improving the response flexibility, scheduling accuracy, and system stability of virtual power plants.

[0006] Therefore, in the multi-timescale demand response control of virtual power plants, the lack of a unified clock and synchronous feedback, as well as the fragmentation of time scales, leads to increased prediction errors and response lags, affecting system stability and control accuracy. Summary of the Invention

[0007] This invention provides a multi-timescale adaptive demand response control method for virtual power plants. Its main purpose is to solve the problems in multi-timescale demand response control of virtual power plants, such as the lack of a unified clock and synchronous feedback, and the fragmentation of time scales, which leads to increased prediction errors and response lags, affecting system stability and control accuracy.

[0008] Firstly, to achieve the above objectives, the present invention provides a virtual power plant multi-timescale adaptive demand response control method, comprising: Collect operational data of initial resource nodes within the target virtual power plant, and construct time feature embedding vectors based on the operational data at different time scales; Cross-scale similarity analysis is performed on the time feature embedding vectors to obtain the correlation strength between each time scale; Distributed clock synchronization and delay compensation are performed on the local clock stamp of the initial resource node to obtain a synchronized resource node; The adjustable capability of the synchronization resource node is analyzed on multiple time scales based on the time feature embedding vector, and the results of the multi-time scale analysis are adaptively updated based on the correlation strength to obtain the target adjustable capability. Based on the target adjustability, the synchronous resource nodes are subjected to distributed collaborative optimization to generate a coordinated optimization control strategy. The demand response of each synchronization resource node in the target virtual power plant is controlled using the aforementioned coordinated optimization control strategy to obtain the synchronization response control result.

[0009] Secondly, the present invention also provides a virtual power plant multi-timescale adaptive demand response control system, the system comprising: The feature embedding vector construction module is used to collect the operation data of the initial resource nodes in the target virtual power plant and construct time feature embedding vectors based on the operation data at different time scales. The cross-scale similarity analysis module is used to perform cross-scale similarity analysis on the time feature embedding vector to obtain the correlation strength between each time scale. The clock synchronization and delay compensation module is used to perform distributed clock synchronization and delay compensation on the local clock stamp of the initial resource node to obtain a synchronized resource node. The multi-timescale analysis module is used to perform multi-timescale analysis on the adjustability of the synchronization resource node based on the time feature embedding vector, and to adaptively update the multi-timescale analysis results based on the correlation strength to obtain the target adjustability. The distributed collaborative optimization module is used to perform distributed collaborative optimization on the synchronization resource nodes according to the target adjustable capability, and generate a coordinated optimization control strategy. The demand response control module is used to control the demand response of each of the synchronization resource nodes of the target virtual power plant using the coordination optimization control strategy, and obtain the synchronization response control result.

[0010] This invention achieves temporal consistency and dynamic correlation mining of data among resource nodes in a virtual power plant through multi-timescale feature embedding and distributed clock synchronization; through cross-scale similarity analysis and adaptive update mechanism, it can accurately assess the adjustability of each node, improve the coordination and real-time performance of demand response; combined with distributed collaborative optimization control strategy, it realizes efficient and unified regulation of multi-source heterogeneous resources, thereby effectively reducing response lag and energy dispatch deviation, and significantly improving the overall stability, flexibility and economy of the virtual power plant. Attached Figure Description

[0011] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a flowchart illustrating a multi-timescale adaptive demand response control method for a virtual power plant according to an embodiment of the present invention. Figure 2 This is a functional block diagram of a virtual power plant multi-timescale adaptive demand response control system provided in an embodiment of the present invention; The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0013] To enable those skilled in the art to better understand the technical solutions of this disclosure, and to fully understand and implement the process of how this disclosure applies technical means to solve technical problems and achieve corresponding technical effects, the technical solutions in the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this disclosure, not all embodiments. The embodiments of this disclosure and the various features within them can be combined with each other without conflict, and the resulting technical solutions are all within the protection scope of this disclosure. All other embodiments obtained by those skilled in the art based on the embodiments of this disclosure without creative effort should fall within the protection scope of this disclosure.

[0014] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this disclosure are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0015] This application provides a virtual power plant multi-timescale adaptive demand response control method, which can be executed by software or hardware installed on terminal devices or server devices. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and demand forecasting platforms.

[0016] Reference Figure 1 The diagram shown is a flowchart illustrating a multi-timescale adaptive demand response control method for a virtual power plant according to an embodiment of the present invention. In this embodiment, the multi-timescale adaptive demand response control method for a virtual power plant includes: S1. Collect resource operation data of the initial resource nodes in the target virtual power plant, and construct time feature embedding vectors based on the resource operation data at different time scales.

[0017] In this embodiment of the invention, the target virtual power plant is a virtual energy entity that can be aggregated and controlled, composed of multiple distributed energy sources (such as photovoltaics, energy storage, electric vehicles, controllable loads, etc.). Initial resource nodes are the distributed energy units or load nodes initially included in the management of the virtual power plant. Before synchronization and optimization, these initial resource nodes operate independently, possessing their own operating data and clock systems. Resource operating data refers to the dynamic data generated by each initial resource node during operation, including power output, load changes, voltage, current, status signals, equipment operating parameters, etc.

[0018] The time scale refers to the temporal resolution or granularity used when analyzing resource operation data, such as second-level (real-time response), minute-level (short-term scheduling), hour-level (intraday optimization), or day-level (long-term prediction). A time feature embedding vector is a multi-dimensional feature representation formed by feature extraction and vectorized encoding of operation data at different time scales.

[0019] Specifically, constructing a time feature embedding vector based on the resource operation data at different time scales includes: Obtain day-ahead, intraday, and real-time scales, and then stratify the resource operation data according to the day-ahead, intraday, and real-time scales respectively to obtain day-ahead scale operation data, intraday scale operation data, and real-time scale operation data. The day-ahead operational data, the intraday operational data, and the real-time operational data are denoised to obtain day-ahead denoised data, intraday denoised data, and real-time denoised data. The day-ahead scale denoised data, the intraday scale denoised data, and the real-time scale denoised data are aggregated into a time-scale denoised dataset. The time-scale denoised dataset is normalized to obtain a normalized dataset; Perform temporal convolution on the normalized dataset to obtain the convolutional feature dataset; Temporal attention-weighted encoding is performed on the convolutional feature dataset to obtain a global temporal feature representation; Variational coding is performed on the global time feature representation, and the generated mean vector is used as the time feature embedding vector.

[0020] In this embodiment of the invention, the day-ahead scale is a data level with a time range from 24 hours prior to the dispatch date, typically used for medium- to long-term forecasting and planned optimization of energy systems. In a virtual power plant, day-ahead scale operating data reflects the periodic load characteristics and predictive output change trends of resources, primarily used to generate the overall operating plan and dispatchable capacity assessment for the following day.

[0021] Intraday scale is a data level covering the entire day's operation (generally measured in hours or 15-minute increments), reflecting the short-term dynamic characteristics of virtual power plant operation. Intraday scale operating data is primarily used for real-time adjustments to day-ahead plans to address weather changes, load fluctuations, or market price changes, ensuring the system's flexibility and adaptability.

[0022] Real-time scale refers to data at the second or minute level, used for real-time monitoring and rapid response control. Real-time scale data reflects instantaneous power fluctuations, frequency offsets, and voltage changes of resource nodes, and is a key input to ensure the safe and stable operation of the system.

[0023] In this embodiment of the invention, during the time series data preprocessing, the smoothness and authenticity of the data are improved by filtering out random interference, abnormal mutations and invalid sampling points, and the interference of acquisition errors and communication delays on the control model is removed. Furthermore, the time-scale denoised dataset, which is a collection of denoised data from each time level (previous day, intraday, real-time), is used to achieve information fusion at different time resolutions, ensuring that the system can establish a unified time-series mapping relationship between long-term trends and short-term fluctuations.

[0024] In this embodiment of the invention, data with different dimensions and amplitudes are unified into the same numerical range (such as 0 to 1 or -1 to 1) through linear or nonlinear mapping, thereby eliminating numerical deviations between different nodes or different types of resources.

[0025] In this embodiment of the invention, the sliding window feature extraction operation performed on the time series extracts feature information representing short-term dependencies by identifying local change trends and dynamic patterns in the time neighborhood, and captures local regular changes in the time series, such as power increase, frequency fluctuation, delay response and other characteristic behaviors.

[0026] In this embodiment of the invention, the convolutional feature dataset is a set of high-dimensional temporal feature representations obtained after temporal convolution processing. Each feature vector contains local dynamic information within different time windows, forming a data structure that can express both local temporal changes and reflect differences in resource status.

[0027] In this embodiment of the invention, weights are assigned to the convolutional feature dataset, and the contribution of the convolutional feature dataset to the global representation is dynamically adjusted according to the importance of features at different times. The influence of features in key periods, such as system peak periods and frequency disturbance periods, is strengthened through the "attention" strategy, thereby improving the model's perception ability and expression accuracy in complex temporal structures.

[0028] In this embodiment of the invention, the global time feature representation is a comprehensive time feature expression obtained by fusing attention-weighted time features. It is no longer limited to a certain time scale, but integrates dynamic features at three levels: day-ahead, intraday, and real-time in a unified vector space, which is used to describe the behavior pattern of the entire virtual power plant system in multiple time dimensions.

[0029] In this embodiment of the invention, global time features are compressed and generalized by probabilistic distribution modeling. Under the premise of maintaining the integrity of feature information, potential statistical regularities are extracted, and the model's adaptability to uncertainty is enhanced by probabilistic representation of the feature space.

[0030] In this embodiment of the invention, the mean vector refers to one of the distribution parameters generated during variational coding. It represents the central expression of global temporal features in the latent space, encompassing not only the average characteristic trend of the time series but also reflecting the comprehensive state under multi-scale temporal features. The mean vector is structured into a fixed-length multi-dimensional vector to serve as a unified expression of the temporal features of each resource node in the virtual power plant.

[0031] This invention models the operational data of initial resource nodes within a target virtual power plant across multiple time scales, enabling the simultaneous capture of dynamic characteristics of resources at different levels, including long-term trends, short-term fluctuations, and instantaneous responses. By constructing time feature embedding vectors, the originally scattered, asynchronous, and inconsistent operational data is transformed into a unified feature representation.

[0032] S2. Perform cross-scale similarity analysis on the time feature embedding vector to obtain the correlation strength between each time scale.

[0033] In this embodiment of the invention, by comparing and analyzing the embedded vectors of time features at different time scales, the dynamic coupling relationships and changing patterns between different time levels are identified. Through this cross-scale similarity analysis, the implicit dependencies and synchronization characteristics between day-ahead, intraday, and real-time data can be revealed.

[0034] In detail, the cross-scale similarity analysis of the time feature embedding vector to obtain the correlation strength between each time scale includes: Calculate the Euclidean distance between the time feature embedding vectors of any two time scales; Based on the Euclidean distance, generate cross-scale similarity values ​​for the temporal feature embedding vectors corresponding to the two time scales; A cross-scale similarity matrix is ​​constructed using the cross-scale similarity values; The cross-scale similarity matrix is ​​normalized to obtain a normalized similarity matrix; The correlation strength at each time scale is calculated based on the normalized similarity matrix.

[0035] In this embodiment of the invention, the spatial differences between time feature embedding vectors at different time scales are compared to reflect the degree of similarity or difference in their temporal behavior. The smaller the distance, the closer the resource operation characteristics at the two time scales are, indicating a high degree of consistency in temporal behavior; the larger the distance, the more significant the deviation in the temporal regularity of the operation characteristics.

[0036] Furthermore, the cross-scale similarity value is a similarity index generated based on the Euclidean distance between the temporal feature embedding vectors of two time scales, used to quantify the strength of dynamic correlation between different time scales. It not only considers the spatial proximity between features but also implicitly reflects the synchronicity and correlation of operational trends at different scales. A higher similarity value indicates more coordinated resource operation characteristics between the two time scales and more consistent response patterns at different time levels.

[0037] In this embodiment of the invention, the cross-scale similarity matrix is ​​a two-dimensional matrix structure constructed from the similarity values ​​between any two time scales. Each element in the matrix represents the degree of similarity between a pair of time scales, thereby forming a map of the interrelationships between time levels as a whole. This not only reveals the coupling patterns between different time levels, but can also be used to identify which scales have the strongest synergy and which scales exhibit response lag or mismatch.

[0038] In this embodiment of the invention, the normalized matrix ensures that all cross-scale similarity values ​​are in the same comparable range (e.g., 0 to 1), thereby guaranteeing that the interrelationships at different time levels have a consistent quantitative standard.

[0039] Furthermore, correlation strength refers to the overall degree of correlation between each time scale and all other time scales. It is an aggregated index after normalizing the cross-scale similarity matrix, reflecting the "centrality" or "synergy" of that time scale in the entire multi-time-level structure. The higher the correlation strength, the more dominant the scale is in the overall system and the stronger its dynamic correlation ability; the lower the correlation strength, the more independent the temporal behavior of that scale or the slower its response.

[0040] In detail, based on the normalized similarity matrix, the resource feature vectors at each time scale are compared pairwise, the similarity distribution at different time scales is calculated, and the degree of similarity aggregation at each time scale is statistically analyzed by combining the normalized matrix weights. This yields the correlation strength, which reflects the degree of dependence between variables at that time scale, and is used to characterize the dynamic coupling relationship at different time scales.

[0041] This invention, through cross-scale similarity analysis of time feature embedding vectors, can deeply reveal the dynamic coupling relationships and response consistency between different time scales (such as day-ahead, intraday, and real-time), thereby achieving a global understanding of the operating characteristics of virtual power plants. By calculating the correlation strength between each time scale, the dominant and lagging time scales can be identified, clarifying the response influence paths at different levels.

[0042] S3. Perform distributed clock synchronization and delay compensation on the local clock stamp of the initial resource node to obtain a synchronized resource node.

[0043] In this embodiment of the invention, the timestamps of data collected by each initial resource node within the virtual power plant are compared using a unified time base to detect the offset between the local clocks of each node and the network transmission delay. Through a distributed clock synchronization algorithm, each node dynamically adjusts its local clock based on mutual communication feedback without centralized control, thereby gradually converging towards consistency. For signal transmission delays still present during the synchronization process, a delay compensation mechanism is used to correct time errors, ensuring the true correspondence of the sampling time.

[0044] Specifically, the step of performing distributed clock synchronization and delay compensation on the local clock stamp of the initial resource node to obtain a synchronized resource node includes: Distributed clock synchronization is performed on the local clock stamp of the initial resource node to obtain the local updated clock stamp; The communication delay between the initial resource node and the communication neighbor node is detected based on the local update clock stamp to obtain a delay estimate. When the estimated delay value is greater than the preset delay threshold, the time alignment error between the initial resource node and the communication neighbor node corresponding to the estimated delay value is corrected to obtain the corrected resource node and the corrected neighbor node. The global clock drift is calculated based on the clock stamps of the corrected resource node and the corrected neighbor node, and it is determined whether the global clock drift is less than a preset drift threshold. When the global clock drift is greater than or equal to the drift threshold, the corrected resource node and the corrected neighbor node are recalibrated to obtain a synchronized resource node; When the global clock drift is less than the drift threshold, the corrected resource node is used as the synchronization resource node.

[0045] In this embodiment of the invention, the local update clock stamp refers to the time stamp information of the initial resource node after being adjusted by distributed clock synchronization. It not only reflects the sampling time of the node itself, but also corrects the local clock deviation during the synchronization process, so that each node is closer to the time base of its neighboring nodes in the network.

[0046] In this embodiment of the invention, a communication neighbor node is another resource node in the virtual power plant network that directly exchanges data or communicates with the target resource node. The delay estimate refers to the communication delay between the initial resource node and the communication neighbor node based on the locally updated clock stamp. It not only quantifies the time required for a signal to be sent and received, but also implicitly reflects the combined impact of network transmission, node processing, and communication load.

[0047] In this embodiment of the invention, the corrected resource node and the corrected neighbor node are nodes whose clocks or timestamps are adjusted through time alignment error correction operations after the delay estimate exceeds a preset threshold. This not only changes the timestamp value, but also logically achieves the time sequence consistency of data between nodes.

[0048] Furthermore, global clock drift, after local correction, measures the degree to which the clocks of each node in the entire virtual power plant network deviate from the global time reference. The magnitude of the drift directly determines whether further correction is needed to ensure that the final synchronization resource nodes maintain high-precision timing alignment in a multi-node environment.

[0049] In detail, the clock difference between each node is calculated based on the clock stamps of the corrected resource node and the corrected neighbor node; the clock difference between multiple nodes is statistically normalized to obtain the overall time offset distribution; then the mean squared error or mean absolute deviation of the relative clock offset of all nodes is calculated based on the overall time offset distribution to obtain the global clock drift degree, which characterizes the global time synchronization consistency and is used to evaluate the overall clock synchronization accuracy of the system.

[0050] In this embodiment of the invention, the synchronization resource node is a set of resource nodes whose timestamps reach the specified accuracy requirements globally after distributed clock synchronization, delay detection, error correction, and necessary recalibration. They are aligned on a unified time axis, providing a reliable and comparable time series basis for subsequent multi-timescale analysis, cross-scale feature extraction, and collaborative control.

[0051] The above steps, from updating the local clock stamp to delay estimation, and then to error correction and drift assessment, form an adaptive multi-level correction closed loop. This process not only handles the instantaneous delay differences between nodes but also ensures global time consistency, compensating for the multi-timescale data misalignment problem caused by node asynchrony in traditional virtual power plants.

[0052] Logically, if only local delay correction is performed without considering global drift, a situation may arise where local time alignment occurs but global drift still exceeds the limit. By introducing global drift calculation and threshold determination, combined with recalibration operations, this embodiment achieves dual protection of local correction and global consistency, ensuring that the final synchronization resource node has a high-precision and unified time reference throughout the entire virtual power plant network, providing a reliable foundation for subsequent multi-timescale analysis and collaborative optimization control.

[0053] In this embodiment of the invention, the step of performing distributed clock synchronization on the local clock stamp of the initial resource node to obtain a locally updated clock stamp includes: Obtain the communication neighbor nodes of each initial resource node, as well as the local clock stamp of the initial resource node and the neighbor clock stamp of the communication neighbor nodes; Broadcast the local clock stamp to the communication neighbor node, and receive the neighbor clock stamp from the communication neighbor node; Based on the received clock information of the neighbor clock stamp, the local clock stamp of the initial resource node is synchronized and updated to obtain the local updated clock stamp.

[0054] In this embodiment of the invention, resource nodes gradually calibrate to a unified clock reference by exchanging time information with each other. Each initial resource node first identifies the communication neighbor nodes with which it directly communicates, then sends its own time information to the communication neighbor nodes, while simultaneously collecting the time data returned by the communication neighbor nodes.

[0055] Furthermore, based on the clock information from the received neighbor clock stamps, the clock difference between the initial resource node's local clock and the neighbor node's clock is determined. Combining the clock differences from multiple neighbor nodes, synchronization algorithms such as weighted average or least mean square error are used to correct the local clock. After clock offset compensation, a local update clock stamp consistent with the neighbor clocks is generated. Through multiple rounds of interaction, the clocks of all nodes gradually converge, ensuring that data acquisition, control commands, and response events within the system occur on a unified timeline.

[0056] This invention uses distributed clock synchronization to coordinate and adjust the local clocks of each initial resource node with those of its communication neighbors, unifying the node timestamps to a consistent reference standard. This can reduce time inconsistencies caused by clock drift, equipment differences, or communication delays, and provide a reliable foundation for data alignment of multi-source heterogeneous resources within a virtual power plant.

[0057] S4. Perform multi-timescale analysis on the adjustability of the synchronization resource node based on the time feature embedding vector, and adaptively update the multi-timescale analysis results based on the correlation strength to obtain the target adjustability.

[0058] In this embodiment of the invention, through this multi-timescale analysis and adaptive update, the final target adjustable capability not only reflects the adjustment potential of each node at different time levels, but also takes into account the cross-scale synergy between nodes, providing a reliable basis for the unified optimization scheduling of virtual power plants.

[0059] In detail, the multi-timescale analysis of the adjustability of the synchronization resource node based on the time feature embedding vector includes: Obtain the resource state variables of the synchronous resource node after delay compensation, and use the time feature embedding vector and the resource state variables to update the preset nonlinear capability analysis model to obtain the autoregressive capability analysis model. The autoregressive capability analysis model is used to calculate the adjustable capability analysis value of the synchronization resource node at each time scale; Obtain the actual call feedback data of the synchronization resource node, and generate the actual value of the adjustable capability based on the actual call feedback data; The adjustable capability deviation for each time scale is calculated based on the difference between the adjustable capability analysis value and the actual adjustable capability value.

[0060] In this embodiment of the invention, the resource state variables after delay compensation are a set of key parameters reflecting the node's operating status recorded after the synchronous resource node has undergone communication delay detection and time alignment correction. These parameters include power output, load level, energy storage capacity, response status, etc. The real-time operating characteristics of the node are preserved, and the time deviation caused by communication and sampling is eliminated through delay compensation, so that the state information of each node truly corresponds on the global time axis.

[0061] In this embodiment of the invention, the nonlinear capability analysis model is an analytical model used to describe the nonlinear characteristics of resource node controllability as state variables change. By capturing the nonlinear characteristics of node capability, it provides a more realistic and refined capability mapping for subsequent autoregressive model construction and multi-timescale analysis. Further, the autoregressive capability analysis model refers to an analytical model that can self-update and predict node controllability after inputting time feature embedding vectors and resource state variables into the nonlinear capability analysis model. The calculation formula is shown below:

[0062] in, This represents the future timescale step size of the i-th synchronization resource node calculated by the autoregressive capability analysis model at timescale s. Adjustable capability analysis values ​​at any time, This represents a nonlinear prediction model. This represents the set of trainable parameters for the model. This represents the m-th temporal feature embedding vector at time scale s. This represents the nth resource state variable at time t after delay compensation.

[0063] In this embodiment of the invention, the adjustable capability analysis value is the theoretical adjustment potential of the synchronization resource node at various time scales, calculated by an autoregressive capability analysis model. It reflects the maximum adjustment range that the node can achieve under physical constraints and dynamic characteristics at different levels, such as day-ahead, intraday, and real-time. The actual adjustable capability value is the actual adjustment capability realized by each synchronization resource node at various time scales, obtained based on the actual call feedback data.

[0064] Furthermore, based on the adjustable capacity analysis value and the corresponding actual adjustable capacity value, the difference between the two at each time scale is calculated as the deviation of the adjustable capacity. By normalizing or weighting the deviation at different time scales, the adjustable capacity deviation at each time scale is obtained, which is used to reflect the degree of consistency between the analysis prediction and the actual execution.

[0065] In detail, the adaptive update of the multi-timescale analysis results based on the correlation strength to obtain the target adjustability includes: When the adjustable capability deviation is greater than a preset deviation threshold, the adjustable capability analysis value is weighted and fused according to the correlation strength to obtain a cross-scale weighted correction value for each time scale. Adaptive fusion of all the aforementioned cross-scale weighted correction values ​​yields the target adjustable capability.

[0066] In this embodiment of the invention, the cross-scale weighted correction value is a correction result generated for each time scale after weighted fusion of the adjustable capability analysis value and the correlation strength. It corrects the capability prediction that deviates from reality and retains the synergistic information between the scales, ensuring that the adjustment potential of each time scale is more realistic and reliable as a whole.

[0067] In this embodiment of the invention, adaptive fusion is a process of integrating all cross-scale weighted correction values ​​according to their importance and dynamic characteristics to generate the final target adjustable capability. It comprehensively considers the contribution, correlation strength, and deviation of each time scale to achieve dynamic optimization and continuous updating of the adjustable capability. Furthermore, the target adjustable capability is the final quantitative result of the adjustable capability after cross-scale weighted correction and adaptive fusion, indicating the actual adjustment potential at different time scales while also taking into account the synergistic consistency across multiple scales.

[0068] If only single-scale corrections are performed, the capabilities at certain time scales may be over-adjusted, compromising overall consistency. Through cross-scale weighted fusion and adaptive integration, this embodiment achieves dynamic balancing and optimized integration of information from multiple time scales, ensuring that the target's adjustable capabilities reflect the true capabilities at each scale while maintaining global consistency.

[0069] This invention comprehensively analyzes the dynamic behavior of synchronous resource nodes at different time scales by utilizing time feature embedding vectors, accurately characterizing the adjustable capabilities of nodes at multiple levels, including day-ahead, intraday, and real-time. By adaptively updating the analysis results with cross-scale correlation strength, the information weight of highly correlated time scales is enhanced, while noise or bias at low-correlation time scales is suppressed, thereby generating more reliable target adjustable capabilities. This approach fully utilizes the dynamic characteristics of multiple time scales while maintaining cross-scale synergy, enabling virtual power plants to achieve precise and stable scheduling and response optimization, improving the overall system operating efficiency and control accuracy.

[0070] S5. Based on the target adjustable capability, perform distributed collaborative optimization on the synchronization resource nodes to generate a coordinated optimization control strategy.

[0071] In this embodiment of the invention, the target adjustable capabilities of each synchronous resource node and the preset target response of the virtual power plant are used as inputs for collaborative optimization under a distributed architecture. Each node independently generates an optimal response scheme based on its own capabilities and local constraints, while exchanging information with neighboring nodes through communication and cooperation to ensure that the scheduling decisions of each node are coordinated globally. Through multiple rounds of iterative adjustments and optimizations, each node ultimately achieves a unified control scheme that satisfies the overall response target while fully utilizing the adjustable capabilities of each node, thereby generating a coordinated optimization control strategy that balances local flexibility and global coordination.

[0072] Specifically, the step of performing distributed collaborative optimization of the synchronization resource nodes based on the target adjustable capability to generate a coordinated optimization control strategy includes: A control optimization function is constructed based on the target adjustable capability and the preset local cost function; The local control quantity of the synchronization resource node is calculated using the control optimization function. Obtain the neighbor control quantity of the communication neighbor node corresponding to the synchronization resource node, and use the neighbor control quantity to iteratively update the local control quantity to obtain the coordination control quantity; The difference between the coordinated control quantity and the local control quantity is taken as the global error; When the global error is less than a preset error threshold, a coordinated optimization control strategy is generated based on the coordinated control quantity.

[0073] In this embodiment of the invention, the control optimization function is an evaluation function established by taking the target adjustable capability and the preset target response as inputs, and is used to measure the merits of the control schemes at each node. The calculation formula is as follows:

[0074] in, This represents the local control variable of the i-th synchronization resource node. This represents the control optimization function for the i-th synchronization resource node. The local cost function represents the actual output power of the i-th synchronization resource node. This represents the weight of the i-th synchronization resource node. This represents the target adjustability of the i-th synchronization resource node at time scale s. This represents the actual output power of the i-th synchronization resource node.

[0075] In this embodiment of the invention, the local control quantity is the control action or adjustment output calculated by the initial resource node under the guidance of the control optimization function, based on its own adjustability and constraints. This achieves the optimal response of the node under local conditions and provides an initial scheme for global coordination. Specifically, according to the control optimization function, the actual output power of the i-th synchronous resource node, the local cost function, and the target adjustability of the synchronous resource node at time scale s are first obtained. Then, these parameters are substituted into the control optimization function to solve for the local control variables, minimizing the weighted sum of the cost function term and the power deviation term. During the optimization process, by adjusting the balance between the economic efficiency and accuracy constraints of the local control quantity, the local control quantity that minimizes the control optimization function is finally obtained, achieving the optimal adjustment output of the synchronous resource node.

[0076] Furthermore, the neighbor control variable is the control output or adjustment action calculated by the communicating neighbor nodes during the same optimization process. It describes the real-time control status of the neighbor nodes and also implies the dynamic dependencies and cooperative relationships between nodes. The coordination control variable is the final control output obtained by combining the local control variable with the neighbor control variable during multiple rounds of iterative updates. It ensures that the control decisions of a single node are highly consistent with the global objective, thereby realizing the overall coordinated control of the virtual power plant.

[0077] In this embodiment of the invention, the global error is a comprehensive deviation index that compares the difference between the coordinated control quantity and the local control quantity across the entire network, reflecting whether the overall response is close to the preset target. The global error is a key parameter for triggering iterative adjustments and judging the convergence of optimization.

[0078] In this embodiment of the invention, the coordinated optimization control strategy is a unified control scheme generated by coordinated control quantities under the condition that the global error meets a preset threshold, providing specific operation instructions for each node at multiple time scales.

[0079] The above steps form a distributed iterative optimization closed loop. First, the local control quantity is calculated, then the neighbor control quantities are used for iterative correction, and finally, convergence is judged by the global error to generate a coordinated optimization control strategy. This process not only ensures local optimization of nodes, but also achieves global coordination through information exchange, making up for the latency and consistency problems of traditional centralized control in distributed resource scheduling.

[0080] Logically, relying solely on local control variables without considering the dynamic information of neighboring nodes may lead to inconsistent responses between nodes or the inability to achieve the global objective. By introducing iterative updates of neighboring control variables and global error assessment, this embodiment achieves dual protection of local optimization and global coordination, ensuring that the final coordinated optimization control strategy is both efficient and reliable, providing solid support for multi-timescale scheduling of virtual power plants.

[0081] This invention uses the target adjustable capabilities of each synchronous resource node and the overall preset response of the virtual power plant as inputs to perform collaborative optimization in a distributed architecture. This allows each node to optimize independently within its own capability constraints while simultaneously achieving collaboration and coordination between nodes through information exchange. This not only fully utilizes the adjustment potential of each node and improves local control efficiency but also ensures the consistency of response and global optimization effect of the entire virtual power plant across multiple time scales. Consequently, it achieves high precision, stability, and reliability in system scheduling, enhancing overall operational efficiency and control flexibility.

[0082] S6. The demand response of each synchronization resource node of the target virtual power plant is controlled using the coordination optimization control strategy to obtain the synchronization response control result.

[0083] In this embodiment of the invention, the generated coordinated optimization control strategy is distributed to each synchronous resource node within the virtual power plant. Each node adjusts its own output or load state according to the strategy instructions to respond to demand signals. During execution, the nodes continuously collect their own operating status and dynamically adjust their responses based on feedback to ensure that local capacity constraints are met while maintaining coordination with the control behavior of neighboring nodes.

[0084] In detail, the step of controlling the demand response of each synchronization resource node of the target virtual power plant using the coordinated optimization control strategy to obtain the synchronization response control result includes: Based on the target adjustability and the coordinated optimization control strategy, control instructions for each of the synchronization resource nodes are generated at multiple time scales. Obtain the demand response tasks of the target virtual power plant, and dynamically allocate the demand response tasks using the control commands to obtain the target response tasks for each of the synchronization resource nodes; The target response tasks are prioritized to obtain an ordered sequence of response tasks; The control commands are sent to the corresponding synchronization resource nodes according to the ordered response task sequence; Based on the received control instructions, the synchronous resource nodes of the target virtual power plant are tasked with executing tasks to obtain the synchronization response control results between multiple nodes.

[0085] In this embodiment of the invention, the multi-timescale control instructions refer to the specific operational instructions generated for each synchronization resource node at different timescales based on the target adjustability and coordinated optimization control strategy. These control instructions not only include adjustment schemes for nodes at various time levels such as day-ahead, intraday, and real-time, but also reflect node capability limitations, response priorities, and global coordination requirements.

[0086] In this embodiment of the invention, the demand response task is an energy regulation or load control task allocated by the virtual power plant according to the overall operating goals, market instructions, or scheduling requirements. It reflects the target response amount of the system in different time periods and also includes constraints, response windows, and adjustable point ranges. Furthermore, the target response task is the specific execution task ultimately received by each synchronous resource node after the overall demand response task of the virtual power plant is allocated through control instructions and node capabilities. This transforms the global response goal into specific tasks operable by each node, enabling multi-node collaboration.

[0087] In this embodiment of the invention, the ordered response task sequence is a sequence of target response tasks of each node ordered according to node capabilities, task priority and response urgency. It reflects the execution priority of tasks, and also considers inter-node coordination constraints, resource utilization efficiency and response timeliness. It guides control commands to be sent and executed in an optimized order, ensuring the coordination and timing consistency of the overall system response.

[0088] In this embodiment of the invention, the synchronous response control result between multiple nodes is a coordinated and consistent control effect formed on the global time axis after each synchronous resource node executes the target response task according to the received control instructions. It reflects the response status of a single node and embodies the collaborative operation status of the entire virtual power plant. Through the synchronous execution of multiple nodes, the virtual power plant as a whole achieves high-precision and reliable response to demand requirements at multiple time scales.

[0089] Logically, relying solely on control command generation without task allocation and priority sorting may lead to node execution conflicts or response delays. Through dynamic allocation, priority sorting, and synchronous execution, this embodiment achieves an organic combination of global goals and local capabilities, ensuring time consistency of multi-node responses, task coordination, and overall system optimization performance.

[0090] This invention distributes coordinated optimization control strategies to each synchronous resource node, enabling the node to dynamically adjust its demand response tasks within its own capabilities. Simultaneously, during execution, global coordination is maintained through information exchange between nodes, fully leveraging the adjustability of each node, improving local response efficiency, and ensuring the overall response synchronization and consistency of the virtual power plant across multiple time scales. This enhances the system's scheduling accuracy, operational stability, and control reliability, achieving efficient and collaborative demand response management.

[0091] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0092] like Figure 2 The diagram shown is a functional block diagram of a virtual power plant multi-timescale adaptive demand response control system provided in an embodiment of the present invention.

[0093] This disclosure provides a virtual power plant multi-timescale adaptive demand response control system, which corresponds one-to-one with the virtual power plant multi-timescale adaptive demand response control method described in the above embodiments. For example... Figure 2 As shown, the virtual power plant multi-timescale adaptive demand response control system 100 includes a feature embedding vector construction module 101, a cross-scale similarity analysis module 102, a clock synchronization and delay compensation module 103, a multi-timescale analysis module 104, a distributed collaborative optimization module 105, and a demand response control module 106. Detailed descriptions of each functional module are as follows: The feature embedding vector construction module 101 is used to collect resource operation data of the initial resource nodes in the target virtual power plant and construct time feature embedding vectors based on the resource operation data at different time scales. The cross-scale similarity analysis module 102 is used to perform cross-scale similarity analysis on the time feature embedding vector to obtain the correlation strength between each time scale. The clock synchronization and delay compensation module 103 is used to perform distributed clock synchronization and delay compensation on the local clock stamp of the initial resource node to obtain a synchronized resource node. The multi-timescale analysis module 104 is used to perform multi-timescale analysis on the adjustability of the synchronization resource node based on the time feature embedding vector, and to adaptively update the multi-timescale analysis results based on the correlation strength to obtain the target adjustability. The distributed collaborative optimization module 105 is used to perform distributed collaborative optimization on the synchronization resource nodes according to the target adjustable capability, and generate a coordinated optimization control strategy. The demand response control module 106 is used to control the demand response of each of the synchronization resource nodes of the target virtual power plant using the coordination optimization control strategy, and obtain the synchronization response control result.

[0094] In one embodiment, the feature embedding vector construction module 101 constructs time feature embedding vectors based on the resource operation data at different time scales, for the following purposes: Obtain day-ahead, intraday, and real-time scales, and then stratify the resource operation data according to the day-ahead, intraday, and real-time scales respectively to obtain day-ahead scale operation data, intraday scale operation data, and real-time scale operation data. The day-ahead operational data, the intraday operational data, and the real-time operational data are denoised to obtain day-ahead denoised data, intraday denoised data, and real-time denoised data. The day-ahead scale denoised data, the intraday scale denoised data, and the real-time scale denoised data are aggregated into a time-scale denoised dataset. The time-scale denoised dataset is normalized to obtain a normalized dataset; Perform temporal convolution on the normalized dataset to obtain the convolutional feature dataset; Temporal attention-weighted encoding is performed on the convolutional feature dataset to obtain a global temporal feature representation; Variational coding is performed on the global time feature representation, and the generated mean vector is used as the time feature embedding vector.

[0095] In one embodiment, the cross-scale similarity analysis module 102 performs cross-scale similarity analysis on the temporal feature embedding vector to obtain the correlation strength between each time scale, for the following purposes: Calculate the Euclidean distance between the time feature embedding vectors of any two time scales; Based on the Euclidean distance, generate cross-scale similarity values ​​for the temporal feature embedding vectors corresponding to the two time scales; A cross-scale similarity matrix is ​​constructed using the cross-scale similarity values; The cross-scale similarity matrix is ​​normalized to obtain a normalized similarity matrix; The correlation strength at each time scale is calculated based on the normalized similarity matrix.

[0096] In one embodiment, the clock synchronization and delay compensation module 103 performs distributed clock synchronization and delay compensation on the local clock stamp of the initial resource node to obtain a synchronized resource node, for the following purposes: Distributed clock synchronization is performed on the local clock stamp of the initial resource node to obtain the local updated clock stamp; The communication delay between the initial resource node and the communication neighbor node is detected based on the local update clock stamp to obtain a delay estimate. When the estimated delay value is greater than the preset delay threshold, the time alignment error between the initial resource node and the communication neighbor node corresponding to the estimated delay value is corrected to obtain the corrected resource node and the corrected neighbor node. The global clock drift is calculated based on the clock stamps of the corrected resource node and the corrected neighbor node, and it is determined whether the global clock drift is less than a preset drift threshold. When the global clock drift is greater than or equal to the drift threshold, the corrected resource node and the corrected neighbor node are recalibrated to obtain a synchronized resource node; When the global clock drift is less than the drift threshold, the corrected resource node is used as the synchronization resource node.

[0097] In one embodiment, the clock synchronization and delay compensation module 103 performs distributed clock synchronization on the local clock stamp of the initial resource node to obtain a locally updated clock stamp, which is used for: Obtain the communication neighbor nodes of each initial resource node, as well as the local clock stamp of the initial resource node and the neighbor clock stamp of the communication neighbor nodes; Broadcast the local clock stamp to the communication neighbor node, and receive the neighbor clock stamp from the communication neighbor node; Based on the received clock information of the neighbor clock stamp, the local clock stamp of the initial resource node is synchronized and updated to obtain the local updated clock stamp.

[0098] In one embodiment, the multi-timescale analysis module 104 performs multi-timescale analysis of the adjustability of the synchronization resource node based on the time feature embedding vector, for the following purposes: Obtain the resource state variables of the synchronous resource node after delay compensation, and use the time feature embedding vector and the resource state variables to update the preset nonlinear capability analysis model to obtain the autoregressive capability analysis model. The autoregressive capability analysis model is used to calculate the adjustable capability analysis value of the synchronization resource node at each time scale; Obtain the actual call feedback data of the synchronization resource node, and generate the actual value of the adjustable capability based on the actual call feedback data; The adjustable capability deviation for each time scale is calculated based on the difference between the adjustable capability analysis value and the actual adjustable capability value.

[0099] In one embodiment, the multi-timescale analysis module 104, when performing adaptive updates of the multi-timescale analysis results based on the correlation strength to obtain target adjustability, is used for: When the adjustable capability deviation is greater than a preset deviation threshold, the adjustable capability analysis value is weighted and fused according to the correlation strength to obtain a cross-scale weighted correction value for each time scale. Adaptive fusion of all the aforementioned cross-scale weighted correction values ​​yields the target adjustable capability.

[0100] In one embodiment, the distributed collaborative optimization module 105 performs distributed collaborative optimization on the synchronization resource nodes based on the target adjustable capability, generating a coordinated optimization control strategy for: A control optimization function is constructed based on the target adjustable capability and the preset local cost function; The local control quantity of the synchronization resource node is calculated using the control optimization function. Obtain the neighbor control quantity of the communication neighbor node corresponding to the synchronization resource node, and use the neighbor control quantity to iteratively update the local control quantity to obtain the coordination control quantity; The difference between the coordinated control quantity and the local control quantity is taken as the global error; When the global error is less than a preset error threshold, a coordinated optimization control strategy is generated based on the coordinated control quantity.

[0101] In one embodiment, the demand response control module 106 executes the demand response control of each of the synchronization resource nodes of the target virtual power plant using the coordinated optimization control strategy, and obtains the synchronization response control result, for the following purposes: Based on the target adjustability and the coordinated optimization control strategy, control instructions for each of the synchronization resource nodes are generated at multiple time scales. Obtain the demand response tasks of the target virtual power plant, and dynamically allocate the demand response tasks using the control commands to obtain the target response tasks for each of the synchronization resource nodes; The target response tasks are prioritized to obtain an ordered sequence of response tasks; The control commands are sent to the corresponding synchronization resource nodes according to the ordered response task sequence; Based on the received control instructions, the synchronous resource nodes of the target virtual power plant are tasked with executing tasks to obtain the synchronization response control results between multiple nodes.

[0102] In this invention, the specific limitations of a virtual power plant multi-timescale adaptive demand response control system can be found in the above-described limitations of a virtual power plant multi-timescale adaptive demand response control method, and will not be repeated here. Each module in the aforementioned virtual power plant multi-timescale adaptive demand response control system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0103] In the embodiments provided by this invention, it should be understood that the disclosed system can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0104] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0105] Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be embraced within the invention. No appended diagram markings in the claims should be construed as limiting the scope of the claims.

[0106] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0107] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0108] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the system can be divided into different functional units or modules to complete all or part of the functions described above.

[0109] In the embodiments provided in this disclosure, it should be understood that the disclosed systems and methods can also be implemented in other ways. The system embodiments described above are merely illustrative; for example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0110] It should be noted that, in this disclosure, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element limited by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0111] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. 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 of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A multi-timescale adaptive demand response control method for a virtual power plant, characterized in that, The method includes: Collect operational data of initial resource nodes within the target virtual power plant and construct a time feature embedding vector; Analyze the cross-scale correlation strength of the aforementioned temporal feature embedding vectors; Distributed clock synchronization and delay compensation are performed on each resource node to obtain synchronized resource nodes; Based on the aforementioned vectors and correlation strengths, the multi-timescale adjustable capability of nodes is analyzed and adaptively updated; including: Obtain the resource status variables after latency compensation for the synchronized resource nodes; By combining the time feature embedding vector and the resource state variables to update the nonlinear capability analysis model, an autoregressive capability analysis model is obtained. The adjustable capability analysis values ​​at each time scale are calculated using the autoregressive capability analysis model. Generate actual values ​​of adjustable capabilities based on actual call feedback data; The adjustable capacity deviation at each time scale is calculated based on the adjustable capacity analysis value and the actual adjustable capacity value. When the adjustable capability deviation is greater than a preset deviation threshold, the adjustable capability analysis value is weighted and fused according to the correlation strength to obtain the cross-scale weighted correction value for each time scale. Adaptive fusion of all the aforementioned cross-scale weighted correction values ​​yields the target's adjustable capability. Based on the updated adjustable capabilities, distributed collaborative optimization is performed to generate a coordinated optimization control strategy. The aforementioned strategy is used to control the demand response of nodes.

2. The virtual power plant multi-timescale adaptive demand response control method as described in claim 1, characterized in that, The process of collecting operational data from initial resource nodes within the target virtual power plant and constructing a time feature embedding vector includes: Acquire operational data of resource nodes across multiple time scales; The runtime data is subjected to hierarchical denoising and normalization processing to obtain a standardized dataset; Temporal features are extracted from the standardized dataset to obtain a convolutional feature dataset; Attention-weighted and variational coding are applied to the convolutional feature dataset to generate the temporal feature embedding vector.

3. The virtual power plant multi-timescale adaptive demand response control method as described in claim 1, characterized in that, The analysis of the cross-scale correlation strength of the time feature embedding vector includes: Calculate the Euclidean distance between the temporal feature embedding vectors corresponding to each time scale; Cross-scale similarity values ​​are generated based on the Euclidean distance, and a cross-scale similarity matrix is ​​constructed; the matrix is ​​normalized; and the correlation strength at each time scale is calculated based on the normalized matrix data.

4. The virtual power plant multi-timescale adaptive demand response control method as described in claim 1, characterized in that, The process of performing distributed clock synchronization and delay compensation on each resource node to obtain synchronized resource nodes includes: Distributed clock synchronization is performed on the initial resource nodes to obtain a local updated clock stamp; communication delay is estimated based on the local updated clock stamp; when the delay estimate exceeds a preset threshold, time alignment error correction between nodes is performed; Calculate the global clock drift of the corrected node, and determine whether to perform recalibration based on the comparison result with the preset threshold. Finally, output the synchronized resource node.

5. The virtual power plant multi-timescale adaptive demand response control method as described in claim 4, characterized in that, The step of performing distributed clock synchronization on the initial resource nodes to obtain the locally updated clock stamp includes: Determine the communication neighbor nodes of each initial resource node; Broadcast the local clock stamp to the communication neighbor node and receive the corresponding neighbor clock stamp; Based on the clock information received from the neighbor's clock stamp, the local clock stamp is updated to complete clock synchronization.

6. The virtual power plant multi-timescale adaptive demand response control method as described in claim 1, characterized in that, The updated adjustable capabilities are subjected to distributed collaborative optimization to generate a coordinated optimization control strategy, including: A control optimization function is constructed based on the target adjustability and the local cost function to calculate the local control quantity of the synchronization resource node; The local control quantity is iteratively updated based on the control quantity of neighboring nodes to obtain the coordinated control quantity; Calculate the global error based on the coordinated control quantity and the local control quantity; When the global error is less than a preset threshold, a coordinated optimization control strategy is generated based on the coordinated control quantity.

7. The virtual power plant multi-timescale adaptive demand response control method as described in claim 1, characterized in that, The strategy described above is used to control the demand response of nodes, including: Based on the target adjustability and coordination optimization control strategy, multi-timescale control instructions are generated for each synchronous resource node; demand response tasks are dynamically allocated based on the control instructions to obtain the target response tasks for each node; the target response tasks are prioritized to generate an ordered response task sequence; control instructions are sent to the corresponding nodes according to the ordered response task sequence; the control instructions are executed to realize multi-node synchronous response control.

8. A virtual power plant multi-timescale adaptive demand response control system, characterized in that, The system includes: The feature embedding vector construction module is used to collect the operating data of the initial resource nodes in the target virtual power plant and construct the time feature embedding vector. A cross-scale similarity analysis module is used to analyze the cross-scale correlation strength of the time feature embedding vector; The clock synchronization and delay compensation module is used to perform distributed clock synchronization and delay compensation on each resource node to obtain synchronized resource nodes. The multi-timescale analysis module is used to analyze and adaptively update the multi-timescale adjustability of nodes based on the vectors and correlation strengths; it includes: Obtain the resource status variables after latency compensation for the synchronized resource nodes; By combining the time feature embedding vector and the resource state variables to update the nonlinear capability analysis model, an autoregressive capability analysis model is obtained. The adjustable capability analysis values ​​at each time scale are calculated using the autoregressive capability analysis model. Generate actual values ​​of adjustable capabilities based on actual call feedback data; The adjustable capacity deviation at each time scale is calculated based on the adjustable capacity analysis value and the actual adjustable capacity value. When the adjustable capability deviation is greater than a preset deviation threshold, the adjustable capability analysis value is weighted and fused according to the correlation strength to obtain the cross-scale weighted correction value for each time scale. Adaptive fusion of all the aforementioned cross-scale weighted correction values ​​yields the target's adjustable capability. The distributed collaborative optimization module is used to perform distributed collaborative optimization based on the updated adjustable capabilities and generate a coordinated optimization control strategy. The demand response control module is used to control the demand response of nodes using the aforementioned strategy.

Images