Method and device for constructing digital twin model of power grid

Abstract

The invention discloses a power grid digital twin model construction method, which specifically comprises the following steps: acquiring a geometric model of power grid equipment, combining with real scene data, fusing power grid physical parameters and representing behaviors and rules of a physical entity, performing power grid information physical modeling, and generating a power grid mechanism model; based on the observation data transmitted into the power grid equipment, a probabilistic graph model is adopted as a data-driven modeling method, power grid data driving modeling is carried out, and a power grid data driving model is generated; performing power grid mechanism-data driving hybrid modeling on the obtained power grid mechanism model and the obtained power grid data driving model by adopting an interface learning framework; and taking the obtained power grid mechanism-data driving hybrid model as a power grid digital twinborn model, and combining constraint conditions of different models to realize power grid digital twinborn model service. The digital twin model of the power grid can be accurately, effectively and quickly constructed, and the model precision is improved.

Description

Technical Field

[0001] The present invention relates to the field of information technology applications in power systems, and in particular to a method and device for constructing a digital twin model of a power grid. Background Art

[0002] Traditional power grid geometric models, real-life data, physical parameters, behaviors and rules exist in various power grid professional businesses, and information is not interoperable. The disadvantage of data-driven modeling is that its inherent analysis accuracy cannot achieve high precision, while machine learning can achieve higher accuracy than physical models in some fields, provided that sufficient high-quality training data is provided. However, machine learning also faces many problems: too much data is required, and it is impossible to obtain enough data in most real scenarios; black box machine learning models may produce results that are inconsistent with physics; machine learning models have poor generalization performance and are basically unreliable. Physical mechanisms have been well applied in many traditional businesses, but for complex physical mechanism phenomena and problems, traditional tools and methods also face many problems, such as overly ideal results, far from the actual situation, high investment, and large amount of calculation.

[0003] The digital twin model of the power grid is the basis for realizing the computing, deduction, evaluation and prediction capabilities of the digital twin. How to accurately, effectively and quickly build the digital twin model of the power grid entity is the most critical step. The massive amount of power equipment, personnel, asset data and complex data structures contained in the physical power grid will cause the dimensionality disaster problem of heterogeneous data. The dimensionality disaster directly brings difficulties to the modeling of digital twins, such as the difficulty of modeling multi-scale physical field coupling models, the complexity of heterogeneous data cleaning, and the complexity of multi-source data fusion calculation. How to combine data-driven technology with existing physical mechanism knowledge for effective and accurate modeling is a major difficulty. Therefore, it is urgent to propose a method for building a digital twin model of the power grid.

[0004] Prior art document 1 (CN116050250A) discloses a method and device for modeling a digital twin model of a power grid based on a hybrid drive model. The method includes: based on the physical laws between the parameters of the power system, setting corresponding boundary conditions to generate a power grid mechanism model; based on reading the input and output information of each element of each power system from the power grid area, training and generating a power grid data-driven model; dynamically weighting the mechanism model and the data-driven model to generate a hybrid drive model as a power grid digital twin model. Its shortcoming is that it only combines the mechanism model and the data-driven model to generate a hybrid drive model by dynamic weighting. Summary of the invention

[0005] In order to solve the shortcomings of the prior art, the present invention provides a method for constructing a digital twin model of a power grid device level to achieve high-precision mapping of the virtual and real with the power physical objects; then, a method for constructing a digital twin analysis model of a distribution network is studied to form a multi-level optimization from initialization device modeling to hybrid modeling; finally, a model construction method based on an interface learning framework and a probabilistic graphical model is studied to construct a power digital twin analysis model to provide underlying model support for the power digital twin operation simulation.

[0006] The present invention adopts the following technical solution.

[0007] A first aspect of the present invention provides a method for constructing a digital twin model of a power grid, the method comprising:

[0008] Step 1: Obtain the geometric model of the power grid equipment and combine it with the real scene data, integrate the physical parameters of the power grid and characterize the behavior and rules of the physical entity, perform power grid information physical modeling, and generate a power grid mechanism model;

[0009] Step 2: Based on the observation data transmitted to the power grid equipment, the probabilistic graphical model is used as a data-driven modeling method to perform power grid data-driven modeling and generate a power grid data-driven model;

[0010] Step 3: Use the interface learning framework to perform grid mechanism-data driven hybrid modeling on the grid mechanism model obtained in step 1 and the grid data driven model obtained in step 2;

[0011] Step 4: Use the power grid mechanism-data-driven hybrid model obtained in step 3 as the power grid digital twin model, and combine the constraints of different models to realize the power grid digital twin model service.

[0012] Preferably, in step 1, the specific process of power grid information-physical modeling is as follows: based on the appearance structure model of the power grid equipment, the physical parameters of the power grid are used to describe the equipment characteristics, and the geographical coordinate position and equipment assets are combined to realize the three-dimensional modeling of multiple device elements of a single device. Through the coupling connection relationship of the elements, the external disturbance response behavior, the internal operation rules, the energy flow trajectory and the information flow transmission relationship as well as the basic natural laws, the connection between the models of individual elements is completed to form a dynamic virtual model set, and the virtual model set is also associated, and is superimposed block by block and layer by layer to form a digital twin full-space model.

[0013] Preferably, a power grid physical parameter database is formed based on power grid physical parameters, which include electrical and non-electrical quantity data collected by various types of sensors, and information collected by various types of equipment and devices, and the information collected by various types of equipment and devices includes: status indications and measurement values.

[0014] Preferably, the power grid data-driven modeling adopts the probabilistic graphical model as the data-driven modeling method, understands the status of the equipment power grid equipment through the incoming observation data, and then analyzes and predicts the information through the probabilistic graphical model to generate intelligent control information, and controls the power grid equipment by publishing the intelligent control information.

[0015] Preferably, the power grid probabilistic graph model includes a Bayesian network, a Markov network and a factor graph, which are used simultaneously to perform data-driven modeling on a power grid data model with Bayesian network characteristics, a power grid data model with Markov network characteristics and a power grid data model with characteristics in a factor graph, and finally obtain a data-driven model including a Bayesian network, a Markov network and a factor graph probabilistic graph model.

[0016] Preferably, the power grid data-driven model with Bayesian network characteristics is connected into a directed acyclic graph according to the Bayesian network, and a conditional probability table is used to describe the task sending conditions, probabilities and related tasks.

[0017] Preferably, the power grid data-driven model with Markov network characteristics is connected into an undirected graph according to the Markov network, and a conditional probability table is used to describe the task sending conditions, probabilities and related tasks.

[0018] Preferably, the power grid data-driven model with factor graph characteristics is connected into an undirected graph according to the factor graph, and a conditional probability table is used to describe the task sending conditions, probabilities and related tasks.

[0019] Preferably, the interface learning framework connects the power grid mechanism model and the data-driven model through the power grid digital twin model interface to obtain a power grid mechanism-data-driven hybrid model.

[0020] Preferably, the power grid digital twin model interface performs interface error correction and extension by directly extending the interface and correcting and extending the interface.

[0021] Preferably, the prediction range is adopted to directly extend the interface to solve the accuracy difference between the power grid mechanism model and the data-driven model;

[0022] The correction extension interface uses a long short-term memory network recurrent neural network to control the difference between the predicted and optimal values ​​of the interface variables, and the accuracy difference between the correction mechanism model and the data-driven model.

[0023] Preferably, the power grid digital twin model service process includes: constructing a hierarchical relationship of the model and clarifying the dependency order of the model; adding constraints in the model service process; and implementing the power grid digital twin model service based on the constructed constraint relationship, model dependency order and power grid digital twin model.

[0024] A second aspect of the present invention provides a power grid digital twin model construction device, which is used to run the power grid digital twin model construction method, including:

[0025] A data acquisition module is used to acquire data for building a digital twin model of the power grid;

[0026] The model building module, based on the power grid digital twin model data, adopts the power grid information physical modeling, power grid data-driven modeling, and power grid mechanism-data-driven hybrid modeling methods to obtain the power grid digital twin model;

[0027] The model service module adopts the power grid digital twin model service method and the power grid digital twin model to provide power grid digital twin model services.

[0028] An electronic device comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program implements a method for constructing a digital twin model of a power grid when loaded into the processor.

[0029] A computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements a method for constructing a digital twin model of a power grid.

[0030] The beneficial effect of the present invention is that, compared with the prior art, the present invention adopts the grid information physical modeling method to realize the full-scale multi-physics field digital twin model of the equipment to achieve high-precision mapping between the virtual and real power physical objects; the probability graph model is used as a data-driven modeling method. The probability graph model is used to characterize the dependency relationship of the random variables of the model at the variable level, reflecting the probability structure of the problem and the difficulty of reasoning, and also provides an operable data structure for the reasoning algorithm; by adopting the grid data-driven modeling method, the accuracy of data-driven modeling can be effectively improved. The model construction method involved in this patent adopts grid information physical modeling, grid data-driven modeling, grid mechanism-data-driven hybrid modeling, and grid digital twin model services to form a multi-level optimization from initialization equipment modeling to hybrid modeling, and can construct a power digital twin analysis model to provide underlying model support for power digital twin operation simulation. It can realize accurate, effective and rapid construction of a grid digital twin model and improve model accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 It is a correlation diagram of a power grid digital twin model construction device in the present invention;

[0032] Figure 2 It is a flow chart of the method for constructing a digital twin model of a power grid in the present invention;

[0033] Figure 3It is a connection diagram of an electronic device in the present invention. DETAILED DESCRIPTION

[0034] In order to make the purpose, technical scheme and advantages of the present invention clearer, the technical scheme of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. The embodiments described in this application are only part of the embodiments of the present invention, not all of them. Based on the spirit of the present invention, other embodiments obtained by ordinary technicians in this field without creative work are all within the scope of protection of the present invention.

[0035] Example 1

[0036] Embodiment 1 of the present invention discloses a method for constructing a digital twin model of a power grid, including information-physical modeling of a power grid, data-driven modeling of a power grid, hybrid modeling of a power grid mechanism-data-driven modeling, and a digital twin model service of a power grid. Information-physical modeling of a power grid realizes the establishment of a physical mechanism and process model of a power grid to reflect a physical entity. The data-driven power grid can bypass the complex physical modeling process and use input and output data to describe the physical process well. Combined with the monitoring data and test data of the power equipment, a data model between the input data and the output data is constructed using a machine learning method, so that the physical phenomena and mechanisms of the operation of the power equipment are characterized by a data model, and the twinning of the equipment is completed. The digital twin model service of the power grid realizes the process of the digital twin model from the unit-level model to the system-level model and then to the complex system-level model service.

[0037] like Figure 1 As shown, the following steps are included:

[0038] Step 1: Obtain the geometric model of the power grid equipment and combine it with the real-life data, integrate the physical parameters of the power grid, and characterize the behaviors and rules of the physical entities, perform information physics modeling of the power grid, and generate a power grid mechanism model.

[0039] Grid information physical modeling realizes the establishment of grid physical mechanism and process models to reflect physical entities.

[0040] The information-physical modeling of the power grid first constructs the geometric model of the equipment and combines it with the real-life data to realize a realistic digital twin model. On the other hand, it is necessary to integrate the physical parameters of the power grid and characterize the behavior and rules of the physical entities to realize a full-scale multi-physics digital twin model of the equipment.

[0041] The physical parameters of the power grid include electrical and non-electrical data collected by various types of sensors, as well as information collected by various types of equipment and devices. The information collected by various types of equipment and devices includes but is not limited to: status indications and measurement values. A physical parameter database of the power grid is formed based on the physical parameters of the power grid.

[0042] The behaviors and rules of physical entities include: the coupling connection relationship of elements, the response behavior to external disturbances, the internal operating rules, the energy flow trajectory, the information flow transmission relationship and the basic laws of nature.

[0043] The specific process of grid information physical modeling is as follows: Based on the appearance structure model of the grid equipment, the grid physical parameter database is used to describe the equipment characteristics, and the geographical coordinate position and equipment assets are combined to realize the three-dimensional modeling of multiple device elements of a single device. Through the behavior and rules of physical entities, the connection between single element models is completed to form a dynamic virtual model set. The virtual model set is also associated, and it is superimposed piece by piece and layer by layer to form a digital twin full-space model, while realizing the digital model expression of the scene, spatial analysis, and accurate expression of physical facilities and functional information.

[0044] Taking photovoltaics as an example, the digital twin modeling of photovoltaic inverters should start from the perspectives of multiple devices such as power electronic components, DC boost components, circuit board materials and capacitors, integrated circuits, etc., and integrate offline experiments, power-on state monitoring, and external environmental influences to form a digital twin database of photovoltaic inverters. The different indicators obtained should be continuously weighted for data pre-driving, and the measurement data of photovoltaic inverters should be segmented and judged for high-dimensional indicators. Finally, the digital twin model should be modeled through random matrices or other mathematical methods.

[0045] For example, the BIM model is used to describe the virtual model set of the equipment, including various sensor devices, the equipment itself, and the control mechanism, and the external line model is connected and coupled on this basis. The digital model of the power grid has built-in massive static description data and dynamic perception data, superimposed characteristic behavior labels, combined with a modular element model set, to support the dynamic three-dimensional demonstration of the physical power grid, implicit law mining and intelligent quantitative simulation, and on this basis to complete the optimization of modeling.

[0046] Step 2: Based on the observation data transmitted to the power grid equipment, the probabilistic graphical model is used as the data-driven modeling method to perform power grid data-driven modeling and generate a data-driven model.

[0047] Grid data-driven can bypass the complex physical modeling process and use input and output data to well describe the physical process combined with monitoring data and test data of power equipment.

[0048] The power grid data-driven modeling adopts the power grid probabilistic graphical model as a data-driven modeling method. The probabilistic graphical model is used to represent the dependency of the model's random variables at the variable level, reflecting the probabilistic structure of the problem and the difficulty of reasoning. The representation of the probabilistic graphical model consists of two parts: the network structure and the parameter model.

[0049] The network structure G = (N, E) consists of a node set N and an edge set E. The nodes in N represent variables in the random variable set X = {X1, ..., Xn}. The edges in the edge set E represent the probabilistic relationship between two variables, which can be directed or undirected.

[0050] The parameter model is a probability distribution model of the network structure. The probabilistic graphical model uses probability representation, reasoning and learning based on the probability distribution model to realize the interaction between the physical model of the device and the digital twin model.

[0051] The state of the equipment grid equipment is obtained through the incoming observation data, and then the information is analyzed and predicted through the probabilistic graphical model, and the intelligent control information is released to control the grid equipment, so as to establish a relationship between the two, thereby realizing the data-driven model to guide the intelligent control of the grid equipment.

[0052] The power grid probabilistic graphical model includes Bayesian networks, Markov networks and factor graphs, which are used simultaneously to perform data-driven modeling on power grid data models with Bayesian network characteristics, power grid data models with Markov network characteristics, and power grid data models with characteristics in factor graphs, and finally a data-driven model of the probabilistic graphical model including Bayesian networks, Markov networks and factor graphs is obtained.

[0053] Bayesian network is a network structure based on directed acyclic graph to describe the dependency relationship between attributes, and uses conditional probability table to describe the joint probability distribution. The power grid data model with Bayesian network characteristics is connected into a parallel or serial task directed acyclic graph according to the Bayesian network, and the conditional probability table is used to describe the task sending conditions, probabilities, and related tasks. The Bayesian network model is mainly used to process power grid data-driven model tasks with Bayesian network characteristics.

[0054] Markov network is a probability model based on undirected graph. After a random variable state sequence is expanded in chronological order, given the current state and all past states, the conditional probability distribution of its future state depends only on the current state. The power grid data model with Markov network characteristics is connected into a parallel or serial task undirected graph according to the Markov network, and the conditional probability table is used to describe the task sending conditions, probabilities, and related tasks. The Markov network model is mainly used to process power grid data-driven model tasks with Markov network characteristics.

[0055] Factoring a global function with multiple variables to obtain the product of several local functions, based on which a bidirectional graph is obtained is called a factor graph. In a factor graph, there are two types of vertices: random variables and factors. Edges are undirected, connecting factors and the random variables they contain. The power grid data model with the characteristics of the factor graph is connected into a parallel or serial task undirected graph according to the factor graph, and a conditional probability table is used to describe the task sending conditions, probabilities, and related tasks. The model in the factor graph is mainly used to process power grid data-driven model tasks with the characteristics of the factor graph.

[0056] Step 3: Use the interface learning framework to perform grid mechanism-data-driven hybrid modeling on the grid mechanism model obtained in step 1 and the grid data-driven model obtained in step 2.

[0057] The hybrid modeling of power grid mechanism-data driven mainly adopts the interface learning framework as the hybrid modeling method of mechanism-data driven. The power grid mechanism model reflects the physical entity in terms of physical mechanism and process; the data driven model uses input and output data to describe the physical process.

[0058] The interface learning framework is used to interact between the power grid mechanism model obtained in multiple steps 1 and the data-driven model obtained in step 2 through the power grid digital twin model interface to obtain a power grid mechanism-data-driven hybrid model, which solves the accuracy difference between the power grid mechanism model and the data-driven model.

[0059] The interaction is to connect the low-fidelity data-driven model describing the power grid with the high-fidelity power grid mechanism model to perform mechanism-data-driven hybrid modeling.

[0060] The interface learning framework uses a data-driven proxy model to represent a low-fidelity description of a part of the domain, physical field or scale, while a high-fidelity full-order model based on a mechanism model is used to handle the rest. The two models are coupled at the interface and connected through information exchange and matching. The coupling at the interface involves a variety of information exchange and matching methods, including boundary condition matching, data interpolation, error correction and parameter transfer.

[0061] Boundary condition matching: At the interface, the boundary conditions of different models need to match each other. This usually involves using the output of the proxy model (such as displacement, temperature, etc.) as the input boundary condition of the full-order model, and feeding the output of the full-order model (such as force, heat flow, etc.) back to the proxy model as a new boundary condition.

[0062] Data interpolation: Since the proxy model and the full-order model may have different resolutions in space or time, data interpolation may be required to ensure data consistency at the interface. This can be achieved through various interpolation methods, such as linear interpolation, spline interpolation, etc.

[0063] Error Correction: To improve the accuracy of the proxy model, the predictions of the proxy model can be error corrected at the interface. This can be done by comparing the exact solution of the full-order model with the predicted solution of the proxy model and calculating a correction term. The correction term is then applied to the predictions of the proxy model to improve its accuracy.

[0064] Parameter transfer: In some cases, the proxy model and the full-order model may need to share certain parameters or variables. These parameters or variables can be transferred at the interface to ensure that both models use the same information during the solution process.

[0065] The power grid digital twin model interface connects the low-fidelity data-driven model describing the power grid with the high-fidelity power grid mechanism model. The power grid digital twin model interface realizes interface error correction or extension by directly extending the interface, correcting and extending the interface, and other methods.

[0066] Directly extending the interface provides a data-driven model to predict interface variables, and adopts a prediction range to resolve the accuracy differences between the power grid mechanism model and the data-driven model.

[0067] The correction and extension interface uses the long-short-term memory network recurrent neural network method to control the difference between the prediction and the optimal value of the interface variables, thereby correcting the accuracy difference between the mechanism model and the data-driven model.

[0068] Step 4: Use the power grid mechanism-data-driven hybrid model obtained in step 3 as the power grid digital twin model, and combine the constraints of different models to realize the power grid digital twin model service.

[0069] The service process of the power grid digital twin model is as follows. First, the hierarchical relationship of the model needs to be constructed and the dependency order of the model needs to be clarified; second, appropriate constraints need to be added in the model service process, and there are certain differences in the constraint relationships that need to be paid attention to and added for models at different levels; finally, based on the constructed constraint relationships, model dependency order, and power grid digital twin model, the power grid digital twin model service is realized. The power grid digital twin model service realizes the process of digital twin model services from unit-level models to system-level models and then to complex system-level models.

[0070] In the model assembly process from parts to components to equipment, it is necessary to build and add constraint relationships such as angular model assembly constraints, contact constraints, offset constraints, etc. between parts. In the model assembly process from equipment to lines to systems, it is necessary to build and add spatial layout relationships and spatial constraint relationships between equipment. In the process from transformer monitoring to substation monitoring, it is necessary to build the spatial layout relationship of the transformer oil temperature and oil quality status data-driven model, the transformer physical model, and the substation physical model, as well as the spatial constraint relationship between different equipment in the substation.

[0071] Example 2

[0072] Based on the same concept as in Example 1, Example 2 of the present invention further provides a device for constructing a digital twin model of a power grid, such as Figure 2 As shown, it includes the following modules:

[0073] A data acquisition module 21 is used to acquire data for building a power grid digital twin model;

[0074] The model building module 22 obtains the digital twin model of the power grid based on the power grid digital twin model data by adopting the power grid information physical modeling, power grid data driven modeling, and power grid mechanism-data driven hybrid modeling methods;

[0075] The model service module 23 adopts the power grid digital twin model service method and the power grid digital twin model to provide a power grid digital twin model service;

[0076] Example 3

[0077] Based on the same concept, an electronic device according to an embodiment of the present invention is as follows: Figure 3 As shown, the electronic device may include a processor 31 and a memory 32, wherein the processor 31 and the memory 32 may be connected via a bus or other means. Figure 3 The example of connecting through bus is taken in the following.

[0078] The processor 31 may be a central processing unit (CPU). The processor 31 may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or a combination of the above chips.

[0079] The memory 32 is a non-transient computer-readable storage medium that can be used to store non-transient software programs, non-transient computer executable programs and modules. The processor 31 executes various functional applications and data processing of the processor by running the non-transient software programs, instructions and modules stored in the memory 32, that is, to implement the power grid business digitalization project data analysis method in the above embodiment.

[0080] The memory 32 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required by at least one function; the data storage area may store data created by the processor 31, etc. In addition, the memory 32 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some embodiments, the memory 32 may optionally include a memory remotely arranged relative to the processor 31, and these remote memories may be connected to the processor 31 via a network. Examples of the above-mentioned network include, but are not limited to, a power grid, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0081] The one or more modules are stored in the memory 32, and when executed by the processor 31, the power grid service digitalization project data analysis method in the embodiment shown in the accompanying drawings is executed.

[0082] The specific details of the above electronic device can be understood by referring to the corresponding related descriptions and effects in the embodiments shown in the accompanying drawings, and will not be repeated here.

[0083] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the storage medium can be a disk, an optical disk, a read-only memory (ROM), a random access memory (RAM), a flash memory, a hard disk drive (HDD) or a solid-state drive (SSD), etc.; the storage medium can also include a combination of the above-mentioned types of memory.

[0084] The beneficial effect of the present invention is that, compared with the prior art, the present invention adopts the grid information physical modeling method to realize the full-scale multi-physics field digital twin model of the equipment to achieve high-precision mapping between the virtual and real power physical objects; the probability graph model is used as a data-driven modeling method. The probability graph model is used to characterize the dependency relationship of the random variables of the model at the variable level, reflecting the probability structure of the problem and the difficulty of reasoning, and also provides an operable data structure for the reasoning algorithm; by adopting the grid data-driven modeling method, the accuracy of data-driven modeling can be effectively improved. The model construction method involved in this patent adopts grid information physical modeling, grid data-driven modeling, grid mechanism-data-driven hybrid modeling, and grid digital twin model services to form a multi-level optimization from initialization equipment modeling to hybrid modeling, and can construct a power digital twin analysis model to provide underlying model support for power digital twin operation simulation. It can realize accurate, effective and rapid construction of a grid digital twin model and improve model accuracy.

Claims

1. A method for constructing a digital twin model of a power grid, characterized in that: The method comprises: Step 1: Obtain the geometric model of the power grid equipment and combine it with the real scene data, integrate the physical parameters of the power grid and characterize the behavior and rules of the physical entity, perform power grid information physical modeling, and generate a power grid mechanism model; Step 2: Based on the observation data transmitted to the power grid equipment, the probabilistic graphical model is used as a data-driven modeling method to perform power grid data-driven modeling and generate a power grid data-driven model; Step 3: Use the interface learning framework to perform grid mechanism-data driven hybrid modeling on the grid mechanism model obtained in step 1 and the grid data driven model obtained in step 2; Step 4: Use the power grid mechanism-data-driven hybrid model obtained in step 3 as the power grid digital twin model, and combine the constraints of different models to realize the power grid digital twin model service.

2. A method for constructing a digital twin model of a power grid according to claim 1, characterized in that: In step 1, the specific process of power grid information-physical modeling is as follows: based on the appearance structure model of the power grid equipment, the physical parameters of the power grid are used to describe the equipment characteristics, and the geographical coordinate position and equipment assets are combined to realize the three-dimensional modeling of multiple device elements of a single device. Through the coupling connection relationship of the elements, the external disturbance response behavior, the internal operation rules, the energy flow trajectory and the information flow transmission relationship as well as the basic natural laws, the connection between the single element models is completed to form a dynamic virtual model set. The virtual model set is also associated, and it is superimposed block by block and layer by layer to form a digital twin full-space model.

3. A method for constructing a digital twin model of a power grid according to claim 2, characterized in that: A power grid physical parameter database is formed based on the physical parameters of the power grid. The power grid physical parameters include electrical and non-electrical quantity data collected by various types of sensors, and information collected by various types of equipment and devices. The information collected by various types of equipment and devices includes: status indications and measurement values.

4. A method for constructing a digital twin model of a power grid according to claim 1, characterized in that: The data-driven modeling of power grid adopts the probabilistic graphical model as the data-driven modeling method. It understands the status of equipment and power grid equipment through the incoming observation data, and then analyzes and predicts the information through the probabilistic graphical model to generate intelligent control information, and controls the power grid equipment by publishing the intelligent control information.

5. A method for constructing a digital twin model of a power grid according to claim 4, characterized in that: The power grid probabilistic graphical model includes Bayesian networks, Markov networks and factor graphs, which are used simultaneously to perform data-driven modeling on power grid data models with Bayesian network characteristics, power grid data models with Markov network characteristics and power grid data models with characteristics in factor graphs, and finally a data-driven model including Bayesian networks, Markov networks and factor graph probabilistic graphical models is obtained.

6. A method for constructing a digital twin model of a power grid according to claim 5, characterized in that: The power grid data-driven model with Bayesian network characteristics is connected into a directed acyclic graph according to the Bayesian network, and the conditional probability table is used to describe the task sending conditions, probabilities and related tasks.

7. A method for constructing a digital twin model of a power grid according to claim 5, characterized in that: The power grid data-driven model with Markov network characteristics is connected into an undirected graph according to the Markov network, and a conditional probability table is used to describe the task sending conditions, probabilities and related tasks.

8. A method for constructing a digital twin model of a power grid according to claim 5, characterized in that: The power grid data-driven model with factor graph characteristics is connected into an undirected graph according to the factor graph, and the conditional probability table is used to describe the task sending conditions, probabilities and related tasks.

9. A method for constructing a digital twin model of a power grid according to claim 1, characterized in that: The interface learning framework connects the power grid mechanism model and the data-driven model through the power grid digital twin model interface to obtain a power grid mechanism-data-driven hybrid model.

10. A method for constructing a digital twin model of a power grid according to claim 9, characterized in that: The interface of the power grid digital twin model is corrected and extended by directly extending the interface and correcting the extended interface.

11. A method for constructing a digital twin model of a power grid according to claim 10, characterized in that: Directly extend the interface to use the prediction range to solve the accuracy difference between the power grid mechanism model and the data-driven model; The correction extension interface uses a long short-term memory network recurrent neural network to control the difference between the predicted and optimal values ​​of the interface variables, and the accuracy difference between the correction mechanism model and the data-driven model.

12. A method for constructing a digital twin model of a power grid according to claim 1, characterized in that: The power grid digital twin model service process includes: building the hierarchical relationship of the model and clarifying the dependency order of the model; adding constraints in the model service process; and realizing the power grid digital twin model service based on the constructed constraint relationship, model dependency order and power grid digital twin model.

13. A power grid digital twin model construction device, used to run a power grid digital twin model construction method according to claims 1-12, characterized in that: include: A data acquisition module is used to acquire data for building a digital twin model of the power grid; The model building module, based on the power grid digital twin model data, adopts the power grid information physical modeling, power grid data-driven modeling, and power grid mechanism-data-driven hybrid modeling methods to obtain the power grid digital twin model; The model service module adopts the power grid digital twin model service method and the power grid digital twin model to provide power grid digital twin model services.

14. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the computer program is loaded into the processor, a method for constructing a digital twin model of a power grid is implemented according to any one of claims 1 to 12.

15. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, a method for constructing a digital twin model of a power grid is implemented according to any one of claims 1 to 12.

Images