Electric power virtual simulation optimization method and system based on digital twinning technology

Abstract

The invention relates to an electric power virtual simulation optimization method based on a digital twin technology, and the method comprises the following steps: S1, deploying sensors at key nodes of an electric power system, and integrating an intelligent electric meter, weather forecast, social media information and market data to form a comprehensive data pool; s2, constructing a digital twinborn model of the power system; s3, optimizing the digital twin physical model based on a Bayesian optimization method and an adversarial neural network; s4, applying a bee colony algorithm to formulate a cooperation strategy, and optimizing transmission and load adjustment of power in different areas; s5, introducing an adaptive neural fuzzy inference system, and further optimizing a cooperation strategy; and S6, based on the optimized digital twinborn model and the operation strategy, performing large-scale distributed simulation by using cloud resources. According to the invention, accurate construction of the electric power simulation model and accurate implementation of electric power virtual simulation are realized.

Description

Technical Field

[0001] The present invention relates to the field of power optimization, and in particular to a power virtual simulation optimization method and system based on digital twin technology. Background Art

[0002] The power system is rapidly transforming to adapt to the rapid growth of energy demand and the widespread use of renewable energy. However, these changes have brought many new challenges, including load fluctuations, complex management of distributed generation access, and increasingly stringent environmental and safety standards. Faced with these challenges, power companies and operators must ensure the stability and efficiency of the system under extreme conditions.

[0003] Therefore, power companies use simulation to help identify potential defects in strategies and operational risks, and verify new control strategies and technology integration solutions. With the support of cloud resources, distributed simulation makes real-time experiments on large-scale and complex systems possible. This capability is crucial for strategy verification and performance evaluation before the deployment of new technologies. Summary of the invention

[0004] In order to solve the above problems, the purpose of the present invention is to provide a power virtual simulation optimization method and system based on digital twin technology, which realizes the precise construction of power simulation models and the precise implementation of large-scale simulations, which not only improves the operating efficiency and management level of the power system, but also significantly enhances its adaptability in variable environments.

[0005] To achieve the above object, the present invention adopts the following technical solutions:

[0006] A power virtual simulation optimization method based on digital twin technology includes the following steps:

[0007] Step S1: Deploy sensors at key nodes of the power system, use IoT technology to transmit the collected data to the central data platform, and integrate smart meters, weather forecasts, social media intelligence and market data to form a comprehensive data pool;

[0008] Step S2: Introduce a hybrid modeling method of physical and data-driven models to build a digital twin model of the power system;

[0009] Step S3: Based on the Bayesian optimization method, the model parameters are adjusted in real time, the deviations are automatically corrected, and the adversarial neural network is used to generate virtual data for model training and performance evaluation to optimize the digital twin physical model;

[0010] Step S4: Apply the bee colony algorithm to perform cross-regional and cross-system comprehensive optimization, formulate a collaborative strategy, and optimize the transmission and load adjustment of electricity in different regions;

[0011] Step S5: introducing an adaptive neuro-fuzzy inference system to further optimize the collaboration strategy;

[0012] Step S6: Based on the optimized digital twin model and operation strategy, use cloud resources to perform large-scale distributed simulation.

[0013] Furthermore, step S1 is specifically as follows:

[0014] Deploy sensors at key nodes of substations, transmission lines, and generators, and build a star network topology of sensors;

[0015] In the Internet of Things, the lightweight protocol CoAP is used for data transmission to transmit the collected data to the central data platform;

[0016] Adopt data lake architecture, use NoSQL database to process unstructured data, build ETL process, extract data from various sources, integrate smart meters, weather forecasts, social media intelligence and market data to form a comprehensive data pool.

[0017] Furthermore, step S2 is specifically as follows:

[0018] Construct physical models, including equivalent circuit models of transmission lines and dynamic models of generators;

[0019] Use historical operation data and sensor output to train supervised learning models (including LSTM and random forest) and build data-driven models;

[0020] The prediction y provided by the physical model physical Generate reference values ​​as a priori constraints for system operation:

[0021]

[0022] Use data-driven models to monitor real-time data and perform anomaly detection on sample output ydata-driven. When the output value is detected to exceed the threshold range Δy, the constraints of the physical model are triggered.

[0023]

[0024] In the event of anomalies being detected, corrections are made in conjunction with prior predictions from the physical model:

[0025]

[0026] Among them, α is the fusion coefficient.

[0027] Furthermore, based on the Bayesian optimization method, the model parameters are adjusted in real time to automatically correct the deviation, specifically:

[0028] Define the objectives of the power system digital twin model, improve the prediction accuracy and dynamic response capability of the power system under different environmental and load conditions, and adaptively adjust model parameters to minimize prediction errors;

[0029] Define the root mean square error as the optimization objective function:

[0030]

[0031] Among them, θ is the parameter of the power system digital twin model; represents the output value predicted by the power system digital twin model; y true is the true value;

[0032] Initialize the digital twin model using existing data and select parameter θ0 as the starting point:

[0033]

[0034] Among them, β is the weight coefficient, μ t (θ) is the predicted mean, σ t (θ) is the prediction standard deviation;

[0035] Through Gaussian process regression, the unknowable distribution of the target function is fitted to predict the expected effect and uncertainty of different parameter combinations; the Gaussian process is updated through the Bayesian formula, and the parameter set θ is selected after each iteration to evaluate and update the parameters.

[0036] Furthermore, the unknowable distribution of the objective function is fitted through Gaussian process regression to predict the expected effects and uncertainties of different parameter combinations. The Gaussian process is updated through the Bayesian formula, and the parameter set θ is selected after each iteration to evaluate and update the parameters, as follows:

[0037] For a new input point θ*, predict its corresponding function value:

[0038] μ(θ * )=k(θ * ,θ) T [K+σ 2 I] -1 y

[0039] σ 2 (θ * )=k(θ * ,θ * )-k(θ * ,θ) T [K+σ 2 I] -1 k(θ * ,θ);

[0040] Among them, μ(θ* ) is the predicted mean, σ 2 (θ * ) is the variance of the prediction, K is the covariance matrix of the training data, σ 2 is the noise variance;

[0041] Calculate the expected improvement EI:

[0042]

[0043] Among them, θ is the current optimal parameter configuration;

[0044] After obtaining new data points, the posterior distribution of the Gaussian process is updated using the Bayesian formula;

[0045] In each iteration, select θ* that maximizes the acquisition function;

[0046]

[0047] Furthermore, adversarial neural networks are used to generate virtual data for model training and performance evaluation to optimize the digital twin physical model, as follows:

[0048] Construct the generator G(z;θ g ) and the discriminator D(x;θ d ), and initialize the generator and discriminator parameters θ g and θ d ;

[0049] Based on the adversarial target formula, a training cycle is performed:

[0050]

[0051] in, represents the expected value of all real data samples; p data Represents the distribution of real data; logD(x) represents the probability that the output of the discriminator for the real data sample xx is the real category; represents the expected value of all generated samples; p z is the distribution of the noise source; G(z) is the sample generated by the generator;

[0052] Discriminator update:

[0053] Generate x by random sampling from real data;

[0054] Sampling z from the noise distribution pz, calculate the data G(z) generated by the generator:

[0055] Calculate the discriminator loss as the sum of the two target values;

[0056] Update the discriminator parameters θd To maximize losses;

[0057] Generator Updates:

[0058] Generate z by sampling from the noise distribution pz;

[0059] Generate samples through the generator G(z) and calculate the loss of the discriminator;

[0060] Update the generator parameters θg to minimize the loss;

[0061] The above steps are iteratively updated until the generator and discriminator reach a balance, and virtual data for model training and performance evaluation are generated based on the trained generator.

[0062] Furthermore, S4 is specifically:

[0063] Establish a fitness function to evaluate and optimize the solution:

[0064] Fitness=w1·f loss (x)+w2·f cost (x)+w3·f balance (x);

[0065] Among them, x is the current solution, which represents the power configuration plan; f loss (x) is the function of transmission loss; f cost (x) is the cost function; f balance (x) is the load balancing function;

[0066] Based on the swarm algorithm, optimize the transmission and load adjustment of electricity in different areas:

[0067] Initialize the population:

[0068] Randomly initialize a set of candidate solutions in the solution space;

[0069] Each solution represents a generation and load dispatch plan.

[0070] Worker bee stage:

[0071] Search for better solutions based on the neighborhood of existing solutions;

[0072] x new =x current +φ(x current -x neighbor );

[0073] Where φ is a random number between [-1, 1];

[0074] Based on the fitness function, calculate the fitness of each solution;

[0075] Update the solution. If the new solution is better than the old one, accept the new solution.

[0076] Scout Bee Phase:

[0077] Randomly select some low-fitness solutions and replace them with new random solutions;

[0078] Ensure that the algorithm can escape from the local optimum and improve the diversity of solutions;

[0079] Update and select:

[0080] Select the solution with the highest fitness and keep it as the current optimal solution, and continue to the next round of search.

[0081] Furthermore, S5 is specifically:

[0082] S51: Define optimization objectives, including improving the operating efficiency and stability of the power system and minimizing system costs and energy consumption; collect historical operating data and real-time measurement data, and clarify input parameters (such as power load, power generation, environmental factors) and output decisions (such as power scheduling and load transfer strategies).

[0083] S52: Build the ANFIS model and select key influencing factors as input, including load demand, meteorological data, and equipment status; and standardize and normalize the data to improve the efficiency and accuracy of model training;

[0084] Based on the Gaussian membership function, the input variables are converted into fuzzy expressions, and the initial parameters are set, including the center and width of the membership function;

[0085] S53: construct a fuzzy rule base;

[0086] S54: Use historical data to train the model and optimize the parameters by combining gradient descent and least squares method;

[0087] Verify the model's adaptability to the test data and adjust the model parameters based on error feedback.

[0088] S55: During the operation phase, input data is collected in real time and input into the trained ANFIS model for reasoning; and optimized operation strategy recommendations are obtained;

[0089] S56: According to the strategy recommendations output by ANFIS, adjust the operating parameters of the power system, monitor the execution effect, and provide real-time feedback to the ANFIS system through sensor and system log analysis.

[0090] A power virtual simulation optimization system based on digital twin technology includes a processor, a memory, and a computer program stored in the memory. When the processor executes the computer program, it specifically executes the steps in the power virtual simulation optimization method based on digital twin technology as described above.

[0091] The present invention has the following beneficial effects:

[0092] 1. The present invention realizes the accurate construction of power simulation model and the accurate implementation of large-scale simulation, which not only improves the operation efficiency and management level of the power system, but also significantly enhances its adaptability in variable environments;

[0093] 2. The present invention enhances the accuracy of the power system digital twin model and its agile response capability to uncertain events by introducing a hybrid modeling method of physical and data-driven models. The power system digital twin model ensures the basic constraints and accurate description of the system, while the data-driven model enhances the system's ability to respond to real-time changes and complex responses. The combination of the two greatly improves the overall performance.

[0094] 3. The present invention combines Bayesian optimization and GAN technology to achieve more accurate and efficient application of digital twin models in power systems. Bayesian optimization continuously optimizes system parameters by minimizing prediction errors, while GAN enhances the learning ability of data-driven models by generating diverse and authoritative virtual data.

[0095] 4. The present invention combines adaptive neural-fuzzy inference system (ANFIS), machine learning, swarm intelligence algorithm and cloud computing containerization technology, which can improve the overall performance and optimization capabilities of the digital twin model. Through collaborative optimization and simulation, it can adapt to the rapidly changing operating environment and always maintain an optimized state. BRIEF DESCRIPTION OF THE DRAWINGS

[0096] Figure 1 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0097] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:

[0098] refer to Figure 1 In this embodiment, a power virtual simulation optimization method based on digital twin technology is provided, comprising the following steps:

[0099] Step S1: Deploy sensors at key nodes of the power system (such as substations, transmission lines, generators, etc.) to collect various physical data such as voltage, current, temperature, humidity, etc. in real time, use IoT technology to transmit the collected data to the central data platform, and integrate smart meters, weather forecasts, social media intelligence and market data to form a comprehensive data pool;

[0100] Step S2: Introduce a hybrid modeling method of physical and data-driven models to build a digital twin model of the power system;

[0101] Step S3: Based on the Bayesian optimization method, the model parameters are adjusted in real time, the deviations are automatically corrected, and the adversarial neural network (GAN) is used to generate virtual data for model training and performance evaluation to optimize the digital twin physical model;

[0102] Step S4: Apply the bee colony algorithm to perform cross-regional and cross-system comprehensive optimization, formulate a collaborative strategy, and optimize the transmission and load adjustment of electricity in different regions;

[0103] Step S5: introducing an adaptive neuro-fuzzy inference system to further optimize the collaboration strategy;

[0104] Step S6: Based on the optimized digital twin model and operation strategy, use cloud resources to perform large-scale distributed simulation.

[0105] In this embodiment, step S1 is specifically as follows:

[0106] Deploy sensors at key nodes of substations, transmission lines, and generators, including temperature sensors, voltage and current sensors, humidity sensors, etc., and build a star network topology of sensors;

[0107] In the Internet of Things, a lightweight protocol CoAP (Constrained Application Protocol) is used for data transmission to transfer the collected data to the central data platform;

[0108] Adopt data lake architecture, use NoSQL databases (such as Cassandra, MongoDB) to process unstructured data, build ETL (extraction, transformation, loading) processes, extract data from various sources, integrate smart meters, weather forecasts, social media intelligence and market data to form a comprehensive data pool.

[0109] In this embodiment, step S2 is specifically as follows:

[0110] Construct physical models, including equivalent circuit models of transmission lines and dynamic models of generators;

[0111] The equivalent circuit model of the transmission line adopts the π network model:

[0112] Z=R+jωL;

[0113] Y = G + jωC;

[0114] Where R is resistance, L is inductance, G is conductance, C is capacitance, and ω is angular frequency;

[0115] The generator dynamic model considers the dynamic behavior in the power system through the dynamic equations of the motor:

[0116] Rotor equation:

[0117] Electromagnetic relationship: V = E-jX s I;

[0118] Where J is the moment of inertia, ω m is the mechanical speed, T m and T e are mechanical and electromagnetic torques respectively, V is the terminal voltage, E is the electromotive force, and X s is the synchronous reactance, I is the current;

[0119] Use historical operation data and sensor output to train supervised learning models (including LSTM and random forest) and build data-driven models;

[0120] The prediction y provided by the physical model physical Generate reference values ​​as a priori constraints for system operation:

[0121]

[0122] Use data-driven models to monitor real-time data and perform anomaly detection on sample output ydata-driven. When the output value is detected to exceed the threshold range Δy, the constraints of the physical model are triggered.

[0123]

[0124] In the event of anomalies being detected, corrections are made in conjunction with prior predictions from the physical model:

[0125]

[0126] Among them, α is the fusion coefficient.

[0127] In this embodiment, based on the Bayesian optimization method, the model parameters are adjusted in real time to automatically correct the deviation, specifically:

[0128] Define the objectives of the power system digital twin model, improve the prediction accuracy and dynamic response capability of the power system under different environmental and load conditions, and adaptively adjust model parameters to minimize prediction errors;

[0129] Define the root mean square error as the optimization objective function:

[0130]

[0131] Among them, θ is the parameter of the power system digital twin model; represents the output value predicted by the power system digital twin model; y true is the true value;

[0132] Initialize the digital twin model using existing data and select parameter θ0 as the starting point:

[0133]

[0134] Among them, β is the weight coefficient, μ t (θ) is the predicted mean, σ t (θ) is the prediction standard deviation;

[0135] Through Gaussian process regression, the unknowable distribution of the target function is fitted to predict the expected effect and uncertainty of different parameter combinations; the Gaussian process is updated through the Bayesian formula, and the parameter set θ is selected after each iteration to evaluate and update the parameters.

[0136] In this embodiment, Gaussian process regression is used to fit the unknowable distribution of the objective function to predict the expected effects and uncertainties of different parameter combinations. The Gaussian process is updated by the Bayesian formula, and a parameter set θ is selected after each iteration to evaluate and update the parameters, as follows:

[0137] For a new input point θ*, predict its corresponding function value:

[0138] μ(θ * )=k(θ * ,θ) T [K+σ 2 I] -1 y

[0139] σ 2 (θ * )=k(θ * ,θ * )-k(θ * ,θ) T [K+σ 2 I] -1 k(θ * ,θ);

[0140] Among them, μ(θ * ) is the predicted mean, σ 2 (θ * ) is the variance of the prediction, K is the covariance matrix of the training data, σ 2 is the noise variance;

[0141] Calculate the expected improvement EI:

[0142]

[0143] Among them, θ is the current optimal parameter configuration;

[0144] After obtaining new data points, the posterior distribution of the Gaussian process is updated using the Bayesian formula;

[0145] In each iteration, select θ* that maximizes the acquisition function;

[0146]

[0147] In this embodiment, an adversarial neural network is used to generate virtual data for model training and performance evaluation to optimize the digital twin physical model, as follows:

[0148] Construct the generator G(z;θ g ) and the discriminator D(x;θ d ), and initialize the generator and discriminator parameters θ g and θ d ;

[0149] Based on the adversarial target formula, a training cycle is performed:

[0150]

[0151] in, represents the expected value of all real data samples; p data Represents the distribution of real data; logD(x) represents the probability that the output of the discriminator for the real data sample xx is the real category; represents the expected value of all generated samples; p z is the distribution of the noise source; G(z) is the sample generated by the generator;

[0152] Discriminator update:

[0153] Generate x by random sampling from real data;

[0154] Sampling z from the noise distribution pz, calculate the data G(z) generated by the generator:

[0155] Calculate the discriminator loss as the sum of the two target values;

[0156] Update the discriminator parameters θ d To maximize losses;

[0157] Generator Updates:

[0158] Generate z by sampling from the noise distribution pz;

[0159] Generate samples through the generator G(z) and calculate the loss of the discriminator;

[0160] Update the generator parameters θg to minimize the loss;

[0161] The above steps are iteratively updated until the generator and discriminator reach a balance, and virtual data for model training and performance evaluation are generated based on the trained generator.

[0162] In this embodiment, S4 is specifically:

[0163] Establish a fitness function to evaluate and optimize the solution:

[0164] Fitness=w1·f loss (x)+w2·f cost (x)+w3·f balance (x);

[0165] Among them, x is the current solution, which represents the power configuration plan; f loss (x) is the function of transmission loss; f cost (x) is the cost function; f balance (x) is the load balancing function;

[0166] Based on the swarm algorithm, optimize the transmission and load adjustment of electricity in different areas:

[0167] Initialize the population:

[0168] Randomly initialize a set of candidate solutions in the solution space;

[0169] Each solution represents a generation and load dispatch plan.

[0170] Worker bee stage:

[0171] Search for better solutions based on the neighborhood of existing solutions;

[0172] x new =x current +φ(x current -x neighbor );

[0173] Where φ is a random number between [-1, 1];

[0174] Based on the fitness function, calculate the fitness of each solution;

[0175] Update the solution. If the new solution is better than the old one, accept the new solution.

[0176] Scout Bee Phase:

[0177] Randomly select some low-fitness solutions and replace them with new random solutions;

[0178] Ensure that the algorithm can escape from the local optimum and improve the diversity of solutions;

[0179] Update and select:

[0180] Select the solution with the highest fitness and keep it as the current optimal solution, and continue to the next round of search.

[0181] In this embodiment, S5 is specifically:

[0182] S51: Define optimization objectives, including improving the operating efficiency and stability of the power system and minimizing system costs and energy consumption; collect historical operating data and real-time measurement data, and clarify input parameters (such as power load, power generation, environmental factors) and output decisions (such as power scheduling and load transfer strategies).

[0183] S52: Build the ANFIS model, select key influencing factors as input, including load demand, meteorological data, and equipment status; and standardize and normalize the data to improve the efficiency and accuracy of model training;

[0184] Based on the Gaussian membership function, the input variables are converted into fuzzy expressions, and the initial parameters are set, including the center and width of the membership function;

[0185] S53: construct a fuzzy rule base;

[0186] Build "if-then" rule sets by combining expert knowledge with historical data analysis.

[0187] Rule Example: If load demand is high and price is low, increase generation.

[0188] S54: Use historical data to train the model and optimize the parameters by combining gradient descent and least squares method;

[0189] Verify the model's adaptability to the test data and adjust the model parameters based on error feedback.

[0190] S55: In the operation phase, input data is collected in real time and input into the trained ANFIS model for reasoning, and optimized operation strategy recommendations are obtained, such as load scheduling optimization and power generation switching decision.

[0191] S56: According to the strategy recommendations output by ANFIS, adjust the operating parameters of the power system, monitor the execution effect, and provide real-time feedback to the ANFIS system through sensor and system log analysis.

[0192] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0193] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0194] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0195] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0196] The above is only a preferred embodiment of the present invention, and does not limit the present invention in other forms. Any technician familiar with the profession may use the above disclosed technical content to change or modify it into an equivalent embodiment with equivalent changes. However, any simple modification, equivalent change and modification made to the above embodiment according to the technical essence of the present invention without departing from the technical solution of the present invention still belongs to the protection scope of the technical solution of the present invention.

Claims

1. A power virtual simulation optimization method based on digital twin technology, characterized in that: The following steps are involved: Step S1: Deploy sensors at key nodes of the power system, use IoT technology to transmit the collected data to the central data platform, and integrate smart meters, weather forecasts, social media intelligence and market data to form a comprehensive data pool; Step S2: Introduce a hybrid modeling method of physical and data-driven models to build a digital twin model of the power system; Step S3: Based on the Bayesian optimization method, the model parameters are adjusted in real time, the deviations are automatically corrected, and the adversarial neural network is used to generate virtual data for model training and performance evaluation to optimize the digital twin physical model; Step S4: Apply the bee colony algorithm to perform cross-regional and cross-system comprehensive optimization, formulate a collaborative strategy, and optimize the transmission and load adjustment of electricity in different regions; Step S5: introducing an adaptive neuro-fuzzy inference system to further optimize the collaboration strategy; Step S6: Based on the optimized digital twin model and operation strategy, use cloud resources to perform large-scale distributed simulation.

2. The power virtual simulation optimization method based on digital twin technology according to claim 1 is characterized in that: The step S1 is specifically as follows: Deploy sensors at key nodes of substations, transmission lines, and generators, and build a star network topology of sensors; In the Internet of Things, the lightweight protocol CoAP is used for data transmission to transmit the collected data to the central data platform; Adopt data lake architecture, use NoSQL database to process unstructured data, build ETL process, extract data from various sources, integrate smart meters, weather forecasts, social media intelligence and market data to form a comprehensive data pool.

3. The power virtual simulation optimization method based on digital twin technology according to claim 1 is characterized in that: The step S2 is specifically as follows: Construct physical models, including equivalent circuit models of transmission lines and dynamic models of generators; Use historical operation data and sensor output to train supervised learning models (including LSTM and random forest) and build data-driven models; The prediction y provided by the physical model physical Generate reference values ​​as a priori constraints for system operation: Use data-driven models to monitor real-time data and perform anomaly detection on sample output ydata-driven. When the output value is detected to exceed the threshold range Δy, the constraints of the physical model are triggered. In the event of anomalies being detected, corrections are made in conjunction with prior predictions from the physical model: Among them, α is the fusion coefficient.

4. The power virtual simulation optimization method based on digital twin technology according to claim 1 is characterized in that: The Bayesian optimization method is used to adjust the model parameters in real time and automatically correct the deviation, specifically: Define the objectives of the power system digital twin model, improve the prediction accuracy and dynamic response capability of the power system under different environmental and load conditions, and adaptively adjust model parameters to minimize prediction errors; Define the root mean square error as the optimization objective function: Among them, θ is the parameter of the power system digital twin model; represents the output value predicted by the power system digital twin model; y true is the true value; Initialize the digital twin model using existing data and select parameter θ0 as the starting point: Among them, β is the weight coefficient, μ t (θ) is the predicted mean, σ t (θ) is the prediction standard deviation; Through Gaussian process regression, the unknowable distribution of the target function is fitted to predict the expected effect and uncertainty of different parameter combinations; the Gaussian process is updated through the Bayesian formula, and the parameter set θ is selected after each iteration to evaluate and update the parameters.

5. The power virtual simulation optimization method based on digital twin technology according to claim 4 is characterized in that: The Gaussian process regression is used to fit the unknowable distribution of the target function to predict the expected effect and uncertainty of different parameter combinations; the Gaussian process is updated through the Bayesian formula, and the parameter set θ is selected after each iteration to evaluate and update the parameters, as follows: For a new input point θ * , predict its corresponding function value: Among them, μ(θ * ) is the predicted mean, σ 2 (θ * ) is the variance of the prediction, K is the covariance matrix of the training data, σ 2 is the noise variance; Calculate the expected improvement EI: Among them, θ + is the current optimal parameter configuration; After obtaining new data points, the posterior distribution of the Gaussian process is updated using the Bayesian formula; In each iteration, we select θ that maximizes the acquisition function * ; 6. The power virtual simulation optimization method based on digital twin technology according to claim 1 is characterized in that: The adversarial neural network is used to generate virtual data for model training and performance evaluation, and optimize the digital twin physical model, as follows: Construct the generator G(z;θ g ) and the discriminator D(x;θ d ), and initialize the generator and discriminator parameters θ g and θ d ; Based on the adversarial target formula, a training cycle is performed: in, represents the expected value of all real data samples; p data Represents the distribution of real data; logD(x) represents the probability that the output of the discriminator for the real data sample xx is the real category; represents the expected value of all generated samples; p z is the distribution of the noise source; G(z) is the sample generated by the generator; Discriminator update: Generate x by random sampling from real data; From the noise distribution p z Sampling generates z, and calculating the data G(z) generated by the generator: Calculate the discriminator loss as the sum of the two target values; Update the discriminator parameters θ d To maximize losses; Generator Updates: From the noise distribution p z The sampling generates z; Generate samples through the generator G(z) and calculate the loss of the discriminator; Update the generator parameters θ g To minimize losses; The above steps are iteratively updated until the generator and discriminator reach a balance, and virtual data for model training and performance evaluation are generated based on the trained generator.

7. The power virtual simulation optimization method based on digital twin technology according to claim 1 is characterized in that: The S4 is specifically: Establish a fitness function to evaluate and optimize the solution: Fitness=w1·f loss (x)+w2·f cost (x)+w3·f balance (x); Among them, x is the current solution, which represents the power configuration plan; f loss (x) is the function of transmission loss; f cost (x) is the cost function; f balance (x) is the load balancing function; Based on the swarm algorithm, optimize the transmission and load adjustment of electricity in different areas: Initialize the population: Randomly initialize a set of candidate solutions in the solution space; Each solution represents a generation and load dispatch plan. Worker bee stage: Search for better solutions based on the neighborhood of existing solutions; x new =x current +φ(x current -x neighbor ); Where φ is a random number between [-1,1]; Based on the fitness function, calculate the fitness of each solution; Update the solution. If the new solution is better than the old one, accept the new solution. Scout Bee Phase: Randomly select some low-fitness solutions and replace them with new random solutions; Ensure that the algorithm can escape from the local optimum and improve the diversity of solutions; Update and select: Select the solution with the highest fitness and keep it as the current optimal solution, and continue to the next round of search.

8. The power virtual simulation optimization method based on digital twin technology according to claim 1 is characterized in that: The S5 is specifically: S51: Define optimization objectives, including improving the operating efficiency and stability of the power system and minimizing system costs and energy consumption; collect historical operating data and real-time measurement data, and clarify input parameters (such as power load, power generation, environmental factors) and output decisions (such as power scheduling and load transfer strategies). S52: Build the ANFIS model, select key influencing factors as input, including load demand, meteorological data, and equipment status; and standardize and normalize the data to improve the efficiency and accuracy of model training; Based on the Gaussian membership function, the input variables are converted into fuzzy expressions, and the initial parameters are set, including the center and width of the membership function; S53: construct a fuzzy rule base; S54: Use historical data to train the model and optimize the parameters by combining gradient descent and least squares method; Verify the model's adaptability to the test data and adjust the model parameters based on error feedback. S55: During the operation phase, input data is collected in real time and input into the trained ANFIS model for reasoning; and optimized operation strategy recommendations are obtained; S56: According to the strategy recommendations output by ANFIS, adjust the operating parameters of the power system, monitor the execution effect, and provide real-time feedback to the ANFIS system through sensor and system log analysis.

9. A power virtual simulation optimization system based on digital twin technology, characterized in that: It includes a processor, a memory, and a computer program stored in the memory. When the processor executes the computer program, it specifically performs the steps in the power virtual simulation optimization method based on digital twin technology as described in any one of claims 1 to 8.

Images