A method for characterizing the flexible resource credibility and responsiveness of building complexes

By employing intelligent measurement data feature dimensionality reduction, differential privacy federated learning, and CPSS architecture, the problems of low physical topology identification efficiency and data silos in flexible resource regulation of building clusters are solved, enabling precise regulation and efficient scheduling of flexible resources.

CN123117612APending Publication Date: 2026-11-17SICHUAN UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611237315.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-14
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing technologies for flexible resource management in building complexes suffer from low efficiency in physical topology identification, data silos, low accuracy in fitting response behavior, and insufficient multidimensional feature characterization, resulting in poor performance in flexible resource management.

Method used

By employing intelligent measurement data feature dimensionality reduction and similarity calculation, combined with differential privacy federated learning and CPSS architecture, a distributed interaction framework is constructed to extract physical, information and social domain features. Combined with a multi-timescale system dynamics model, a dynamic characterization set of the flexible resource response capability of building complexes is generated.

Benefits of technology

It enables precise mapping of flexible resources within building complexes while protecting user privacy, improving control efficiency and response accuracy, and providing scientific scheduling decision support.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123117612A_ABST
    Figure CN123117612A_ABST
Patent Text Reader

Abstract

This invention relates to the field of integrated energy demand response technology and discloses a method for reliable cognition and responsiveness characterization of flexible resources in building clusters. The method includes: acquiring voltage time series data; performing feature dimensionality reduction and similarity calculation on the voltage time series data to obtain physical space topology verification results for flexible resources; constructing a distributed aggregator-user feature information interaction framework; training user electricity price response models locally on each aggregator; integrating the CPSS architecture; and extracting physical domain operational boundary features, information domain communication delay features, and social domain psychological preference features for flexible resources based on the physical space topology verification results and response behavior baseline models; and generating a dynamic characterization set of flexible resource responsiveness for building clusters covering time, space, capacity, and probability dimensions. The advantages of this invention are that it achieves highly reliable, multi-dimensional, and accurate assessment of adjustable potential, providing reliable support for improving the resilience of urban power grids and the efficient aggregation and scheduling of flexible resources.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of integrated energy demand response technology, and in particular to a method for characterizing the flexible resource reliability perception and response capabilities of building complexes. Background Technology

[0002] Faced with the dual challenges of the accelerated construction of new power systems and the frequent occurrence of extreme weather events such as high temperatures, urban power grids often encounter severe supply-demand imbalances. Against this backdrop, fully leveraging the regulatory potential of the vast and flexible resources within building complexes, such as air conditioners, electric vehicles, and energy storage, has become a key means of alleviating power grid supply-demand imbalances and enhancing the resilience of urban power grids.

[0003] The efficient allocation of flexible resources relies heavily on accurate mapping of their physical locations and a comprehensive understanding of response patterns. However, existing technologies face numerous bottlenecks in practical applications. On one hand, the identification of traditional physical topologies (such as household-transformer relationships) largely depends on manual investigation, which is inefficient and prone to errors. This makes it difficult for aggregators to accurately map resources to distribution nodes, compromising the effectiveness of control commands. On the other hand, user electricity price response data is scattered across various aggregators, creating severe data silos due to privacy concerns and business competition. Traditional centralized models cannot obtain complete global data, resulting in low accuracy in fitting response behavior and failing to reflect the true group response patterns of users in a region.

[0004] Furthermore, existing flexible resource assessment models suffer from significant deficiencies in characterizing multidimensional features. Most existing methods are based on physical device parameters, lacking a fusion of "social-physical" characterization within the CPSS (cyber-physical-social system) framework. This neglects social domain characteristics such as user comfort preferences, bounded rationality in decision-making, and response fatigue, leading to a significant discrepancy between theoretically calculated capacity and actual available capacity. Simultaneously, existing research often isolates the time dimension, ignoring the evolutionary coupling of users' long-term energy-saving investments, medium-term contract signings, and short-term electricity consumption habits across multiple cycles. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method for characterizing the flexible resource credibility and responsiveness of building complexes.

[0006] The objective of this invention is achieved through the following technical solution: a method for characterizing the credible cognition and response capabilities of flexible resources in a building complex, the method comprising the following steps:

[0007] S1. Based on intelligent measurement data, obtain the voltage time series of users in the building cluster transformer area. Use the symbolic aggregation approximation method and clustering algorithm to perform feature dimensionality reduction and similarity calculation on the voltage time series to obtain the physical space topology verification results of flexible resources.

[0008] S2. Construct a distributed framework for interaction between aggregators and user feature information. Adopt a federated learning algorithm with differential privacy to train user electricity price response models locally on each aggregator. Then, form a regional user response behavior baseline model through parameter secure interaction and aggregation.

[0009] S3 integrates the CPSS architecture, based on the physical space topology verification results and response behavior baseline model, extracts the physical domain operation boundary features and information domain communication delay features of flexible resources, and introduces the user willingness function to extract the psychological preference features of the social domain.

[0010] S4. Based on the extracted physical domain operation boundary features, information domain communication delay features, and social domain psychological preference features, combined with a multi-timescale system dynamics model, a dynamic characterization set of flexible resource response capabilities of building complexes covering time, space, capacity, and probability dimensions is generated.

[0011] Specifically, the feature dimensionality reduction and similarity calculation for the voltage time series are as follows:

[0012] ;

[0013] In the formula, Original voltage time series Length; The length of the sequence after dimensionality reduction; For the first in the original sequence One data point; For the first segment after aggregation The average value of each interval.

[0014] Specifically, the federated learning algorithm incorporating differential privacy is used to train the user electricity price response model locally at each aggregator. Its objective function is expressed as:

[0015] ;

[0016] In the formula, The objective function is... Number the aggregator; For aggregators Mean squared error loss function trained on local data; For aggregators Local model parameters; These are the parameters for the global integrated model issued by the distribution network operator. This is the regularization penalty parameter;

[0017] The gradient update parameters obtained from training the local user electricity price response model are pruned and Gaussian noise is added, specifically as follows:

[0018] ;

[0019] In the formula, The gradient of the original model; Set the model gradient sensitivity clipping threshold; With a mean of 0 and a variance of Gaussian noise; It is the identity matrix; This represents the gradient update amount for the final encrypted upload.

[0020] Specifically, the process of introducing a user willingness function to extract psychological preference features in the social domain includes:

[0021] The perceived strength of the incentive price is calculated using a quantification function:

[0022] ;

[0023] In the formula, The perceived intensity of the incentive coefficient for users; For perception coefficient; This refers to the actual excitation coefficient signal issued.

[0024] Calculate the overall response probability:

[0025] ;

[0026] In the formula, The combined response probability of economic and psychological factors; The probability of a normally distributed response influenced by economic incentives; The probability of a response influenced by psychological factors; and These are the weighting coefficients for economic factors and psychological factors, respectively.

[0027] Specifically, the generation of a dynamic characterization set of the flexible resource response capabilities of building complexes, covering time, space, capacity, and probability dimensions, includes:

[0028] Calculate the static physical capacity boundary:

[0029] ;

[0030] In the formula, for The theoretical maximum response capacity of the entire building complex during the time period; The total number of flexible resource devices participating in the response; For the first The device is The power of the load that can be moved during a given time period; For the first The device is Reduced load power during certain time periods;

[0031] Calculate the actual dynamic capacity of intention:

[0032] ;

[0033] In the formula, This represents the final, actual usable dynamic response capacity. For the overall response probability; This is the response loss coefficient caused by communication and control delays in the information domain.

[0034] Specifically, the system dynamics model uses stock and flow feedback equations to update the response potential in the medium to long term, expressed as:

[0035] ;

[0036] In the formula, for The total flexible resource response potential of the time system; This represents the initial potential reserve of the response. Add a rate variable to the potential; This represents the potential decay rate variable.

[0037] Specifically, the potential increase rate variable is expressed as:

[0038] ;

[0039] In the formula, For a moment The normalized policy-positive incentive subsidy index; For a moment The equivalent response capacity base calculated from the newly added flexible equipment investment in the region; and The system input conversion weight coefficients are respectively the subsidy index and the investment base;

[0040] The potential decay rate variable is expressed as:

[0041] ;

[0042] In the formula, The overall aging and depreciation rate of flexible resource equipment; The cumulative index of users' subjective response fatigue; and These are the capacity decay sensitivity coefficients corresponding to the fatigue index and aging depreciation rate, respectively.

[0043] Specifically, the subjective response fatigue accumulation index is cumulatively corrected based on the actual response behavior under the day-to-day / real-time short timescale:

[0044] ;

[0045] In the formula, Forgetting factors related to user fatigue; The calculation period between adjacent macroscopic time nodes; This represents the actual usable dynamic response capacity over a short timescale. This represents the maximum adjustable potential of static theory over a short timescale.

[0046] Between adjacent macroscopic time nodes, using the response potential stock of the previous macroscopic time node as the initial value, the difference between the potential increase rate variable and the potential decay rate variable within the macroscopic calculation period is accumulated over time to obtain the response potential stock of the current macroscopic time node, and the response potential stock of the current macroscopic time node is used as the result of medium- and long-term potential evolution.

[0047] The medium- and long-term potential evolution results calculated through the stock and flow feedback equations are used as macro-boundary constraints. The actual dynamic willingness capacity at the day-to-day and real-time short time scales is dynamically input and corrected, thereby forming a multi-time-scale closed-loop linkage evolution mechanism in which the medium- and long-term potential stock feedback adjusts the short-term willingness capacity and the short-term actual response behavior accumulates to correct the medium- and long-term fatigue.

[0048] The present invention has the following advantages:

[0049] 1. This invention constructs a differential privacy-fed near-end learning framework, which, under the premise of strictly protecting the privacy of underlying user data of multiple aggregators and breaking down data silos, achieves global secure learning and highly reliable cognition of massive user interaction behaviors in a region; at the same time, it uses symbolic aggregation approximation and improved clustering algorithms to automatically verify the underlying physical and electrical topology, thus getting rid of the inefficiency and error of traditional manual inspection.

[0050] 2. This invention deeply integrates the CPSS perspective, not only accurately quantifying the reduction in actual response probability due to psychological factors such as bounded rationality and loss aversion of users in the social domain through Weber-Fechner's law, but also coupling long-term equipment investment replacement, medium-term contract default fatigue, and short-term real-time willingness through a closed-loop system dynamics model, achieving spatiotemporal dynamic evolution analysis of flexible resources across multiple time scales. The implementation of this method effectively overcomes the shortcomings of traditional assessments that rely solely on physical equipment parameters and blindly overestimate adjustable potential, restoring a more realistic and usable dynamic response capacity that better reflects real-world business scenarios under extreme events or tight balance conditions. This provides scientific and solid decision support for improving the resilience of urban power grids and the precise aggregation and scheduling of flexible resources. Attached Figure Description

[0051] Figure 1 This is a schematic diagram of the characterization method of the present invention;

[0052] Figure 2 A diagram illustrating the principles of physical space topology cognition;

[0053] Figure 3 Diagram of the differential privacy-federated learning interaction framework;

[0054] Figure 4 A structural diagram of CPSS feature decoupling and psychological mapping;

[0055] Figure 5 This is a diagram showing the dynamic evolution of a multi-timescale system. Detailed Implementation

[0056] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only for explaining the invention and are not intended to limit the invention; that is, the described embodiments are merely some embodiments of the invention, and not all embodiments. The components of the embodiments of the invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0057] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0058] It should be noted that relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, 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 defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0059] The present invention will be further described below with reference to the accompanying drawings, but the scope of protection of the present invention is not limited to the following description.

[0060] like Figures 1 to 5 As shown, a method for characterizing the flexible resource trustworthiness and responsiveness of a building complex includes the following steps:

[0061] S1. Based on intelligent measurement data, obtain the voltage time series of users in the building cluster transformer area. Use the symbolic aggregation approximation method and clustering algorithm to perform feature dimensionality reduction and similarity calculation on the voltage time series to obtain the physical space topology verification results of flexible resources.

[0062] This embodiment specifically includes:

[0063] Voltage data from users is continuously collected through the intelligent metering system (AMI) in the building complex's transformer substations to construct a voltage time series. To eliminate differences in user units and electricity consumption benchmarks, the original voltage time series is Z-score standardized. The specific formula is as follows:

[0064] ;

[0065] In the formula, These are the standardized voltage sequence data points; and Original sequences The mean and standard deviation;

[0066] After processing, a standardized voltage sequence is obtained. .

[0067] In this embodiment, to reduce the computational complexity caused by high-dimensional voltage time series, the piecewise aggregation approximation (PAA) method is used to normalize the sequence. Perform dimensionality reduction. Reduce the length to... The sequence is divided into equal parts. Each fragment contains [number] segments. There are 10 data points. Calculate the average value of the data within each segment, and use this average value as the representative point for that segment. The specific calculation is as follows:

[0068] ;

[0069] In the formula, Original voltage time series Length; The length of the sequence after dimensionality reduction; For the first in the original sequence One data point; For the first segment after aggregation The average value of each interval. This transforms the original long sequence into a sequence of length [missing value]. Dimensionality reduction sequence This significantly reduces the dimensionality of the data while preserving the core morphological features of the sequence.

[0070] This embodiment, based on PAA dimensionality reduction, further introduces the Symbolic Aggregation Approximation (SAX) method to convert continuous numerical sequences into discrete string sequences, effectively eliminating redundant information. Based on the characteristics of the standard normal distribution, a set of breakpoint sequences is determined. The amplitude space of the vertical axis is divided into A region of equal probability. Based on the dimensionality reduction sequence. Each element The range of amplitudes that the data falls into is mapped to corresponding symbols in a predefined alphabet set. Through this discretization process, complex numerical sequences are transformed into symbol sequences that are easy to compute and compare.

[0071] After completing the symbolization conversion, it is necessary to quantify the similarity between different user sequences. This embodiment calculates the similarity between string sequences based on the edit distance metric. By comparing the minimum cost required to achieve consistency in the symbol sequences corresponding to two users through insertion, deletion, or replacement operations, the similarity index between sequences is calculated. Based on this, all flexible resource users participating in demand response within the building complex are traversed, and the similarity is calculated pairwise, thereby constructing a similarity matrix to characterize the global user electrical distance relationships.

[0072] Finally, the constructed similarity matrix is ​​input into an improved K-means clustering algorithm for unsupervised clustering analysis. The algorithm clusters users with high similarity (i.e., users with highly consistent voltage fluctuation characteristics) into the same cluster, with each cluster corresponding to a specific distribution transformer or power supply phase. The final output clustering result represents the physical spatial connectivity of flexible resources within the building complex. This data-driven approach breaks through the efficiency bottleneck of traditional manual inspection, automatically verifies and corrects errors and omissions in the underlying physical topology, and achieves the goal of reliable understanding of the physical domain of flexible resources, laying the foundation for subsequent efficient aggregation and scheduling.

[0073] S2. Construct a distributed framework for interaction between aggregators and user feature information. Adopt a federated learning algorithm with differential privacy to train user electricity price response models locally on each aggregator. Then, form a regional user response behavior baseline model through parameter secure interaction and aggregation.

[0074] This step specifically includes:

[0075] A feature information interaction architecture is constructed, consisting of an upper-level distribution network operator and multiple lower-level distributed building cluster flexible resource aggregators. Each aggregator manages only the local historical energy consumption data and electricity price response records of users within its jurisdiction, and conducts distributed collaborative training without sharing the original energy consumption data. (This is in the context of the global...) During each iteration, the distribution network operator distributes global integrated model parameters to each aggregator. ;

[0076] Each aggregator receives the global integrated model parameters Subsequently, the user electricity price response model was trained using local privacy data. To address the heterogeneity of user response behavior and the non-independent and identically distributed nature of data among aggregators in different regions, the local network model was used in the first... In the training objective function during each iteration, a distance penalty with the global integrated model is introduced as a proximal term to limit the excessive difference between the local model and the global model. The specific formula for its local optimization objective function is as follows:

[0077] ;

[0078] In the formula, The objective function is... Number the aggregator; For aggregators Mean squared error loss function trained on local data; For aggregators Local model parameters; These are the parameters for the global integrated model issued by the distribution network operator. This is a regularization penalty parameter used to limit the difference between the local model and the global model from being too large.

[0079] After local model training is complete, each aggregator performs a differential privacy mechanism before uploading parameters to prevent privacy data from being leaked inversely through gradients. Specifically, the original gradients obtained from local model training are pruned to control the maximum impact of individual data on the overall gradient, thereby constraining the sensitivity of the model's gradients. Subsequently, Gaussian noise is added to the pruned gradients to achieve differential privacy protection, specifically:

[0080] ;

[0081] In the formula, The gradient of the original model; Set the model gradient sensitivity clipping threshold; With a mean of 0 and a variance of Gaussian noise; It is the identity matrix; This represents the gradient update amount for the final encrypted upload.

[0082] Finally, each aggregator will update the gradient using differential privacy encryption. The data is uploaded to the distribution network operator's server. Upon receiving the local gradient parameters, the server performs secure interaction and collaborative aggregation, updating and forming a global comprehensive model. This process iterates through multiple rounds until the model converges, ultimately outputting a baseline model of regional user response behavior. This process, without sharing raw energy consumption data and breaking down data privacy barriers, eliminates the data silo effect among multiple aggregators, achieving a comprehensive and reliable understanding of user interaction behavior within these data silos.

[0083] S3 integrates the CPSS architecture, based on the physical space topology verification results and response behavior baseline model, extracts the physical domain operation boundary features and information domain communication delay features of flexible resources, and introduces the user willingness function to extract the psychological preference features of the social domain.

[0084] In this embodiment, S3 specifically refers to:

[0085] The Converged Information-Physical-Social System (CPSS) architecture decouples the response characteristics of flexible resources in multiple dimensions. Combining the physical space topology of the distribution area verified by S1, the physical domain operational boundary characteristics of flexible resources are extracted. Specifically, based on the physical ledgers and rated parameters of various flexible resources (such as air conditioning loads, electric vehicles, and distributed energy storage), the real-time operational boundaries of each device in the physical space are defined, including maximum scalable power, load shifting periods, energy storage charging and discharging power limits, and safe operating range of state of charge (SOC). This constructs the theoretical maximum adjustable potential boundary at the underlying physical device level, ensuring physical security during the execution of response scheduling commands.

[0086] Based on the distributed feature information interaction framework constructed by S2, information domain communication delay features are extracted. By quantifying the network communication time and control calculation delay generated during the multi-round model parameter secure interaction aggregation, differential privacy noise addition, and final demand response control command issuance between distribution network operators and distributed aggregators, and between aggregators and various terminal users in the distribution area, the response deviation caused by information interaction lag is decoupled and transformed into a response loss coefficient caused by information domain communication and control delay, which is used to correct the loss of theoretical physical boundary in actual communication environment;

[0087] Based on the regional user response behavior baseline model established in S2, a user willingness function is introduced to extract user psychological preference features in the social domain, thereby eliminating response bias caused by the neglect of user subjective willingness in traditional purely physical models. Specifically, the extraction of social domain features involves calculating the perceived intensity of user's incentive price using a quantification function and calculating the comprehensive response probability, as follows:

[0088] Using a perceptual intensity quantification formula based on the Weber-Fechner law, the user's psychological sensitivity to power grid excitation signals is calculated:

[0089] ;

[0090] In the formula, The perceived intensity of the incentive coefficient for users; For perception coefficient; This refers to the actual excitation coefficient signal issued.

[0091] By deeply integrating the bounded rationality theory of economics and the behavioral decision-making theory of psychology, and combining economic and psychological factors, a comprehensive response probability formula is constructed to calculate the comprehensive response probability:

[0092] ;

[0093] In the formula, The combined response probability of economic and psychological factors; The response probability, which follows a normal distribution and is influenced by economic incentives, represents the nominal response probability of a user under different incentive intensities. The response probability is influenced by psychological factors, and it is jointly determined by the user's loss aversion factor and the accumulated public opinion influence factor or fatigue generated from long-term participation in demand response. and These are the weighting coefficients for economic and psychological factors, respectively. Through the dynamic decoupling and organic coupling of the physical domain's operational boundary, the information domain's communication delay, and the social domain's psychological preferences, a highly reliable understanding of the flexible resource panoramic response characteristics of the building complex is achieved.

[0094] S4. Based on the extracted physical domain operation boundary features, information domain communication delay features, and social domain psychological preference features, combined with a multi-timescale system dynamics model, a dynamic characterization set of flexible resource response capabilities of building complexes covering time, space, capacity, and probability dimensions is generated.

[0095] In this embodiment, S4 specifically includes:

[0096] At short timescales such as day-to-day and real-time, the physical domain operational boundary features decoupled in step S3 are aggregated in terms of spatial and capacity dimensions. The static theoretical maximum adjustable potential of the entire building complex under physical boundary constraints within the time period is calculated. The static physical capacity boundary is calculated as follows:

[0097] ;

[0098] In the formula, for The theoretical maximum response capacity of the entire building complex during the time period; The total number of flexible resource devices participating in the response; For the first The device is The power of the load that can be moved during a given time period; For the first The device is Reduced load power during certain time periods;

[0099] To eliminate the disconnect between traditional purely physical capacity boundaries and actual user response behavior, the aforementioned static physical capacity boundary is deeply coupled with the comprehensive response probability output by S3, which includes social and subjective psychological factors, and the control delay loss generated by the information domain. This dynamically characterizes the actual dynamic willingness capacity, taking into account social psychology and information delay factors. The formula for characterizing the actual dynamic willingness capacity is:

[0100] ;

[0101] In the formula, This represents the final, actual usable dynamic response capacity. For the overall response probability; This represents the response loss coefficient caused by communication and control delays in the information domain. It enables accurate dynamic assessment of the flexible resource capacity and probabilistic dimension of building clusters at the day-to-day and real-time scales.

[0102] Based on this, a system dynamics (SD) model with a medium- to long-term time scale is introduced to construct stock and flow feedback equations, which are used to dynamically update and evolve the total response potential of the building complex in the medium- to long-term macro dimension. The medium- to long-term response potential state update formula is as follows:

[0103] ;

[0104] In the formula, for The total flexible resource response potential of the time system; This represents the initial potential reserve of the response. Add a rate variable to the potential, which is jointly determined by the positive correlation function relationship between policy incentive subsidies, new investment in flexible equipment, etc. The potential decay rate variable is jointly determined by the positive correlation function relationship between the cumulative subjective response fatigue caused by users' long-term and frequent participation in the response and the equipment aging and depreciation rate.

[0105] To achieve a precise quantitative mapping between macroscopic environmental factors and underlying physical and social conditions, this embodiment adds a rate variable to the potential in the system dynamics model. and potential decay rate variable Multi-factor decoupling and refined modeling were performed.

[0106] Among them, the potential increase rate variable Primarily driven by positive incentives from the external environment, it is determined by a combination of quantitative policy subsidy intensity and the scale of new equipment investment. The specific calculation formula is as follows:

[0107] ;

[0108] In the formula: For a moment The normalized policy-positive incentive subsidy index; For a moment The equivalent response capacity base calculated from the investment in new flexible equipment (such as new energy storage facilities, electric vehicle charging piles, etc.) within the region; and The system input conversion weight coefficients corresponding to the subsidy index and investment base are respectively obtained by fitting and calibrating the regional historical long-term operation data.

[0109] The potential decay rate variable Primarily influenced by both internal physical equipment wear and tear and user subjective fatigue, it is determined by a combination of quantified equipment aging depreciation and subjective response fatigue. The specific calculation formula is as follows:

[0110] ;

[0111] In the formula: The overall aging depreciation rate of flexible resource equipment decreases non-linearly over time. The cumulative index of user subjective response fatigue represents the user's tendency to become bored or refuse due to long-term and frequent participation in demand response; and These are the capacity decay sensitivity coefficients corresponding to the fatigue index and aging depreciation rate, respectively.

[0112] 3. The dynamic closed loop between short-term behavior and medium- to long-term fatigue

[0113] To overcome the shortcomings of traditional static evaluation models, this invention constructs a model based on the cumulative correction of long-term fatigue based on short-term actual response behavior. The dynamic update mechanism has an update state equation expressed as:

[0114] ;

[0115] In the formula: The memory forgetting factor for user fatigue (value range is within) (between), representing the user's ability to recover and forget from past response fatigue over time; The calculation period between adjacent macroscopic time nodes; This represents the actual usable dynamic response capacity over a short timescale. This represents the maximum adjustable potential of static theory over a short timescale. (Ratio) Characterized the time period The actual response depth of the internal users.

[0116] In the specific calculations, the medium- to long-term evolution process is divided into successively adjacent macroscopic time nodes. For the current macroscopic time node... First, the user's subjective response fatigue level is updated using the actual response behavior of the previous macroeconomic cycle in the day-to-day and real-time phases. In conjunction with the equipment aging and depreciation rate Calculate the potential decay rate Meanwhile, based on the policy positive incentive subsidy index and the equivalent response capacity base of newly added flexible equipment The rate of increase in computing potential .

[0117] Based on this, by accumulating the difference between the potential increase rate and the potential decay rate over time, we obtain:

[0118] ;

[0119] Calculated This represents the medium- to long-term potential evolution of the current macroeconomic cycle. This result is mapped to the macroeconomic capacity boundaries for each day-to-real-time period based on the online status and availability of flexible resources in each short timeframe. Furthermore, the actual dynamic capacity intention, determined by the theoretical maximum response capacity, the overall response probability, and communication control losses at the short-term timescale, is subject to a limiting correction.

[0120] As one implementation method, the modified actual dynamic willingness capacity can be expressed as:

[0121] ;

[0122] In the formula: The actual dynamic willingness capacity in the $t$th period after the medium-to-long-term potential boundary adjustment; For the first The maximum adjustable potential of static theory over a given period; For the first The overall response probability over a given time period; This is the response loss coefficient caused by communication and control delays in the information domain; Macro time node The corresponding medium- and long-term potential evolution results; To transform medium- and long-term potential assets into the first The availability coefficient for short-term capacity mapping is determined by the online status, physical operating boundary, and availability ratio of flexible resources in the corresponding time period.

[0123] The actual response results formed on a short timescale are used to update the user's subjective response fatigue in the next macro cycle, thus forming a multi-timescale closed-loop linkage evolution mechanism of "medium- and long-term potential stock constraining short-term willingness capacity, and short-term actual response behavior feedback correcting medium- and long-term fatigue".

[0124] Finally, the system dynamics model uses the calculated medium- and long-term potential evolution results as macroscopic boundary constraints through the aforementioned integral equations. This is dynamically input into and corrects the day-to-day / real-time short-term dynamic capacity characterization, thus forming a multi-timescale closed-loop linkage evolution mechanism where "medium- and long-term potential stock feedback adjusts short-term capacity, and short-term actual response behavior cumulatively corrects medium- and long-term fatigue." Ultimately, the system outputs a complete dynamic characterization set of response capabilities covering time, space, capacity, and probability dimensions, providing distribution network operators or aggregators with highly reliable, spatiotemporally linked, and dynamically evolving demand response scheduling decision-making basis.

[0125] The above description is merely a preferred embodiment of the present invention and does not constitute any limitation on the present invention. Any person skilled in the art can make many possible variations and modifications to the technical solution of the present invention, or modify it into equivalent embodiments, without departing from the scope of the present invention. Therefore, any modifications, equivalent changes, and alterations made to the above embodiments based on the technology of the present invention without departing from the scope of the present invention are within the protection scope of the present invention.

Claims

1. A method for characterizing the credible cognition and response capabilities of flexible resources in a building complex, characterized by: The method includes the following steps: S1. Based on intelligent measurement data, obtain the voltage time series of users in the building cluster transformer area. Use the symbolic aggregation approximation method and clustering algorithm to perform feature dimensionality reduction and similarity calculation on the voltage time series to obtain the physical space topology verification results of flexible resources. S2. Construct a distributed framework for interaction between aggregators and user feature information. Adopt a federated learning algorithm with differential privacy to train user electricity price response models locally on each aggregator. Then, form a regional user response behavior baseline model through parameter secure interaction and aggregation. S3 integrates the CPSS architecture, based on the physical space topology verification results and response behavior baseline model, extracts the physical domain operation boundary features and information domain communication delay features of flexible resources, and introduces the user willingness function to extract the psychological preference features of the social domain. S4. Based on the extracted physical domain operation boundary features, information domain communication delay features, and social domain psychological preference features, combined with a multi-timescale system dynamics model, a dynamic characterization set of flexible resource response capabilities of building complexes covering time, space, capacity, and probability dimensions is generated.

2. The method for characterizing the flexible resource credibility and responsiveness of a building complex according to claim 1, characterized in that: Feature dimensionality reduction and similarity calculation for voltage time series are as follows: ; In the formula, Original voltage time series Length; The length of the sequence after dimensionality reduction; For the first in the original sequence One data point; For the first segment after aggregation The average value of each interval.

3. The method for characterizing the flexible resource credibility and responsiveness of a building complex according to claim 1, characterized in that: The federated learning algorithm, which incorporates differential privacy, trains the user electricity price response model locally at each aggregator. Its objective function is expressed as: ; In the formula, The objective function is... Number the aggregator; For aggregators Mean squared error loss function trained on local data; For aggregators Local model parameters; These are the parameters for the global integrated model issued by the distribution network operator. This is the regularization penalty parameter; The gradient update parameters obtained from training the local user electricity price response model are pruned and Gaussian noise is added, specifically as follows: ; In the formula, The gradient of the original model; Set the model gradient sensitivity clipping threshold; With a mean of 0 and a variance of Gaussian noise; It is the identity matrix; This represents the gradient update amount for the final encrypted upload.

4. The method for characterizing the flexible resource credibility and responsiveness of building complexes according to claim 1, characterized in that: The method of introducing a user willingness function to extract psychological preference features in the social domain includes: The perceived strength of the incentive price is calculated using a quantification function: ; In the formula, The perceived intensity of the incentive coefficient for users; For perception coefficient; This refers to the actual excitation coefficient signal issued. Calculate the overall response probability: ; In the formula, The combined response probability of economic and psychological factors; The probability of a normally distributed response influenced by economic incentives; The probability of a response influenced by psychological factors; and These are the weighting coefficients for economic factors and psychological factors, respectively.

5. The method for characterizing the flexible resource credibility and responsiveness of a building complex according to claim 1, characterized in that: The generation of the dynamic characterization set of the flexible resource response capability of building complexes, covering the dimensions of time, space, capacity, and probability, specifically includes: Calculate the static physical capacity boundary: ; In the formula, for The theoretical maximum response capacity of the entire building complex during the time period; The total number of flexible resource devices participating in the response; For the first The device is The power of the load that can be moved during a given time period; For the first The device is Reduced load power during certain time periods; Calculate the actual dynamic capacity of intention: ; In the formula, This represents the final, actual usable dynamic response capacity. For the overall response probability; This is the response loss coefficient caused by communication and control delays in the information domain.

6. The method for characterizing the flexible resource credibility and responsiveness of a building complex according to claim 1, characterized in that: The system dynamics model uses stock and flow feedback equations to update the response potential in the medium to long term, expressed as: ; In the formula, for The total flexible resource response potential of the time system; This represents the initial potential reserve of the response. Add a rate variable to the potential; This represents the potential decay rate variable.

7. The method for characterizing the credible cognition and response capabilities of flexible resources in a building complex according to claim 6, characterized in that: The potential rate of increase variable is expressed as: ; In the formula, For a moment The normalized policy-positive incentive subsidy index; For a moment The equivalent response capacity base calculated from the newly added flexible equipment investment in the region; and The system input conversion weight coefficients are respectively the subsidy index and the investment base; The potential decay rate variable is expressed as: ; In the formula, The overall aging and depreciation rate of flexible resource equipment; The cumulative index of users' subjective response fatigue; and These are the capacity decay sensitivity coefficients corresponding to the fatigue index and aging depreciation rate, respectively.

8. The method for characterizing the flexible resource credibility and response capability of a building complex according to claim 7, characterized in that: The subjective response fatigue accumulation index is cumulatively corrected based on the actual response behavior under the day-to-day / real-time short timescale. ; In the formula, Forgetting factors related to user fatigue; The calculation period between adjacent macroscopic time nodes; This represents the actual usable dynamic response capacity over a short timescale. This represents the maximum adjustable potential of static theory over a short timescale. Specifically, between adjacent macroscopic time nodes, the response potential stock of the previous macroscopic time node is used as the initial value. The difference between the potential increase rate variable and the potential decay rate variable is accumulated over time to obtain the response potential stock of the current macroscopic time node, which is then used as the medium-to-long-term potential evolution result. The medium-to-long-term potential evolution result is mapped to the capacity upper limit of each corresponding day-to-real-time period, and the boundary correction is performed on the actual dynamic willingness capacity calculated under the short time scale. The calculated medium- and long-term potential evolution results are used as macro-boundary constraints. The actual dynamic willingness capacity at the day-to-day and real-time short time scales is dynamically input and corrected, thereby forming a multi-time-scale closed-loop linkage evolution mechanism in which the medium- and long-term potential stock feedback adjusts the short-term willingness capacity and the short-term actual response behavior accumulates to correct the medium- and long-term fatigue.