Automatic power grid dispatching system based on digital twinning

Abstract

The invention relates to power grid dispatching, in particular to an automatic power grid dispatching system based on digital twinning, which comprises a data perception and fusion module, acquires multi-source heterogeneous data of a power grid in real time, constructs a unified data lake and knowledge graph and provides high-quality input for digital twinning modeling; the digital twinning modeling module is used for constructing equipment-level, unit-level and system-level digital twinning models of the power grid and realizing real-time synchronous mapping of a physical power grid and a virtual model; the optimal dispatching module is used for solving the power grid dispatching model by adopting a cordyceps sinensis optimization algorithm and generating an optimal power grid dispatching strategy; the decision execution and feedback module is used for converting the optimal power grid dispatching strategy into a corresponding dispatching instruction and issuing the dispatching instruction to an AGC / AVC system, and verifying the dispatching effect in real time through a digital twin model to form closed-loop control; according to the technical scheme provided by the invention, the defect that the global optimal scheduling strategy is difficult to accurately and quickly find in the prior art can be effectively overcome.

Description

TECHNICAL FIELD

[0001] The present application relates to power grid dispatching, in particular to a power grid automatic dispatching system based on digital twinning. BACKGROUND

[0002] In the field of power grid operation management, achieving accurate, efficient and reliable dispatching is the key to ensuring stable operation of the power system, improving energy utilization efficiency and meeting diversified power demand. With the continuous expansion of the power grid scale and the increasing level of intelligentization, the power grid structure is becoming increasingly complex, and the data involved in the operation process is extensive in source and diverse in type. How to effectively collect, integrate and utilize these massive data has become the primary challenge faced by power grid dispatching.

[0003] Traditional power grid dispatching systems have limitations in data processing, making it difficult to deeply integrate and uniformly manage multi-source heterogeneous data, and unable to provide comprehensive and high-quality data support for dispatching decisions. At the same time, traditional power grid dispatching models often fail to accurately reflect the real-time operating state of complex power grids, and there is a lack of effective real-time synchronization mapping mechanism between physical power grids and virtual models, making it difficult for power grid dispatching strategies to closely match the actual operation of the power grid.

[0004] In terms of solving power grid dispatching strategies, many existing algorithms are difficult to achieve an ideal balance between global search and local development capabilities. When faced with complex power grid dispatching problems, it is easy to fall into local optimization, and it is difficult to accurately and quickly find a globally optimal dispatching strategy, thereby affecting the economy and stability of power grid operation.

[0005] In addition, traditional power grid dispatching systems lack effective verification feedback mechanisms after decision execution, making it difficult to assess dispatching effects in real time and dynamically adjust power grid dispatching strategies, unable to form a closed-loop control, and difficult to adapt to changing operating conditions and demands during power grid operation. Therefore, it is of great practical significance to develop a new type of power grid automatic dispatching system that can overcome the above shortcomings. SUMMARY

[0006] (I) Technical problems solved In view of the above-mentioned shortcomings of the prior art, the present application provides a power grid automatic dispatching system based on digital twinning, which can effectively overcome the defect that it is difficult to accurately and quickly find a globally optimal dispatching strategy.

[0007] (II) Technical solutions To achieve the above purpose, the present application is realized by the following technical solutions: The power grid automatic dispatching system based on digital twinning comprises: The data perception and fusion module collects multi-source heterogeneous data of the power grid in real time, and constructs a unified data lake and knowledge graph to provide high-quality input for digital twin modeling. The digital twin modeling module constructs device-level, unit-level, and system-level digital twin models of the power grid to realize real-time synchronous mapping of the physical power grid and the virtual model. The optimal scheduling module uses the Cordyceps optimization algorithm to solve the power grid scheduling model and generates the optimal power grid scheduling strategy. The decision execution and feedback module converts the optimal power grid scheduling strategy into corresponding scheduling instructions and issues them to the AGC / AVC system, and verifies the scheduling effect in real time through the digital twin model to form a closed-loop control. The Cordyceps optimization algorithm simulates the parasitic behavior of Cordyceps by constructing a two-stage search framework that balances global exploration and local development: In the global exploration stage, the fluctuation propulsion operator and optimal parasitic behavior are used to simulate the diffusion behavior of Cordyceps in the horizontal dimension and the selection of optimal growth conditions, enhancing the global search capability and avoiding falling into local optima. In the local development stage, the spiral ascent operator and re-parasitic behavior are used to simulate the focused search of Cordyceps in the vertical dimension and the competitive parasitism of high-quality hosts, achieving fine search and accelerating convergence to the global optimum.

[0008] Preferably, the data perception and fusion module collects multi-source heterogeneous data of the power grid in real time, and constructs a unified data lake and knowledge graph to provide high-quality input for digital twin modeling, including: Real-time collection of multi-source heterogeneous data of the power grid, including SCADA measurement data, device status data, and weather data; Through data cleaning, time series alignment, and feature extraction, a unified data lake and knowledge graph are constructed to provide high-quality input for digital twin modeling.

[0009] Preferably, the LSTM network is used for feature extraction, and the hidden layer of the LSTM network is represented by the following formula: ; Where h(t) and h(t-1) are the hidden states at time t and time t-1, respectively, [h(t-1), x(t)] represents an input vector formed by concatenating the hidden state h(t-1) at time t-1 and the input x(t) at time t, W f is a weight matrix, b f is a bias term, and Sigmoid is a sigmoid activation function.

[0010] Preferably, the digital twin modeling module constructs a three-level digital twin model of the power grid at the device level, unit level and system level, realizes real-time synchronous mapping of the physical power grid and the virtual model, including: A finite element model of the device is built using ANSYS Twin Builder, and real-time synchronization of data between the edge computing node and the cloud is realized using the OpenFMB protocol to construct a device-level digital twin model; Based on the device-level digital twin model, units are combined according to device functions, the interaction logic and data flow between devices in the unit are clarified, and a unit-level digital twin model is constructed; Based on graph computing technology, a power grid topology model is constructed, and a system-level digital twin model is constructed by using node voltage equations to describe electrical relationships to support multi-physical field coupling simulation.

[0011] Preferably, the optimal scheduling module uses the Cordyceps optimization algorithm to solve the power grid scheduling model to generate an optimal power grid scheduling strategy, including: S1, determining the objective function of power grid scheduling; S2, determining the constraint condition of power grid scheduling; S3, combining the objective function and constraint condition of power grid scheduling, and the three-level digital twin model of the power grid at the device level, unit level and system level, to construct a power grid scheduling model; S4, using the Cordyceps optimization algorithm to solve the power grid scheduling model to generate an optimal power grid scheduling strategy that meets the constraint condition.

[0012] Preferably, S4 uses the Cordyceps optimization algorithm to solve the power grid scheduling model to generate an optimal power grid scheduling strategy that meets the constraint condition, including: S41, randomly generating an initial population of a certain size within the search space, each Cordyceps individual in the initial population representing a potential solution, and initializing parameters; S42, switching between global exploration phase and local development phase through dynamic probability threshold control to avoid search rigidity caused by fixed probability, if entering global exploration phase, jumping to S43; if entering local development phase, jumping to S44; S43, in the global exploration phase, through the fluctuation propulsion operator and the optimal parasitic behavior, simulating the diffusion behavior of Cordyceps in the horizontal dimension and the selection of optimal growth conditions, enhancing the global search ability, avoiding falling into local optimum, and entering S45; S44, in the local development phase, through the spiral rising operator and the re-parasitic behavior, simulating the focused search of Cordyceps in the vertical dimension and the competitive parasitism of high-quality hosts, realizing fine search and accelerating convergence to global optimum, and entering S45; S45, evaluate all Ophiocordyceps individual in current population by target function of grid dispatching, calculate corresponding fitness value, record and update global optimal solution; S46, judge whether iteration termination condition is met, if not, return to S42, otherwise take current global optimal solution as optimal grid dispatching strategy.

[0013] Preferably, in S43, in global exploration stage, through fluctuation propulsion operator and optimal parasitic behavior, simulate diffusion behavior of Ophiocordyceps in horizontal dimension and selection of optimal growth condition, enhance global search ability, and avoid falling into local optimum, including: Each time of iteration, each Ophiocordyceps individual in current population selects fluctuation propulsion operator with 50% probability and optimal parasitic behavior with 50% probability: Fluctuation propulsion operator is expressed by the following formula: ; Wherein, X i (k) and X i (k+1) are positions of Ophiocordyceps individual i in kth and k+1th iteration, Guide individual to move to high-quality area through global optimal solution, X gbest is position of global optimal solution, rand(1, D) represents a random vector of 1*D, subscript indicates different random vectors, each element is uniformly distributed in [0, 1], control step length of individual moving to global optimal solution, and D is dimension of decision variable; Expand search range through population boundary, avoid falling into local optimum, X best (k) is position of current global optimal solution in kth iteration, is search coefficient, control step length of individual moving to population boundary, 2.2 is basic search strength, ensure that algorithm has enough driving force to expand search range, rand(0, 1) represents a random number uniformly distributed in [0, 1), and subscript indicates different random numbers.

[0014] Preferably, the optimal parasitic behavior is expressed by the following formula: ; Wherein, is adaptive step length, balance global exploration and local development, Reflect relative distance between current global optimal solution and Ophiocordyceps individual i, if Ophiocordyceps individual i is far away from current global optimal solution, the ratio is larger, and adaptive step length is increased to enhance global search, otherwise adaptive step length is reduced to carry out fine search, and the exponential Decaying with the increase of iteration number k, gradually suppressing large step jumps, Rapidly decreasing with the increase of iteration number k, ensuring that the step size tends to be stable in the later stage, avoiding violent disturbance.

[0015] Preferably, in the local development stage in S44, through the spiral ascent operator and the re-parasitic behavior, the focused search of Ophiocordyceps in the vertical dimension and the competitive parasitism of high-quality hosts are simulated, fine search is realized, and convergence to the global optimum is accelerated, including: Each Ophiocordyceps individual in the current population selects the spiral ascent operator with a probability of 50% and the re-parasitic behavior with a probability of 50% at each iteration: The spiral ascent operator is expressed by the following formula: ; Wherein, lets is a dynamic step size, controlling the step size of the individual moving towards the current global optimal solution, 2 is an initial reference step size, ensuring that the individual has sufficient mobility in the early stage, Reflecting the relative distance between the current global optimal solution and the Ophiocordyceps individual i, if the Ophiocordyceps individual i is far away from the current global optimal solution, the ratio is larger, the dynamic step size lets is increased to strengthen global search, otherwise the dynamic step size lets is reduced for fine search, the exponential Gradually converging with the increase of iteration number k, avoiding falling into local optimum too early.

[0016] Preferably, the re-parasitic behavior is expressed by the following formula: ; Wherein, 3.7 is a step size scaling factor, adjusting the overall step size range.

[0017] (Three) beneficial effects Compared with the prior art, the power grid automatic dispatching system based on digital twinning provided by the present application has the following beneficial effects: 1) Precise mapping modeling, improving the accuracy of dispatching The data perception and fusion module collects power grid multi-source heterogeneous data in real time, constructs a unified data lake and knowledge graph, and provides high-quality input for digital twinning modeling; the digital twinning modeling module constructs three-level digital twinning models of equipment, units and systems of the power grid, realizes real-time synchronous mapping of the physical power grid and the virtual model, and such precise mapping relationship enables the power grid dispatching system to accurately and timely grasp the actual operation state of the power grid, avoids dispatching deviation caused by inaccurate data or inaccurate model, and thus generates an optimal power grid dispatching strategy that better meets the actual needs of the power grid, greatly improving the accuracy of power grid dispatching; 2) Two-stage optimization, enhancing algorithm performance The optimization scheduling module solves the power grid scheduling model by using the cordyceps optimization algorithm. The algorithm simulates the parasitic behavior of cordyceps, and constructs a two-stage search framework of dynamic balance global exploration and local development: in the global exploration stage, the fluctuation propulsion operator and the optimal parasitic behavior are used to enhance the global search ability and avoid falling into local optimum; in the local development stage, the spiral rising operator and the re-parasitic behavior are used to realize fine search and accelerate convergence to the global optimum. This unique algorithm design enables the power grid scheduling model to find the optimal solution that meets the constraint conditions more efficiently, significantly improves the search efficiency and convergence speed compared with traditional algorithms, and enhances the algorithm performance. 3) Dynamic feedback loop to improve system adaptability The decision execution and feedback module converts the optimal power grid scheduling strategy into corresponding scheduling instructions and issues them to the AGC / AVC system, and at the same time verifies the scheduling effect in real time through the digital twin model to form a closed-loop control. This closed-loop control mechanism enables the system to adjust the scheduling strategy in a timely manner according to the actual scheduling effect, adapt to the changing working conditions and demands in the process of power grid operation, for example, when a sudden fault or load fluctuation occurs in the power grid, the system can quickly adjust the scheduling instructions through the closed-loop control mechanism to ensure the stable operation of the power grid and improve the adaptability of the system to complex power grid environments and the ability to respond to unexpected events. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiment or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can obtain other drawings according to these drawings without creative labor.

[0019] Figure 1 The system schematic diagram of the present application; Figure 2 The flowchart of the present application is shown in the figure. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be described clearly and completely in the following with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0021] The specific function modules (as shown in Figure 1 The system function modules include: The data perception and fusion module collects multi-source heterogeneous data of the power grid in real time, and constructs a unified data lake and knowledge graph to provide high-quality input for digital twin modeling. The digital twin modeling module constructs three-level digital twin models of the power grid at the device level, unit level, and system level, and realizes real-time synchronous mapping of the physical power grid and the virtual model. The optimal scheduling module uses the cordyceps optimization algorithm to solve the power grid scheduling model and generates the optimal power grid scheduling strategy. The decision execution and feedback module converts the optimal power grid scheduling strategy into corresponding scheduling instructions and issues them to the AGC / AVC system, and verifies the scheduling effect in real time through the digital twin model to form a closed-loop control.

[0022] In the technical solution of the present application, the cordyceps optimization algorithm simulates the parasitic behavior of cordyceps by constructing a two-stage search framework of dynamic balance global exploration and local development: In the global exploration stage, the fluctuation advancing operator and the optimal parasitic behavior are used to simulate the diffusion behavior of cordyceps in the horizontal dimension and the selection of the optimal growth conditions, to enhance the global search capability and avoid falling into local optimum. In the local development stage, the spiral rising operator and the re-parasitic behavior are used to simulate the focused search of cordyceps in the vertical dimension and the competitive parasitism of high-quality hosts, to realize fine search and accelerate convergence to the global optimum.

[0023] I. Data perception and fusion module The data perception and fusion module collects multi-source heterogeneous data of the power grid in real time, and constructs a unified data lake and knowledge graph to provide high-quality input for digital twin modeling, including: Real-time collection of multi-source heterogeneous data of the power grid, including SCADA measurement data, device state data and weather data; Through data cleaning, time series alignment and feature extraction, a unified data lake and knowledge graph are constructed to provide high-quality input for digital twin modeling.

[0024] Specifically, the LSTM network is used for feature extraction, and the hidden layer of the LSTM network is represented by the following formula: ; Where h(t) and h(t-1) are the hidden states at time t and time t-1, respectively, and [h(t-1), x(t)] represents an input vector formed by concatenating the hidden state h(t-1) at time t-1 and the input x(t) at time t, W fis a weight matrix, b f is a bias term, denotes a Sigmoid activation function.

[0025] II. Digital Twin Modeling Module The digital twin modeling module constructs a three-level digital twin model of the power grid at the device level, unit level, and system level, realizes real-time synchronous mapping of the physical power grid and the virtual model, and includes: A finite element model of the device is built using ANSYS Twin Builder, real-time synchronization of data between the edge computing node and the cloud is achieved using the OpenFMB protocol, and a device-level digital twin model is constructed; Based on the device-level digital twin model, units are combined according to device functions, the interaction logic and data flow between devices within the unit are clarified, and a unit-level digital twin model is constructed; A power grid topology model is constructed based on graph computing technology, a node voltage equation is used to describe the electrical relationship, and a system-level digital twin model is constructed to support multi-physical field coupling simulation.

[0026] The above technical solution, the data perception and fusion module collects power grid multi-source heterogeneous data in real time, constructs a unified data lake and knowledge graph, and provides high-quality input for digital twin modeling; the digital twin modeling module constructs a three-level digital twin model of the power grid at the device level, unit level, and system level, realizes real-time synchronous mapping of the physical power grid and the virtual model, and this precise mapping relationship enables the power grid dispatching system to accurately and timely grasp the actual operating state of the power grid, avoids dispatching deviations caused by inaccurate data or inaccurate models, and thus generates an optimal power grid dispatching strategy that better meets the actual needs of the power grid, greatly improving the accuracy of power grid dispatching.

[0027] III. Optimization Dispatching Module The optimization dispatching module uses the Cordyceps optimization algorithm to solve the power grid dispatching model, generates an optimal power grid dispatching strategy, and includes: S1, determining the objective function of power grid dispatching; S2, determining the constraint conditions of power grid dispatching; S3, combining the objective function and constraint conditions of power grid dispatching, and the three-level digital twin model of the power grid at the device level, unit level, and system level, constructing a power grid dispatching model; S4, using the Cordyceps optimization algorithm to solve the power grid dispatching model, generating an optimal power grid dispatching strategy that meets the constraint conditions.

[0028] Specifically, in S4, the Cordyceps optimization algorithm is used to solve the power grid dispatching model, and an optimal power grid dispatching strategy that meets the constraint conditions is generated, as shown in Figure 2 , including: S41, randomly generate an initial population of a certain size in the search space, each individual of the initial population represents a potential solution, and initialize parameters; S42, control the switching between the global exploration stage and the local development stage through the dynamic probability threshold, avoid search rigidity caused by fixed probability, if entering the global exploration stage, jump to S43, if entering the local development stage, jump to S44; S43, in the global exploration stage, simulate the diffusion behavior of O. tsu-taengensis in the horizontal dimension and the selection of optimal growth conditions through the fluctuation propulsion operator and optimal parasitic behavior, enhance the global search ability, avoid falling into local optimum, and enter S45; S44, in the local development stage, simulate the focused search of O. tsu-taengensis in the vertical dimension and the competitive parasitism of high-quality hosts through the spiral rising operator and re-parasitic behavior, realize fine search, and accelerate convergence to the global optimum, and enter S45; S45, evaluate all O. tsu-taengensis individuals in the current population using the objective function of power grid dispatching, calculate the corresponding fitness value, record and update the global optimal solution; S46, judge whether the iteration termination condition is met, if the iteration termination condition is not met, return to S42, otherwise take the current global optimal solution as the optimal power grid dispatching strategy.

[0029] 1) In the global exploration stage in S43, simulate the diffusion behavior of O. tsu-taengensis in the horizontal dimension and the selection of optimal growth conditions through the fluctuation propulsion operator and optimal parasitic behavior, enhance the global search ability, avoid falling into local optimum, including: Each time the iteration, each O. tsu-taengensis individual in the current population selects the fluctuation propulsion operator with a probability of 50% and the optimal parasitic behavior with a probability of 50%: The fluctuation propulsion operator is represented by the following formula: ; Where, X i (k), X i (k+1) are the positions of O. tsu-taengensis individual i at the kth and k+1th iterations, Move the individual to the high-quality area guided by the global optimal solution, X gbest is the position of the global optimal solution, rand(1, D) represents a random vector of 1*D, the subscript indicates different random vectors, each element is uniformly distributed in the range of [0, 1], control the step size of the individual moving to the global optimal solution, D is the dimension of the decision variable; Expand the search range by the population boundary, avoid falling into local optimum, X best (k) is the position of the current global optimal solution at the kth iteration, For searching coefficient, the step length of controlling individual moving to the boundary of population, 2.2 is the basic search intensity, ensuring the algorithm has enough driving force to expand the search range, rand(0, 1) represents a random number uniformly distributed in the range of [0, 1), and the subscript indicates different random numbers.

[0030] 2) The optimal parasitic behavior is expressed as follows: ; wherein, is the adaptive step length, balancing global exploration and local exploitation, reflects the relative distance between the current global optimal solution and the Cordyceps individual i. If the Cordyceps individual i is far away from the current global optimal solution, the ratio is larger, and the adaptive step length increases to strengthen global search, otherwise the adaptive step length decreases to conduct fine search, the index decays with the increase of iteration number k, gradually suppressing large step jumping, rapidly decreases with the increase of iteration number k, ensuring that the step length tends to be stable in the later period, avoiding violent disturbance.

[0031] 3) In the local exploitation stage in S44, through the spiral climbing operator and the re-parasitic behavior, the focused search of Cordyceps in the vertical dimension and the competitive parasitism of high-quality hosts are simulated, fine search is realized, and convergence to the global optimum is accelerated, including: At each iteration, each Cordyceps individual in the current population selects the spiral climbing operator with a probability of 50% and the re-parasitic behavior with a probability of 50%: The spiral climbing operator is expressed as follows: ; wherein, lets is the dynamic step length, controlling the step length of individual moving to the current global optimal solution, 2 is the initial reference step length, ensuring that the individual has enough moving ability in the early stage, reflects the relative distance between the current global optimal solution and the Cordyceps individual i. If the Cordyceps individual i is far away from the current global optimal solution, the ratio is larger, and the dynamic step length lets increases to strengthen global search, otherwise the dynamic step length lets decreases to conduct fine search, the index gradually converges with the increase of iteration number k, avoiding falling into local optimum too early.

[0032] 4) The re-parasitic behavior is expressed as follows: ; wherein, 3.7 is the step length scaling factor, adjusting the overall step length range.

[0033] The above technical solution is that the optimization scheduling module uses the cordyceps optimization algorithm to solve the power grid scheduling model. The algorithm simulates the parasitic behavior of cordyceps, constructs a two-stage search framework of dynamic balance global exploration and local development: in the global exploration stage, the fluctuation advancing operator and the optimal parasitic behavior are used to enhance the global search ability and avoid falling into local optimum; in the local development stage, the spiral rising operator and the re-parasitic behavior are used to realize fine search and accelerate convergence to the global optimum. This unique algorithm design enables the power grid scheduling model to find the optimal solution that meets the constraint conditions more efficiently, significantly improves the search efficiency and convergence speed compared with traditional algorithms, and enhances the algorithm performance.

[0034] IV. Decision execution and feedback module The decision execution and feedback module converts the optimal power grid scheduling strategy into corresponding scheduling instructions and issues them to the AGC / AVC system, verifies the scheduling effect in real time through the digital twin model, and forms a closed-loop control.

[0035] The above technical solution is that the decision execution and feedback module converts the optimal power grid scheduling strategy into corresponding scheduling instructions and issues them to the AGC / AVC system, and verifies the scheduling effect in real time through the digital twin model, forming a closed-loop control. This closed-loop control mechanism enables the system to adjust the scheduling strategy in a timely manner according to the actual scheduling effect, adapt to the changing working conditions and demands in the process of power grid operation, for example, when a sudden fault or load fluctuation occurs in the power grid, the system can quickly adjust the scheduling instructions through the closed-loop control mechanism to ensure the stable operation of the power grid and improve the adaptability of the system to complex power grid environments and the ability to respond to unexpected events.

[0036] The above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements for some technical features; and these modifications or replacements will not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A power grid automation dispatching system based on digital twinning, characterized in that: The application comprises: a data perception and fusion module that collects multi-source heterogeneous data of a power grid in real time, and constructs a unified data lake and knowledge graph to provide high-quality input for digital twin modeling; a digital twin modeling module that constructs device-level, unit-level, and system-level three-level digital twin models of the power grid to realize real-time synchronous mapping of the physical power grid and the virtual model; an optimal scheduling module that uses the Cordyceps optimization algorithm to solve the power grid scheduling model to generate an optimal power grid scheduling strategy; a decision execution and feedback module that converts the optimal power grid scheduling strategy into corresponding scheduling instructions and issues them to the AGC / AVC system, and verifies the scheduling effect in real time through the digital twin model to form a closed-loop control; wherein the Cordyceps optimization algorithm simulates the parasitic behavior of Cordyceps to construct a two-stage search framework of dynamic balance global exploration and local development: in the global exploration stage, the fluctuation propulsion operator and the optimal parasitic behavior are used to simulate the diffusion behavior of Cordyceps in the horizontal dimension and the selection of optimal growth conditions, thereby enhancing the global search capability and avoiding falling into local optimum; in the local development stage, the spiral ascent operator and the re-parasitic behavior are used to simulate the focused search of Cordyceps in the vertical dimension and the competitive parasitism of high-quality hosts, thereby realizing fine search and accelerating convergence to the global optimum.

2. The digital-twin-based power grid automation dispatching system according to claim 1, wherein: The data perception and fusion module collects multi-source heterogeneous data of a power grid in real time, and constructs a unified data lake and knowledge graph to provide high-quality input for digital twin modeling, including: real-time collection of multi-source heterogeneous data of a power grid, including SCADA measurement data, device status data, and weather data; through data cleaning, time series alignment, and feature extraction, a unified data lake and knowledge graph are constructed to provide high-quality input for digital twin modeling.

3. The digital-twin-based power grid automation dispatching system of claim 2, wherein: LSTM network is used for feature extraction, and the hidden layer of the LSTM network is expressed by the following formula: ； where h(t) and h(t-1) are the hidden states at time t and t-1, respectively, [h(t-1), x(t)] represents an input vector formed by concatenating the hidden state h(t-1) at time t-1 and the input x(t) at time t, W f is a weight matrix, and b f is a bias term, denotes a Sigmoid activation function.

4. The digital-twin-based power grid automation dispatching system of claim 1, wherein: The digital twin modeling module constructs device-level, unit-level, and system-level three-level digital twin models of the power grid to realize real-time synchronous mapping of the physical power grid and the virtual model, including: ANSYS Twin Builder is used to build a finite element model of the device, and OpenFMB protocol is used to realize real-time synchronization of data between the edge computing node and the cloud to construct a device-level digital twin model; based on the device-level digital twin model, units are combined according to device functions, the interaction logic and data flow between devices in the unit are clarified, and a unit-level digital twin model is constructed; a power grid topology model is constructed based on graph computing technology, and a node voltage equation is used to describe the electrical relationship to construct a system-level digital twin model to support multi-physical field coupling simulation.

5. The digital-twin-based power grid automation dispatching system of claim 1, wherein: The optimal scheduling module uses the Cordyceps optimization algorithm to solve the power grid scheduling model to generate an optimal power grid scheduling strategy, including: S1, determining the objective function of power grid scheduling; S2, determining the constraint conditions of power grid scheduling; S3, combining the objective function and constraint conditions of power grid scheduling, and the device-level, unit-level, and system-level three-level digital twin models of the power grid, to construct a power grid scheduling model; S4, using the Cordyceps optimization algorithm to solve the power grid scheduling model to generate an optimal power grid scheduling strategy that meets the constraint conditions.

6. The digital-twin-based power grid automation dispatching system according to claim 5, wherein: The winter worm summer grass optimization algorithm is used in S4 to solve the power grid scheduling model to generate an optimal power grid scheduling strategy that meets the constraint conditions, including: S41, an initial population of a certain size is randomly generated in the search space, each winter worm summer grass individual in the initial population represents a potential solution, and the parameters are initialized; S42, the switching between the global exploration stage and the local development stage is controlled by the dynamic probability threshold to avoid search rigidity caused by fixed probability, if the global exploration stage is entered, jump to S43, if the local development stage is entered, jump to S44; S43, in the global exploration stage, the diffusion behavior of winter worm summer grass in the horizontal dimension and the selection of optimal growth conditions are simulated through the fluctuation propulsion operator and the optimal parasitic behavior to enhance the global search ability and avoid falling into local optimum, and enter S45; S44, in the local development stage, the focused search of winter worm summer grass in the vertical dimension and the competitive parasitism of high-quality hosts are simulated through the spiral ascent operator and the re-parasitic behavior to realize fine search and accelerate convergence to the global optimum, and enter S45; S45, the objective function of power grid scheduling is used to evaluate all winter worm summer grass individuals in the current population, calculate the corresponding fitness value, record and update the global optimal solution; S46, determine whether the iteration termination condition is met, if the iteration termination condition is not met, return to S42, otherwise the current global optimal solution is taken as the optimal power grid scheduling strategy.

7. The digital-twin-based power grid automation dispatching system according to claim 6, wherein: In S43, in the global exploration stage, the diffusion behavior of winter worm summer grass in the horizontal dimension and the selection of optimal growth conditions are simulated through the fluctuation propulsion operator and the optimal parasitic behavior to enhance the global search ability and avoid falling into local optimum, including: Each winter worm summer grass individual in the current population selects the fluctuation propulsion operator with a probability of 50% and the optimal parasitic behavior with a probability of 50% at each iteration: The fluctuation propulsion operator is represented by the following formula: ； where X i (k), X i (k+1) are the positions of the individual i at the kth, k+1th iteration, The global optimal solution guides individuals to move to high-quality areas, X gbest is the position of the global optimal solution, rand(1, D) represents a random vector of 1*D, the subscript indicates different random vectors, and each element is uniformly distributed in the range [0, 1], controls the step size of the individual moving towards the global optimal solution, and D is the dimension of the decision variable. By expanding the search range through population boundary, the algorithm avoids falling into local optimum, X best (k) is the position of the current global optimal solution at the kth iteration, is the search coefficient, which controls the step size of the individual moving to the population boundary, 2.2 is the basic search strength, which ensures that the algorithm has enough driving force to expand the search range, rand(0,1) represents a uniformly distributed random number in the range [0,1), and the subscript indicates different random numbers.

8. The digital-twin-based power grid automation dispatching system according to claim 7, wherein: The optimal parasitic behavior is represented by the following formula: ； wherein, is an adaptive step size balancing global exploration and local exploitation, reflects the relative distance between the current global optimum and the Cordyceps individual i, if the Cordyceps individual i is far away from the current global optimum, the ratio is larger, the adaptive step size is increased to strengthen the global search, otherwise the adaptive step size is decreased to conduct a fine search, the exponential decays with the increase of the iteration number k, gradually inhibiting the large step jump, rapidly decreases with the increase of the iteration number k, ensuring that the step size tends to be stable in the later stage, avoiding violent disturbance.

9. The digital-twin-based power grid automation dispatching system according to claim 8, wherein: In S44, in the local development stage, the focused search of winter worm summer grass in the vertical dimension and the competitive parasitism of high-quality hosts are simulated through the spiral ascent operator and the re-parasitic behavior to realize fine search and accelerate convergence to the global optimum, including: Each winter worm summer grass individual in the current population selects the spiral ascent operator with a probability of 50% and the re-parasitic behavior with a probability of 50% at each iteration: The spiral ascent operator is represented by the following formula: ； wherein lets is a dynamic step length, controlling the step length of the individual to move towards the current global optimal solution, 2 is an initial reference step length, ensuring that the individual has enough moving ability in the early stage, reflects the relative distance between the current global optimal solution and the cordyceps individual i, if the cordyceps individual i is far away from the current global optimal solution, the ratio is larger, the dynamic step length lets is increased to strengthen the global search, otherwise the dynamic step length lets is reduced to carry out fine search, the exponential gradually converges with the increase of the iteration number k, avoiding falling into local optimum too early.

10. The digital-twin-based power grid automation dispatching system according to claim 9, wherein: The re-parasitic behavior is represented by the following formula: ； Wherein, 3.7 is the step scaling factor, which adjusts the overall step range.

Images