Power grid evolution behavior modeling method and system based on dynamic digital twinning

Abstract

The invention provides a power grid evolution behavior modeling method and system based on dynamic digital twinning, and relates to the technical field of intelligent power grid digital twinning. According to the method, interaction characteristics of power station output, user load response and an energy storage scheduling strategy are captured in real time through a digital twin interface, behavior coupling factors containing complementarity and conflict weight are formed, and a power grid response relation and a multi-scale demand elastic curved surface are constructed by using an implicit tensor fusion technology and nonlinear projection. And through the steps of virtual behavior anchor point disturbance and the like, offset is separated to identify an unbalance region, an evolution trajectory is generated, an optimal collaborative path is extracted, and the weight of a behavior coupling factor is dynamically adjusted, so that rapid convergence of a global collaborative stability domain is realized. The method covers a behavior coupling factor generation module and the like, supports various operations, effectively solves many problems in power grid resource interaction, and improves dynamic adaptability and cooperation efficiency of power grid evolution modeling.

Description

Technical Field

[0001] The present invention relates to the technical field of smart grid digital twins, and specifically to a method and system for modeling grid evolution behavior based on dynamic digital twins. Background Art

[0002] As energy systems evolve toward intelligence, the application of digital twin technology in power grid modeling is deepening, primarily used for mirroring and state deduction between physical systems and virtual space. Existing technologies often construct local simulation models based on single-dimensional data or approximate the interactions between multiple entities through linear superposition.

[0003] However, as the penetration of distributed resources in the power grid increases and interaction scenarios become more complex, the spatiotemporal behavior of multiple entities becomes highly dynamic and asymmetric, and the limitations of existing technologies become increasingly apparent. These issues make it difficult for the power grid to achieve efficient coordination and dynamic optimization in spatiotemporal dimensions when dealing with fluctuations in the presence of high renewable energy and user-side demand. Summary of the Invention

[0004] To achieve the above objectives, the present invention provides a method and system for modeling power grid evolution behavior based on dynamic digital twins; In a first aspect, an embodiment of the present invention provides a method for modeling power grid evolution behavior based on dynamic digital twins; Power plant output characteristics, user load response patterns, and energy storage operator dispatch strategies are captured in real time through a digital twin interface and abstracted into behavioral coupling factors. These behavioral coupling factors extract the potential complementarity and conflict weights of resource interactions between entities using implicit tensor fusion technology. Furthermore, time-series data on power plant output, user load, and energy storage strategies are fed into a three-dimensional implicit tensor model to decompose output fluctuation cycle characteristics, load elasticity distribution patterns, and energy storage regulation rules, and to extract cross-entity interaction characteristics. A dynamic mask matrix is ​​generated based on historical grid conflict events to prune the power plant-energy storage resource competition paths while retaining the power plant-user supply and demand complementary paths. The difference between power plant regulation delay and user response delay is calculated, and the original matching degree is dynamically weakened using an exponential decay function to generate an effective complementary coefficient. The effective complementary coefficient is then dual-channel fused with the real-time regulation difference between power plants and energy storage, with complementary terms weighted and conflict terms filtered by thresholds to form a coupling matrix. Sudden output changes or load distortions are detected in real time, triggering weight adjustments to enhance the complementary path weights and suppress conflict terms. The pruning rules and decay parameters are then modified through digital twin feedback.

[0005] Furthermore, a grid response relationship is constructed based on the behavioral coupling factor. The power station's output fluctuations are mapped into a user-side demand elasticity surface through nonlinear projection, and a grid decision space is formed with the energy storage charging and discharging strategy. Specifically, the fundamental frequency and high-frequency components of the power station's output are mapped to the user-side price-sensitive area and overload risk area through non-orthogonal projection using the output cycle characteristics and load elasticity pattern, generating a multi-scale demand elasticity surface to capture the distorted diffusion path of output fluctuations on the user side. According to the energy storage regulation rule, the charging and discharging rate and capacity margin are discretized into spatiotemporal gradient units, and a buffer association matrix is ​​established between them and the supply-demand imbalance area in the elastic surface, where the association strength is regulated by the pruned complementary path. Based on the effective complementary coefficient and historical conflict rate, the conflict suppression factor and real-time complementarity are dynamically integrated to perform a weighted reconstruction of the initial connection strength. The demand elasticity surface and the energy storage buffer matrix are expanded and fused according to the implicit tensor basis, and the modified dynamic strength is embedded as an orthogonal dimension constraint to form a grid decision space containing an asymmetric coupling chain. Its orthogonality ensures the analyzability of subsequent virtual behavioral anchor point perturbation trajectories and the stability domain deviation measurement.

[0006] Furthermore, by introducing virtual behavioral anchor points to perturb the grid response relationship, a grid evolution model is generated; the virtual behavioral anchor points are based on the orthogonal coupling chain of the grid decision space, extracting the breakpoints of resource interaction in the historical evolution, and combining the complementary path weight attenuation extreme value and the conflict path mutation threshold to cross-locate the spatiotemporal heterogeneous anchor points; the anchor points with the largest deviation from the real-time complementarity are screened, and virtual output or load disturbances are injected into their associated nodes through digital twins to induce local resonance of the coupling chain; the post-disturbance trajectory offset is separated by the independence of orthogonal basis vectors to identify new breakpoints that are not covered by historical conflicts; the new breakpoints are fed back to the complementary path pruning rules, the conflict threshold is dynamically updated, and the convergence of the corrected coupling chain is verified. If the threshold is not reached, a secondary disturbance is triggered until the evolution trajectory meets the chain resonance constraint.

[0007] Furthermore, local resonance of the coupling chain is triggered by virtual behavioral anchor point disturbances to generate a model evolution trajectory that reflects the dynamics of resource interaction; the optimal collaborative path is extracted based on the deviation between the trajectory and the target stable domain, and the deviation is defined as a composite function of resource mismatch, collaborative delay cost and evolutionary inertia resistance; specifically, the following steps are involved: decomposing the resonance response of the coupling chain in the orthogonal decision space, dynamically adjusting the connection weights of the power station-user-energy storage nodes according to the trajectory offset, and reconstructing the disturbance propagation path; using orthogonal basis vectors to separate the evolutionary trajectory of each node after the disturbance, calculating the superposition of the power station output deviation, user load distortion rate and energy storage regulation hysteresis, and extracting the implicit imbalance area; mapping the implicit imbalance area to the coupling matrix, dynamically updating the effective coefficient of the complementary path and the conflict term filtering rules, and reversely correcting the elastic surface mapping parameters through the digital twin interface; verifying the convergence of the corrected coupling chain under the stable domain deviation metric, and triggering a secondary disturbance of the new anchor point if the threshold is not reached.

[0008] Furthermore, the optimal collaborative path is injected into the adversarial evolution channel of the digital twin, and the collaborative deviation is detected through the chain resonance of virtual and real behaviors, and the fusion weight of the behavioral coupling factor is dynamically adjusted until the deviation is lower than the preset tolerance threshold; specifically, it includes: separating the spatiotemporal coupling components in the trajectory offset caused by the virtual behavior anchor point disturbance based on the independence of the orthogonal basis vectors, and constructing the response vector of the chain resonance; combining the boundary constraints of the target stability domain, performing a tensor inner product operation on the response vector and the dynamic strength of the coupling matrix to generate a collaborative deviation index; if the collaborative deviation index exceeds the preset tolerance threshold, the fusion weight of the behavioral coupling factor is dynamically adjusted, and the dynamic mask matrix and the complementary coefficient attenuation parameter are inversely corrected through implicit tensor basis expansion to enhance the complementary path weight and suppress the conflict term, thereby triggering the secondary response of the local resonance of the coupling chain.

[0009] In a second aspect, an embodiment of the present invention provides a power grid evolution behavior modeling system based on dynamic digital twins; The system includes a behavior coupling factor generation module, a power grid evolution behavior model module, a collaborative path optimization module and a feedback control module; the behavior coupling factor generation module captures the power station output characteristics, user load response patterns and energy storage operator scheduling strategies in real time through a digital twin interface, and uses implicit tensor fusion technology to extract the complementarity and conflict weights of resource interactions between entities; the power grid evolution behavior model module constructs a power grid response relationship based on the behavior coupling factor, generates a power grid decision space that fuses the demand elasticity surface and the energy storage buffer matrix through nonlinear projection, and dynamically corrects the node connection strength; the collaborative path optimization module triggers local resonance of the coupling chain through virtual behavior anchor point perturbations, generates a model evolution trajectory and extracts the optimal collaborative path; the feedback control module injects the optimal collaborative path into the adversarial evolution channel, detects collaborative deviations and dynamically adjusts the fusion weights to ensure that the evolution trajectory converges to the target stability domain.

[0010] Furthermore, the grid evolution behavior model module captures the distorted diffusion path of output fluctuations through non-orthogonal projection technology, and establishes a buffer correlation matrix in combination with the energy storage spatiotemporal gradient unit to dynamically regulate the correlation strength of the supply and demand imbalance area; the collaborative path optimization module uses orthogonal basis vectors to separate the disturbance trajectory offset, extract the implicit imbalance area and reversely correct the elastic surface parameters; the feedback control module quantifies the collaborative deviation through tensor inner product operation, triggers the dynamic adjustment of the fusion weight and the secondary response of the chain resonance, and achieves rapid convergence of the global collaborative stability domain.

[0011] This invention uses dynamic digital twin technology to quantify the complementary and conflicting relationships between power grid entities in real time. Combining virtual behavioral anchor point perturbations with a chain resonance mechanism, it overcomes the limitations of traditional static modeling methods in dynamic evolution path optimization and collaborative stability control. Compared to existing technologies, this invention provides a method and system for modeling power grid evolutionary behavior based on dynamic digital twins, which has the following beneficial effects: (1) This paper realizes the refined extraction of multi-entity resource interaction features and the active suppression of conflicting paths through implicit tensor fusion and dynamic mask pruning technology.

[0012] (2) Based on the perturbation mechanism of orthogonal decision space and virtual behavior anchor points, the present invention enhances the dynamic recognition ability of the evolutionary model for new imbalance areas.

[0013] (3) The present invention ensures the global coordinated stability of the grid evolution behavior under complex disturbances through chain resonance constraints and deviation-driven feedback adjustment; the method and system provide operational technical support for the efficient configuration of grid resources and real-time optimization of evolution paths. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0015] The technical solutions in the embodiments of the present invention will be described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0016] Example 1: The embodiment of the present invention provides a specific implementation method of a power grid evolution behavior modeling method based on dynamic digital twins. Taking a regional smart grid as an example, the grid includes distributed photovoltaic power stations, smart user load clusters and distributed energy storage systems.

[0017] Photovoltaic output is affected by weather changes and exhibits 24-hour base frequency periodic fluctuations and high-frequency minute-level disturbances. User loads are divided into price-sensitive areas and overload risk areas. Distributed energy storage systems need to dynamically adjust the charging and discharging rates and capacity margins to balance the contradiction between supply and demand. Traditional modeling methods cannot capture the asymmetry of dynamic interactions among multiple entities due to static weight allocation, resulting in low collaborative efficiency.

[0018] This embodiment uses dynamic digital twin technology to achieve closed-loop control from data-driven modeling to global optimization. The specific implementation process is as follows: PV output time series data is captured in real time through a digital twin interface and decomposed into a 24-hour fundamental frequency component and a 5-minute high-frequency perturbation component. For example, the fundamental frequency fluctuates by ±20% on sunny days, while the instantaneous high-frequency perturbation can reach ±50% on rainy days. The user load response curve is divided into a price-sensitive zone and an overload risk zone. The load elasticity coefficient in the price-sensitive zone is 0.8, allowing a 5-minute response delay. The load elasticity coefficient in the overload risk zone is 0.2, with a response delay of only 10 seconds. The distributed energy storage system scheduling strategy includes a charge and discharge rate threshold of 2 MW / min and dynamic capacity margin adjustment rules, with an upper limit of 80% for off-peak capacity. After the three types of data are input into a three-dimensional implicit tensor model, the output fluctuation characteristics are decomposed along the time, space, and frequency domains to extract the load elasticity distribution pattern and analyze the energy storage regulation rules. Cross-entity interaction residual features are extracted using an implicit tensor fusion algorithm. For example, when PV output drops by 50% within 10 minutes, the load in the user's price-sensitive zone is not adjusted synchronously due to the response delay, resulting in an overcharge-triggered protective shutdown of the energy storage system.

[0019] A dynamic mask matrix is ​​generated based on historical conflict events to prune PV-energy storage resource competition paths. For example, conflicting paths where energy storage charge and discharge rates exceed limits during high-frequency PV output disturbances are eliminated, retaining complementary paths that actively regulate loads during PV baseband output troughs and in price-sensitive user areas. By calculating the difference between a 30-second PV regulation delay and a 5-minute user response delay, an exponential decay function is defined to dynamically correct the initial matching degree using a time window sliding matching mechanism. The decay function's time constant is set to 10 minutes, dynamically reducing the initial matching degree of 0.9 to an effective complementarity coefficient of 0.75. The energy storage regulation discrepancy is monitored in real time. When the actual charge and discharge rate deviates from the target by more than 15%, a conflict item filtering mechanism is triggered. Dual-channel fusion combines the effective complementarity coefficient with the conflict item at a weight of 7:3 to generate a spatiotemporal coupling matrix. When a sudden PV output change exceeding 40% is detected or the user load distortion rate exceeds 20%, the complementary path weight is automatically increased to 8:2, the conflict filtering threshold is simultaneously tightened to 10%, and the decay function's time constant is optimized to 5 minutes through a feedback loop.

[0020] A grid response relationship is constructed based on a spatiotemporal coupling matrix. Nonlinear projection is used to map the PV output fundamental frequency to user price-sensitive areas. For example, 24-hour fundamental frequency fluctuations are associated with user off-peak periods, enhancing resilience and allowing delayed load response to match output troughs. High-frequency, five-minute disturbance components are mapped to user overload risk areas, triggering protection mechanisms to prevent voltage overshoot. The energy storage system is discretized into spatiotemporal gradient units based on a charge and discharge rate of 2 MW / min and an 80% capacity margin, with each five-minute unit representing a time unit. A buffer correlation matrix is ​​then established with the supply-demand imbalance areas on the elastic surface.

[0021] The association strength in the buffer association matrix is ​​regulated by the pruned complementary path, with an effective complementary coefficient of 0.75 corresponding to a buffer strength of 0.6. The historical conflict rate of 5 times / day is dynamically integrated with the real-time complementarity, with the conflict suppression factor weight set to 0.4 and the real-time complementarity weight set to 0.6. The initial connection strength of 0.6 is weighted and reconstructed to 0.52. The demand elasticity surface and the energy storage buffer matrix are expanded and integrated through an implicit tensor basis, and orthogonal dimensional constraints are embedded to form the grid decision space. The orthogonal basis vectors independently represent the output fluctuation, load elasticity, and energy storage regulation dimensions, ensuring that the evolution trajectory of each dimension after the virtual behavior anchor point disturbance can be separated and analyzed.

[0022] When locating virtual behavior anchor points within the orthogonal decision space, the target anchor points are screened by combining historical evolution breakpoint data and real-time complementarity deviations. The orthogonal decision space is constructed through implicit tensor basis expansion, and its core consists of three independent orthogonal basis vectors. The first basis vector is used to characterize the frequency domain characteristics of the fluctuating energy distribution of the output, the second basis vector maps the user load elasticity gradient, and the third basis vector describes the spatiotemporal regulation capability of the energy storage. Specifically, the first basis vector is generated by decomposing the fundamental frequency component and high-frequency disturbance component of the photovoltaic output time series data. For example, the energy distribution of the fundamental frequency 24-hour periodic fluctuation and the high-frequency 5-minute disturbance is projected onto the spatiotemporal dimension to quantify the spectral characteristics of the dynamic changes in the power output. The second basis vector is formed by non-orthogonal projection of the elastic coefficient difference and response delay time between the user price-sensitive zone and the overload risk zone, reflecting the hierarchical characteristics of the load-side elastic response. The third basis vector is generated by orthogonal decomposition based on the discretization rules of the energy storage charge and discharge rate threshold, capacity margin, and spatiotemporal gradient unit, describing the spatiotemporal constraints of the energy storage regulation.

[0023] When an anchor point deviates from the target value of 0.8 by 6.25% due to the complementary coefficient of 0.75, a virtual disturbance is injected into the node of its buffer correlation matrix; for example, the photovoltaic output is simulated to drop by another 10%, or the load delay response time in the price-sensitive area of ​​the user is forced to be extended to 8 minutes, triggering local resonance of the coupling chain; the trajectory offset after the disturbance is separated and analyzed by orthogonal basis vectors: the first basis vector shows that the force fluctuation energy distribution is offset by 12%, reflecting the degree of diffusion of high-frequency disturbance energy in the frequency domain dimension; the second basis vector detects that the voltage fluctuation rate in the overload risk area has increased to 5.5%, reflecting the deterioration of voltage stability due to insufficient load elasticity; the third basis vector reveals that the energy storage regulation lag is 7%. , which is caused by the deviation of charging and discharging rates under capacity margin constraints; residual analysis further identifies new breakpoints, such as a certain energy storage node being unable to respond to high-frequency disturbances due to insufficient capacity margin, resulting in an increase of 1.5MW in the supply-demand gap; this breakpoint is fed back into the dynamic mask matrix, a new energy storage capacity verification condition is added, and the complementary coefficient is reconstructed to 0.78; if the evolution trajectory deviates from the target stable domain, resulting in a resource mismatch of more than 5%, a secondary disturbance is triggered, and the elasticity coefficient of the overload risk zone is increased to 0.25 along the second basis vector to enhance voltage stability. The energy storage charge and discharge rate is optimized to 2.3MW / min along the third basis vector to break through the spatiotemporal gradient constraint until the voltage fluctuation rate drops to 4.8%.

[0024] After injecting the optimized collaborative path into the adversarial evolution channel of the digital twin, a chain resonance detection model is constructed based on the independence of orthogonal basis vectors; the first chain resonance vector is composed of the tensor product of the output fluctuation energy distribution represented by the first basis vector and the user voltage fluctuation rate mapped by the second basis vector, quantifying the nonlinear coupling effect of the output disturbance on the voltage stability of the user side; the second chain vector is generated by the time domain convolution of the energy storage regulation hysteresis described by the third basis vector and the load response delay time in the second basis vector, reflecting the cumulative impact of the coordinated failure of energy storage regulation hysteresis and load response delay on the supply and demand gap; combined with the boundary conditions of the target stability domain, the composite constraint indicators of resource mismatch ≤5%, voltage fluctuation rate ≤5%, and collaborative delay cost ≤3% are defined; the current The synergy deviation index is 5.2%, of which the first chain vector contributes 3.1% and the second chain vector contributes 2.1%. After exceeding the preset tolerance threshold, the feedback control engine is activated to dynamically adjust the behavioral coupling factor weights: the complementary path weight is increased from 8:2 to 9:1 to enhance the regulation priority of the output-load complementary path, the conflict deviation rate threshold is tightened from 10% to 8% to suppress the activation probability of the energy storage-output conflict path, and the implicit tensor basis is expanded and refined to a 3-minute granularity for improved regulation time resolution. After two chain resonance iterations, the output fluctuation energy offset is reduced to 10%, the voltage fluctuation rate is stabilized at 4.7%, the energy storage regulation hysteresis is optimized to 4.5%, the synergy deviation index converges to 2.8%, and the global stability domain is achieved.

[0025] Example 2: This embodiment of the present invention provides a power grid evolution behavior modeling system based on dynamic digital twins. The system consists of four core modules, including a behavior coupling factor generation module, a power grid evolution behavior model module, a collaborative path optimization module, and a feedback control module. Each module interacts with the chain control logic through a standardized data interface. The specific technical implementation is as follows: The behavioral coupling factor generation module collects PV output time series data, user load response curves, and energy storage scheduling strategies in real time through the digital twin interface. The module has a built-in three-dimensional implicit tensor processor that performs frequency domain decomposition on PV output data, extracts the energy distribution characteristics of the fundamental frequency component and the high-frequency disturbance component, and simultaneously analyzes the elasticity coefficient of the user load in the price-sensitive zone and the overload risk zone. Dynamic adjustment parameters are generated based on the energy storage charge and discharge rate threshold and capacity margin rules. Cross-entity interaction features are extracted through an implicit tensor fusion algorithm. For example, the correlation analysis between PV output sag events and user response delays is used. A dynamic mask matrix is ​​generated by combining the historical conflict event library, pruning the resource competition paths between PV and energy storage, and retaining the complementary paths during output troughs and user-active adjustments. The output end of the behavioral coupling factor generation module transmits the spatiotemporal coupling matrix to the grid evolution behavior model module through a standardized JSON interface. The data fields include the output spectrum distribution, load elasticity gradient, and energy storage adjustment parameters.

[0026] After receiving the spatiotemporal coupling matrix, the grid evolution behavior model module uses a nonlinear projection engine to map the PV output fundamental frequency to the user's price-sensitive area, and associates the high-frequency disturbance component to the overload risk area, thereby constructing a multi-scale demand elasticity surface. This surface quantifies the transmission effect of output fluctuations on user-side voltage stability and load distortion. For example, high-frequency disturbances cause the voltage fluctuation rate in the overload risk area to increase by 3%. The energy storage system constructs a buffer association matrix based on the spatiotemporal gradient unit, and its capacity margin and charge and discharge rate dynamically match the supply and demand imbalance area of ​​the elastic surface. The elastic surface and the buffer matrix are merged into the grid decision space through orthogonal constraints. The three orthogonal basis vectors independently characterize the spectral energy distribution of the output, the load elasticity stratification characteristics, and the spatiotemporal regulation capability of the energy storage. The grid decision space data is transmitted to the collaborative path optimization module via a high-speed binary stream to ensure low-latency analysis of multi-dimensional data.

[0027] The collaborative path optimization module locates virtual behavioral anchor points based on the grid decision space and uses the historical evolution breakpoint library to screen high-risk anchor points with complementarity deviation exceeding the threshold. It also simulates output sudden drop or load delay scenarios through a virtual disturbance injection mechanism. For example, a 10% drop in PV output triggers local resonance in the coupling chain. The trajectory offset after the disturbance is parsed into output spectrum offset, voltage fluctuation increase, and energy storage regulation hysteresis via orthogonal basis vectors. After residual analysis identifies new breakpoints, it dynamically updates the conflict path filtering rules of the mask matrix and feeds the corrected parameters back to the preceding behavioral coupling factor generation module. If the resource mismatch exceeds the limit, the collaborative path optimization module optimizes the response capability of the overload risk zone along the load elasticity gradient and increases the charge and discharge rate threshold along the energy storage regulation dimension. The optimal collaborative path is generated and transmitted to the feedback control module via a RESTful API.

[0028] The feedback control module injects collaborative paths into the adversarial evolution channel to construct a chain resonance detection model. The first chain resonance vector is generated by the tensor product of the output spectrum energy and the voltage fluctuation, quantifying the nonlinear impact of power disturbances on user-side stability. The second chain vector reveals the cumulative effect of collaborative failure through the convolution operation of the energy storage regulation hysteresis and the load response delay. The feedback control module calculates the collaborative deviation index based on the boundary conditions of the target stability domain, triggering dynamic weight adjustment and conflict threshold optimization. For example, the weight of the complementary path is increased to 9:1 and the granularity of the energy storage spatiotemporal gradient is refined to 3 minutes. After two iterative optimizations, the convergence time of the system's global stability domain is shortened, the voltage fluctuation rate suppression efficiency is improved, and the resource mismatch is stabilized.

[0029] The system's modules achieve seamless interaction through standardized interfaces. The behavioral coupling factor generation module and the power grid evolution model module encapsulate data in JSON format. Decision space data is transmitted via binary streams. Feedback control instructions are reversely synchronized via a RESTful API. The three-dimensional implicit tensor processor supports frequency-domain and space-time joint analysis, the orthogonalization engine achieves multi-dimensional data decoupling, and the chain resonance model quantifies dynamic game effects through tensor operations.

[0030] The present invention aims to achieve a systematic breakthrough through dynamic digital twin technology to address the problems of traditional power grid modeling methods in dynamic interaction scenarios, such as rough modeling of multiple single-entity game, delayed response to high-frequency disturbances, and insufficient control of supply and demand imbalance risks. Traditional models are difficult to characterize the dynamic asymmetric interaction characteristics of photovoltaic output, user load and energy storage regulation due to static coupling and linearization processing, resulting in limited collaborative efficiency. The system extracts multi-entity interaction characteristics and dynamically prunes conflict paths through a behavioral coupling factor generation module, significantly reducing the risk of resource mismatch. Based on the orthogonal decision space, it realizes independent analysis of output spectrum, load elasticity and energy storage regulation dimensions, accurately locates high-frequency disturbance conduction paths, and improves dynamic response sensitivity. In addition, through modular architecture and standardized interface design, the system realizes full-process control from data acquisition to collaborative optimization, effectively improving the adaptability and operational economy of the power grid in complex dynamic scenarios, and providing reliable technical support for the efficient absorption and safe operation of a high proportion of renewable energy.

[0031] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for modeling power grid evolution behavior based on dynamic digital twins, characterized by: The following steps are involved: S1. Capture power plant output characteristics, user load response patterns, and energy storage operator dispatch strategies in real time through a digital twin interface, abstracting them into behavioral coupling factors. These behavioral coupling factors encompass the potential complementarity and conflict weights of resource interactions between entities. S2. A grid response relationship is constructed based on behavioral coupling factors. Power plant output fluctuations are mapped to a user-side demand elasticity surface through nonlinear projection. This is combined with energy storage charging and discharging strategies to form a grid decision space. The connection strength between nodes in this grid response relationship is inversely modified by the historical conflict rate and real-time complementarity of entity behaviors. S3. Generate a grid evolution model by introducing a virtual behavior anchor point to perturb the grid response relationship; S4. Perturb the grid response relationship through virtual behavioral anchors, triggering local resonance in the coupling chain and generating a model evolution trajectory that reflects the dynamics of resource interaction. Extract the optimal coordination path based on the deviation between the model evolution trajectory and the target stability domain, where the deviation is defined as a composite function of resource mismatch, coordination delay cost, and evolutionary inertia resistance. S5. Inject the optimal collaborative path into the adversarial evolution channel of the digital twin, detect collaborative deviations through virtual-real behavior chain resonance, and dynamically adjust the fusion weights of the behavioral coupling factors until the deviation is below the preset tolerance threshold; Among them, the behavioral coupling factor is optimized through the coupling-driven optimization of implicit tensor fusion and grid response relationship, realizing the asymmetric complementary configuration of resources of power stations, users and energy storage operators in the time and space dimensions, and the grid decision space is constrained by chain resonance of decision verification, so that the behavior of the grid evolution model converges to the global collaborative stability domain.

2. The method for modeling power grid evolution behavior based on dynamic digital twins according to claim 1, characterized in that: The process of abstracting the behavioral coupling factor in the S1 step is as follows: S11. Input the time series data of power station output, user load, and energy storage strategy into the three-dimensional implicit tensor model, decompose the output fluctuation period characteristics, extract the load elasticity distribution pattern, and analyze the energy storage regulation rules; Extract the cross-entity interaction features implicit in the three-way decomposition residuals; S12. Generate a dynamic mask matrix based on historical grid conflict events, prune the power plant-energy storage resource competition paths in the interaction features, and retain the power plant-user supply and demand complementary paths; S13. In the complementary path, the difference between the power plant regulation delay and the user response delay is calculated, and the original matching degree is dynamically weakened by an exponential decay function to generate an effective complementary coefficient; S14. Perform dual-channel fusion of the effective complementary coefficient and the real-time regulation difference between the power station and the energy storage system, perform weighted superposition of the complementary terms, and perform threshold filtering of the conflicting terms to generate a coupling matrix. S15. Real-time detection of sudden output changes or load distortions triggers weight adjustments, enhances the weights of complementary paths and suppresses conflicting terms, and modifies pruning rules and attenuation parameters through digital twin feedback.

3. The method for modeling power grid evolution behavior based on dynamic digital twins according to claim 1, characterized in that: The process of constructing the power grid response relationship in step S2 is as follows: S21. Based on the output cycle characteristics and load elasticity patterns, the power station output fundamental frequency and high-frequency components are mapped to the user-side price-sensitive areas and overload risk areas through non-orthogonal projection. This generates a multi-scale demand elasticity surface to capture the distortion and diffusion path of output fluctuations on the user side. S22. Based on the energy storage regulation rules, the charge and discharge rates and capacity margins are discretized into spatiotemporal gradient units, and a buffer association matrix is ​​established between them and the supply-demand imbalance region on the elastic surface. The association strength is regulated by the complementary paths pruned in step S1, and conflicting paths are eliminated through threshold filtering. S23. Based on the effective complementarity coefficient and historical conflict rate, the conflict suppression factor and real-time complementarity are dynamically integrated to perform weighted reconstruction of the initial connection strength; S24. Expand and fuse the demand elasticity surface and the energy storage buffer matrix according to the implicit tensor basis in step S1, embed the modified dynamic intensity as an orthogonal dimension constraint, and form a power grid decision space containing an asymmetric coupling chain. Its orthogonality ensures the analyzability of the subsequent virtual behavior anchor point disturbance trajectory and the stability domain deviation measurement.

4. The method for modeling power grid evolution behavior based on dynamic digital twins according to claim 1, characterized in that: The virtual behavioral anchor points in step 3 are orthogonal coupling chains based on the power grid decision space. The breakpoints of resource interactions in historical evolution are extracted, and the spatiotemporal heterogeneous anchor points are located by combining the extreme value of complementary path weight attenuation and the conflict path mutation threshold. The anchor points with the largest deviation from the real-time complementarity are selected, and virtual output or load disturbances are injected into their associated nodes through digital twins to induce local resonance in the coupling chain. The post-disturbance trajectory offset is separated through the independence of orthogonal basis vectors to identify new breakpoints not covered by historical conflicts. The new breakpoints are fed back to the complementary path pruning rule, and the conflict threshold is dynamically updated. The convergence of the corrected coupling chain is verified under the deviation metric of the power grid evolution stability domain.

5. The method for modeling power grid evolution behavior based on dynamic digital twins according to claim 4 is characterized by: The construction process of the power grid evolution model is as follows: S31. Based on virtual disturbances, the resonant response of the coupling chain is decomposed in the orthogonal decision space. The connection weights of the power station, user, and energy storage nodes are dynamically adjusted according to the trajectory offset to reconstruct the disturbance propagation path. S32. Use orthogonal basis vectors to separate the evolutionary trajectory of each node after the disturbance, calculate the superposition of power station output deviation, user load distortion rate, and energy storage regulation hysteresis, and extract the hidden imbalance area not covered by conflict pruning; S33. Map the implicit imbalance area to the coupling matrix, dynamically update the complementary path effectiveness coefficient and conflict term filtering rules, and reversely correct the elastic surface mapping parameters through the digital twin interface; S34. Verify the convergence of the modified coupling chain under the stability domain deviation metric. If the threshold is not reached, trigger the secondary perturbation of the new anchor point.

6. The method for modeling power grid evolution behavior based on dynamic digital twins according to claim 5, characterized in that: The identification of the hidden imbalance region in step S32 includes: S321. Perform time-domain convolution on the power station output deviation and the conflict weight in the behavior coupling factor to generate a conflict correlation vector; simultaneously, perform reverse gradient matching on the user load distortion rate and the complementary coefficient to generate a complementary correlation vector; S322. Based on the trajectory deviation direction after the virtual behavior anchor point disturbance, the conflict correlation vector and the complementary correlation vector are dynamically superimposed according to the basis vectors of the orthogonal power grid decision space to generate a correlation matrix containing the spatiotemporal game characteristics, and abnormal paths are filtered through the historical conflict paths in the dynamic mask matrix; S323. If the rate of increase of the conflict weight in the correlation matrix exceeds the dynamic attenuation threshold of the complementary coefficient, it is determined to be a strategic conflict zone between the power station and the energy storage; if the distortion rate of the complementary correlation vector on the user-side basis vector exceeds the overload threshold of the elastic surface, it is determined to be a resource mismatch zone between the user and the energy storage; S324. Align the judgment results with the pruned complementary paths in time and space, remove the areas that repeat historical conflict events, generate a dynamic map containing only new imbalances, and reversely correct the mapping weights of the supply and demand imbalance areas in the elastic surface and the update step size of the coupling matrix.

7. The method for modeling power grid evolution behavior based on dynamic digital twins according to claim 1, characterized in that: The model evolution trajectory is triggered by virtual behavior anchor point disturbances and represents the dynamic state transition sequence of the power grid evolution model after being subjected to virtual output or load disturbances within the power grid decision space. The generation process is as follows: the trajectory offset after the disturbance is separated by orthogonal basis vectors, and combined with the local resonance response of the coupling chain, a spatiotemporal evolution path is formed that includes power station output deviation, user load distortion, and energy storage regulation lag. The optimal collaborative path is extracted based on the deviation minimization principle. The evolution trajectory is converged to the target stability domain by reversely correcting the complementary path weights and conflict thresholds.

8. The method for modeling power grid evolution behavior based on dynamic digital twins according to claim 7, characterized in that: The optimal collaborative path is injected into the adversarial evolution channel of the digital twin. Based on the independence of the orthogonal basis vectors, the spatiotemporal coupling components in the trajectory offset caused by the virtual behavior anchor point disturbance are separated, and the response vector of the chain resonance is constructed. Combined with the boundary constraints of the target stability domain, the response vector and the dynamic strength of the coupling matrix are subjected to a tensor inner product operation to generate a collaborative deviation index. If the collaborative deviation index exceeds the preset tolerance threshold, the fusion weight of the behavioral coupling factor is dynamically adjusted, and the dynamic mask matrix and the complementary coefficient attenuation parameters are inversely corrected through implicit tensor basis expansion to enhance the complementary path weight and suppress the conflict term, thereby triggering the secondary response of the local resonance of the coupling chain.

9. A power grid evolution behavior modeling system based on dynamic digital twins, capable of implementing the method described in any one of claims 18, characterized in that: The system comprises: Behavioral Coupling Factor Generation Module: This module captures power plant output characteristics, user load response patterns, and energy storage operator dispatch strategies in real time through the digital twin interface and abstracts them into behavioral coupling factors. These behavioral coupling factors include the potential complementarity and conflict weights of resource interactions between entities. Grid Evolution Behavior Model Module: This module constructs a grid response relationship based on behavioral coupling factors. It maps power station output fluctuations into a user-side demand elasticity surface through nonlinear projection. This, combined with energy storage charging and discharging strategies, forms a grid decision space. The connection strength between nodes in this grid response relationship is inversely modified by the historical conflict rate and real-time complementarity of entity behaviors. Collaborative Path Optimization Module: Generates a grid evolution model by introducing virtual behavioral anchor points to perturb the grid response relationship. The optimal collaborative path is extracted based on the deviation between the model evolution trajectory and the target stability domain, where the deviation is defined as a composite function of resource mismatch, collaborative delay cost, and evolutionary inertia resistance. Feedback control module: Injects the optimal collaborative path into the adversarial evolution channel of the digital twin, detects collaborative deviations through virtual-real behavior chain resonance, and dynamically adjusts the fusion weight of the behavioral coupling factor until the deviation is lower than the preset tolerance threshold.

Images