Optical-storage park-level micro-grid multi-priority target regulation method and system

Abstract

This invention provides a multi-priority target control method and system based on a photovoltaic-storage park-level microgrid, belonging to the field of power grid technology. This invention collects microgrid data in real time and uses a fuzzy algorithm to dynamically identify current priority targets, ensuring that the control strategy can quickly respond to real-time changes in the microgrid. Furthermore, by synchronously updating the digital twin and combining it with a federated learning algorithm to continuously retrain and optimize the intelligent control model, the model adapts to complex and changing operating environments, enhancing the adaptability and robustness of the control strategy. Finally, a cooperative game theory algorithm is used to generate a control scheme that comprehensively considers the interests of all stakeholders, ensuring the acceptability and collaborative efficiency of the control scheme. This invention achieves dynamic adaptive control, solving the technical problems of mismatch between traditional microgrid control strategies and actual conditions, and low reliability. It improves the real-time performance and accuracy of microgrid control strategies, ensuring the continuity and reliability of microgrid power supply.

Description

Technical Field

[0001] This invention relates to the field of power grid technology, and in particular to a multi-priority target control method and system based on a photovoltaic-storage park-level microgrid. Background Technology

[0002] With the ongoing energy transition and the widespread application of distributed energy, park-level microgrids have received widespread attention as an important solution for improving power supply reliability and optimizing energy utilization efficiency. Photovoltaic-storage park-level microgrids, by integrating photovoltaic power generation, energy storage systems, and controllable loads, can achieve multiple objectives such as dynamic capacity expansion, peak shaving and valley filling, and reverse current prevention when the park's power supply capacity is insufficient, effectively alleviating power distribution pressure and improving energy self-sufficiency.

[0003] However, traditional microgrid control strategies are mostly based on fixed priority objectives, such as safety, backflow prevention, dynamic capacity expansion, and economic operation. They employ hierarchical optimization methods to solve for and determine decisions step by step. While the calculated control scheme may satisfy the priority objectives, the actual operating conditions of a microgrid are complex and variable due to multiple factors, including fluctuations in distributed energy output, energy storage charging and discharging behavior, and random load changes. Fixed-priority control strategies struggle to dynamically respond to system state changes, easily leading to a mismatch between the strategy and actual operating conditions. This can cause system power oscillations, even safety hazards, and affect power supply reliability.

[0004] For example, it lacks sufficient response flexibility when facing complex scenarios such as extreme weather, grid failures, and sudden changes in energy storage status. In addition, traditional strategies rely on predictive models trained offline based on historical data, which are difficult to reflect the real-time operating status of the microgrid in a timely manner. This leads to deviations between the formulated control strategies and the actual situation, further weakening the system's adaptability and robustness. Summary of the Invention

[0005] This invention provides a multi-priority target control method and system based on a photovoltaic-storage park-level microgrid, which solves the technical problems of mismatch between traditional microgrid control strategies and actual conditions and low reliability, improves the real-time performance of microgrid control strategies, and ensures the power supply reliability of the microgrid.

[0006] In a first aspect, the present invention provides a multi-priority target control method for a photovoltaic-storage park-level microgrid. This method includes: real-time acquisition of internal operating data and external environmental data of the microgrid; based on the internal operating data and external environmental data, using fuzzy computing to evaluate the current operating status and identify operating scenarios, and determining the current priority target; synchronously updating the digital twin of the microgrid based on the internal operating data and external environmental data; generating simulation data based on the digital twin, and based on the simulation data, combining a federated learning algorithm to retrain and optimize the parameters of the intelligent control model, obtaining an optimized intelligent control model; and based on the current priority target and the optimized intelligent control model, using a cooperative game theory algorithm to balance the optimization targets of each entity and quantify their contributions, generating a control scheme for the microgrid.

[0007] Secondly, embodiments of the present invention provide a multi-priority target control device based on a photovoltaic-storage park-level microgrid. This control device includes a communication module and a processing module. The communication module is used to collect internal operating data and external environmental data of the microgrid in real time. The processing module is used to evaluate the current operating status and identify operating scenarios based on the internal operating data and external environmental data using a fuzzy computing method, and determine the current priority target. Based on the internal operating data and external environmental data, it synchronously updates the digital twin of the microgrid. Based on the digital twin, it generates simulation data, and based on the simulation data, combined with a federated learning algorithm, it retrains and optimizes the parameters of the intelligent control model to obtain an optimized intelligent control model. Based on the current priority target and the optimized intelligent control model, it uses a cooperative game theory algorithm to balance the optimization targets of each subject and quantify their contributions to generate a control scheme for the microgrid.

[0008] Thirdly, embodiments of the present invention provide a multi-priority target control system based on a photovoltaic-storage park-level microgrid. The control system includes an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor is used to call and run the computer program stored in the memory to perform the steps of the method as described in the first aspect and any possible implementation thereof.

[0009] Fourthly, embodiments of the present invention provide a computer-readable storage medium storing a computer program, characterized in that, when the computer program is executed by a processor, it implements the steps of the method as described in the first aspect and any possible implementation thereof.

[0010] This invention provides a multi-priority target control method and system for a photovoltaic-storage park-level microgrid. By collecting internal operational data and external environmental data in real time and employing fuzzy computing to dynamically identify operating scenarios and current priority targets, this invention ensures that the control strategy can quickly respond to real-time changes in system status. Furthermore, by synchronously updating the digital twin and combining it with a federated learning algorithm to continuously retrain and optimize the intelligent control model using simulation data, the model can continuously learn and adapt to complex and changing operating environments, enhancing the adaptability and robustness of the control strategy. Finally, based on the current priority targets and the optimized intelligent control model, a cooperative game theory algorithm is used to generate a control scheme that comprehensively considers the interests of all stakeholders, ensuring the acceptability and collaborative efficiency of the control scheme in actual implementation. This invention, through data-driven and model self-evolution, achieves a leap from static, rigid control to dynamic, adaptive control, solving the technical problems of mismatch between traditional microgrid control strategies and actual conditions, and low reliability. It improves the real-time performance and accuracy of microgrid control strategies, ensuring the continuity and reliability of microgrid power supply. Attached Figure Description

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

[0012] Figure 1 This is a flowchart illustrating a multi-priority target control method for a photovoltaic-storage park-level microgrid provided in an embodiment of the present invention;

[0013] Figure 2 This is a schematic diagram of a multi-priority target control device based on a photovoltaic-storage park-level microgrid provided in an embodiment of the present invention;

[0014] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0015] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of the invention. However, those skilled in the art will understand that the invention can be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods are omitted so as not to obscure the description of the invention with unnecessary detail.

[0016] In the description of this invention, unless otherwise stated, " / " means "or". For example, A / B can mean A or B. The term "and / or" in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone. Furthermore, "at least one" and "more than one" refer to two or more. The terms "first," "second," etc., do not limit the quantity or order of execution, and "first," "second," etc., do not necessarily imply differences.

[0017] In the embodiments of this application, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design that is described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a specific manner to facilitate understanding.

[0018] Furthermore, the terms "comprising" and "having," and any variations thereof, used in the description of this application are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or modules is not limited to the steps or modules listed, but may optionally include other steps or modules not listed, or may optionally include other steps or modules inherent to such process, method, product, or device.

[0019] To make the objectives, technical solutions, and advantages of the present invention clearer, the following description will be provided in conjunction with the accompanying drawings and specific embodiments.

[0020] like Figure 1 As shown, this embodiment of the invention provides a multi-priority target control method based on a photovoltaic-storage park-level microgrid. The method includes steps S101-S105.

[0021] S101. Real-time acquisition of internal operation data and external environmental data of the microgrid.

[0022] In some embodiments, internal operating data includes bus voltage, system frequency, load power, energy storage system charging and discharging power, and state of charge. External environmental data includes irradiance, ambient temperature, and grid time-of-use pricing signals.

[0023] S102. Based on internal operating data and external environment data, use fuzzy computing methods to assess the current operating status and identify operating scenarios, and determine the current priority target.

[0024] As one possible implementation, step S102 can be specifically implemented as steps S1021-S1025.

[0025] S1021. Based on internal operating data and external environment data, feature extraction is performed to obtain multiple input features.

[0026] In some embodiments, the input features include the real-time state of charge of the energy storage system, the current net load power, the grid time-of-use tariff level for the current period, and the short-term volatility of photovoltaic output.

[0027] For example, the real-time state of charge of the energy storage system is directly read from the battery management system and normalized; the current net load power is calculated as the real-time difference between the load power and the predicted photovoltaic power, and normalized according to the transformer's rated capacity; the grid time-of-use price level for the current period is mapped to discrete values ​​{peak: 0.8, flat: 0.5, valley: 0.2} based on the peak, flat, and valley electricity price periods published by the grid; and the short-term volatility of photovoltaic output is calculated as the ratio of its standard deviation to its mean based on the photovoltaic output data sequence of the most recent 15 minutes.

[0028] S1022. Based on multiple input features and a preset membership function, determine the membership degree of the fuzzy linguistic variable corresponding to each input feature.

[0029] For example, embodiments of the present invention can define a set of fuzzy linguistic variables for each input feature, wherein: the fuzzy set for state of charge is {too low, low, normal, high, too high}, using a trapezoidal membership function; the fuzzy set for net load power is {severe deficit, deficit, balanced, surplus, severe surplus}, using a triangular membership function; the fuzzy set for electricity price level is {valley, flat, peak}, using a trapezoidal membership function; the fuzzy set for volatility is {stable, slight fluctuation, severe fluctuation}, using a Gaussian membership function; and the membership value of the input feature to each fuzzy linguistic variable is calculated using the corresponding membership function.

[0030] S1023. Based on the membership degree of the fuzzy linguistic variables corresponding to each input feature and the preset fuzzy rule base, reasoning is performed to obtain a fuzzy set describing the weights of each priority target.

[0031] In some embodiments, a fuzzy rule base is used to describe the mapping relationship between fuzzy combinations of input features and weights of priority targets.

[0032] For example, the fuzzy rule base contains 81 rules in the form of "IF-THEN", which are trained using domain expert knowledge and historical operational data. The Mamdani fuzzy inference method is used to perform synthesis operations on the input fuzzy set. The matching fuzzy rules are activated by the rule form "IF state of charge is A AND net load power is B AND electricity price level is C AND volatility is D THEN safety weight is E, economic weight is F, and green weight is G". The output of all activated rules is maximized to obtain the output fuzzy set describing the weight of each priority target.

[0033] S1024. Aggregate the priority targets in the fuzzy set, and use the centroid method to perform defuzzification calculation to obtain the weights of each priority target.

[0034] In some embodiments, priority objectives include security and stability objectives, economic operation objectives, and green and efficient objectives.

[0035] For example, in this embodiment of the invention, the area centroid of the output fuzzy set of each priority target in the domain can be calculated; the precise value of the weight of each target can be obtained through integration, and the calculation formula is as follows: ;in, The weight of the i-th priority target. Let x be the membership function, and let x be the possible values ​​of the weights of each priority target. Output the weight vector [ws, we, wg] for the safety and stability target, the economic operation target, and the green and efficient target, and satisfy ws + we + wg = 1.

[0036] S1025. Determine the current priority target based on the weight of each priority target.

[0037] For example, embodiments of the present invention can compare the numerical values ​​of each element in the weight vector; select the target with the largest weight value as the current priority target; and when the weights are equal, make a decision according to the preset default priority order (safety and stability > economic operation > green and efficient).

[0038] S103. Based on internal operating data and external environmental data, the digital twin of the microgrid is updated synchronously.

[0039] In some embodiments, the digital twin of a microgrid is a virtual model that is completely corresponding to the physical microgrid, synchronized in real time, and has high-fidelity simulation capabilities; it is a data-driven, dynamically evolving, and predictive "living" digital clone.

[0040] For example, the digital twin continuously injects internal operating data (bus voltage, load power, energy storage SOC, etc.) and external environmental data (light intensity, ambient temperature, etc.) to ensure that the virtual model is synchronized with the physical entity.

[0041] For example, the digital twin includes a photovoltaic array model, an energy storage system model, a line impedance model, and a load model. The photovoltaic array model simulates the output characteristics under different light and temperature conditions. The energy storage system model accurately simulates the battery's charge / discharge efficiency, internal resistance changes, and aging characteristics. The line impedance model accurately calculates network losses and voltage drops. The load model simulates the dynamic behavior of different types of loads.

[0042] As one possible implementation, step S103 can be specifically implemented as steps S1031-S1033.

[0043] S1031. Based on internal operating data and external environmental data, the initial state and boundary conditions of the digital twin are synchronized to obtain a synchronized digital twin, so that the static topology and dynamic operating state of the synchronized digital twin are synchronized with the real-time operating state of the microgrid.

[0044] For example, static topology refers to the connection relationships of electrical devices in a microgrid, including structural information that does not change frequently over time, such as circuit breaker status, bus connection methods, and network topology. Dynamic operating status refers to electrical quantities in a microgrid that change rapidly over time, including real-time operating parameters such as voltage, current, power, and frequency.

[0045] For example, embodiments of the present invention can map the real-time operating status of the physical microgrid (including the voltage of each node, branch power, energy storage SOC, and photovoltaic inverter operating point) to the initial conditions of the simulation model in the digital twin; set real-time environmental data (including light intensity, ambient temperature, and humidity) as the external input of the digital twin simulation environment; ensure the consistency of the static topology by comparing the circuit breaker status in the digital twin with the actual status of the physical system; and use the real-time collected system frequency, voltage amplitude, and other core electrical quantities as the boundary conditions for the power flow calculation of the digital twin.

[0046] S1032. Based on internal operating data and external environment data, as well as the synchronized digital twin, state estimation and parameter identification are used to dynamically calibrate the parameters of key models in the digital twin to obtain the calibrated digital twin.

[0047] In some embodiments, key models include a photovoltaic array model, an energy storage system model, and a line impedance model.

[0048] For example, state estimation: a technique that uses redundant real-time measurement data to deduce the complete and accurate operating state of a system through mathematical algorithms. Parameter identification: the process of estimating the internal parameters of a model based on the system's input and output data using optimization algorithms.

[0049] For example, embodiments of the present invention may employ weighted least squares method, based on real-time measured data (voltage, power), to estimate complete state quantities of the system such as node voltage phase angle and generator output that cannot be directly measured; applying recursive least squares method, based on real-time operating data sequences, to identify the dynamic parameters of key models: photovoltaic array model: identifying its equivalent series resistance, parallel resistance, and photovoltaic current source parameters; energy storage system model: identifying its equivalent internal resistance, capacity decay coefficient, and charge / discharge efficiency curve parameters; line impedance model: identifying the temperature characteristic coefficients of its resistance and reactance parameters; establishing a parameter error feedback correction mechanism, automatically triggering parameter re-identification when the deviation between the digital twin output and the measured data of the physical system exceeds a threshold.

[0050] S1033. Based on the calibrated digital twin, recalculate the implicit state variables and performance indicators of the microgrid to obtain the updated digital twin.

[0051] In some embodiments, implicit state variables include the system's real-time short-circuit capacity and the voltage stability margin of critical nodes. Performance metrics include the actual available capacity and health status of the energy storage system.

[0052] For example, implicit state variables are state quantities that cannot be directly measured but can reflect the inherent characteristics of a system, such as stability margin and short-circuit capability. Voltage stability margin is a quantitative indicator that measures the safe distance between the current operating state of the system and the voltage collapse point.

[0053] For example, embodiments of the present invention can use digital twins to calculate the short-circuit current of each node under N-1 fault conditions, thereby obtaining the short-circuit capacity distribution of the nodes; and use the continuous power flow method to calculate the power distance of the critical node from the current operating point to the voltage collapse point, thereby obtaining the voltage stability margin.

[0054] For example, the actual usable capacity is calculated based on the real-time internal resistance and temperature characteristics of the energy storage system, and its maximum chargeable and dischargeable capacity under the current environment is calculated. The health status is assessed by establishing a comprehensive evaluation function based on the capacity decay rate and the internal resistance growth rate to evaluate the remaining lifespan and health of the energy storage system. The calculated implicit state variables and performance indicators are updated to the state database of the digital twin to provide accurate basic data for subsequent simulations.

[0055] S104. Based on the digital twin, generate simulation data, and based on the simulation data, combine the federated learning algorithm to retrain and optimize the parameters of the intelligent control model to obtain the optimized intelligent control model.

[0056] As one possible implementation, step S104 can be specifically implemented as steps S1041-S1044.

[0057] S1041. Based on the current real-time state of the digital twin, set the initial conditions for the simulation.

[0058] In some embodiments, the current real-time state of a digital twin refers to the complete operating state of the digital twin in line with the physical microgrid, including all electrical quantities, device status, and control modes.

[0059] For example, electrical state initialization: set the voltage amplitude (0.95-1.05 pu), phase angle (-π to π radians), energy storage SOC (0.2-0.9), and photovoltaic inverter operating point (0-1 pu) of each node in the digital twin to the simulation start state; control parameter synchronization: map the currently running controller parameters (including PQ control, VF control, and droop control coefficient) to the control model of the digital twin; network topology confirmation: based on the real-time circuit breaker status, confirm the connection relationship of the simulation network and establish a node-branch association matrix; environmental baseline setting: set the current light intensity (0-1500 W / m²) as the baseline. 2 The ambient temperature (-20 to 50℃) serves as the baseline environmental condition for the simulation.

[0060] S1042. Based on ultra-short-term prediction data and preset typical scenarios, set the boundary conditions for the simulation.

[0061] In some embodiments, the boundary conditions include a light intensity variation curve, a load power variation curve, and a grid electricity price signal.

[0062] In some embodiments, ultra-short-term forecast data refers to high temporal resolution forecast data for the next 0-4 hours, which is typically generated using numerical weather prediction combined with machine learning algorithms.

[0063] For example, the time-series boundary conditions are as follows: Illumination intensity variation curve: Based on numerical weather forecast data, a 15-minute resolution illumination intensity prediction sequence is generated; Load power variation curve: Using the ARIMA time series prediction model, load power prediction data at 5-minute intervals is generated; Grid electricity price signal: According to the time-of-use pricing policy, the electricity price change time series within the simulation period is set; Scenario boundary conditions are as follows: Typical daily scenario: Includes standard daily curves for different weather types such as sunny, cloudy, and rainy days; Special event scenario: Includes special operating conditions such as holiday load characteristics and equipment maintenance plans; Constraint boundary conditions: Set operating limits such as voltage safety constraints, frequency constraints, and equipment capacity constraints.

[0064] S1043. Based on the initial conditions and boundary conditions, apply random perturbations within a preset range to the key model parameters in the digital twin, and simulate various typical scenarios for simulation calculation.

[0065] In some embodiments, key model parameters include photovoltaic module efficiency, energy storage internal resistance, and line parameters; typical scenarios include photovoltaic sudden drop scenarios, load surge scenarios, and equipment failure scenarios.

[0066] In some embodiments, the random disturbance within a preset range is a statistically consistent random change applied to the standard values ​​of the model parameters to simulate uncertainties such as equipment aging and measurement errors.

[0067] For example, parameter perturbation settings: photovoltaic module efficiency: randomly perturbed according to a normal distribution within ±5% of the rated value; energy storage internal resistance: randomly perturbed according to a uniform distribution within ±10% of the nominal value; line parameters: perturbed considering the temperature effect coefficient within ±8% of the design value.

[0068] Example, typical scenario construction: Photovoltaic sudden drop scenario: Simulating the change in light intensity from 1000W / m² within 1 minute. 2 Decreased to 200W / m 2 Extreme situations; load surge scenario: simulates a 50% increase in power demand for a critical load within 30 seconds; equipment failure scenario: simulates fault conditions such as energy storage converter failure and single-phase grounding of the line.

[0069] For example, parallel simulation execution: multi-threading technology is used to run simulation calculations for multiple scenarios simultaneously; the simulation duration of each scenario covers a complete rolling optimization cycle (usually 4-24 hours); the simulation step size is set to 10ms-1s according to the dynamic process requirements.

[0070] S1044. During the simulation process, dynamic response data of the microgrid system simulated by the digital twin is collected and determined as simulation data for each typical scenario.

[0071] In some embodiments, the simulation data includes energy storage charging and discharging power sequences, bus voltage fluctuation sequences, system power deficit sequences, and optimization objective function values.

[0072] In some embodiments, a typical scenario refers to a set of representative operating conditions that significantly impact system operation, determined based on historical operating data and risk assessment. Dynamic response data refers to complete process data showing the changes of each state variable over time under external stimuli or internal disturbances, reflecting the system's dynamic characteristics and stability.

[0073] For example, power dynamic sequences include: energy storage charging and discharging power sequence: recording the second-level changes in active and reactive power of energy storage; photovoltaic output sequence: recording the minute-level fluctuation characteristics of photovoltaic array output power; electrical measurement sequences: bus voltage fluctuation sequence: recording the transient and steady-state changes of voltage at PCC point and important load nodes; system frequency sequence: recording the frequency dynamic response process during islanded operation; operating status sequence: system power deficit sequence: recording the magnitude and duration of power deficit when supply and demand are unbalanced; and optimization objective function value: recording the achievement indicators of safety, economy, and green goals under different scenarios.

[0074] As one possible implementation, step S104 can be specifically implemented as steps A1-A4.

[0075] A1. Based on simulation data and historical operation data of microgrids, construct an incremental training sample set.

[0076] For example, simulation data is spatiotemporally aligned and fused with historical operational data, with the historical data spanning at least one year and including typical operating conditions for all four seasons. Sample feature engineering: Input feature dimensions include 28-dimensional feature vectors such as time-series power data (PV, load, energy storage), electrical parameters (voltage, current, frequency), environmental parameters (illuminance, temperature), and equipment status (SOC, health). Output labels include 12-dimensional control vectors such as optimal energy storage scheduling commands, PV reduction rates, and voltage regulation commands. Sample weight allocation: Based on the timeliness and importance of samples, more recent data are assigned higher weights, with weight coefficients calculated based on time decay.

[0077] A2. Determine the loss function with the optimization objective of minimizing the overall system operating cost and the penalty term for violating security constraints.

[0078] For example, the economic operating cost item Lecon = electricity purchase cost + equipment operation and maintenance cost + energy storage loss cost; where the electricity purchase cost is calculated based on time-of-use electricity pricing, the equipment operation and maintenance cost adopts a linear depreciation model, and the energy storage loss cost is based on a lifetime model established by the Rainflow counting method.

[0079] For example, the safety constraint penalty term is: Lsafe = β1·max(0,|V|-Vlim) 2 + β2·max(0,|f|-f_lim) 2 + β3·max(0,SOC-SOC_lim) 2 A quadratic penalty function is used to handle safety constraints such as voltage over-limit, frequency deviation, and SOC over-limit.

[0080] For example, the comprehensive loss function is: Ltotal = Lecon + λ·Lsafe + γ·Lreg; where Lreg is the L2 regularization term, and λ and γ are hyperparameters, whose optimal values ​​are determined through cross-validation.

[0081] A3. Based on the incremental training sample set and the loss function, the gradient descent algorithm is used to incrementally train the intelligent regulation model to obtain the updated intelligent regulation model and the updated model parameters.

[0082] A4. Based on the updated model parameters of the intelligent control model, the federated learning algorithm is used to optimize the model and obtain the optimized intelligent control model.

[0083] In some embodiments, federated learning algorithms are distributed machine learning frameworks that allow multiple participants to jointly train machine learning models without sharing the original data, and to achieve knowledge sharing by exchanging model parameters.

[0084] For example, step A4 can be specifically implemented as steps A41-A47.

[0085] A41. Based on the updated model parameters of the intelligent control model, homomorphic encryption processing is performed to obtain encrypted model parameters.

[0086] In some embodiments, homomorphic encryption is a special encryption technique that allows mathematical operations to be performed directly on the ciphertext, and the decrypted result is identical to the result of performing the same operation on the plaintext. This is used in federated learning to protect the privacy of the model parameters of each node.

[0087] For example, embodiments of the present invention may employ a partially homomorphic encryption algorithm to support addition and multiplication operations on floating-point models, ensuring that model aggregation calculations can be completed in ciphertext state; a public-private key pair is generated by the federated learning cloud platform, the public key is distributed to each microgrid node for encryption, and the private key is securely stored by the platform for final decryption; the model parameters are converted from floating-point vectors to integer field representations supported by the encryption algorithm, and fixed-point encoding is used to maintain computational accuracy; the model parameters are grouped and batch-encrypted to reduce the computational overhead of encryption operations and improve processing efficiency.

[0088] A42. Homomorphic encryption is performed on multiple incremental samples corresponding to the simulation data in the incremental training sample set, the number of incremental samples, and the training rounds in the incremental training process to obtain metadata.

[0089] For example, embodiments of the present invention can establish a unified metadata format, including core information such as data statistical features, training quality indicators, and node identity identifiers; homomorphically encrypt statistical quantities such as sample quantity and data distribution characteristics to ensure that data privacy is not leaked; encrypt and store key parameters in the training process, including quality indicators such as loss function convergence and gradient change trends; and add digital signatures and timestamps to each metadata data packet to ensure data integrity and timeliness.

[0090] A43. Based on the metadata of multiple microgrid nodes on the federated learning platform, calculate the proportion of training samples and determine the weighted average coefficient of each microgrid node.

[0091] In some embodiments, the weighted average coefficient is a weight value calculated based on factors such as the data quality and quantity of each node during the federated learning aggregation process, used to determine the degree of contribution of different nodes to the global model.

[0092] For example, embodiments of the present invention can calculate data credibility weights based on the training data quality indicators provided by each node; comprehensively consider factors such as sample quantity, data freshness, and training effect, and adopt a weight calculation model that integrates multiple indicators; establish a contribution accumulation mechanism based on the performance of each node in historical federated learning, and give higher weights to nodes that continuously provide high-quality updates; identify abnormal nodes through statistical analysis methods, and automatically reduce or eliminate the influence of abnormal nodes when calculating weights.

[0093] A44. A weighted average algorithm based on homomorphic encryption, using weighted average coefficients, securely aggregates the encryption model parameters of multiple microgrid nodes to obtain global update parameters.

[0094] In some embodiments, secure aggregation is the process of aggregating model updates from multiple nodes into a global model while protecting individual privacy, ensuring that the original data or model parameters are not leaked during the entire aggregation process.

[0095] For example, embodiments of the present invention can directly perform weighted average calculations on the model parameters of each node in an encrypted state, keeping the parameters encrypted throughout the process; employing high-precision numerical calculation methods to ensure numerical stability during the encrypted operation process and avoid precision loss; verifying the correctness of the aggregation process through zero-knowledge proof technology to prevent malicious nodes from destroying the aggregation results; supporting incremental aggregation of some nodes, where newly added nodes only need to perform differential aggregation with the global model, reducing computational overhead.

[0096] A45. Decrypt the global update parameters to obtain the global model parameters.

[0097] In some embodiments, global update parameters refer to the set of parameters of the next-generation global model generated after aggregating the model updates of all participating nodes through federated learning, which includes the knowledge contributions of all nodes.

[0098] A46. Based on the global model parameters and the intelligent control model, the model performance is verified, and the verification results are obtained.

[0099] A47. If the verification result is successful, then the optimized intelligent control model is determined based on the global model parameters.

[0100] For example, embodiments of the present invention can use a private key to decrypt ciphertext in the secure environment of a federated learning cloud platform and restore it to plaintext model parameters; evaluate the performance indicators of the global model on a standard test set, including tuning accuracy, response speed, stability, etc.; verify the compatibility of the new global model with the local systems of each node to ensure that the model can be loaded and executed correctly; and establish a sound version management mechanism to support a quick rollback to the previous stable version when the performance of the new model does not meet the requirements.

[0101] S105. Based on the current priority objectives and the optimized intelligent control model, a cooperative game algorithm is used to balance the optimization objectives of each subject and quantify their contribution to generate a microgrid control scheme.

[0102] As one possible implementation, step S105 can be specifically implemented as steps S1051-S1058.

[0103] S1051. Establish the interest function of each subject.

[0104] In some embodiments, the subjects include microgrid operators, energy storage investors, photovoltaic owners, and users, and the benefit function includes the network loss cost function of the microgrid operator, the energy storage depreciation cost function of the energy storage investor, the curtailment loss function of the photovoltaic owner, and the electricity satisfaction function of the user.

[0105] For example, the network loss cost function for grid operators: a network loss quantification model based on real-time power flow calculation results, considering line loss changes under different operating modes, and establishing a loss cost mapping relationship related to network topology and load distribution; the energy storage depreciation cost function for energy storage investors: using a lifetime decay model based on cycle number and depth, establishing a mathematical relationship between charging and discharging power and equipment depreciation cost, including initial investment amortization and operation and maintenance costs; the curtailment loss function for photovoltaic owners: based on the difference between predicted photovoltaic output and actual consumption, combined with feed-in tariffs and subsidy policies, quantifying the direct economic losses and reduced revenue caused by curtailment; and the electricity satisfaction function for users: establishing a dual evaluation system for power supply reliability and power quality, including a model of the impact of indicators such as power outage duration, voltage qualification rate, and frequency stability on user satisfaction.

[0106] S1052. Based on the interest functions of each subject, and with the goal of minimizing the overall operating cost of the system, construct the characteristic function of the cooperative alliance.

[0107] In some embodiments, a cooperative alliance may include multiple alliances consisting of one or more entities.

[0108] In some embodiments, the characteristic function of a cooperative alliance, in cooperative game theory, is a function used to define the maximum total benefit that any possible combination of entities (alliance) can obtain through cooperation, and is a core tool for analyzing the value of cooperation and the distribution of benefits.

[0109] For example, for each possible combination of entities in an alliance, the maximum total revenue that the alliance can achieve through internal coordination optimization is calculated, taking into account the resource complementarity and synergy within the alliance; an independent operating cost benchmark is established to calculate the individual operating costs of each entity without participating in the cooperation, serving as a reference for the distribution of cooperation revenue; security constraints and technical limitations of system operation are integrated into the characteristic function to ensure the feasibility and practicality of alliance revenue calculation; and the parameters and constraints of the characteristic function are dynamically updated according to the real-time operating status of the system and changes in the external environment.

[0110] S1053. Based on the optimized intelligent control model, and with the current priority target as a constraint, determine the preliminary control scheme that satisfies the current priority target.

[0111] For example, embodiments of the present invention can utilize a well-trained intelligent control model to generate a preliminary control strategy that takes into account multiple optimization objectives under the current operating state; perform simulation verification of the preliminary scheme through a digital twin to evaluate its safety and robustness under different operating conditions; calculate the operating cost of the preliminary control scheme and analyze the changes in the benefits of each subject in the scheme; and conduct sensitivity analysis on key parameters to evaluate the adaptability and stability of the scheme under different scenarios.

[0112] S1054. Based on the preliminary control plan and the characteristic function of the alliance, calculate the benefit distribution vector of each alliance in the cooperative alliance.

[0113] In some embodiments, an element in the benefit allocation vector represents the benefit function value of the corresponding subject in a cooperative state.

[0114] S1055. Iterate through each entity and calculate the marginal contribution of each entity to each alliance.

[0115] In some embodiments, marginal contribution refers to the additional revenue a participant brings to a particular alliance when joining that alliance, and is an important indicator for measuring the participant's contribution to the alliance's value.

[0116] S1056. Based on the marginal contribution of each entity to each alliance, perform a weighted summation to calculate the Shapley value of each entity.

[0117] In some embodiments, the Shapley value is used to characterize the overall contribution of the agents in the microgrid optimization process. The Shapley value is a classic solution in cooperative game theory, calculated based on the marginal contribution of each participant to all possible alliances. It is considered one of the fairest ways to distribute benefits, satisfying axioms such as efficiency, symmetry, and additivity.

[0118] For example, embodiments of the present invention can employ a systematic method to calculate the marginal contribution value of each entity to all possible alliances, i.e., the additional benefit brought by the entity joining the alliance; based on all marginal contribution values, the Shapley value of each entity is calculated according to the theoretical framework of cooperative game theory to ensure the completeness and fairness of the calculation; the accuracy of the Shapley value calculation is ensured through multiple verification methods, including monotonicity test, efficiency verification, etc.; and a differentiated weight allocation mechanism is established according to the actual contribution of each entity to reflect the principle of more work, more pay.

[0119] S1057. Using the Shapley value of each subject as a weight, the optimization objectives of each subject are weighted to construct a comprehensive objective function.

[0120] In some embodiments, the comprehensive objective function is a single objective function formed by weighting and integrating multiple optimization objectives of different dimensions and subjects according to their importance and contribution, and is used to generate an optimization scheme that comprehensively considers the interests of all parties.

[0121] S1058. Using the preliminary control scheme as the initial solution and the current priority target as the constraint, the comprehensive objective function is solved to obtain the microgrid control scheme.

[0122] For example, embodiments of the present invention can weight and merge the optimization objectives of each subject according to their contribution weights to form a unified comprehensive objective function; the comprehensive optimization fully considers the technical constraints of system operation and the interest protection requirements of each subject; a distributed optimization algorithm suitable for large-scale systems is adopted to ensure the efficiency and convergence of the optimization process; through multiple iterative optimizations and feasibility verifications, the final control scheme is ensured to meet both the system operation requirements and the interests of each subject.

[0123] This invention provides a multi-priority target control method for a photovoltaic-storage park-level microgrid. By collecting internal operational data and external environmental data in real time, and employing fuzzy computing to dynamically identify operating scenarios and current priority targets, the control strategy can quickly respond to real-time changes in system status. Furthermore, by synchronously updating the digital twin and combining it with a federated learning algorithm to continuously retrain and optimize the intelligent control model using simulation data, the model can continuously learn and adapt to complex and ever-changing operating environments, enhancing the adaptability and robustness of the control strategy. Finally, based on the current priority targets and the optimized intelligent control model, a cooperative game theory algorithm is used to generate a control scheme that comprehensively considers the interests of all stakeholders, ensuring the acceptability and collaborative efficiency of the control scheme in actual implementation. This invention, through data-driven and model self-evolution, achieves a leap from static, rigid control to dynamic, adaptive control, solving the technical problems of mismatch between traditional microgrid control strategies and actual conditions, and low reliability. It improves the real-time performance and accuracy of microgrid control strategies, ensuring the continuity and reliability of microgrid power supply.

[0124] Optionally, the multi-priority target control method based on photovoltaic-storage park-level microgrid provided in this embodiment of the invention further includes steps S201-S205 after step S105.

[0125] S201. Based on the updated digital twin, simulate the microgrid fault set and extreme weather scenarios to calculate the real-time resilience index of the microgrid.

[0126] In some embodiments, real-time resilience metrics include expected load shedding, average recovery time, and maximum disturbance resilience. Real-time resilience metrics are comprehensive indicators used to quantify a microgrid's ability to withstand disturbances, absorb impacts, and quickly restore power supply. They comprehensively reflect the system's resilience from three dimensions: load assurance, recovery speed, and tolerance limits.

[0127] In some embodiments, fault sets and extreme weather scenarios are pre-designed sets of systematic test cases that cover a variety of possible combinations of equipment failures and severe weather conditions, used to verify the robustness of control schemes under abnormal operating conditions.

[0128] For example, embodiments of the present invention can use Monte Carlo simulation to statistically calculate the expected load reduction under hundreds of fault scenarios, comprehensively considering factors such as fault probability, load importance level, and recovery time; based on equipment fault repair models and system reconfiguration strategies, the entire process from the occurrence of a fault to the complete restoration of power supply to the core load is simulated, and the average recovery time is statistically calculated; through continuous power flow analysis and transient stability simulation, the maximum power generation loss or load surge that the system can withstand before collapse is determined; after normalizing the three core indicators, different weights are assigned to them in combination with historical operating data to form a comprehensive resilience evaluation index.

[0129] S202, based on the microgrid-based control scheme and the updated digital twin, the execution results under various disturbance scenarios are simulated and evaluated.

[0130] S203. Based on real-time elasticity indicators and execution results under various disturbance scenarios, determine the risk level of the control plan.

[0131] In some embodiments, the risk level includes low risk, medium risk, and high risk. Risk level: Based on the results of digital twin simulation, it is a qualitative classification of the vulnerability and potential consequences of the control scheme under disturbed environments, providing an intuitive basis for preventive decision-making.

[0132] For example, embodiments of the present invention can establish a multi-level disturbance scenario library including equipment failure, meteorological disasters, and market fluctuations, with each scenario equipped with a complete time-series development model; inject the control scheme to be evaluated into a digital twin, conduct batch simulations under various disturbance scenarios, and record the dynamic response process of the system; analyze the simulation results to identify the weak links and potential cascading failure paths of the control scheme under specific disturbances; and establish a risk matrix based on the degree of deterioration of elasticity indicators and the severity of failure consequences, classifying the control scheme into three levels: low risk, medium risk, and high risk.

[0133] S204. If the risk level is medium or high, the resilience enhancement mode is activated, the weight of the safety and stability target is set to the maximum value, the cooperative game calculation is re-performed, and a backup control scheme for the microgrid is generated.

[0134] In some embodiments, the elastic enhancement mode is a special operating state that is automatically triggered when the system faces a high operational risk. In this mode, the system will temporarily prioritize power supply security and continuity, and correspondingly reduce the importance of conventional optimization objectives such as economy.

[0135] S205. If the internal operating data of the microgrid is detected to exceed the preset risk threshold, the microgrid control scheme will be switched to the backup control scheme.

[0136] In some embodiments, a preset risk threshold is defined as a safety limit for key parameters pre-set according to the system's safe operation requirements. When real-time monitoring data exceeds these thresholds, it indicates that the system has entered a potentially dangerous state and risk mitigation measures need to be taken immediately.

[0137] For example, embodiments of the present invention can adjust the weight coefficient of the security and stability objective to a dominant position within a cooperative game calculation framework, ensuring that resilience becomes the primary consideration in scheme generation; add strict resilience constraints to the optimization model, including backup capacity requirements, network reconfiguration capabilities, and critical load protection levels; based on a cooperative game model with enhanced security weights, quickly generate backup control schemes that emphasize system resilience; and establish an automatic switching logic based on real-time risk monitoring, so that when system operating parameters reach a preset risk threshold, a seamless switch from the conventional scheme to the resilience-enhanced scheme is completed within a single control cycle.

[0138] Thus, this embodiment of the invention upgrades the traditional "passive response" of microgrids to "active defense" by introducing a risk simulation and resilience enhancement mechanism based on digital twins. It can accurately identify potential risks before strategy execution and automatically generate highly resilient backup plans, ensuring that the system can quickly restore power supply under extreme disturbances, significantly improving the power supply reliability and disaster survivability of the microgrid.

[0139] Optionally, the multi-priority target control method based on photovoltaic-storage park-level microgrid provided in this embodiment of the invention further includes steps S301-S306 after step S105.

[0140] S301, Monitor the actual operating data of the microgrid.

[0141] In some embodiments, actual operating data include bus voltage, system frequency, net load power, energy storage charging and discharging power, and energy storage SOC.

[0142] For example, embodiments of the present invention can use a SCADA system to collect core electrical measurement data such as bus voltage amplitude and phase, system frequency dynamics, and power flow direction of each branch at a frequency of seconds; record in real time the accurate values ​​of charging and discharging power of the energy storage system, the state of charge change curve, the operating point of the power converter, and other equipment operating parameters; synchronously collect external environmental parameters such as light intensity, ambient temperature, and humidity, and establish a correlation mapping with electrical response; and adopt multiple verification mechanisms to ensure the accuracy and integrity of the collected data, including data rationality checks, timestamp alignment, and abnormal data identification.

[0143] S302. Based on the control scheme and the updated digital twin, extrapolation and prediction are performed to obtain the predicted operation data of the microgrid.

[0144] S303. Based on actual operating data and predicted operating data, perform deviation calculations to determine the deviation data.

[0145] In some embodiments, the deviation data includes the deviation values ​​of each running data.

[0146] In some embodiments, actual operating data refers to real operating parameters directly collected from the physical microgrid through sensing and measurement devices. It reflects the objective operating state of the system and serves as the benchmark for verifying model accuracy. Predicted operating data refers to the estimated future state of the system calculated using a digital twin based on the current control scheme and system model. It reflects the model's understanding and predictive ability regarding the system's dynamic characteristics. Deviation data refers to the quantified difference between actual observations and model predictions. It reflects the degree of matching between the model and the real system and is a key feedback signal driving model improvement.

[0147] For example, in this embodiment of the invention, a calibrated digital twin is used to simulate the dynamic response of the system over several future scheduling cycles, with the current control scheme as input. A complete system prediction trajectory is generated, including voltage distribution, power flow, frequency change, and equipment status. The predicted data is time-series aligned and compared point-by-point with the actual operating data to calculate the absolute deviation and relative error of each electrical measurement parameter. The statistical characteristics and time-series patterns of the deviation data are analyzed to identify systematic deviations and random fluctuations, and to pinpoint the main sources of model mismatch.

[0148] S304. Based on the deviation data, update the incremental training sample set by considering the system state of the microgrid, the control scheme, and the external conditions when the deviation data is generated.

[0149] In some embodiments, the incremental training sample set is a newly added sample set that is continuously accumulated on the basis of the original training data. It is specifically used to record the model prediction bias and related contextual information, and to support the model in learning new experiences without forgetting the original knowledge.

[0150] For example, in this embodiment of the invention, each prediction deviation is associated and encapsulated with a complete system state snapshot at the time of its occurrence, including comprehensive information such as network topology, device parameters, control mode, and environmental conditions; a sample value evaluation system is established to screen high-value samples based on indicators such as deviation significance, scenario representativeness, and data quality; a management strategy combining first-in-first-out and importance weighting is adopted to maintain a moderately sized and highly representative incremental sample set; and the sample ratio of different operating scenarios is balanced through oversampling and undersampling techniques to ensure the balance of training data.

[0151] S305. Based on the updated training sample set, the intelligent regulation model is retrained to obtain a real-time updated intelligent regulation model.

[0152] In some embodiments, the real-time updated intelligent control model is a control decision model that continuously evolves by learning the latest operating experience and has the dynamic optimization capability to adapt to the time-varying characteristics of the system and changes in the environment.

[0153] S306. Based on a real-time updated intelligent control model, perform multi-priority target control of photovoltaic-storage park-level microgrids.

[0154] For example, in embodiments of the present invention, when incremental samples accumulate to a set scale or significant model degradation occurs, the model retraining process is automatically triggered; fine-tuning and optimization are performed on the existing model parameters, retaining existing knowledge while quickly adapting to new system characteristics; the training objectives simultaneously consider performance improvement and prediction bias reduction, achieving the dual goals of precise control and accurate prediction; and comprehensive performance testing is performed on the updated model in the digital twin to ensure the effectiveness and security of model improvement.

[0155] Thus, this embodiment of the invention establishes a closed-loop online learning mechanism of "monitoring-comparison-learning," enabling the intelligent control model to continuously learn from actual operational deviations. This mechanism effectively solves the problem of model aging and inaccuracy, ensuring that the control strategy always remains consistent with the characteristics of the real system, significantly improving the model's adaptability and control accuracy, and providing core technical support for the long-term stable and optimized operation of microgrids.

[0156] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0157] The following are device embodiments of the present invention. For details not described in detail, please refer to the corresponding method embodiments described above.

[0158] Figure 2 This diagram illustrates a multi-priority target control device for a photovoltaic-storage park-level microgrid provided by an embodiment of the present invention. The control device 400 includes a communication module 401 and a processing module 402.

[0159] The communication module 401 is used to collect internal operating data and external environmental data of the microgrid in real time.

[0160] The processing module 402 is used to evaluate the current operating status and identify the operating scenario based on internal operating data and external environmental data, and to determine the current priority target using fuzzy computing methods; to synchronously update the digital twin of the microgrid based on internal operating data and external environmental data; to generate simulation data based on the digital twin, and to retrain and optimize the intelligent control model based on the simulation data and in combination with a federated learning algorithm, thereby obtaining an optimized intelligent control model; and to generate a microgrid control scheme based on the current priority target and the optimized intelligent control model, using a cooperative game algorithm to balance the optimization targets of each subject and quantify their contributions.

[0161] Figure 3This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. The electronic device 500 includes: a processor 501, a memory 502, and a computer program 503 stored in the memory 502 and executable on the processor 501. When the processor 501 executes the computer program 503, it implements the steps in the above-described method embodiments. Alternatively, when the processor 501 executes the computer program 503, it implements the functions of each module / unit in the above-described device embodiments.

[0162] For example, the computer program 503 may be divided into one or more modules / units, which are stored in the memory 502 and executed by the processor 501 to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program 503 in the electronic device 500.

[0163] The processor 501 may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0164] The memory 502 can be an internal storage unit of the electronic device 500, such as a hard disk or memory of the electronic device 500. The memory 502 can also be an external storage device of the electronic device 500, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc., equipped on the electronic device 500. Furthermore, the memory 502 can include both internal and external storage units of the electronic device 500. The memory 502 is used to store the computer program and other programs and data required by the terminal. The memory 502 can also be used to temporarily store data that has been output or will be output.

[0165] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A multi-priority target control method based on a photovoltaic-storage park-level microgrid, characterized in that, include: Real-time acquisition of internal operating data and external environmental data of the microgrid; Based on the internal operating data and external environmental data, a fuzzy computing method is used to assess the current operating status and identify the operating scenario, determining the current priority target. This includes: extracting features from the internal operating data and external environmental data to obtain multiple input features, including the real-time state of charge of the energy storage system, the current net load power, the current time-of-use electricity price level, and the short-term volatility of photovoltaic output; determining the membership degree of each input feature corresponding to a fuzzy linguistic variable based on the multiple input features and a preset membership function; performing inference based on the membership degree of each input feature corresponding to the fuzzy linguistic variable and a preset fuzzy rule base to obtain a fuzzy set describing the weights of each priority target, whereby the fuzzy rule base describes the mapping relationship between the fuzzy combination of input features and the weights of priority targets; aggregating each priority target in the fuzzy set and using the centroid method for defuzzification calculation to obtain the weights of each priority target; the priority targets include safety and stability targets, economic operation targets, and green and efficient targets; and determining the current priority target based on the weights of each priority target. Based on the internal operating data and external environmental data, the digital twin of the microgrid is updated synchronously; Based on the digital twin, simulation data is generated, and based on the simulation data, combined with the federated learning algorithm, the intelligent control model is retrained and the parameters are optimized to obtain the optimized intelligent control model. Based on the current priority objectives and the optimized intelligent control model, a cooperative game theory algorithm is used to balance the optimization objectives of each subject and quantify their contributions to generate a microgrid control scheme.

2. The multi-priority target control method based on a photovoltaic-storage park-level microgrid according to claim 1, characterized in that, The process of synchronously updating the digital twin of the microgrid based on the internal operating data and external environmental data includes: Based on the internal operating data and external environmental data, the initial state and boundary conditions of the digital twin are synchronized to obtain a synchronized digital twin, so that the static topology and dynamic operating state of the synchronized digital twin are synchronized with the real-time operating state of the microgrid. Based on the internal operating data and external environmental data, as well as the synchronized digital twin, state estimation and parameter identification are used to dynamically calibrate the parameters of key models in the digital twin to obtain a calibrated digital twin; the key models include a photovoltaic array model, an energy storage system model, and a line impedance model. Based on the calibrated digital twin, the implicit state variables and performance indicators of the microgrid are recalculated to obtain an updated digital twin; the implicit state variables include the system's real-time short-circuit capacity and the voltage stability margin of key nodes; the performance indicators include the actual available capacity and health status of the energy storage system.

3. The multi-priority target control method based on photovoltaic-storage park-level microgrids according to claim 1, characterized in that, The generation of simulation data based on the digital twin includes: Based on the current real-time state of the digital twin, the initial conditions for the simulation are set; Based on ultra-short-term forecast data and preset typical scenarios, simulation boundary conditions are set; the boundary conditions include light intensity variation curves, load power variation curves, and grid electricity price signals; Based on the initial conditions and the boundary conditions, random perturbations within a preset range are applied to the key model parameters in the digital twin, and simulation calculations are performed for various typical scenarios. The key model parameters include photovoltaic module efficiency, energy storage internal resistance, and line parameters. The typical scenarios include photovoltaic sudden drop scenario, load surge scenario, and equipment failure scenario. During the simulation process, dynamic response data of the microgrid system simulated by the digital twin is collected and determined as simulation data for each typical scenario; the simulation data includes energy storage charging and discharging power sequence, bus voltage fluctuation sequence, system power deficit sequence and optimization objective function value.

4. The multi-priority target control method based on a photovoltaic-storage park-level microgrid according to claim 1, characterized in that, The intelligent regulation model is retrained and its parameters optimized based on simulation data and a federated learning algorithm to obtain an optimized intelligent regulation model, including: Based on the simulation data and the historical operation data of the microgrid, an incremental training sample set is constructed. The loss function is determined with the optimization objective of minimizing the overall system operating cost and the penalty term for violating security constraints; Based on the incremental training sample set and the loss function, the gradient descent algorithm is used to incrementally train the intelligent regulation model to obtain the updated intelligent regulation model and the updated model parameters of the intelligent regulation model. Based on the updated model parameters of the intelligent control model, a federated learning algorithm is used to optimize the model, resulting in an optimized intelligent control model.

5. The multi-priority target control method based on photovoltaic-storage park-level microgrids according to claim 4, characterized in that, The model parameters updated based on the intelligent control model are then optimized using a federated learning algorithm to obtain the optimized intelligent control model, including: Based on the updated model parameters of the intelligent control model, homomorphic encryption processing is performed to obtain encrypted model parameters. Homomorphic encryption is performed on multiple incremental samples corresponding to the simulation data in the incremental training sample set, the number of incremental samples, and the training rounds in the incremental training process to obtain metadata; Based on the metadata of multiple microgrid nodes on the federated learning platform, the proportion of training samples is calculated to determine the weighted average coefficient of each microgrid node. The weighted average algorithm based on homomorphic encryption uses the weighted average coefficients to securely aggregate the encryption model parameters of multiple microgrid nodes to obtain global update parameters. The global update parameters are decrypted to obtain the global model parameters; Based on the global model parameters and the intelligent control model, the model performance is verified, and the verification results are obtained. If the verification result is successful, the optimized intelligent control model is determined based on the global model parameters.

6. The multi-priority target control method based on a photovoltaic-storage park-level microgrid according to claim 1, characterized in that, Based on the current priority objective and the optimized intelligent control model, a cooperative game theory algorithm is used to balance the optimization objectives of each entity and quantify their contributions to generate a microgrid control scheme, including: Establish benefit functions for each entity, including microgrid operators, energy storage investors, photovoltaic owners, and users. The benefit functions include the network loss cost function of microgrid operators, the energy storage depreciation cost function of energy storage investors, the curtailment loss function of photovoltaic owners, and the electricity satisfaction function of users. Based on the interest functions of each entity, and with the goal of minimizing the overall operating cost of the system, a characteristic function for the cooperative alliance is constructed; the cooperative alliance includes multiple alliances composed of one or more entities. Based on the optimized intelligent control model, and taking the current priority target as a constraint, a preliminary control scheme that satisfies the current priority target is determined. Based on the preliminary control scheme and the characteristic function of the cooperative alliance, the interest distribution vector of each alliance in the cooperative alliance is calculated, and an element in the interest distribution vector represents the interest function value of the corresponding subject in the cooperative state. Iterate through each entity and calculate the marginal contribution of each entity to each alliance; Based on the marginal contribution of each entity to each alliance, a weighted sum is performed to calculate the Shapley value of each entity; the Shapley value is used to characterize the comprehensive contribution of the entity in the microgrid optimization process. By weighting the Shapley values ​​of each subject, the optimization objectives of each subject are weighted to construct a comprehensive objective function; Using the preliminary control scheme as the initial solution and the current priority target as the constraint, the comprehensive objective function is solved to obtain the control scheme of the microgrid.

7. The multi-priority target control method based on a photovoltaic-storage park-level microgrid according to any one of claims 1 to 6, characterized in that, After generating a microgrid control scheme based on the current priority target and the optimized intelligent control model, using a cooperative game algorithm to balance the optimization targets of each subject and quantify their contributions, the process further includes: Based on the updated digital twin, a microgrid fault set and extreme weather scenario are simulated to calculate the real-time resilience index of the microgrid. The real-time resilience index includes the expected load reduction value, the average recovery time, and the maximum disturbance immunity. Based on the microgrid control scheme and the updated digital twin, the execution results under various disturbance scenarios are simulated and evaluated. Based on the real-time elasticity index and the execution results under the various disturbance scenarios, the risk level of the control scheme is determined, and the risk level includes low risk, medium risk and high risk. If the risk level is medium or high, the resilience enhancement mode is activated, the weight of the safety and stability target is set to the maximum value, the cooperative game calculation is re-performed, and a backup control scheme for the microgrid is generated. If the internal operating data of the microgrid is detected to exceed the preset risk threshold, the microgrid's control scheme will be switched to the backup control scheme.

8. The multi-priority target control method based on a photovoltaic-storage park-level microgrid according to any one of claims 1 to 6, characterized in that, The method further includes: Monitor the actual operating data of the microgrid; the actual operating data includes bus voltage, system frequency, net load power, energy storage charging and discharging power, and energy storage SOC; Based on the aforementioned control scheme and the updated digital twin, extrapolation and prediction are performed to obtain the predicted operating data of the microgrid. Based on the actual operating data and the predicted operating data, deviation calculation is performed to determine the deviation data; the deviation data includes the deviation value of each operating data. Based on the deviation data, the system state of the microgrid, the control scheme, and the external conditions at the time the deviation data was generated are used to update the incremental training sample set. Based on the updated training sample set, the intelligent regulation model is retrained to obtain a real-time updated intelligent regulation model. Based on a real-time updated intelligent control model, multi-priority target control is carried out in the photovoltaic-storage park-level microgrid.

9. A multi-priority target control system based on a photovoltaic-storage park-level microgrid, characterized in that, The control system includes an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor is used to call and run the computer program stored in the memory to perform the method as described in any one of claims 1 to 8.

Images