A construction method of a digital power system and related devices

Abstract

The present application discloses a method for constructing a digital power system and related devices, which can be used in the field of power systems. In this method, first, digital characteristic production factors and multiple functional operators in the power system are obtained; then, based on the control logic of each of the multiple functional operators and the storage form of the characteristic production factors included, corresponding neural network models are configured for the multiple functional operators respectively; then, based on the corresponding relationship among the functional operators, the characteristic production factors, and the neural network models, and at least one of the functional operators, an intelligent agent is constructed; finally, based on the multiple intelligent agents, a digital power system is constructed. Thus, the functional operators are compressed into intelligent agents and a digital power system with intelligent agents as the core is constructed, and multiple types of intelligent agents are added to the digital power system, thereby realizing the digital transformation of the power system and enabling the scientific and convenient construction of a digital power system.

Description

Technical Field

[0001] This application relates to the technical field of power systems, and particularly to a method for constructing a digital power system and related devices. Background Art

[0002] With the advancement of the industrialization process and the progress and development of science and technology, the importance of accurately understanding and keenly controlling the power system has become increasingly prominent.

[0003] Currently, due to the diversity of power plant categories connected to the power grid and factors such as the uncertainty of power generation in new energy power plants such as photovoltaic and hydropower, the construction of the power system faces complex and changeable challenges, which also greatly affects the reliability and stability of power supply in the power system. At present, it is urgent to combine emerging fields such as new power and intelligent hydropower services to construct an intelligent and digital power system that contributes to the reform of the domestic power market and global power trading. Summary of the Invention

[0004] Based on the above problems, this application provides a method for constructing a digital power system and related devices, which can construct intelligent agents with rich functions through characteristic production factors, and construct a digital power system through intelligent agents with diverse functions.

[0005] The embodiments of this application disclose the following technical solutions:

[0006] In a first aspect, an embodiment of this application provides a method for constructing a digital power system, and the method includes:

[0007] Obtain digital characteristic production factors and multiple functional operators in the power system;

[0008] Based on the control logic of each of the multiple functional operators and the storage form of the characteristic production factors included, configure corresponding neural network models for the multiple functional operators respectively;

[0009] Based on the correspondence relationship among the functional operators, characteristic production factors, and neural network models, and at least one of the functional operators, construct an intelligent agent;

[0010] Based on the multiple intelligent agents, construct a digital power system.

[0011] Optionally, the constructing a digital power system based on the multiple intelligent agents includes:

[0012] Based on the association relationship among the multiple intelligent agents, construct a digital power system by nesting and calling among the multiple intelligent agents.

[0013] Optionally, obtaining digital characteristic production factors in the power system includes:

[0014] Obtain production factor data in the power system;

[0015] Based on hierarchical feature extraction, data alignment, and data normalization, perform feature extraction on the production factor data to obtain feature production factors.

[0016] Optionally, the obtaining of digitalized feature production factors and multiple functional operators in the power system includes:

[0017] Obtain digitalized feature production factors and multiple functional functions in the power system;

[0018] Based on the feature production factors, correct multiple functional functions through deep learning and reinforcement learning to obtain multiple functional operators centered on the functional functions.

[0019] Optionally, after obtaining multiple functional operators centered on the functional functions by correcting multiple functional functions through deep learning and reinforcement learning based on the feature production factors, the method further includes:

[0020] Update multiple functional operators through the backpropagation algorithm of the loss function.

[0021] Optionally, the feature production factors include at least one of grid feature production factors and power plant feature production factors.

[0022] Optionally, based on the control logics of multiple functional operators and the storage forms of the included feature production factors, configure corresponding neural network models for multiple functional operators, including:

[0023] If the control logic of the functional operator is information flow to the local domain, and the storage form of the feature production factors included in the functional operator is stored in a rule network, configure a convolutional neural network for the functional operator;

[0024] If the control logic of the functional operator is information sequence flow, and the storage form of the feature production factors included in the functional operator is data input in order, configure a recurrent neural network for the functional operator;

[0025] If the control logic of the functional operator is information flow along fixed edges, and the storage of the feature production factors included in the functional operator is in a fixed graph structure, configure a graph neural network for the functional operator;

[0026] If in the control logic of the functional operator, the information flow is dynamically controlled by a neural network, and the storage form of the feature production elements included in the functional operator is stored in an unordered set, then an attention neural network is configured for the functional operator.

[0027] In a second aspect, an embodiment of the present application provides a construction device for a digital power system, the device including: an acquisition module, a configuration module, an agent construction module, and a system construction module;

[0028] The acquisition module is used to acquire digital feature production elements and a plurality of functional operators in the power system;

[0029] The configuration module is used to respectively configure corresponding neural network models for the plurality of functional operators based on the control logic of each of the plurality of functional operators and the storage form of the feature production elements included;

[0030] The agent construction module is used to construct an agent based on the correspondence relationship between the functional operator, the feature production element, and the neural network model, and at least one of the functional operators;

[0031] The system construction module is used to construct a digital power system based on the plurality of agents.

[0032] In a third aspect, an embodiment of the present application provides a construction device for a digital power system, the device including: a memory and a processor;

[0033] The memory is used to store program code and transmit the program code to the processor;

[0034] The processor is used to execute the steps of the construction method of the digital power system according to any one of the embodiments in the first aspect according to the program code.

[0035] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, in which a computer program is stored. When the computer program runs on the construction device of the digital power system, the construction device of the digital power system executes the steps of the construction method of the digital power system according to any one of the embodiments in the first aspect.

[0036] In a fifth aspect, an embodiment of the present application provides a digital power system, and the digital power system is established by the construction method of the digital power system according to any one of the embodiments in the first aspect.

[0037] Compared with the prior art, the present application has the following beneficial effects:

[0038] The embodiment of the present application provides a method for constructing a digital power system. In this method, first, digital characteristic production factors and multiple functional operators in the power system are obtained; then, based on the control logic of each of the multiple functional operators and the storage form of the characteristic production factors included, corresponding neural network models are configured for the multiple functional operators respectively; then, based on the correspondence between the functional operators, the characteristic production factors, and the neural network models, and at least one of the functional operators, an intelligent agent is constructed; finally, based on the multiple intelligent agents, a digital power system is constructed. Thus, the functional operators are compressed into intelligent agents, and a digital power system with intelligent agents as the core is constructed. The intelligent agents corresponding to various different power function requirements are added to the digital power system, and each intelligent agent is used as a power function implementation unit in the digital power system, thereby realizing the digital transformation of the power system and scientifically and conveniently constructing a digital power system. Description of the Drawings

[0039] 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 required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0040] Figure 1 It is a flowchart of a method for constructing a digital power system provided by an embodiment of the present application;

[0041] Figure 2 It is a schematic diagram of constructing an operator through deep learning and reinforcement learning provided by an embodiment of the present application;

[0042] Figure 3 It is a schematic diagram of the relationship between a function and an operator provided by an embodiment of the present application;

[0043] Figure 4 It is a flowchart of constructing an intelligent agent based on data provided by an embodiment of the present application;

[0044] Figure 5 It is a structural diagram of data in the construction of a digital power system provided by an embodiment of the present application;

[0045] Figure 6 It is a schematic diagram of a device for constructing a digital power system provided by an embodiment of the present application;

[0046] Figure 7 It is a structural diagram of a device for constructing a digital power system provided by an embodiment of the present application. Detailed Embodiments

[0047] A construction method and related devices of a digital power system provided by this application can be used in the field of power systems. The above is only an example and does not limit the application field of a construction method and related devices of a digital power system provided by this application.

[0048] The terms "first", "second", "third", "fourth", etc. in the specification, claims and drawings description of this application are used to distinguish different objects, rather than to limit a specific order.

[0049] In the embodiments of this application, words such as "as an example" or "for example" are used to represent examples, illustrations or explanations. Any embodiment or design solution described as "as an example" or "for example" in the embodiments of this application should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Rather, the use of words such as "as an example" or "for example" is intended to present relevant concepts in a specific way.

[0050] The terms used in the implementation part of this application are only used to explain the specific embodiments of this application, rather than to limit this application.

[0051] The following introduces some inventive concept ideas involved in the technical solutions of this application:

[0052] Symbiosis: At the data level, after digitizing a class of analogous transactions, they have certain associated and connected properties under a certain abstraction. Symbiosis symbiotically transfers the production factors in nature or the information for producing and controlling production factors into data, making it possible for internal association and transmission in data. Symbiosis also enables data feature abstraction to exist in different operators or agents, and different operators and agents can coexist, and the characteristics also exist in multiple operators and agents. Essentially, different types of transactions may have the characteristics of being scattered in form but similar in essence. Finding the essence and connotation of symbiosis can find the same nature. In the technical solutions of this application, the inventor, based on the accumulated knowledge of power systems, uses symbiosis to symbiotically transfer the concepts in the neural network into the power system, realizing the ingenious application of artificial intelligence technology in the power field.

[0053] Migration: Transfer the characteristics of a class of transactions to the characteristics of another related class of transactions, so as to generate an association and enable the two classes of transactions to be interchangeable at the data asset level. By using the nesting and invocation of agents, the migration function can be realized, enabling multi-functional agents to require not too many iterations, and using operators to implement specific functions. Using migration enables most functions in the power system to be carried from one agent to another agent.

[0054] Emergence: When the data scale is large enough and there are enough logics within the intelligent body, the phenomenon that the data and logics can regenerate and iterate with each other. It is the demarcation point between intelligence and wisdom. When artificial intelligence develops to the emergence stage, wisdom will be generated. Recall that humans use tools to change productivity. Tools are production materials. First in ancient times, and then continuously expand the input data scale to continuously promote the development of productivity, invent production tools, and then summarize and abstract laws through disciplines such as mathematics, physics, and chemistry, and finally obtain logics and formula theorems; In modern society, the development of informatics has changed the carrier of the third industrial revolution of mankind, generating electricity, computers, etc. These are also tools for iterating information in informatics. However, in the fourth industrial revolution, artificial intelligence fundamentally upgrades computers, etc. from simple to complex. Using neural networks and statistical principles, it approaches the production function again, enabling it to directly drive from data in a simple and straightforward manner. In the next 10 years, the development of intelligent bodies and their internal logics is equivalent to the development of production tools in previous societies. Human understanding is always deepening step by step. Many phenomena that cannot be explained have internal connections and deeper theorems and concepts. However, the high-order abstraction of data exists and is easier to obtain. The key is to analyze the relevant logics within the intelligent body to be natural, more like waiting for a certain opportunity.

[0055] Intelligent body: An intelligent body is a unit with one or more functions. It makes the power system and artificial intelligence more closely combined. The intelligent body has a protective effect on the power system and the outside world, similar to a cell, but there are smaller kinetic units such as operators inside. The intelligent body has special data characteristics in different scenarios, which has a more profound significance for generating emergence functions and studying the digital nature of the power system. The intelligent body is a core component of artificial intelligence in the power industry, similar to the thermal control logic diagram or relay protection involving logic in the power industry during the third industrial revolution. For example, the logic inside the "five-prevention" of a switch station can be symbolically represented. The logic inside the intelligent body can also be symbolically represented, which is used to more accurately and deterministically represent the key logics and controls within the power system based on big data and large models. The general logic of the intelligent body conforms to the key theorems and common sense within the power system, conforms to the energy conversion formula and related rules. Its research objects are production characteristic elements and control production elements, such as illuminance, wind speed, water level, heat consumption, load rate, system inertia, frequency, etc. The interface of the intelligent body is directly connected to the input layer or output layer (output function) of deep learning and reinforcement learning, and is also connected to the intermediate layer when needed. In the technical solution of this application, the first half of the neural network abstracts the production elements in nature into calibrated transmitted data, enhancing the parallel computing ability, that is, the fast ability. The operator function and intelligent body in the second half of the output reflect the ability of mathematical abstraction and the physical laws in the industry, and can be combined and reconstructed, enabling the intelligent body to have the capabilities and functions of various industries.

[0056] To enable those skilled in the art to better understand the solution of this application, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of this application. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in this application without creative efforts belong to the scope of protection of this application.

[0057] Refer to Figure 1 , which is a flowchart of a method for constructing a digital power system provided by an embodiment of this application. The method includes:

[0058] S101: Obtain the digital characteristic production factors and multiple functional operators in the power system.

[0059] Specifically, the production factor data reflecting natural characteristics in the power system can be obtained first, such as production factor data such as illuminance, coal type and quality, and water level; then, for the large sample data in the production factor data, feature extraction can be performed through the method of hierarchical feature extraction to obtain characteristic production factors. For the small sample data in the production factor data, feature extraction can be performed through the methods of data alignment and data normalization to obtain characteristic production factor data.

[0060] In the embodiments of this application, the characteristic production factors include power plant characteristic production factors and power grid characteristic production factors. Among them, the power plant characteristic production factors can include, but are not limited to, characteristic production factors such as illuminance, coal type and quality, and water level; the power grid characteristic production factors can include, but are not limited to, the real part of the voltage of each node, the imaginary part of the voltage, the active power flowing into each node, the reactive power flowing into each node, the system inertia, the system frequency, the ratio of the system power change rate to the system frequency change rate, and other characteristic production factors.

[0061] In the embodiments of this application, the digital characteristics of the characteristic production factors at each level are extracted. For example, the digital characteristics of the low-order characteristic production factors are extracted on the power plant side, and the digital characteristics of the high-order control production factors are extracted on the power grid side, such as extracting the characteristics of the system inertia to the frequency change rate and the frequency change rate to the power change rate. For the low-order characteristic production factors and high-order characteristic production factors, different means are adopted for digital characteristic extraction. For example, for the digital characteristics of low-order production factors such as water level and photovoltaic illuminance, digital characteristics can be extracted by involving clustering or the Euclidean distance method, while the digital characteristics of high-order control production factors require modeling or more advanced mathematical tools to extract.

[0062] Functions in the data structure reflect electricity prices, active and reactive power, inertia and frequency, electricity-carbon, security and stability, and network combination disconnection. They are embodied through operators and are parallel computing function units that implement specific functions in a power system within an intelligent body, such as electricity prices, system inertia, and frequency. As long as it is a function, it can exist in the form of an operator. For example Figure 2 , this figure is a schematic diagram of constructing an operator through deep learning and reinforcement learning provided by an embodiment of the present application. Based on characteristic production factors, multiple functional functions are corrected through deep learning and reinforcement learning, and multiple functional operators are updated through the backpropagation algorithm of the loss function to obtain multiple functional operators with functional functions as the core.

[0063] Taking the power grid load and electricity price function computing power group as an example, the mapping function as the kernel of the electricity price operator can be expressed as: CERm = U * N / E(1 - AUX). CERm is the grid compensation coefficient for paying for 1 kWh of electricity. U represents the coal price, which is also the fuel cost subsidy, in dollars per ton. E represents the heat consumption rate of the unit for power generation, that is, the heat consumption per kWh of electricity of the unit, in kcal / kWh. N represents the specific heat index, in kcal / kWh. AUX represents the auxiliary power rate. This formula reflects the relationship between the grid compensation coefficient for paying for 1 kWh of electricity and characteristic production factors such as coal price, unit power generation heat consumption, specific heat index, and auxiliary efficiency.

[0064] Deep learning has specific algorithms or logics in extracting data features, making the data features more regular in subsequent extraction, facilitating the establishment of ultra-high-level mechanisms or data structures based on data cleaning using data characteristics; it keeps the information loss of data less during data transmission, while ensuring the data calculation and transmission speed. In reinforcement learning, through mechanisms such as rewards, the correctness of the learned logic is continuously verified and prioritized, so that the logic is generated through continuous comparison and calibration.

[0065] The encapsulation of data features can use binary trees or stacks, enabling the data features to be more appropriately reflected or easily shared. Among them, binary trees can be used to implement power grid disconnection and grid connection, and practical data tools are not limited to functions, but also include graph theory, etc.; the structure and order characteristics of the stack, where the last in is the first out, conform to the clearing logic of the power grid and can also achieve the dynamic orderliness of data. For operations on high-order production factors such as electricity price clearing, if high-speed calibration of data parallel computing is required, multiple stacks or hierarchical processing are needed, which increases the pressure of parallel computing but reduces the time of logical judgment, and is a typical case driven by data.

[0066] Figure 3This is a schematic diagram of the relationship between functions and operators provided by the embodiments of the present application. The production factors of the power grid are reflected in the operators and belong to the power grid. For example, in functional operators, there are electricity price operators, such as electricity price functions, power grid electricity price break-even points, and internal control operators, and the operators corresponding to electricity charges in the final function. The power grid constructs power grid electricity price operators and simultaneously establishes power grid electricity price break-even points according to the digital composition of different types of power plant units in the power plant (including their power plant electricity price break-even points) and in combination with the actual clearing function of the power grid. The following will describe Figure 4 some concepts involved in

[0067] 1. Input operator: Technical parameter operator (corresponding to the technical solutions in various embodiments, quantifying and digitalizing various production factors such as photovoltaic irradiance, water inflow and water level of hydropower, and coal type and quality of thermal power as input operators. For example, coal parameters participate in coal-electricity linkage). The operators in this part, and the constructed parameter operators come from the first-level production factor indicators in nature, and then are directly associated with subsequent power generation and electricity prices.

[0068] Starting from illuminance GTI, daily power generation, and PR daily efficiency, analyze. According to the basic calculation of photovoltaic power generation and the statistical daily efficiency, the electricity quantity parameters in this input operator change due to factors such as irradiance and the improved efficiency η0 of technology.

[0069] 2. Discrete quantity continuousization operator: Normalization operator, feature extraction operator (used to measure which mathematical tools are used for data feature vectors, and characterize the 9-dimensional volatility, trend, and variability characteristics of time series).

[0070] For the normalization operator, there are various methods for extracting data features. Specifically, there are range features, standard deviation features, coefficient of variation features, and geometric mean features for volatility; centroid features, correlation coefficient features, and median difference features for trend; and waveform entropy features and second-order central moment features for variability.

[0071] 3. Functional operator, an operator that reflects the specific functions of the power system, including power generation control operator, electricity price operator, internal control operator, special working condition operator, and safety operator. Construct an operator cluster to form a multi-functional intelligent agent, and the multi-intelligent agent constructs the core of the digital power system. It is the main object to be studied in the power profession. For various production factors, on the generator and power grid sides: active and reactive power, frequency, voltage, etc. are used as the initial research objects, and it is extended to the research of higher-order material flow and physical mechanisms.

[0072] 4. Related operators of the function

[0073] Functions are divided into input functions, transfer functions, electricity price functions, and final functions. The final function can include: unit water consumption, hydropower assets, cost per kilowatt-hour, and unit heat capacity index of the unit.

[0074] In the embodiments of the present application, the information of the production factors in nature is symbiotically integrated into data, making it possible for the data to be associated and transmitted internally. Symbiosis also enables abstract data features to exist in different operators or agents, and different operators and agents can coexist simultaneously.

[0075] S102: Based on the control logics of multiple functional operators and the storage forms of the characteristic production factors they contain, configure corresponding neural network models for the multiple functional operators respectively.

[0076] Mathematical tools can adopt one of convolutional neural networks, recurrent neural networks, graph neural networks, or attention networks according to the data and information flow characteristics of the object, etc. The characteristics of the above various neural networks are as follows: 1. Convolutional neural network: Regular network data, information flows to the local area; 2. Recurrent neural network: Data is input in sequence, information flows in a sequence; 3. Graph neural network: Data is in a fixed graph structure, information flows along fixed edges; 4. Attention network: Data is in an unordered set, and the information flow is dynamically controlled by the neural network. These neural networks target the logical core: operators and agents. In the process of constructing a digital power system, the data characteristics and information characteristics can be analyzed separately from the data feature extraction to the rules of information sequence flow (physical and equipment mechanisms), and the appropriate neural network models can be selected to make the algorithms of artificial intelligence highly compatible with the power system.

[0077] Specifically, if the control logic of the functional operator is that the information flows to the local area, and the storage form of the characteristic production factors contained in the functional operator is stored in a regular network, then configure a convolutional neural network for the functional operator; if the control logic of the functional operator is that the information flows in a sequence, and the storage form of the characteristic production factors contained in the functional operator is that the data is input in sequence, then configure a recurrent neural network for the functional operator; if the control logic of the functional operator is that the information flows along fixed edges, and the storage of the characteristic production factors contained in the functional operator is in a fixed graph structure, then configure a graph neural network for the functional operator; if in the control logic of the functional operator, the information flow is dynamically controlled by the neural network, and the storage form of the characteristic production factors contained in the functional operator is stored in an unordered set, then configure an attention neural network for the functional operator.

[0078] S103: Based on the correspondence relationship among the functional operator, the characteristic production factor, and the neural network model, and at least one functional operator, construct an agent.

[0079] Specifically, functional operators are the basic components for constructing agents; feature production factors involve elements that can extract key information from data, which are crucial for agents to understand the environment and make decisions; neural network models, as a computational model that mimics the structure and function of the human brain, can identify complex patterns and relationships by learning a large amount of data. Thus, with at least one functional operator as the core, combining the feature production factors used in the functional operator and the neural network model configured for the functional operator in step S102, an agent that realizes specific functions can be constructed.

[0080] When using mathematical tools, the mathematical tools can be determined specifically in combination with the data scale and sample size. The following Figure 4 introduces an example process for constructing an agent based on data in this application. As Figure 4 shown, this process includes:

[0081] (1) Collect production factor data.

[0082] (2) Determine whether the data scale belongs to small-sample data or large-sample data. If it belongs to large-sample data, step (3) is omitted. If it belongs to small-sample data, step (3) is executed.

[0083] (3) Data fitting, using mathematical tools or formulas such as data alignment or data normalization to regress the data to a certain convergence value.

[0084] (4) Data cleaning, deleting abnormal data in the production factor data after fitting or processing it in a preset manner.

[0085] (5) Extract data features from the production factors to form feature production factors.

[0086] (6) Agent construction, writing key processes or constraints according to the internal logic of the power industry, laying out deep learning and reinforcement learning interfaces for the constructed agent, receiving the data extracted from the data features, and further performing deep learning or reinforcement learning according to its features and classifications.

[0087] (7) Create an output layer output function for the results of deep learning or reinforcement learning, and correct the intermediate results of each layer through the backpropagation algorithm.

[0088] (8) Generalize the output function into control logic, bring it into the agent, and nest the agent and its key internal data elements into the control logic of the actuator to achieve data-driven.

[0089] S104: Construct a digital power system based on multiple agents.

[0090] Specifically, a digital power system can be constructed based on the association relationships among multiple agents by nesting and invoking among the multiple agents.

[0091] Migration can be achieved through the nesting and invocation among multiple agents. When the data volume, operators, and agents reach a sufficient scale, emergence occurs, enabling the construction of a digital power system centered on multi-functional agents. The multi-functional agents do not require too many iterations, but use operators to achieve specific functions. Migration enables the functions in the digital power system to be transferred from one agent to another.

[0092] From the above description, it can be seen that in this application, the construction process of the digital power system is as follows: (1) Co-evolve internal information from data into production factors; (2) Then, integrate into operators through data-driven to achieve a specific function; (3) Compress the operator group into agents to enable the agents to have multiple specific functions; (4) Through the mutual nesting and invocation of agents, migration is achieved; (5) When the data volume, operators, and agents reach a sufficient scale, emergence occurs, enabling the construction of a digital power system centered on multi-functional agents.

[0093] In the embodiment of this application, first, obtain the digital characteristic production factors and multiple functional operators in the power system; then, based on the control logics of the multiple functional operators and the storage forms of the included characteristic production factors, configure corresponding neural network models for the multiple functional operators respectively; then, based on the correspondence relationships among the functional operators, characteristic production factors, and neural network models, and at least one of the functional operators, construct agents; finally, based on the multiple agents, construct a digital power system. Thus, the functional operators are compressed into agents, and a digital power system centered on agents is constructed. The agents corresponding to various different power function requirements are added to the digital power system, and each agent is used as a power function implementation unit in the digital power system, thereby realizing the digital transformation of the power system and enabling the scientific and convenient construction of a digital power system.

[0094] Figure 5 This is the structural diagram of the data in the construction of the digital power system provided by the embodiment of this application, focusing on the data aspect. Combining Figure 5As shown in the figure, digital power includes four main components that progress step by step: the vectorization of power data, the vectorization of scheduling, the digitization of all production factors, and the digitization of the entire industrial chain. Specifically, in the process of extracting characteristic production factors, the vectorization of power data is realized by vectorizing safety data such as protection setting values, vectorizing production factor data, and vectorizing other data; the vectorization of scheduling is achieved by establishing scheduling algorithms such as power generation scheduling algorithms and power plant scheduling algorithms, and using artificial intelligence to process more complex scheduling requirements; the digitization of all production factors is realized by processing production factor data belonging to small sample data through mathematical tools such as data alignment and data normalization, making the production factor data continuous and normalized, functionalizing production and settlement numbers, and establishing functional functions such as electricity price functions and scheduling functions; from data elements, data vectors, data functionalization to electricity prices, and then to scheduling algorithms, all directly affect the digitization of the entire industrial chain.

[0095] See Figure 6 , which is a schematic diagram of a device for constructing a digital power system provided by an embodiment of the present application. The device includes: an acquisition module 601, a configuration module 602, an agent construction module 603, and a system construction module 604;

[0096] The acquisition module 601 is used to acquire digital characteristic production factors and multiple functional operators in the power system;

[0097] The configuration module 602 is used to configure corresponding neural network models for multiple functional operators respectively based on the control logic of each functional operator and the storage form of the included characteristic production factors;

[0098] The agent construction module 603 is used to construct an agent based on the corresponding relationship among the functional operators, characteristic production factors, and neural network models, and at least one functional operator;

[0099] The system construction module 604 is used to construct a digital power system based on multiple agents.

[0100] Thus, the functional operators are compressed into agents, and a digital power system with agents as the core is constructed. The agents corresponding to various different power function requirements are added to the digital power system, and each agent is used as a power function implementation unit in the digital power system, thereby realizing the digital transformation of the power system and scientifically and conveniently constructing a digital power system.

[0101] Optionally, the system construction module 604 is specifically used to construct a digital power system by nesting and calling multiple agents based on the association relationship among the multiple agents.

[0102] Optionally, the obtaining module 601 is specifically configured to: obtain the production factor data in the power system;

[0103] Based on hierarchical feature extraction, data alignment, and data normalization, extract features from the production factor data to obtain feature production factors.

[0104] Optionally, the obtaining module 601 is specifically configured to: obtain the digital feature production factors and multiple functional functions in the power system; based on the feature production factors, correct the multiple functional functions through deep learning and reinforcement learning to obtain multiple functional operators with the functional functions as the core.

[0105] Optionally, the obtaining module 601 is further configured to: update the multiple functional operators through the backpropagation algorithm of the loss function.

[0106] Optionally, the configuration module 602 is further configured to: if the control logic of the functional operator is information flowing to the local domain, and the storage form of the feature production factors included in the functional operator is stored in the rule network, configure a convolutional neural network for the functional operator; if the control logic of the functional operator is information sequence flow, and the storage form of the feature production factors included in the functional operator is data input in sequence, configure a recurrent neural network for the functional operator; if the control logic of the functional operator is information flowing along fixed edges, and the storage form of the feature production factors included in the functional operator is stored in a fixed graph structure, configure a graph neural network for the functional operator; if in the control logic of the functional operator, the information flow is dynamically controlled by a neural network, and the storage form of the feature production factors included in the functional operator is stored in an unordered set, configure an attention neural network for the functional operator.

[0107] See Figure 7 , which is a structural diagram of a device for constructing a digital power system provided by an embodiment of the present application. The device includes: a memory 701 and a processor 702.

[0108] The memory 701: is used to store program codes and transmit the program codes to the processor.

[0109] The processor 702: is used to execute the steps of the above method for constructing a digital power system according to the instructions in the program code.

[0110] In addition, the present application further provides a computer-readable storage medium, in which computer instructions are stored. When the computer instructions run on the device for constructing a digital power system, the device for constructing a digital power system executes the steps of the above method for constructing a digital power system.

[0111] In addition, the present application also provides a digital power system, which is established by the above-mentioned method for constructing a digital power system.

[0112] It should be noted that each embodiment in this specification is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other. Each embodiment focuses on the differences from other embodiments. In particular, for the embodiments of the device and system, since they are basically similar to the method embodiments, the description is relatively simple. For the relevant parts, reference can be made to the partial description of the method embodiments. The device and system embodiments described above are only illustrative. The units described as separate components may or may not be physically separated, and the components indicated as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative efforts.

[0113] As described above, it is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present application should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A construction method of a digital power system, characterized in that, The method includes: Obtaining digital characteristic production factors and multiple functional operators in the power system; specifically including: obtaining production factor data reflecting natural characteristics in the power system; extracting features from the production factor data of large samples through a hierarchical feature extraction method, and extracting features from the production factor data of small samples through data alignment and data normalization methods to obtain characteristic production factor data; obtaining multiple functional functions in the power system; based on the characteristic production factors, correcting the multiple functional functions through deep learning and reinforcement learning to obtain multiple functional operators with the functional functions as the core; Based on the control logic of each of the multiple functional operators and the storage form of the characteristic production factors included, configuring corresponding neural network models for the multiple functional operators respectively; specifically including: if the control logic of the functional operator is that the information flow is to the local domain and the storage form of the characteristic production factors included in the functional operator is stored in a rule network, then configuring a convolutional neural network for the functional operator; if the control logic of the functional operator is that the information sequence flows and the storage form of the characteristic production factors included in the functional operator is that the data is input in order, then configuring a recurrent neural network for the functional operator; if the control logic of the functional operator is that the information flows along fixed edges and the characteristic production factors included in the functional operator are stored in a fixed graph structure, then configuring a graph neural network for the functional operator; if in the control logic of the functional operator, the information flow is dynamically controlled by a neural network and the storage form of the characteristic production factors included in the functional operator is stored in an unordered set, then configuring an attention neural network for the functional operator; Based on the correspondence relationship among the functional operators, the characteristic production factors, and the neural network models, and at least one of the functional operators, constructing an agent; Based on the multiple agents, constructing a digital power system.

2. The method according to claim 1, wherein The constructing a digital power system based on the multiple agents includes: Based on the association relationship among the multiple agents, constructing a digital power system by performing nesting and calling among the multiple agents.

3. The method according to claim 1, characterized in that After obtaining multiple functional operators with the functional functions as the core by correcting the multiple functional functions through deep learning and reinforcement learning based on the characteristic production factors, the method further includes: Updating the multiple functional operators through the backpropagation algorithm of the loss function.

4. The method according to claim 1, characterized in that, The characteristic production factors include at least one of grid characteristic production factors and power plant characteristic production factors.

5. A construction device for a digital power system, characterized in that The device includes: an obtaining module, a configuration module, an agent construction module, and a system construction module; The obtaining module is used to obtain digital feature production factors and multiple functional operators in the power system; specifically including: obtaining production factor data reflecting natural features in the power system; extracting features from the production factor data of large samples through a hierarchical feature extraction method, and extracting features from the production factor data of small samples through data alignment and data normalization methods to obtain feature production factor data; obtaining multiple functional functions in the power system; based on the feature production factors, correcting the multiple functional functions through deep learning and reinforcement learning to obtain multiple functional operators with the functional functions as the core; The configuration module is used to configure corresponding neural network models for the multiple functional operators respectively based on the control logics of the multiple functional operators and the storage forms of the included feature production factors; specifically including: if the control logic of the functional operator is that the information flows to the local area, and the storage form of the feature production factors included in the functional operator is stored in the rule network, then configure a convolutional neural network for the functional operator; if the control logic of the functional operator is that the information sequence flows, and the storage form of the feature production factors included in the functional operator is that the data is input in sequence, then configure a recurrent neural network for the functional operator; if the control logic of the functional operator is that the information flows along the fixed edges, and the feature production factors included in the functional operator are stored in the fixed graph structure, then configure a graph neural network for the functional operator; if in the control logic of the functional operator, the information flow is dynamically controlled by the neural network, and the storage form of the feature production factors included in the functional operator is stored in the unordered set, then configure an attention neural network for the functional operator; The agent construction module is used to construct an agent based on the correspondence relationship among the functional operators, the feature production factors and the neural network models, and at least one of the functional operators; The system construction module is used to construct a digital power system based on the multiple agents; 6. A construction device for a digital power system, characterized in that, The device includes: a memory and a processor; The memory is used to store program codes and transmit the program codes to the processor; The processor is used to execute the steps of the construction method of the digital power system according to any one of claims 1-4 based on the program codes; 7. A computer-readable storage medium, characterized in that, A computer program is stored in the computer-readable storage medium. When the computer program runs on the construction device of the digital power system, the construction device of the digital power system executes the steps of the construction method of the digital power system according to any one of claims 1-4; 8. A digital power system, characterized in that, The digital power system is established by the construction method of the digital power system according to any one of claims 1-4.

Images